Physicists at MIT have developed a new way to probe inside an atom’s nucleus, using the atom’s own electrons as “messengers” within a molecule.

In a study appearing today in the journal Science, the physicists precisely measured the energy of electrons whizzing around a radium atom that had been paired with a fluoride atom to make a molecule of radium monofluoride. They used the environments within molecules as a sort of microscopic particle collider, which contained the radium atom’s electrons and encouraged them to briefly penetrate the atom’s nucleus.

Typically, experiments to probe the inside of atomic nuclei involve massive, kilometers-long facilities that accelerate beams of electrons to speeds fast enough to collide with and break apart nuclei. The team’s new molecule-based method offers a table-top alternative to directly probe the inside of an atom’s nucleus.

Within molecules of radium monofluoride, the team measured the energies of a radium atom’s electrons as they pinged around inside the molecule. They discerned a slight energy shift and determined that electrons must have briefly penetrated the radium atom’s nucleus and interacted with its contents. As the electrons winged back out, they retained this energy shift, providing a nuclear “message” that could be analyzed to sense the internal structure of the atom’s nucleus.

The team’s method offers a new way to measure the nuclear “magnetic distribution.” In a nucleus, each proton and neutron acts like a small magnet, and they align differently depending on how the nucleus’ protons and neutrons are spread out. The team plans to apply their method to precisely map this property of the radium nucleus for the first time. What they find could help to answer one of the biggest mysteries in cosmology: Why do we see much more matter than antimatter in the universe?

“Our results lay the groundwork for subsequent studies aiming to measure violations of fundamental symmetries at the nuclear level,” says study co-author Ronald Fernando Garcia Ruiz, who is the Thomas A. Franck Associate Professor of Physics at MIT. “This could provide answers to some of the most pressing questions in modern physics.”

The study’s MIT co-authors include Shane Wilkins, Silviu-Marian Udrescu, and Alex Brinson, along with collaborators from multiple institutions including the Collinear Resonance Ionization Spectroscopy Experiment (CRIS) at CERN in Switzerland, where the experiments were performed.

Molecular trap

According to scientists’ best understanding, there must have been almost equal amounts of matter and antimatter when the universe first came into existence. However, the overwhelming majority of what scientists can measure and observe in the universe is made from matter, whose building blocks are the protons and neutrons within atomic nuclei.

This observation is in stark contrast to what our best theory of nature, the Standard Model, predicts, and it is thought that additional sources of fundamental symmetry violation are required to explain the almost complete absence of antimatter in our universe. Such violations could be seen within the nuclei of certain atoms such as radium.

Unlike most atomic nuclei, which are spherical in shape, the radium atom’s nucleus has a more asymmetrical configuration, similar to a pear. Scientists predict that this pear shape could significantly enhance their ability to sense the violation of fundamental symmetries, to the extent that they may be potentially observable.

“The radium nucleus is predicted to be an amplifier of this symmetry breaking, because its nucleus is asymmetric in charge and mass, which is quite unusual,” says Garcia Ruiz, whose group has focused on developing methods to probe radium nuclei for signs of fundamental symmetry violation.

Peering inside the nucleus of a radium atom to investigate fundamental symmetries is an incredibly tricky exercise.

“Radium is naturally radioactive, with a short lifetime and we can currently only produce radium monofluoride molecules in tiny quantities,” says study lead author Shane Wilkins, a former postdoc at MIT. “We therefore need incredibly sensitive techniques to be able measure them.”

The team realized that by placing a radium atom in a molecule, they could contain and amplify the behavior of its electrons.

“When you put this radioactive atom inside of a molecule, the internal electric field that its electrons experience is orders of magnitude larger compared to the fields we can produce and apply in a lab,” explains Silviu-Marian Udrescu PhD ’24, a study co-author. “In a way, the molecule acts like a giant particle collider and gives us a better chance to probe the radium’s nucleus.”

Energy shift

In their new study, the team first paired radium atoms with fluoride atoms to create molecules of radium monofluoride. They found that in this molecule, the radium atom’s electrons were effectively squeezed, increasing the chance for electrons to interact with and briefly penetrate the radium nucleus.

The team then trapped and cooled the molecules and sent them through a system of vacuum chambers, into which they also sent lasers, which interacted with the molecules. In this way the researchers were able to precisely measure the energies of electrons inside each molecule.

When they tallied the energies, they found that the electrons appeared to have a slightly different energy compared to what physicists expect if they did not penetrate the nucleus. Although this energy shift was small — just a millionth of the energy of the laser photon used to excite the molecules — it gave unambiguous evidence of the molecules’ electrons interacting with the protons and neutrons inside the radium nucleus.

“There are many experiments measuring interactions between nuclei and electrons outside the nucleus, and we know what those interactions look like,” Wilkins explains. “When we went to measure these electron energies very precisely, it didn’t quite add up to what we expected assuming they interacted only outside of the nucleus. That told us the difference must be due to electron interactions inside the nucleus.”

“We now have proof that we can sample inside the nucleus,” Garcia Ruiz says. “It’s like being able to measure a battery’s electric field. People can measure its field outside, but to measure inside the battery is far more challenging. And that’s what we can do now.”

Going forward, the team plans to apply the new technique to map the distribution of forces inside the nucleus. Their experiments have so far involved radium nuclei that sit in random orientations inside each molecule at high temperature. Garcia Ruiz and his collaborators would like to be able to cool these molecules and control the orientations of their pear-shaped nuclei such that they can precisely map their contents and hunt for the violation of fundamental symmetries.

“Radium-containing molecules are predicted to be exceptionally sensitive systems in which to search for violations of the fundamental symmetries of nature,” Garcia Ruiz says. “We now have a way to carry out that search.”

This research was supported, in part, by the U.S. Department of Energy. 

If you think of a single atom as a grain of sand, then a wavelength of visible light — which is a thousand times larger than the atom’s width — is comparable to an ocean wave. The light wave can dwarf an atom, missing it entirely as it passes by. This gulf in size has long made it impossible for scientists to see and resolve individual atoms using optical microscopes alone.

Only recently have scientists found ways to break this “diffraction limit,” to see features that are smaller than the wavelength of light. With new techniques known as super-resolution microscopy, scientists can see down to the scale of a single molecule.

And yet, individual atoms have still been too small for optical microscopes — which are much simpler and less expensive than super-resolution techniques — to distinguish, until now.

In an open-access paper appearing today in Nature Communications, MIT scientists present a new computational method that enables optical microscopes to resolve individual atoms and zero in on their exact locations in a crystal structure.

The team’s new “discrete grid imaging technique,” or DIGIT, is a computational imaging approach that scientists can apply to optical data to calculate the most probable location of individual atoms based on a very important clue: the material’s known atomic configuration. As long as scientists have an idea of what a material’s physical atomic layout should be, they can use this layout as a sort of map to determine where specific atoms or features must be located.

“It’s like you know there’s a seating chart,” says lead author Yuqin “Sophia” Duan, a graduate student in MIT’s Department of Electrical Engineering and Computer Science (EECS). “Previous methods could tell you what section an atom is in. But now we can take this seating chart as prior knowledge, and can pinpoint exactly which seat the atom is in.”

With DIGIT, the team can now pinpoint individual atoms with a resolution of 0.178 angstroms. (One angstrom is one-tenth of a nanometer, which is less than half the width of a single atom). The technique enables optical microscopes to localize atomic-scale features in any material that has a known atomic pattern, such as crystalline materials or certain proteins with repeating molecular chains.

The team says the method could help guide the design of quantum devices, which often require placing individual atoms precisely within a crystal. Beyond quantum technologies, DIGIT can also provide new insights into how defects and impurities shape the behavior of advanced materials — from semiconductors to superconductors.

Duan’s co-authors at MIT are Qiushi Gu, Hanfeng Wang, Yong Hu, Kevin Chen, Matthew Trusheim, and EECS Professor Dirk Englund.

Grid support

Scientists can image features smaller than a nanometer, and sometimes as small as a single atom, but not with optical microscopes. In these cases, they use transmission or scanning electron microscopes, which send high-energy beams of electrons into a sample to generate an image based on the pattern in which the electrons scatter. These electron-based methods produce highly detailed, near-atomic-scale images, but they require imaging in a vacuum and at high energies, and only work in ultrathin, synthetic, or solid-state materials. Electron-based imaging methods are too harsh for more delicate living specimens.

In contrast, optical microscopes work at lower energies, in ambient conditions, and are safe to apply to biological samples. But they cannot discern features past the diffraction limit. Essentially, a microscope is unable to see features that are smaller than half the wavelength of visible light (about 200 to 300 nanometers) that a microscope sends in to probe a sample. Atoms, then, have long eluded optical microscopes.

In 2014, however, the Nobel Prize in Chemistry was awarded to developers of a technique to overcome the diffraction limit. Super-resolution microscopy works by shining laser light on a sample at a specific frequency that is known to resonate with a feature of interest, such as a certain molecule. When that molecule resonates, it effectively announces its presence in the material. With this optical manipulation, scientists can visualize features as small as 10 nanometers, on the scale of a single molecule.

Duan and Englund looked to resolve even smaller features by combining super-resolution techniques with statistical analysis and knowledge of materials that has often been overlooked.

“One thing that gets ignored in imaging optical systems is the physical configuration of your system,” Duan says. “For example, if you want to visualize defects in a diamond system, these defects can only be at certain positions, since they have to follow the grid of the atomic diamond structure. In proteins, there are some structures that grow in an organized grid, and their location must be somewhere along that physical grid.”

The researchers suspected that if they had a reasonably accurate map of a material’s atomic structure (imagine the ball-and-stick models of molecules in a chemistry classroom), they might use such maps as a template and try out many different orientations and rotation angles to find the closest match to whatever features are initially visualized using super-resolution microscopy.

“No one has ever done this before, to include the physical constraints or system information into the resolution technique,” Duan says.

Blurriness, collapsed

To test their idea, the researchers worked with a sample of diamond — a crystal whose microstructure is well-understood and resembles an organized grid, or lattice, of repeating carbon atoms. The researchers blindly knocked out some carbon atoms in the lattice and replaced them with silicon atoms using facilities at MIT.nano. Their goal was to identify and determine the precise locations of the errant silicon atoms.

To do so, they first used established techniques of super-resolution microscopy to probe the diamond sample, using lasers set to specific wavelengths at frequencies known to resonate with the silicon atoms but not the carbon atoms. With this technique, researchers produced images that depicted the silicon atoms, but only as a uniform blur.

The team then applied DIGIT to further resolve the picture. Knowing that diamond in general has a grid-like configuration of carbon atoms, the researchers took this configuration as a map, or seating chart of sorts, and assumed that any silicon atoms that took the place of a carbon atom must sit within the grid, which has a known spacing between atoms.

“Because the silicon atoms are substituting carbon atoms in the lattice, that means they must obey some integer multiple of the atomic spacing of the crystal lattice, separating any two silicon atoms,” Englund says. “That prior knowledge makes the localization different than if you add a purely amorphous material.”

The researchers essentially simulated many possibilities of orientations and rotation angles of the diamond lattice, superimposed on the blurry image of atoms that the super-resolution microscopy technique produced.

“The trick is that, in certain materials, atoms aren’t spread out randomly — they sit on a grid inside a crystal,” Duan explains. “We used that prior knowledge to sharpen the microscope’s picture. Once we factored in that ‘atomic grid,’ the blurriness collapsed, and we could pinpoint exact positions.”

In the end, they found the technique could pinpoint the location of individual silicon atoms within the diamond lattice, with a precision of 0.178 angstroms — the sharpest resolution of any optical-based imaging technique. The team has made the DIGIT code available on GitHub for anyone to apply to their optical measurements, provided their sample of interest has a well-understood atomic structure. Then, they hope that scientists will start to see much finer and detailed features and processes using light.

“It’s a big step — it takes optical microscopes into the realm of atomic scale, something people thought only electron microscopes or X-rays could do,” Duan says. “That opens up a whole new way of studying materials and biology.”

The degree to which someone trusts the information depicted in a chart can depend on their assumptions about who made the data visualization, according to a pair of studies by MIT researchers.

For instance, if someone infers that a graph about a controversial topic like gun violence was produced by an organization they feel is in opposition with their beliefs or political views, they may discredit the information or dismiss the visualization all together.

The researchers found that even the clearest visualizations often communicate more than the data they explicitly depict, and can elicit strong judgments from viewers about the social contexts, identities, and characteristics of those who made the chart.

Readers make these assessments about the social context of a visualization primarily from its design features, like the color palette or the way information is arranged, rather than the underlying data. Often, these inferences are unintended by the designers.

Qualitative and quantitative studies revealed that these social inferences aren’t restricted to certain subgroups, nor are they caused by limited data literacy.

The researchers consolidate their findings into a framework that scientists and communicators can use to think critically about how design choices might affect these social assumptions. Ultimately, they hope this work leads to better strategies for scientific communication.

“If you are scrolling through social media and you see a chart, and you immediately dismiss it as something an influencer has produced just to get attention, that shapes your entire experience with the chart before you even dig into the data. We’ve shown in these papers that visualizations do more than just communicate the data they are depicting — they also communicate other social signals,” says Arvind Satyanarayan, an associate professor in the MIT Department of Electrical Engineering and Computer Science (EECS) and member of the Computer Science and Artificial Intelligence Laboratory (CSAIL) and co-senior author of this research.

He is joined on the paper by co-lead authors Amy Rae Fox, a former CSAIL postdoc, and Michelle Morgenstern, a current postdoc in MIT’s anthropology program; and co-senior author Graham M. Jones, professor of anthropology. Two related papers on this research will be presented at the IEEE Visualization Conference.

Charts as social artifacts

During the height of the Covid-19 pandemic, social media was awash in charts from organizations like the World Health Organization and Centers for Disease Control and Prevention, which were designed to convey information about the spread of disease.

The MIT researchers studied how these visualizations were being used to discuss the pandemic. They found that some citizen scientists were using the underlying data to make visualizations of their own, challenging the findings of mainstream science.

“This was an unexpected discovery as, previously, citizen scientists were typically aligned with mainstream scientists. It took us a few years to figure out how to study this phenomenon more deeply,” Satyanarayan says.

Most research into data visualization studies how charts communicate data. Instead, the researchers wanted to explore visualizations from a social and linguistic perspective to assess the information they convey beyond the data.

Linguistic anthropologists have found that, while language allows people to communicate ideas, it also holds social meaning beyond the words people use. For instance, an accent or dialect can indicate that someone is part of a particular community.

By “pointing” to certain social meanings, identities, and characteristics, language serves what is known as a socio-indexical function.

“We wanted to see if things in the visual language of data communication might point to certain institutions, or the kinds of people in those institutions, that carry a meaning that could be unintended by the makers of the visualization,” Jones says.

To do this, the researchers conducted an initial, qualitative study of users on the social media platform Tumblr. During one-on-one interviews, the researchers showed users a variety of real visualizations from online sources, as well as modified visualizations where they removed the textual information, like titles and axes labels.

Stripping out the textual information was particularly important, since it mimics the way people often interact with online visualizations.

“Our engagement with social media is a few quick seconds. People aren’t taking the time to read the title of a chart or look at the data very carefully,” Satyanarayan says.

The interviews revealed that users made detailed inferences about the people or organizations who created the visualizations based on what they called “vibes,” design elements, like colors or the use of certain graphics. These inferences in turn impacted their trust in the data.

For instance, after seeing a chart with the flags of Georgia and Texas and a graph with two lines in red and black, but no text, one user said, “This kind of looks like something a Texas Republican (legislator) would put on Twitter or on their website, or as part of a campaign presentation.”

A quantitative approach

Building on this initial work, the researchers used the same methodology in three quantitative studies involving surveys sent to larger groups of people from a variety of backgrounds.

They found the same phenomenon: People make inferences about the social context of a visualization based on its design, which can lead to misunderstandings about, and mistrust in, the data it depicts.

For instance, users felt some visualizations were so neatly arranged they believed them to be advertisements, and therefore not trustworthy. In another example, one user dismissed a chart by a Pulitzer-prize winning designer because they felt the hand-drawn graphical style indicated it was made by “some female Instagram influencer who is just trying to look for attention.”

“If that is the first reaction someone has to a chart, it is going to massively impact the degree to which they trust it,” Satyanarayan says.

Moreover, when the researchers reintroduced text in the visualizations from which it had been removed, users still made these social inferences.

Typically, in data visualization, the solution to such a problem would be to create clearer charts or educate people about data literacy. But this research points to a completely different kind of data literacy, Jones says.

“It is not erroneous for people to be drawing these inferences. It requires a lot of cultural knowledge about where visualizations come from, how they are made, and how they circulate. Drawing these inferences is a feature, not a bug, of the way we use signs,” he says.

From these results, they created a classification framework to organize the social inferences users made and the design elements that contributed to them. They hope the typology serves as a tool designers can use to develop more effective visualizations, as well as a starting point for additional studies.

Moving forward, the researchers want to continue exploring the role of data visualizations as social artifacts, perhaps by drilling down on each design feature they identified in the typology. They also want to expand the scope of their study to include visualizations in research papers and scientific journals.

“Part of the value of this work is a methodological contribution to render a set of phenomena amenable to experimental study. But this work is also important because it showcases an interdisciplinary cross-pollination that is powerful and unique to MIT,” Jones says.

This work was supported, in part, by MIT METEOR and PFPFEE fellowships, an Amar G. Bose Fellowship, an Alfred P. Sloan Fellowship, and the National Science Foundation.

Nervous system functions, from motion to perception to cognition, depend on the active zones of neural circuit connections, or “synapses,” sending out the right amount of their chemical signals at the right times. By tracking how synaptic active zones form and mature in fruit flies, researchers at The Picower Institute for Learning and Memory at MIT have revealed a fundamental model for how neural activity during development builds properly working connections.

Understanding how that happens is important, not only for advancing fundamental knowledge about how nervous systems develop, but also because many disorders such as epilepsy, autism, or intellectual disability can arise from aberrations of synaptic transmission, says senior author Troy Littleton, the Menicon Professor in The Picower Institute and MIT’s Department of Biology. The new findings, funded in part by a 2021 grant from the National Institutes of Health, provide insights into how active zones develop the ability to send neurotransmitters across synapses to their circuit targets. It’s not instant or predestined, the study shows. It can take days to fully mature, and that is regulated by neural activity.

If scientists can fully understand the process, Littleton says, then they can develop molecular strategies to intervene to tweak synaptic transmission when it’s happening too much or too little in disease.

“We’d like to have the levers to push to make synapses stronger or weaker, that’s for sure,” Littleton says. “And so knowing the full range of levers we can tug on to potentially change output would be exciting.”

Littleton Lab research scientist Yuliya Akbergenova led the study published Oct. 14 in the Journal of Neuroscience.

How newborn synapses grow up

In the study, the researchers examined neurons that send the neurotransmitter glutamate across synapses to control muscles in the fly larvae. To study how the active zones in the animals matured, the scientists needed to keep track of their age. That hasn’t been possible before, but Akbergenova overcame the barrier by cleverly engineering the fluorescent protein mMaple, which changes its glow from green to red when zapped with 15 seconds of ultraviolet light, into a component of the glutamate receptors on the receiving side of the synapse. Then, whenever she wanted, she could shine light and all the synapses already formed before that time would glow red, and any new ones that formed subsequently would glow green.

With the ability to track each active zone’s birthday, the authors could then document how active zones developed their ability to increase output over the course of days after birth. The researchers actually watched as synapses were built over many hours by tagging each of eight kinds of proteins that make up an active zone. At first, the active zones couldn’t transmit anything. Then, as some essential early proteins accumulated, they could send out glutamate spontaneously, but not if evoked by electrical stimulation of their host neuron (simulating how that neuron might be signaled naturally in a circuit). Only after several more proteins arrived did active zones possess the mature structure for calcium ions to trigger the fusion of glutamate vesicles to the cell membrane for evoked release across the synapse.

Activity matters

Of course, construction does not go on forever. At some point, the fly larva stops building one synapse and then builds new ones further down the line as the neuronal axon expands to keep up with growing muscles. The researchers wondered whether neural activity had a role in driving that process of finishing up one active zone and moving on to build the next.

To find out, they employed two different interventions to block active zones from being able to release glutamate, thereby preventing synaptic activity. Notably, one of the methods they chose was blocking the action of a protein called Synaptotagmin 1. That’s important because mutations that disrupt the protein in humans are associated with severe intellectual disability and autism. Moreover, the researchers tailored the activity-blocking interventions to just one neuron in each larva because blocking activity in all their neurons would have proved lethal.

In neurons where the researchers blocked activity, they observed two consequences: the neurons stopped building new active zones and instead kept making existing active zones larger and larger. It was as if the neuron could tell the active zone wasn’t releasing glutamate and tried to make it work by giving it more protein material to work with. That effort came at the expense of starting construction on new active zones.

“I think that what it’s trying to do is compensate for the loss of activity,” Littleton says.

Testing indicated that the enlarged active zones the neurons built in hopes of restarting activity were functional (or would have been if the researchers weren’t artificially blocking them). This suggested that the way the neuron sensed that glutamate wasn’t being released was therefore likely to be a feedback signal from the muscle side of the synapse. To test that, the scientists knocked out a glutamate receptor component in the muscle, and when they did, they found that the neurons no longer made their active zones larger.

Littleton says the lab is already looking into the new questions the discoveries raise. In particular: What are the molecular pathways that initiate synapse formation in the first place, and what are the signals that tell an active zone it has finished growing? Finding those answers will bring researchers closer to understanding how to intervene when synaptic active zones aren’t developing properly.

In addition to Littleton and Akbergenova, the paper’s other authors are Jessica Matthias and Sofya Makeyeva.

In addition to the National Institutes of Health, The Freedom Together Foundation provided funding for the study.

When it comes to artificial intelligence, MIT and IBM were there at the beginning: laying foundational work and creating some of the first programs — AI predecessors — and theorizing how machine “intelligence” might come to be.

Today, collaborations like the MIT-IBM Watson AI Lab, which launched eight years ago, are continuing to deliver expertise for the promise of tomorrow’s AI technology. This is critical for industries and the labor force that stand to benefit, particularly in the short term: from $3-4 trillion of forecast global economic benefits and 80 percent productivity gains for knowledge workers and creative tasks, to significant incorporations of generative AI into business processes (80 percent) and software applications (70 percent) in the next three years.

While industry has seen a boom in notable models, chiefly in the past year, academia continues to drive the innovation, contributing most of the highly cited research. At the MIT-IBM Watson AI Lab, success takes the form of 54 patent disclosures, an excess of 128,000 citations with an h-index of 162, and more than 50 industry-driven use cases. Some of the lab’s many achievements include improved stent placement with AI imaging techniques, slashing computational overhead, shrinking models while maintaining performance, and modeling of interatomic potential for silicate chemistry.

“The lab is uniquely positioned to identify the ‘right’ problems to solve, setting us apart from other entities,” says Aude Oliva, lab MIT director and director of strategic industry engagement in the MIT Schwarzman College of Computing. “Further, the experience our students gain from working on these challenges for enterprise AI translates to their competitiveness in the job market and the promotion of a competitive industry.”

“The MIT-IBM Watson AI Lab has had tremendous impact by bringing together a rich set of collaborations between IBM and MIT’s researchers and students,” says Provost Anantha Chandrakasan, who is the lab’s MIT co-chair and the Vannevar Bush Professor of Electrical Engineering and Computer Science. “By supporting cross-cutting research at the intersection of AI and many other disciplines, the lab is advancing foundational work and accelerating the development of transformative solutions for our nation and the world.”

Long-horizon work

As AI continues to garner interest, many organizations struggle to channel the technology into meaningful outcomes. A 2024 Gartner study finds that, “at least 30% of generative AI projects will be abandoned after proof of concept by the end of 2025,” demonstrating ambition and widespread hunger for AI, but a lack of knowledge for how to develop and apply it to create immediate value.

Here, the lab shines, bridging research and deployment. The majority of the lab’s current-year research portfolio is aligned to use and develop new features, capacities, or products for IBM, the lab’s corporate members, or real-world applications. The last of these comprise large language models, AI hardware, and foundation models, including multi-modal, bio-medical, and geo-spatial ones. Inquiry-driven students and interns are invaluable in this pursuit, offering enthusiasm and new perspectives while accumulating domain knowledge to help derive and engineer advancements in the field, as well as opening up new frontiers for exploration with AI as a tool.

Findings from the AAAI 2025 Presidential panel on the Future of AI Research support the need for contributions from academia-industry collaborations like the lab in the AI arena: “Academics have a role to play in providing independent advice and interpretations of these results [from industry] and their consequences. The private sector focuses more on the short term, and universities and society more on a longer-term perspective.”

Bringing these strengths together, along with the push for open sourcing and open science, can spark innovation that neither could achieve alone. History shows that embracing these principles, and sharing code and making research accessible, has long-term benefits for both the sector and society. In line with IBM and MIT’s missions, the lab contributes technologies, findings, governance, and standards to the public sphere through this collaboration, thereby enhancing transparency, accelerating reproducibility, and ensuring trustworthy advances.

The lab was created to merge MIT’s deep research expertise with IBM’s industrial R&D capacity, aiming for breakthroughs in core AI methods and hardware, as well as new applications in areas like health care, chemistry, finance, cybersecurity, and robust planning and decision-making for business.

Bigger isn't always better

Today, large foundation models are giving way to smaller, more task-specific models yielding better performance. Contributions from lab members like Song Han, associate professor in the MIT Department of Electrical Engineering and Computer Science (EECS), and IBM Research’s Chuang Gan help make this possible, through work such as once-for-all and AWQ. Innovations such as these improve efficiency with better architectures, algorithm shrinking, and activation-aware weight quantization, letting models like language processing run on edge devices at faster speeds and reduced latency.

Consequently, foundation, vision, multimodal, and large language models have seen benefits, allowing for the lab research groups of Oliva, MIT EECS Associate Professor Yoon Kim, and IBM Research members Rameswar Panda, Yang Zhang, and Rogerio Feris to build on the work. This includes techniques to imbue models with external knowledge and the development of linear attention transformer methods for higher throughput, compared to other state-of-the-art systems. 

Understanding and reasoning in vision and multimodal systems has also seen a boon. Works like “Task2Sim” and “AdaFuse” demonstrate improved vision model performance if pre-training takes place on synthetic data, and how video action recognition can be boosted by fusing channels from past and current feature maps.

As part of a commitment to leaner AI, the lab teams of Gregory Wornell, the MIT EECS Sumitomo Electric Industries Professor in Engineering, IBM Research’s Chuang Gan, and David Cox, VP for foundational AI at IBM Research and the lab’s IBM director, have shown that model adaptability and data efficiency can go hand in hand. Two approaches, EvoScale and Chain-of-Action-Thought reasoning (COAT), enable language models to make the most of limited data and computation by improving on prior generation attempts through structured iteration, narrowing in on a better response. COAT uses a meta-action framework and reinforcement learning to tackle reasoning-intensive tasks via self-correction, while EvoScale brings a similar philosophy to code generation, evolving high-quality candidate solutions. These techniques help to enable resource-conscious, targeted, real-world deployment.

“The impact of MIT-IBM research on our large language model development efforts cannot be overstated,” says Cox. “We’re seeing that smaller, more specialized models and tools are having an outsized impact, especially when they are combined. Innovations from the MIT-IBM Watson AI Lab help shape these technical directions and influence the strategy we are taking in the market through platforms like watsonx.”

For example, numerous lab projects have contributed features, capabilities, and uses to IBM’s Granite Vision, which provides impressive computer vision designed for document understanding, despite its compact size. This comes at a time when there’s a growing need for extraction, interpretation, and trustworthy summarization of information and data contained in long formats for enterprise purposes.

Other achievements that extend beyond direct research on AI and across disciplines are not only beneficial, but necessary for advancing the technology and lifting up society, concludes the 2025 AAAI panel.

Work from the lab’s Caroline Uhler and Devavrat Shah — both Andrew (1956) and Erna Viterbi Professors in EECS and the Institute for Data, Systems, and Society (IDSS) — along with IBM Research’s Kristjan Greenewald, transcends specializations. They are developing causal discovery methods to uncover how interventions affect outcomes, and identify which ones achieve desired results. The studies include developing a framework that can both elucidate how “treatments” for different sub-populations may play out, like on an ecommerce platform or mobility restrictions on morbidity outcomes. Findings from this body of work could influence the fields of marketing and medicine to education and risk management.

“Advances in AI and other areas of computing are influencing how people formulate and tackle challenges in nearly every discipline. At the MIT-IBM Watson AI Lab, researchers recognize this cross-cutting nature of their work and its impact, interrogating problems from multiple viewpoints and bringing real-world problems from industry, in order to develop novel solutions,” says Dan Huttenlocher, MIT lab co-chair, dean of the MIT Schwarzman College of Computing, and the Henry Ellis Warren (1894) Professor of Electrical Engineering and Computer Science.

A significant piece of what makes this research ecosystem thrive is the steady influx of student talent and their contributions through MIT’s Undergraduate Research Opportunities Program (UROP), MIT EECS 6A Program, and the new MIT-IBM Watson AI Lab Internship Program. Altogether, more than 70 young researchers have not only accelerated their technical skill development, but, through guidance and support by the lab’s mentors, gained knowledge in AI domains to become emerging practitioners themselves. This is why the lab continually seeks to identify promising students at all stages in their exploration of AI’s potential.

“In order to unlock the full economic and societal potential of AI, we need to foster ‘useful and efficient intelligence,’” says Sriram Raghavan, IBM Research VP for AI and IBM chair of the lab. “To translate AI promise into progress, it’s crucial that we continue to focus on innovations to develop efficient, optimized, and fit-for-purpose models that can easily be adapted to specific domains and use cases. Academic-industry collaborations, such as the MIT-IBM Watson AI Lab, help drive the breakthroughs that make this possible.”

Since launching in 2008, the MIT Global Seed Funds (GSF) program has awarded roughly $30 million to more than 1,300 high-impact faculty research projects across the world, spurring consequential collaborations on topics that include swine-fever vaccines, deforestation of the Amazon, the impact of “coral mucus” on the Japanese island of Okinawa, and the creation of an AI-driven STEM-education lab within Nigeria’s oldest university.

Administered by the MIT Center for International Studies (CIS) and open to MIT faculty and principal investigators, GSF boasts a unique funding structure consisting of both a general fund for unrestricted geographical use and more than 20 different specific funds for individual universities, regions, and countries.

GSF projects often tackle critical challenges that require international solutions, culminating in patents, policy changes, and published papers in journals such as Nature and Science. Some faculty-led projects from this year include Professor Hugh Herr’s modular crutches for people with disabilities in Sierra Leone, Research Scientist Paolo Santi’s large-language models to predict energy consumption in grocery stores, and Professor Ernest Fraenkel’s development of mRNA therapies for the neurodegenerative disease amyotrophic lateral sclerosis (ALS).

GSF Assistant Director Justin Leahey, who is managing director of the MIT-Germany and MIT-Switzerland programs, says that GSF has expanded exponentially over the years, including most recently into the Czech Republic, Norway, Slovakia, and — starting in fall 2025 — Hungary. This year there were a grand total of roughly 300 research proposals submitted for consideration, with many of the accepted proposals including the active participation of students at both the graduate and undergraduate level.

Central to GSF’s work is “reciprocal exchange” — the concept of collaborators in and out of MIT sharing their work and exchanging ideas in an egalitarian way, rather than bringing a one-sided approach to different research challenges. Frequent collaborator Raffaella Gozzelino, a neurology researcher and principal investigator at NOVA Medical School in Portugal who works closely with Jacquin Niles, an MIT professor of biological engineering, says that research is more impactful “when specialized knowledge integrates local realities and reveals potential solutions to national challenges,” and views the spirit of reciprocal exchange as something that revolves around “sharing knowledge and co-creating solutions that empower one another and build bridges across borders.”

For Cindy Xie ’24, MCP ’25, her master’s thesis emerged from the first-ever GSF-supported research internship in Cape Verde, where she worked with Niles and Gozzelino to explore the impact of climate change on anemia in the country of 500,000 people, focusing specifically on its largest island of Santiago. Xie says that she was struck by the intertwined intersectional nature of the issues of nutrition, climate, and infection in Santiago, home to the nation’s capital city of Praia. For example, Xie and Gozzelino’s team found that respondents perceived a rise in costs of fresh produce over time, exacerbated by drought and unpredictable agricultural conditions, which in turn impacted existing nutritional deficiencies and increased residents’ susceptibility to mosquito-borne diseases.

“Though this multidisciplinary research lens is challenging in terms of actual project implementation, it was meaningful in that it generated insights and connections across fields that allow our research to be better contextualized within the experiences of the communities that it impacts,” Xie says.

Gozzelino says that it has been meaningful to witness how scientific research can transcend academic boundaries and generate real impact. She says that, by examining the effects of climate change on infectious diseases and nutrition in Cape Verde, the team will be able to build a framework that can directly inform public policy.

“Contributing to a project that underscores the importance of integrating scientific knowledge into decision-making will safeguard vulnerable populations and make them feel included in the society they belong,” Gozzelino says. “This collaboration has revealed the enormous potential of international partnerships to strengthen local research capacity and address global challenges.”

During her time in Cape Verde working with Xie and Gozzelino, Amulya Aluru ’23, MEng ’24 got to meet with 20 local officials and connect with new people in a wide range of roles across the country, helping her “recognize the power of interpersonal relationships and collaboration” in public health research. She says that the structure of the GSF grant gave her the unique experience of having mentors and coworkers in three different countries, spanning Cape Verde, the United States, and Portugal.

Aluru says that this kind of cross-pollination “enabled me to strengthen my research with different perspectives and challenged me to approach my work in a way that I’d never done before, with a more global mindset.”

Xie similarly expresses her deep appreciation for the long-term relationships she has built through the project and the linkages between Santiago and Boston, which itself is home to one of the world’s largest Cape Verdean diasporas. “As a student, this was a valuable experience to inform the approaches to collaboration that I would like to implement in my own future work,” Xie says.

More broadly, Gozzelino sees GSF grants like the Cape Verde one as being not simply a vehicle for financial support, but “a catalyst for turning partnerships into long-term impactful collaborations, demonstrating how global networks can aid the development of human capital.”

GSF’s long history of reaching across departments and borders has led to multiple meaningful academic collaborations that have since come to span continents — and decades. In 2015, Professor Jörn Dunkel — an applied mathematician at MIT — kicked off work on a data-sharing repository for bacterial biofilms with the interdisciplinary German microbiologist Knut Drescher, then a professor of biophysics at Philipps-Universität Marburg in Germany. Dunkel and Drescher have since co-authored more than 15 papers together in publications like Nature Physics and Science Advances alongside their teams of graduate students and postdocs, even with Drescher having moved locations and crossed country lines to Switzerland as a faculty member at the University of Basel’s Biozentrum Center for Molecular Life Sciences.

“Our collaboration often creates great synergy by combining my team’s experiments with the theory from Jörn’s team,” says Drescher. “It is a great joy to see his perspective on the experimental systems we are working on. He is able to really understand and engage with experimental biological data, identifying patterns in seemingly distant biological systems.”

In explaining the CIS initiative’s success, Leahey points to the synergistic, academically eclectic, cross-disciplinary nature of the program. “[GSF] is a research fund that doesn’t ‘fund research’ in the conventional sense,” he says. “It seeds early-stage collaboration and lets people explore.”

The MIT Global Seed Funds applications are now open, with a deadline of Dec. 16.

It’s hard to keep up with the ever-changing trends of the fashion world. What’s “in” one minute is often out of style the next season, potentially causing you to re-evaluate your wardrobe.

Staying current with the latest fashion styles can be wasteful and expensive, though. Roughly 92 million tons of textile waste are produced annually, including the clothes we discard when they go out of style or no longer fit. But what if we could simply reassemble our clothes into whatever outfits we wanted, adapting to trends and the ways our bodies change?

A team of researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and Adobe are attempting to bring eco-friendly, versatile garments to life. Their new “Refashion” software system breaks down fashion design into modules — essentially, smaller building blocks — by allowing users to draw, plan, and visualize each element of a clothing item. The tool turns fashion ideas into a blueprint that outlines how to assemble each component into reconfigurable clothing, such as a pair of pants that can be transformed into a dress.

With Refashion, users simply draw shapes and place them together to develop an outline for adaptable fashion pieces. It’s a visual diagram that shows how to cut garments, providing a straightforward way to design things like a shirt with an attachable hood for rainy days. One could also create a skirt that can then be reconfigured into a dress for a formal dinner, or maternity wear that fits during different stages of pregnancy.

“We wanted to create garments that consider reuse from the start,” says Rebecca Lin, MIT Department of Electrical Engineering and Computer Science (EECS) PhD student, CSAIL and Media Lab researcher, and lead author on a paper presenting the project. “Most clothes you buy today are static, and are discarded when you no longer want them. Refashion instead makes the most of our garments by helping us design items that can be easily resized, repaired, or restyled into different outfits.”

Modules à la mode

The researchers conducted a preliminary user study where both designers and novices explored Refashion and were able to create garment prototypes. Participants assembled pieces such as an asymmetric top that could be extended into a jumpsuit, or remade into a formal dress, often within 30 minutes. These results suggest that Refashion has the potential to make prototyping garments more approachable and efficient. But what features might contribute to this ease of use?

Its interface first presents a simple grid in its “Pattern Editor” mode, where users can connect dots to outline the boundaries of a clothing item. It’s essentially drawing rectangular panels and specifying how different modules will connect to each other.

Users can customize the shape of each component, create a straight design for garments (which might be useful for less form-fitting items, like chinos) or perhaps tinkering with one of Refashion’s templates. A user can edit pre-designed blueprints for things like a T-shirt, fitted blouse, or trousers.

Another, more creative route is to change the design of individual modules. One can choose the “pleat” feature to fold a garment over itself, similar to an accordion, for starters. It’s a useful way to design something like a maxi dress. The “gather” option adds an artsy flourish, where a garment is crumpled together to create puffy skirts or sleeves. A user might even go with the “dart” module, which removes a triangular piece from the fabric. It allows for shaping a garment at the waist (perhaps for a pencil skirt) or tailor to the upper body (fitted shirts, for instance).

While it might seem that each of these components needs to be sewn together, Refashion enables users to connect garments through more flexible, efficient means. Edges can be seamed together via double-sided connectors such as metal snaps (like the buttons used to close a denim jacket) or Velcro dots. A user could also fasten them in pins called brads, which have a pointed side that they stick through a hole and split into two “legs” to attach to another surface; it’s a handy way to secure, say, a picture on a poster board. Both connective methods make it easy to reconfigure modules, should they be damaged or a “fit check” calls for a new look.

As a user designs their clothing piece, the system automatically creates a simplified diagram of how it can be assembled. The pattern is divided into numbered blocks, which is dragged onto different parts of a 2D mannequin to specify the position of each component. The user can then simulate how their sustainable clothing will look on 3D models of a range of body types (one can also upload a model).

Finally, a digital blueprint for sustainable clothing can extend, shorten, or combine with other pieces. Thanks to Refashion, a new piece could be emblematic of a potential shift in fashion: Instead of buying new clothes every time we want a new outfit, we can simply reconfigure existing ones. Yesterday’s scarf could be today’s hat, and today’s T-shirt could be tomorrow’s jacket.

“Rebecca’s work is at an exciting intersection between computation and art, craft, and design,” says MIT EECS professor and CSAIL principal investigator Erik Demaine, who advises Lin. “I’m excited to see how Refashion can make custom fashion design accessible to the wearer, while also making clothes more reusable and sustainable.”

Constant change

While Refashion presents a greener vision for the future of fashion, the researchers note that they’re actively improving the system. They intend to revise the interface to support more durable items, stepping beyond standard prototyping fabrics. Refashion may soon support other modules, like curved panels, as well. The CSAIL-Adobe team may also evaluate whether their system can use as few materials as possible to minimize waste, and whether it can help “remix” old store-bought outfits.

Lin also plans to develop new computational tools that help designers create unique, personalized outfits using colors and textures. She’s exploring how to design clothing by patchwork — essentially, cutting out small pieces from materials like decorative fabrics, recycled denim, and crochet blocks and assembling them into a larger item.

“This is a great example of how computer-aided design can also be key in supporting more sustainable practices in the fashion industry,” says Adrien Bousseau, a senior researcher at Inria Centre at Université Côte d'Azur who wasn’t involved in the paper. “By promoting garment alteration from the ground up, they developed a novel design interface and accompanying optimization algorithm that helps designers create garments that can undergo a longer lifetime through reconfiguration. While sustainability often imposes additional constraints on industrial production, I am confident that research like the one by Lin and her colleagues will empower designers in innovating despite these constraints.”

Lin wrote the paper with Adobe Research scientists Michal Lukáč and Mackenzie Leake, who is the paper’s senior author and a former CSAIL postdoc. Their work was supported, in part, by the MIT Morningside Academy for Design, an MIT MAKE Design-2-Making Mini-Grant, and the Natural Sciences and Engineering Research Council of Canada. The researchers presented their work recently at the ACM Symposium on User Interface Software and Technology.

Before cells can divide, they first need to replicate all of their chromosomes, so that each of the daughter cells can receive a full set of genetic material. Until now, scientists had believed that as division occurs, the genome loses the distinctive 3D internal structure that it typically forms.

Once division is complete, it was thought, the genome gradually regains that complex, globular structure, which plays an essential role in controlling which genes are turned on in a given cell.

However, a new study from MIT shows that in fact, this picture is not fully accurate. Using a higher-resolution genome mapping technique, the research team discovered that small 3D loops connecting regulatory elements and genes persist in the genome during cell division, or mitosis.

“This study really helps to clarify how we should think about mitosis. In the past, mitosis was thought of as a blank slate, with no transcription and no structure related to gene activity. And we now know that that’s not quite the case,” says Anders Sejr Hansen, an associate professor of biological engineering at MIT. “What we see is that there’s always structure. It never goes away.”

The researchers also discovered that these regulatory loops appear to strengthen when chromosomes become more compact in preparation for cell division. This compaction brings genetic regulatory elements closer together and encourages them to stick together. This may help cells “remember” interactions present in one cell cycle and carry it to the next one.

“The findings help to bridge the structure of the genome to its function in managing how genes are turned on and off, which has been an outstanding challenge in the field for decades,” says Viraat Goel PhD ’25, the lead author of the study.

Hansen and Edward Banigan, a research scientist in MIT’s Institute for Medical Engineering and Science, are the senior authors of the paper, which appears today in Nature Structural and Molecular Biology. Leonid Mirny, a professor in MIT’s Institute for Medical Engineering and Science and the Department of Physics, and Gerd Blobel, a professor at the Perelman School of Medicine at the University of Pennsylvania, are also authors of the study.

A surprising finding

Over the past 20 years, scientists have discovered that inside the cell nucleus, DNA organizes itself into 3D loops. While many loops enable interactions between genes and regulatory regions that may be millions of base pairs away from each other, others are formed during cell division to compact chromosomes. Much of the mapping of these 3D structures has been done using a technique called Hi-C, originally developed by a team that included MIT researchers and was led by Job Dekker at the University of Massachusetts Chan Medical School. To perform Hi-C, researchers use enzymes to chop the genome into many small pieces and biochemically link pieces that are near each other in 3D space within the cell’s nucleus. They then determine the identities of the interacting pieces by sequencing them.

However, that technique doesn’t have high enough resolution to pick out all specific interactions between genes and regulatory elements such as enhancers. Enhancers are short sequences of DNA that can help to activate the transcription of a gene by binding to the gene’s promoter — the site where transcription begins.

In 2023, Hansen and others developed a new technique that allows them to analyze 3D genome structures with 100 to 1,000 times greater resolution than was previously possible. This technique, known as Region-Capture Micro-C (RC-MC), uses a different enzyme that cuts the genome into small fragments of similar size. It also focuses on a smaller segment of the genome, allowing for high-resolution 3-D mapping of a targeted genome region.

Using this technique, the researchers were able to identify a new kind of genome structure that hadn’t been seen before, which they called “microcompartments.” These are tiny highly connected loops that form when enhancers and promoters located near each other stick together.

In that paper, experiments revealed that these loops were not formed by the same mechanisms that form other genome structures, but the researchers were unable to determine exactly how they do form. In hopes of answering that question, the team set out to study cells as they undergo cell division. During mitosis, chromosomes become much more compact, so that they can be duplicated, sorted, and divvied up between two daughter cells. As this happens, larger genome structures called A/B compartments and topologically associating domains (TADs) disappear completely.

The researchers believed that the microcompartments they had discovered would also disappear during mitosis. By tracking cells through the entire cell division process, they hoped to learn how the microcompartments appear after mitosis is completed.

“During mitosis, it has been thought that almost all gene transcription is shut off. And before our paper, it was also thought that all 3D structure related to gene regulation was lost and replaced by compaction. It’s a complete reset every cell cycle,” Hansen says.

However, to their surprise, the researchers found that microcompartments could still be seen during mitosis, and in fact they become more prominent as the cell goes through cell division.

“We went into this study thinking, well, the one thing we know for sure is that there’s no regulatory structure in mitosis, and then we accidentally found structure in mitosis,” Hansen says.

Using their technique, the researchers also confirmed that larger structures such as A/B compartments and TADs do disappear during mitosis, as had been seen before.

“This study leverages the unprecedented genomic resolution of the RC-MC assay to reveal new and surprising aspects of mitotic chromatin organization, which we have overlooked in the past using traditional 3C-based assays. The authors reveal that, contrary to the well-described dramatic loss of TADs and compartmentalization during mitosis, fine-scale “microcompartments” — nested interactions between active regulatory elements — are maintained or even transiently strengthened,” says Effie Apostolou, an associate professor of molecular biology in medicine at Weill Cornell Medicine, who was not involved in the study.

A spike in transcription

The findings may offer an explanation for a spike in gene transcription that usually occurs near the end of mitosis, the researchers say. Since the 1960s, it had been thought that transcription ceased completely during mitosis, but in 2016 and 2017, a few studies showed that cells undergo a brief spike of transcription, which is quickly suppressed until the cell finishes dividing.

In their new study, the MIT team found that during mitosis, microcompartments are more likely to be found near the genes that spike during cell division. They also discovered that these loops appear to form as a result of the genome compaction that occurs during mitosis. This compaction brings enhancers and promoters closer together, allowing them to stick together to form microcompartments.

Once formed, the loops that constitute microcompartments may activate gene transcription somewhat by accident, which is then shut off by the cell. When the cell finishes dividing, entering a state known as G1, many of these small loops become weaker or disappear.

“It almost seems like this transcriptional spiking in mitosis is an undesirable accident that arises from generating a uniquely favorable environment for microcompartments to form during mitosis,” Hansen says. “Then, the cell quickly prunes and filters many of those loops out when it enters G1.”

Because chromosome compaction can also be influenced by a cell’s size and shape, the researchers are now exploring how variations in those features affect the structure of the genome and in turn, gene regulation.

“We are thinking about some natural biological settings where cells change shape and size, and whether we can perhaps explain some 3D genome changes that previously lack an explanation,” Hansen says. “Another key question is how does the cell then pick what are the microcompartments to keep and what are the microcompartments to remove when you enter G1, to ensure fidelity of gene expression?”

The research was funded in part by the National Institutes of Health, a National Science Foundation CAREER Award, the Gene Regulation Observatory of the Broad Institute, a Pew-Steward Scholar Award for Cancer Research, the Mathers Foundation, the MIT Westaway Fund, the Bridge Project of the Koch Institute and Dana-Farber/Harvard Cancer Center, and the Koch Institute Support (core) Grant from the National Cancer Institute.

Say a person takes their French Bulldog, Bowser, to the dog park. Identifying Bowser as he plays among the other canines is easy for the dog-owner to do while onsite.

But if someone wants to use a generative AI model like GPT-5 to monitor their pet while they are at work, the model could fail at this basic task. Vision-language models like GPT-5 often excel at recognizing general objects, like a dog, but they perform poorly at locating personalized objects, like Bowser the French Bulldog.    

To address this shortcoming, researchers from MIT, the MIT-IBM Watson AI Lab, the Weizmann Institute of Science, and elsewhere have introduced a new training method that teaches vision-language models to localize personalized objects in a scene.

Their method uses carefully prepared video-tracking data in which the same object is tracked across multiple frames. They designed the dataset so the model must focus on contextual clues to identify the personalized object, rather than relying on knowledge it previously memorized.

When given a few example images showing a personalized object, like someone’s pet, the retrained model is better able to identify the location of that same pet in a new image.

Models retrained with their method outperformed state-of-the-art systems at this task. Importantly, their technique leaves the rest of the model’s general abilities intact.

This new approach could help future AI systems track specific objects across time, like a child’s backpack, or localize objects of interest, such as a species of animal in ecological monitoring. It could also aid in the development of AI-driven assistive technologies that help visually impaired users find certain items in a room.

“Ultimately, we want these models to be able to learn from context, just like humans do. If a model can do this well, rather than retraining it for each new task, we could just provide a few examples and it would infer how to perform the task from that context. This is a very powerful ability,” says Jehanzeb Mirza, an MIT postdoc and senior author of a paper on this technique.

Mirza is joined on the paper by co-lead authors Sivan Doveh, a postdoc at Stanford University who was a graduate student at Weizmann Institute of Science when this research was conducted; and Nimrod Shabtay, a researcher at IBM Research; James Glass, a senior research scientist and the head of the Spoken Language Systems Group in the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL); and others. The work will be presented at the International Conference on Computer Vision.

An unexpected shortcoming

Researchers have found that large language models (LLMs) can excel at learning from context. If they feed an LLM a few examples of a task, like addition problems, it can learn to answer new addition problems based on the context that has been provided.

A vision-language model (VLM) is essentially an LLM with a visual component connected to it, so the MIT researchers thought it would inherit the LLM’s in-context learning capabilities. But this is not the case.

“The research community has not been able to find a black-and-white answer to this particular problem yet. The bottleneck could arise from the fact that some visual information is lost in the process of merging the two components together, but we just don’t know,” Mirza says.

The researchers set out to improve VLMs abilities to do in-context localization, which involves finding a specific object in a new image. They focused on the data used to retrain existing VLMs for a new task, a process called fine-tuning.

Typical fine-tuning data are gathered from random sources and depict collections of everyday objects. One image might contain cars parked on a street, while another includes a bouquet of flowers.

“There is no real coherence in these data, so the model never learns to recognize the same object in multiple images,” he says.

To fix this problem, the researchers developed a new dataset by curating samples from existing video-tracking data. These data are video clips showing the same object moving through a scene, like a tiger walking across a grassland.

They cut frames from these videos and structured the dataset so each input would consist of multiple images showing the same object in different contexts, with example questions and answers about its location.

“By using multiple images of the same object in different contexts, we encourage the model to consistently localize that object of interest by focusing on the context,” Mirza explains.

Forcing the focus

But the researchers found that VLMs tend to cheat. Instead of answering based on context clues, they will identify the object using knowledge gained during pretraining.

For instance, since the model already learned that an image of a tiger and the label “tiger” are correlated, it could identify the tiger crossing the grassland based on this pretrained knowledge, instead of inferring from context.

To solve this problem, the researchers used pseudo-names rather than actual object category names in the dataset. In this case, they changed the name of the tiger to “Charlie.”

“It took us a while to figure out how to prevent the model from cheating. But we changed the game for the model. The model does not know that ‘Charlie’ can be a tiger, so it is forced to look at the context,” he says.

The researchers also faced challenges in finding the best way to prepare the data. If the frames are too close together, the background would not change enough to provide data diversity.

In the end, finetuning VLMs with this new dataset improved accuracy at personalized localization by about 12 percent on average. When they included the dataset with pseudo-names, the performance gains reached 21 percent.

As model size increases, their technique leads to greater performance gains.

In the future, the researchers want to study possible reasons VLMs don’t inherit in-context learning capabilities from their base LLMs. In addition, they plan to explore additional mechanisms to improve the performance of a VLM without the need to retrain it with new data.

“This work reframes few-shot personalized object localization — adapting on the fly to the same object across new scenes — as an instruction-tuning problem and uses video-tracking sequences to teach VLMs to localize based on visual context rather than class priors. It also introduces the first benchmark for this setting with solid gains across open and proprietary VLMs. Given the immense significance of quick, instance-specific grounding — often without finetuning — for users of real-world workflows (such as robotics, augmented reality assistants, creative tools, etc.), the practical, data-centric recipe offered by this work can help enhance the widespread adoption of vision-language foundation models,” says Saurav Jha, a postdoc at the Mila-Quebec Artificial Intelligence Institute, who was not involved with this work.

Additional co-authors are Wei Lin, a research associate at Johannes Kepler University; Eli Schwartz, a research scientist at IBM Research; Hilde Kuehne, professor of computer science at Tuebingen AI Center and an affiliated professor at the MIT-IBM Watson AI Lab; Raja Giryes, an associate professor at Tel Aviv University; Rogerio Feris, a principal scientist and manager at the MIT-IBM Watson AI Lab; Leonid Karlinsky, a principal research scientist at IBM Research; Assaf Arbelle, a senior research scientist at IBM Research; and Shimon Ullman, the Samy and Ruth Cohn Professor of Computer Science at the Weizmann Institute of Science.

This research was funded, in part, by the MIT-IBM Watson AI Lab.

A decade-plus collaboration between MIT’s AgeLab and the Toyota Motor Corporation is recognized as a key contributor to advancements in automotive safety and human-machine interaction. Through the AgeLab at the MIT Center for Transportation and Logistics (CTL), researchers have collected and analyzed vast real-world driving datasets that have helped inform Toyota’s vehicle design and safety systems.

Toyota recently marked the completion of its 100th project through the Collaborative Safety Research Center (CSRC), celebrating MIT’s role in shaping technologies that enhance driver-assistance features and continue to forge the path for automated mobility. A key foundation for the 100th project is CSRC’s ongoing support for MIT CTL’s Advanced Vehicle Technology (AVT) Consortium.

Real-world data, real-world impact

“AVT was conceptualized over a decade ago as an academic-industry partnership to promote shared investment in real-world, naturalistic data collection, analysis, and collaboration — efforts aimed at advancing safer, more convenient, and more comfortable automobility,” says Bryan Reimer, founder and co-director of AVT. “Since its founding, AVT has drawn together over 25 organizations — including vehicle manufacturers, suppliers, insurers, and consumer research groups — to invest in understanding how automotive technologies function, how they influence driver behavior, and where further innovation is needed. This work has enabled stakeholders like Toyota to make more-informed decisions in product development and deployment.”

“CSRC’s 100th project marks a significant milestone in our collaboration,” Reimer adds. “We deeply value CSRC’s sustained investment, and commend the organization’s commitment to global industry impact and the open dissemination of research to advance societal benefit.”

“Toyota, through its Collaborative Safety Research Center, is proud to be a founding member of the AVT Consortium,” says Jason Hallman, senior manager of Toyota CSRC. “Since 2011, CSRC has collaborated with researchers such as AVT and MIT AgeLab on projects that help inform future products and policy, and to promote a future safe mobility society for all. The AVT specifically has helped us to study the real-world use of several vehicle technologies now available.”

Among these technologies are lane-centering assistance and adaptive cruise control — widely-used technologies that benefit from an understanding of how drivers interact with automation. “AVT uniquely combines vehicle and driver data to help inform future products and highlight the interplay between the performance of these features and the drivers using them,” says Josh Domeyer, principal scientist at CSRC.

Influencing global standards and Olympic-scale innovation

Insights from MIT’s pedestrian-driver interaction research with CSRC also helped shape Toyota’s automated vehicle communication systems. “These data helped develop our foundational understanding that drivers and pedestrians use their movements to communicate during routine traffic encounters,” said Domeyer. “This concept informed the deployment of Toyota’s e-Palette at the Tokyo Olympics, and it has been captured as a best practice in an ISO standard for automated driving system communication.”

The AVT Consortium's naturalistic driving datasets continue to serve as a foundation for behavioral safety strategies. From identifying moments of distraction to understanding how drivers multitask behind the wheel, the work is guiding subtle but impactful design considerations.

“By studying the natural behaviors of drivers and their contexts in the AVT datasets, we hope to identify new ways to encourage safe habits that align with customer preferences,” Domeyer says. “These can include subtle nudges, or modifications to existing vehicle features, or even communication and education partnerships outside of Toyota that reinforce these safe driving habits.”

Professor Yossi Sheffi, director of MIT CTL, comments, “This partnership exemplifies the impact of MIT collaborative research on industry to make real, practical innovation possible.” 

A model for industry-academic collaboration

Founded in 2015, the AVT Consortium brings together automotive manufacturers, suppliers, and insurers to accelerate research in driver behavior, safety, and the transition toward automated systems. The consortium’s interdisciplinary approach — integrating engineering, human factors, and data science — has helped generate one of the world’s most unique and actionable real-world driving datasets.

As Toyota celebrates its research milestone, MIT reflects on a partnership that exemplifies the power of industry-academic collaboration to shape safer, smarter mobility.

The David and Lucile Packard Foundation has announced that two MIT affiliates have been named 2025 Packard Fellows for Science and Engineering. Darcy McRose, the Thomas D. and Virginia W. Cabot Career Development Assistant Professor in the MIT Department of Civil and Environmental Engineering, has been honored, along with Mehtaab Sawhney ’20, PhD ’24, a graduate of the Department of Mathematics who is now at Columbia University. 

The honorees are among 20 junior faculty named among the nation’s most innovative early-career scientists and engineers. Each Packard Fellow receives an unrestricted research grant of $875,000 over five years to support their pursuit of pioneering research and bold new ideas.

“I’m incredibly grateful and honored to be awarded a Packard Fellowship,” says McRose. “It will allow us to continue our work exploring how small molecules control microbial communities in soils and on plant roots, with much-appreciated flexibility to follow our imagination wherever it leads us.”

McRose and her lab study secondary metabolites — small organic molecules that microbes and plants release into soils. Often known as antibiotics, these compounds do far more than fight infections; they can help unlock soil nutrients, shape microbial communities around plant roots, and influence soil fertility.

“Antibiotics made by soil microorganisms are widely used in medicine, but we know surprisingly little about what they do in nature,” explains McRose. “Just as healthy microbiomes support human health, plant microbiomes support plant health, and secondary metabolites can help to regulate the microbial community, suppressing pathogens and promoting beneficial microbes.” 

Her lab integrates techniques from genetics, chemistry, and geosciences to investigate how these molecules shape interactions between microbes and plants in soil — one of Earth’s most complex and least-understood environments. By using secondary metabolites as experimental tools, McRose aims to uncover the molecular mechanisms that govern processes like soil fertility and nutrient cycling that are foundational to sustainable agriculture and ecosystem health.

Studying antibiotics in the environments where they evolved could also yield new strategies for combating soil-borne pathogens and improving crop resilience. “Soil is a true scientific frontier,” McRose says. “Studying these environments has the potential to reveal fascinating, fundamental insights into microbial life — many of which we can’t even imagine yet.”

A native of California, McRose earned her bachelor’s and master’s degrees from Stanford University, followed by a PhD in geosciences from Princeton University. Her graduate thesis focused on how bacteria acquire trace metals from the environment. Her postdoctoral research on secondary metabolites at Caltech was supported by multiple fellowships, including the Simons Foundation Marine Microbial Ecology Postdoctoral Fellowship, the L’Oréal USA For Women in Science Fellowship, and a Division Fellowship from Biology and Biological Engineering at Caltech.

McRose joined the MIT faculty in 2022. In 2025, she was named a Sloan Foundation Research Fellow in Earth System Science and awarded the Maseeh Excellence in Teaching Award.

Past Packard Fellows have gone on to earn the highest honors, including Nobel Prizes in chemistry and physics, the Fields Medal, Alan T. Waterman Awards, Breakthrough Prizes, Kavli Prizes, and elections to the National Academies of Science, Engineering, and Medicine. Each year, the foundation reviews 100 nominations for consideration from 50 invited institutions. The Packard Fellowships Advisory Panel, a group of 12 internationally recognized scientists and engineers, evaluates the nominations and recommends 20 fellows for approval by the Packard Foundation Board of Trustees.

To help mitigate climate change, companies are using bioreactors to grow algae and other microorganisms that are hundreds of times more efficient at absorbing CO2 than trees. Meanwhile, in the pharmaceutical industry, cell culture is used to manufacture biologic drugs and other advanced treatments, including lifesaving gene and cell therapies.

Both processes are hampered by cells’ tendency to stick to surfaces, which leads to a huge amount of waste and downtime for cleaning. A similar problem slows down biofuel production, interferes with biosensors and implants, and makes the food and beverage industry less efficient.

Now, MIT researchers have developed an approach for detaching cells from surfaces on demand, using electrochemically generated bubbles. In an open-access paper published in Science Advances, the researchers demonstrated their approach in a lab prototype and showed it could work across a range of cells and surfaces without harming the cells.

“We wanted to develop a technology that could be high-throughput and plug-and-play, and that would allow cells to attach and detach on demand to improve the workflow in these industrial processes,” says Professor Kripa Varanasi, senior author of the study. “This is a fundamental issue with cells, and we’ve solved it with a process that can scale. It lends itself to many different applications.”

Joining Varanasi on the study are co-first authors Bert Vandereydt, a PhD student in mechanical engineering, and former postdoc Baptiste Blanc.

Solving a sticky problem

The researchers began with a mission.

“We’ve been working on figuring out how we can efficiently capture CO2 across different sources and convert it into valuable products for various end markets,” Varanasi says. “That’s where this photobioreactor and cell detachment comes into the picture.”

Photobioreactors are used to grow carbon-absorbing algae cells by creating tightly controlled environments involving water and sunlight. They feature long, winding tubes with clear surfaces to let in the light algae need to grow. When algae stick to those surfaces, they block out the light, requiring cleaning.

“You have to shut down and clean up the entire reactor as frequently as every two weeks,” Varanasi says. “It’s a huge operational challenge.”

The researchers realized other industries have similar problem due to many cells’ natural adhesion, or stickiness. Each industry has its own solution for cell adhesion depending on how important it is that the cells survive. Some people scrape the surfaces clean, while others use special coatings that are toxic to cells.

In the pharmaceutical and biotech industries, cell detachment is typically carried out using enzymes. However, this method poses several challenges — it can damage cell membranes, is time-consuming, and requires large amounts of consumables, resulting in millions of liters of biowaste.

To create a better solution, the researchers began by studying other efforts to clear surfaces with bubbles, which mainly involved spraying bubbles onto surfaces and had been largely ineffective.

“We realized we needed the bubbles to form on the surfaces where we don’t want these cells to stick, so when the bubbles detach it creates a local fluid flow that creates shear stress at the interface and removes the cells,” Varanasi explains.

Electric currents generate bubbles by splitting water into hydrogen and oxygen. But previous attempts at using electricity to detach cells were hampered because the cell culture mediums contain sodium chloride, which turns into bleach when combined with an electric current. The bleach damages the cells, making it impractical for many applications.

“The culprit is the anode — that’s where the sodium chloride turns to bleach,” Vandereydt explained. “We figured if we could separate that electrode from the rest of the system, we could prevent bleach from being generated.”

To make a better system, the researchers built a 3-square-inch glass surface and deposited a gold electrode on top of it. The layer of gold is so thin it doesn’t block out light. To keep the other electrode separate, the researchers integrated a special membrane that only allows protons to pass through. The set up allowed the researchers to send a current through without generating bleach.

To test their setup, they allowed algae cells from a concentrated solution to stick to the surfaces. When they applied a voltage, the bubbles separated the cells from the surfaces without harming them.

The researchers also studied the interaction between the bubbles and cells, finding the higher the current density, the more bubbles were created and the more algae was removed. They developed a model for understanding how much current would be needed to remove algae in different settings and matched it with results from experiments involving algae as well as cells from ovarian cancer and bones.

“Mammalian cells are orders of magnitude more sensitive than algae cells, but even with those cells, we were able to detach them with no impact to the viability of the cell,” Vandereydt says.

Getting to scale

The researchers say their system could represent a breakthrough in applications where bleach or other chemicals would harm cells. That includes pharmaceutical and food production.

“If we can keep these systems running without fouling and other problems, then we can make them much more economical,” Varanasi says.

For cell culture plates used in the pharmaceutical industry, the team envisions their system comprising an electrode that could be robotically moved from one culture plate to the next, to detach cells as they’re grown. It could also be coiled around algae harvesting systems.

“This has general applicability because it doesn’t rely on any specific biological or chemical treatments, but on a physical force that is system-agnostic,” Varanasi says. “It’s also highly scalable to a lot of different processes, including particle removal.”

Varanasi cautions there is much work to be done to scale up the system. But he hopes it can one day make algae and other cell harvesting more efficient.

“The burning problem of our time is to somehow capture CO₂ in a way that’s economically feasible,” Varanasi says. “These photobioreactors could be used for that, but we have to overcome the cell adhesion problem.”

The work was supported, in part, by Eni S.p.A through the MIT Energy Initiative, the Belgian American Educational Foundation Fellowship, and the Maria Zambrano Fellowship.

People tend to think of quantum materials — whose properties arise from quantum mechanical effects — as exotic curiosities. But some quantum materials have become a ubiquitous part of our computer hard drives, TV screens, and medical devices. Still, the vast majority of quantum materials never accomplish much outside of the lab.

What makes certain quantum materials commercial successes and others commercially irrelevant? If researchers knew, they could direct their efforts toward more promising materials — a big deal since they may spend years studying a single material.

Now, MIT researchers have developed a system for evaluating the scale-up potential of quantum materials. Their framework combines a material’s quantum behavior with its cost, supply chain resilience, environmental footprint, and other factors. The researchers used their framework to evaluate over 16,000 materials, finding that the materials with the highest quantum fluctuation in the centers of their electrons also tend to be more expensive and environmentally damaging. The researchers also identified a set of materials that achieve a balance between quantum functionality and sustainability for further study.

The team hopes their approach will help guide the development of more commercially viable quantum materials that could be used for next generation microelectronics, energy harvesting applications, medical diagnostics, and more.

“People studying quantum materials are very focused on their properties and quantum mechanics,” says Mingda Li, associate professor of nuclear science and engineering and the senior author of the work. “For some reason, they have a natural resistance during fundamental materials research to thinking about the costs and other factors. Some told me they think those factors are too ‘soft’ or not related to science. But I think within 10 years, people will routinely be thinking about cost and environmental impact at every stage of development.”

The paper appears in Materials Today. Joining Li on the paper are co-first authors and PhD students Artittaya Boonkird, Mouyang Cheng, and Abhijatmedhi Chotrattanapituk, along with PhD students Denisse Cordova Carrizales and Ryotaro Okabe; former graduate research assistants Thanh Nguyen and Nathan Drucker; postdoc Manasi Mandal; Instructor Ellan Spero of the Department of Materials Science and Engineering (DMSE); Professor Christine Ortiz of the Department of DMSE; Professor Liang Fu of the Department of Physics; Professor Tomas Palacios of the Department of Electrical Engineering and Computer Science (EECS); Associate Professor Farnaz Niroui of EECS; Assistant Professor Jingjie Yeo of Cornell University; and PhD student Vsevolod Belosevich and Assostant Professor Qiong Ma of Boston College.

Materials with impact

Cheng and Boonkird say that materials science researchers often gravitate toward quantum materials with the most exotic quantum properties rather than the ones most likely to be used in products that change the world.

“Researchers don’t always think about the costs or environmental impacts of the materials they study,” Cheng says. “But those factors can make them impossible to do anything with.”

Li and his collaborators wanted to help researchers focus on quantum materials with more potential to be adopted by industry. For this study, they developed methods for evaluating factors like the materials’ price and environmental impact using their elements and common practices for mining and processing those elements. At the same time, they quantified the materials’ level of “quantumness” using an AI model created by the same group last year, based on a concept proposed by MIT professor of physics Liang Fu, termed quantum weight.

“For a long time, it’s been unclear how to quantify the quantumness of a material,” Fu says. “Quantum weight is very useful for this purpose. Basically, the higher the quantum weight of a material, the more quantum it is.”

The researchers focused on a class of quantum materials with exotic electronic properties known as topological materials, eventually assigning over 16,000 materials scores on environmental impact, price, import resilience, and more.

For the first time, the researchers found a strong correlation between the material’s quantum weight and how expensive and environmentally damaging it is.

“That’s useful information because the industry really wants something very low-cost,” Spero says. “We know what we should be looking for: high quantum weight, low-cost materials. Very few materials being developed meet that criteria, and that likely explains why they don’t scale to industry.”

The researchers identified 200 environmentally sustainable materials and further refined the list down to 31 material candidates that achieved an optimal balance of quantum functionality and high-potential impact.

The researchers also found that several widely studied materials exhibit high environmental impact scores, indicating they will be hard to scale sustainably. “Considering the scalability of manufacturing and environmental availability and impact is critical to ensuring practical adoption of these materials in emerging technologies,” says Niroui.

Guiding research

Many of the topological materials evaluated in the paper have never been synthesized, which limited the accuracy of the study’s environmental and cost predictions. But the authors say the researchers are already working with companies to study some of the promising materials identified in the paper.

“We talked with people at semiconductor companies that said some of these materials were really interesting to them, and our chemist collaborators also identified some materials they find really interesting through this work,” Palacios says. “Now we want to experimentally study these cheaper topological materials to understand their performance better.”

“Solar cells have an efficiency limit of 34 percent, but many topological materials have a theoretical limit of 89 percent. Plus, you can harvest energy across all electromagnetic bands, including our body heat,” Fu says. “If we could reach those limits, you could easily charge your cell phone using body heat. These are performances that have been demonstrated in labs, but could never scale up. That’s the kind of thing we’re trying to push forward."

This work was supported, in part, by the National Science Foundation and the U.S. Department of Energy.

Earthquakes often bring to mind images of destruction, of the Earth breaking open and altering landscapes. But after an earthquake, the area around it undergoes a period of post-seismic deformation, where areas that didn’t break experience new stress as a result of the sudden change in the surroundings. Once it has adjusted to this new stress, it reaches a state of recovery.

Geologists have often thought that this recovery period was a smooth, continuous process. But MIT research published recently in Science has found evidence that while healing occurs quickly at shallow depths — roughly above 10 km — deeper depths recover more slowly, if at all.

“If you were to look before and after in the shallow crust, you wouldn’t see any permanent change. But there’s this very permanent change that persists in the mid-crust,” says Jared Bryan, a graduate student in the MIT Department of Earth, Atmospheric and Planetary Sciences (EAPS) and lead author on the paper.

The paper’s other authors include EAPS Professor William Frank and Pascal Audet from the University of Ottawa.

Everything but the quakes

In order to assemble a full understanding of how the crust behaves before, during, and after an earthquake sequence, the researchers looked at seismic data from the 2019 Ridgecrest earthquakes in California. This immature fault zone experienced the largest earthquake in the state in 20 years, and tens of thousands of aftershocks over the following year. They then removed seismic data created by the sequence and only looked at waves generated by other seismic activity around the world to see how their paths through the Earth changed before and after the sequence.

“One person’s signal is another person’s noise,” says Bryan. They also used general ambient noise from sources like ocean waves and traffic that are also picked up by seismometers. Then, using a technique called a receiver function, they were able to see the speed of the waves as they traveled and how it changed due to conditions in the Earth such as rock density and porosity, much in the same way we use sonar to see how acoustic waves change when they interact with objects. With all this information, they were able to construct basic maps of the Earth around the Ridgecrest fault zone before and after the sequence.

What they found was that the shallow crust, extending about 10 km into the Earth, recovered over the course of a few months. In contrast, deeper depths in the mid-crust didn’t experience immediate damage, but rather changed over the same timescale as shallow depths recovered.

“What was surprising is that the healing in the shallow crust was so quick, and then you have this complementary accumulation occurring, not at the time of the earthquake, but instead over the post-seismic phase,” says Bryan.

Balancing the energy budget

Understanding how recovery plays out at different depths is crucial for determining how energy is spent during different parts of the seismic process, which includes activities such as the release of energy as waves, the creation of new fractures, or energy being stored elastically in the surrounding areas. Altogether, this is collectively known as the energy budget, and it is a useful tool for understanding how damage accumulates and recovers over time.

What remains unclear is the timescales at which deeper depths recover, if at all. The paper presents two possible scenarios to explain why that might be: one in which the deep crust recovers over a much longer timescale than they observed, or one where it never recovers at all.

“Either of those are not what we expected,” says Frank. “And both of them are interesting.”

Further research will require more observations to build out a more detailed picture to see at what depth the change becomes more pronounced. In addition, Bryan wants to look at other areas, such as more mature faults that experience higher levels of seismic activity, to see if it changes the results.

“We’ll let you know in 1,000 years whether it’s recovered,” says Bryan.

Born in Palermo, Sicily, Giorgio Rizzo spent his childhood curious about the natural world. “I have always been fascinated by nature and how plants and animals can adapt and survive in extreme environments,” he says. “Their highly tuned biochemistry, and their incredible ability to create ones of the most complex and beautiful structures in chemistry that we still can’t even achieve in our laboratories.”

As an undergraduate student, he watched as a researcher mounted a towering chromatography column layered with colorful plant chemicals in a laboratory. When the researcher switched on a UV light, the colors turned into fluorescent shades of blue, green, red and pink. “I realized in that exact moment that I wanted to be the same person, separating new unknown compounds from a rare plant with potential pharmaceutical properties,” he recalls.

These experiences set him on a path from a master’s degree in organic chemistry to his current work as a postdoc in the MIT Department of Civil and Environmental Engineering, where he focuses on developing sustainable fertilizers and studying how rare earth elements can boost plant resilience, with the aim of reducing agriculture’s environmental impact.

In the lab of MIT Professor Benedetto Marelli, Rizzo studies plant responses to environmental stressors, such as heat, drought, and prolonged UV irradiation. This includes developing new fertilizers that can be applied as seed coating to help plants grow stronger and enhance their resistance.

“We are working on new formulations of fertilizers that aim to reduce the huge environmental impact of classical practices in agriculture based on NPK inorganic fertilizers,” Rizzo explains. Although they are fundamental to crop yields, their tendency to accumulate in soil is detrimental to the soil health and microbiome living in it. In addition, producing NPK (nitrogen, phosphorus, and potassium) fertilizers is one of the most energy-consuming and polluting chemical processes in the world.

“It is mandatory to reshape our conception of fertilizers and try to rely, at least in part, on alternative products that are safer, cheaper, and more sustainable,” he says.

Recently, Rizzo was awarded a Kavanaugh Fellowship, a program that gives MIT graduate students and postdocs entrepreneurial training and resources to bring their research from the lab to the market. “This prestigious fellowship will help me build a concrete product for a company, adding more value to our research,” he says.

Rizzo hopes their work will help farmers increase their crop yields without compromising soil quality or plant health. A major barrier to adopting new fertilizers is cost, as many farmers rely heavily on each growing season’s output and cannot risk investing in products that may underperform compared to traditional NPK fertilizers. The fertilizers being developed in the Marelli Lab address this challenge by using chitin and chitosan, abundant natural materials that make them far less expensive to produce, which Rizzo hopes will encourage farmers to try them.

“Through the Kavanaugh Fellowship, I will spend this year trying to bring the technology outside the lab to impact the world and meet the need for farmers to support their prosperity,” he says.

Mentorship has been a defining part of his postdoc experience. Rizzo describes Professor Benedetto Marelli as “an incredible mentor” who values his research interests and supports him through every stage of his work. The lab spans a wide range of projects — from plant growth enhancement and precision chemical delivery to wastewater treatment, vaccine development for fish, and advanced biochemical processes. “My colleagues created a stimulant environment with different research topics,” he notes. He is also grateful for the work he does with international institutions, which has helped him build a network of researchers and academics around the world.

Rizzo enjoys the opportunity to mentor students in the lab and appreciates their curiosity and willingness to learn. “It is one of the greatest qualities you can have as a scientist because you must be driven by curiosity to discover the unexpected,” he says.

He describes MIT as a “dynamic and stimulating experience,” but also acknowledges how overwhelming it can be. “You will feel like a small fish in a big ocean,” he says. “But that is exactly what MIT is: an ocean full of opportunities and challenges that are waiting to be solved.”

Beyond his professional work, Rizzo enjoys nature and the arts. An avid reader, he balances his scientific work with literature and history. “I never read about science-related topics — I read about it a lot already for my job,” he says. “I like classic literature, novels, essays, history of nations, and biographies. Often you can find me wandering in museums’ art collections.” Classical art, Renaissance, and Pre-Raphaelites are his favorite artistic currents.

Looking ahead, Rizzo hopes to shift his professional pathway toward startups or companies focused on agrotechnical improvement. His immediate goal is to contribute to initiatives where research has a direct, tangible impact on everyday life.

“I want to pursue the option of being part of a spinout process that would enable my research to have a direct impact in everyday life and help solve agricultural issues,” he adds.

Oct. 16 is World Food Day, a global campaign to celebrate the founding of the Food and Agriculture Organization 80 years ago, and to work toward a healthy, sustainable, food-secure future. More than 670 million people in the world are facing hunger. Millions of others are facing rising obesity rates and struggle to get healthy food for proper nutrition. 

World Food Day calls on not only world governments, but business, academia, the media, and even the youth to take action to promote resilient food systems and combat hunger. This year, the Abdul Latif Jameel Water and Food Systems Laboratory (J-WAFS) is spotlighting an MIT researcher who is working toward this goal by studying food and water systems in the Global South.

J-WAFS seed grants provide funding to early-stage research projects that are unique to prior work. In an 11th round of seed grant funding in 2025, 10 MIT faculty members received support to carry out their cutting-edge water and food research. Ali Aouad PhD ’17, assistant professor of operations management at the MIT Sloan School of Management, was one of those grantees. “I had searched before joining MIT what kind of research centers and initiatives were available that tried to coalesce research on food systems,” Aouad says. “And so, I was very excited about J-WAFS.” 

Aouad gathered more information about J-WAFS at the new faculty orientation session in August 2024, where he spoke to J-WAFS staff and learned about the program’s grant opportunities for water and food research. Later that fall semester, he attended a few J-WAFS seminars on agricultural economics and water resource management. That’s when Aouad knew that his project was perfectly aligned with the J-WAFS mission of securing humankind’s water and food.

Aouad’s seed project focuses on food subsidies. With a background in operations research and an interest in digital platforms, much of his work has centered on aligning supply-side operations with heterogeneous customer preferences. Past projects include ones on retail and matching systems. “I started thinking that these types of demand-driven approaches may be also very relevant to important social challenges, particularly as they relate to food security,” Aouad says. Before starting his PhD at MIT, Aouad worked on projects that looked at subsidies for smallholder farmers in low- and middle-income countries. “I think in the back of my mind, I've always been fascinated by trying to solve these issues,” he noted.

His seed grant project, Optimal subsidy design: Application to food assistance programs, aims to leverage data on preferences and purchasing habits from local grocery stores in India to inform food assistance policy and optimize the design of subsidies. Typical data collection systems, like point-of-sales, are not as readily available in India’s local groceries, making this type of data hard to come by for low-income individuals. “Mom-and-pop stores are extremely important last-mile operators when it comes to nutrition,” he explains. 

For this project, the research team gave local grocers point-of-sale scanners to track purchasing habits. “We aim to develop an algorithm that converts these transactions into some sort of ‘revelation’ of the individuals’ latent preferences,” says Aouad. “As such, we can model and optimize the food assistance programs — how much variety and flexibility is offered, taking into account the expected demand uptake.” He continues, “now, of course, our ability to answer detailed design questions [across various products and prices] depends on the quality of our inference from  the data, and so this is where we need more sophisticated and robust algorithms.”

Following the data collection and model development, the ultimate goal of this research is to inform policy surrounding food assistance programs through an “optimization approach.” Aouad describes the complexities of using optimization to guide policy. “Policies are often informed by domain expertise, legacy systems, or political deliberation. A lot of researchers build rigorous evidence to inform food policy, but it’s fair to say that the kind of approach that I’m proposing in this research is not something that is commonly used. I see an opportunity for bringing a new approach and methodological tradition to a problem that has been central for policy for many decades.” 

The overall health of consumers is the reason food assistance programs exist, yet measuring long-term nutritional impacts and shifts in purchase behavior is difficult. In past research, Aouad notes that the short-term effects of food assistance interventions can be significant. However, these effects are often short-lived. “This is a fascinating question that I don’t think we will be able to address within the space of interventions that we will be considering. However, I think it is something I would like to capture in the research, and maybe develop hypotheses for future work around how we can shift nutrition-related behaviors in the long run.”

While his project develops a new methodology to calibrate food assistance programs, large-scale applications are not promised. “A lot of what drives subsidy mechanisms and food assistance programs is also, quite frankly, how easy it is and how cost-effective it is to implement these policies in the first place,” comments Aouad. Cost and infrastructure barriers are unavoidable to this kind of policy research, as well as sustaining these programs. Aouad’s effort will provide insights into customer preferences and subsidy optimization in a pilot setup, but replicating this approach on a real scale may be costly. Aouad hopes to be able to gather proxy information from customers that would both feed into the model and provide insight into a more cost-effective way to collect data for large-scale implementation.

There is still much work to be done to ensure food security for all, whether it’s advances in agriculture, food-assistance programs, or ways to boost adequate nutrition. As the 2026 seed grant deadline approaches, J-WAFS will continue its mission of supporting MIT faculty as they pursue innovative projects that have practical and real impacts on water and food system challenges.

Manufacturing better batteries, faster electronics, and more effective pharmaceuticals depends on the discovery of new materials and the verification of their quality. Artificial intelligence is helping with the former, with tools that comb through catalogs of materials to quickly tag promising candidates.

But once a material is made, verifying its quality still involves scanning it with specialized instruments to validate its performance — an expensive and time-consuming step that can hold up the development and distribution of new technologies.

Now, a new AI tool developed by MIT engineers could help clear the quality-control bottleneck, offering a faster and cheaper option for certain materials-driven industries.

In a study appearing today in the journal Matter, the researchers present “SpectroGen,” a generative AI tool that turbocharges scanning capabilities by serving as a virtual spectrometer. The tool takes in “spectra,” or measurements of a material in one scanning modality, such as infrared, and generates what that material’s spectra would look like if it were scanned in an entirely different modality, such as X-ray. The AI-generated spectral results match, with 99 percent accuracy, the results obtained from physically scanning the material with the new instrument.

Certain spectroscopic modalities reveal specific properties in a material: Infrared reveals a material’s molecular groups, while X-ray diffraction visualizes the material’s crystal structures, and Raman scattering illuminates a material’s molecular vibrations. Each of these properties is essential in gauging a material’s quality and typically requires tedious workflows on multiple expensive and distinct instruments to measure.

With SpectroGen, the researchers envision that a diversity of measurements can be made using a single and cheaper physical scope. For instance, a manufacturing line could carry out quality control of materials by scanning them with a single infrared camera. Those infrared spectra could then be fed into SpectroGen to automatically generate the material’s X-ray spectra, without the factory having to house and operate a separate, often more expensive X-ray-scanning laboratory.

The new AI tool generates spectra in less than one minute, a thousand times faster compared to traditional approaches that can take several hours to days to measure and validate.

“We think that you don’t have to do the physical measurements in all the modalities you need, but perhaps just in a single, simple, and cheap modality,” says study lead Loza Tadesse, assistant professor of mechanical engineering at MIT. “Then you can use SpectroGen to generate the rest. And this could improve productivity, efficiency, and quality of manufacturing.”

The study was led by Tadesse, with former MIT postdoc Yanmin Zhu serving as first author.

Beyond bonds

Tadesse’s interdisciplinary group at MIT pioneers technologies that advance human and planetary health, developing innovations for applications ranging from rapid disease diagnostics to sustainable agriculture.

“Diagnosing diseases, and material analysis in general, usually involves scanning samples and collecting spectra in different modalities, with different instruments that are bulky and expensive and that you might not all find in one lab,” Tadesse says. “So, we were brainstorming about how to miniaturize all this equipment and how to streamline the experimental pipeline.”

Zhu noted the increasing use of generative AI tools for discovering new materials and drug candidates, and wondered whether AI could also be harnessed to generate spectral data. In other words, could AI act as a virtual spectrometer?

A spectroscope probes a material’s properties by sending light of a certain wavelength into the material. That light causes molecular bonds in the material to vibrate in ways that scatter the light back out to the scope, where the light is recorded as a pattern of waves, or spectra, that can then be read as a signature of the material’s structure.

For AI to generate spectral data, the conventional approach would involve training an algorithm to recognize connections between physical atoms and features in a material, and the spectra they produce. Given the complexity of molecular structures within just one material, Tadesse says such an approach can quickly become intractable.

“Doing this even for just one material is impossible,” she says. “So, we thought, is there another way to interpret spectra?”

The team found an answer with math. They realized that a spectral pattern, which is a sequence of waveforms, can be represented mathematically. For instance, a spectrum that contains a series of bell curves is known as a “Gaussian” distribution, which is associated with a certain mathematical expression, compared to a series of narrower waves, known as a “Lorentzian” distribution, that is described by a separate, distinct algorithm. And as it turns out, for most materials infrared spectra characteristically contain more Lorentzian waveforms, while Raman spectra are more Gaussian, and X-ray spectra is a mix of the two.

Tadesse and Zhu worked this mathematical interpretation of spectral data into an algorithm that they then incorporated into a generative AI model.

“It’s a physics-savvy generative AI that understands what spectra are,” Tadesse says. “And the key novelty is, we interpreted spectra not as how it comes about from chemicals and bonds, but that it is actually math — curves and graphs, which an AI tool can understand and interpret.”

Data co-pilot

The team demonstrated their SpectroGen AI tool on a large, publicly available dataset of over 6,000 mineral samples. Each sample includes information on the mineral’s properties, such as its elemental composition and crystal structure. Many samples in the dataset also include spectral data in different modalities, such as X-ray, Raman, and infrared. Of these samples, the team fed several hundred to SpectroGen, in a process that trained the AI tool, also known as a neural network, to learn correlations between a mineral’s different spectral modalities. This training enabled SpectroGen to take in spectra of a material in one modality, such as in infrared, and generate what a spectra in a totally different modality, such as X-ray, should look like.

Once they trained the AI tool, the researchers fed SpectroGen spectra from a mineral in the dataset that was not included in the training process. They asked the tool to generate a spectra in a different modality, based on this “new” spectra. The AI-generated spectra, they found, was a close match to the mineral’s real spectra, which was originally recorded by a physical instrument. The researchers carried out similar tests with a number of other minerals and found that the AI tool quickly generated spectra, with 99 percent correlation.

“We can feed spectral data into the network and can get another totally different kind of spectral data, with very high accuracy, in less than a minute,” Zhu says.

The team says that SpectroGen can generate spectra for any type of mineral. In a manufacturing setting, for instance, mineral-based materials that are used to make semiconductors and battery technologies could first be quickly scanned by an infrared laser. The spectra from this infrared scanning could be fed into SpectroGen, which would then generate a spectra in X-ray, which operators or a multiagent AI platform can check to assess the material’s quality.

“I think of it as having an agent or co-pilot, supporting researchers, technicians, pipelines and industry,” Tadesse says. “We plan to customize this for different industries’ needs.”

The team is exploring ways to adapt the AI tool for disease diagnostics, and for agricultural monitoring through an upcoming project funded by Google. Tadesse is also advancing the technology to the field through a new startup and envisions making SpectroGen available for a wide range of sectors, from pharmaceuticals to semiconductors to defense.

More than 300 million people worldwide are living with rare disorders — many of which have a genetic cause and affect the brain and nervous system — yet the vast majority of these conditions lack an approved therapy. Because each rare disorder affects fewer than 65 out of every 100,000 people, studying these disorders and creating new treatments for them is especially challenging.

Thanks to a generous philanthropic gift from Ana Méndez ’91 and Rajeev Jayavant ’86, EE ’88, SM ’88, MIT is now poised to fill gaps in this research landscape. By establishing the Rare Brain Disorders Nexus — or RareNet — at MIT's McGovern Institute for Brain Research, the alumni aim to convene leaders in neuroscience research, clinical medicine, patient advocacy, and industry to streamline the lab-to-clinic pipeline for rare brain disorder treatments.

“Ana and Rajeev’s commitment to MIT will form crucial partnerships to propel the translation of scientific discoveries into promising therapeutics and expand the Institute’s impact on the rare brain disorders community,” says MIT President Sally Kornbluth. “We are deeply grateful for their pivotal role in advancing such critical science and bringing attention to conditions that have long been overlooked.”

Building new coalitions

Several hurdles have slowed the lab-to-clinic pipeline for rare brain disorder research. It is difficult to secure a sufficient number of patients per study, and current research efforts are fragmented, since each study typically focuses on a single disorder (there are more than 7,000 known rare disorders, according to the World Health Organization). Pharmaceutical companies are often reluctant to invest in emerging treatments due to a limited market size and the high costs associated with preparing drugs for commercialization.

Méndez and Jayavant envision that RareNet will finally break down these barriers. “Our hope is that RareNet will allow leaders in the field to come together under a shared framework and ignite scientific breakthroughs across multiple conditions. A discovery for one rare brain disorder could unlock new insights that are relevant to another,” says Jayavant. “By congregating the best minds in the field, we are confident that MIT will create the right scientific climate to produce drug candidates that may benefit a spectrum of uncommon conditions.”

Guoping Feng, the James W. (1963) and Patricia T. Poitras Professor in Neuroscience and associate director of the McGovern Institute, will serve as RareNet’s inaugural faculty director. Feng holds a strong record of advancing studies on therapies for neurodevelopmental disorders, including autism spectrum disorders, Williams syndrome, and uncommon forms of epilepsy. His team’s gene therapy for Phelan-McDermid syndrome, a rare and profound autism spectrum disorder, has been licensed to Jaguar Gene Therapy and is currently undergoing clinical trials. “RareNet pioneers a unique model for biomedical research — one that is reimagining the role academia can play in developing therapeutics,” says Feng.

RareNet plans to deploy two major initiatives: a global consortium and a therapeutic pipeline accelerator. The consortium will form an international network of researchers, clinicians, and patient groups from the outset. It seeks to connect siloed research efforts, secure more patient samples, promote data sharing, and drive a strong sense of trust and goal alignment across the RareNet community. Partnerships within the consortium will support the aim of the therapeutic pipeline accelerator: to de-risk early lab discoveries and expedite their translation to clinic. By fostering more targeted collaborations — especially between academia and industry — the accelerator will prepare potential treatments for clinical use as efficiently as possible.

MIT labs are focusing on four uncommon conditions in the first wave of RareNet projects: Rett syndrome, prion disease, disorders linked to SYNGAP1 mutations, and Sturge-Weber syndrome. The teams are working to develop novel therapies that can slow, halt, or reverse dysfunctions in the brain and nervous system.

These efforts will build new bridges to connect key stakeholders across the rare brain disorders community and disrupt conventional research approaches. “Rajeev and I are motivated to seed powerful collaborations between MIT researchers, clinicians, patients, and industry,” says Méndez. “Guoping Feng clearly understands our goal to create an environment where foundational studies can thrive and seamlessly move toward clinical impact.”

“Patient and caregiver experiences, and our foreseeable impact on their lives, will guide us and remain at the forefront of our work,” Feng adds. “For far too long has the rare brain disorders community been deprived of life-changing treatments — and, importantly, hope. RareNet gives us the opportunity to transform how we study these conditions, and to do so at a moment when it’s needed more than ever.”

Scientists at MIT and elsewhere have discovered extremely rare remnants of “proto Earth,” which formed about 4.5 billion years ago, before a colossal collision irreversibly altered the primitive planet’s composition and produced the Earth as we know today. Their findings, reported today in the journal Nature Geosciences, will help scientists piece together the primordial starting ingredients that forged the early Earth and the rest of the solar system.

Billions of years ago, the early solar system was a swirling disk of gas and dust that eventually clumped and accumulated to form the earliest meteorites, which in turn merged to form the proto Earth and its neighboring planets.

In this earliest phase, Earth was likely rocky and bubbling with lava. Then, less than 100 million years later, a Mars-sized meteorite slammed into the infant planet in a singular “giant impact” event that completely scrambled and melted the planet’s interior, effectively resetting its chemistry. Whatever original material the proto Earth was made from was thought to have been altogether transformed.

But the MIT team’s findings suggest otherwise. The researchers have identified a chemical signature in ancient rocks that is unique from most other materials found in the Earth today. The signature is in the form of a subtle imbalance in potassium isotopes discovered in samples of very old and very deep rocks. The team determined that the potassium imbalance could not have been produced by any previous large impacts or geological processes occurring in the Earth presently.

The most likely explanation for the samples’ chemical composition is that they must be leftover material from the proto Earth that somehow remained unchanged, even as most of the early planet was impacted and transformed.

“This is maybe the first direct evidence that we’ve preserved the proto Earth materials,” says Nicole Nie, the Paul M. Cook Career Development Assistant Professor of Earth and Planetary Sciences at MIT. “We see a piece of the very ancient Earth, even before the giant impact. This is amazing because we would expect this very early signature to be slowly erased through Earth’s evolution.”

The study’s other authors include Da Wang of Chengdu University of Technology in China, Steven Shirey and Richard Carlson of the Carnegie Institution for Science in Washington, Bradley Peters of ETH Zürich in Switzerland, and James Day of Scripps Institution of Oceanography in California.

A curious anomaly

In 2023, Nie and her colleagues analyzed many of the major meteorites that have been collected from sites around the world and carefully studied. Before impacting the Earth, these meteorites likely formed at various times and locations throughout the solar system, and therefore represent the solar system’s changing conditions over time. When the researchers compared the chemical compositions of these meteorite samples to Earth, they identified among them a “potassium isotopic anomaly.”

Isotopes are slightly different versions of an element that have the same number of protons but a different number of neutrons. The element potassium can exist in one of three naturally-occurring isotopes, with mass numbers (protons plus neutrons) of 39, 40, and 41, respectively. Wherever potassium has been found on Earth, it exists in a characteristic combination of isotopes, with potassium-39 and potassium-41 being overwhelmingly dominant. Potassium-40 is present, but at a vanishingly small percentage in comparison.

Nie and her colleagues discovered that the meteorites they studied showed balances of potassium isotopes that were different from most materials on Earth. This potassium anomaly suggested that any material that exhibits a similar anomaly likely predates Earth’s present composition. In other words, any potassium imbalance would be a strong sign of material from the proto Earth, before the giant impact reset the planet’s chemical composition.

“In that work, we found that different meteorites have different potassium isotopic signatures, and that means potassium can be used as a tracer of Earth’s building blocks,” Nie explains.

“Built different”

In the current study, the team looked for signs of potassium anomalies not in meteorites, but within the Earth. Their samples include rocks, in powder form, from Greenland and Canada, where some of the oldest preserved rocks are found. They also analyzed lava deposits collected from Hawaii, where volcanoes have brought up some of the Earth’s earliest, deepest materials from the mantle (the planet’s thickest layer of rock that separates the crust from the core).

“If this potassium signature is preserved, we would want to look for it in deep time and deep Earth,” Nie says.

The team first dissolved the various powder samples in acid, then carefully isolated any potassium from the rest of the sample and used a special mass spectrometer to measure the ratio of each of potassium’s three isotopes. Remarkably, they identified in the samples an isotopic signature that was different from what’s been found in most materials on Earth.

Specifically, they identified a deficit in the potassium-40 isotope. In most materials on Earth, this isotope is already an insignificant fraction compared to potassium’s other two isotopes. But the researchers were able to discern that their samples contained an even smaller percentage of potassium-40. Detecting this tiny deficit is like spotting a single grain of brown sand in a bucket rather than a scoop full of of yellow sand.

The team found that, indeed, the samples exhibited the potassium-40 deficit, showing that the materials “were built different,” says Nie, compared to most of what we see on Earth today.

But could the samples be rare remnants of the proto Earth? To answer this, the researchers assumed that this might be the case. They reasoned that if the proto Earth were originally made from such potassium-40-deficient materials, then most of this material would have undergone chemical changes — from the giant impact and subsequent, smaller meteorite impacts — that ultimately resulted in the materials with more potassium-40 that we see today. 

The team used compositional data from every known meteorite and carried out simulations of how the samples’ potassium-40 deficit would change following impacts by these meteorites and by the giant impact. They also simulated geological processes that the Earth experienced over time, such as the heating and mixing of the mantle. In the end, their simulations produced a composition with a slightly higher fraction of potassium-40 compared to the samples from Canada, Greenland, and Hawaii. More importantly, the simulated compositions matched those of most modern-day materials.

The work suggests that materials with a potassium-40 deficit are likely leftover original material from the proto Earth.

Curiously, the samples’ signature isn’t a precise match with any other meteorite in geologists’ collections. While the meteorites in the team’s previous work showed potassium anomalies, they aren’t exactly the deficit seen in the proto Earth samples. This means that whatever meteorites and materials originally formed the proto Earth have yet to be discovered.

“Scientists have been trying to understand Earth’s original chemical composition by combining the compositions of different groups of meteorites,” Nie says. “But our study shows that the current meteorite inventory is not complete, and there is much more to learn about where our planet came from.”

This work was supported, in part, by NASA and MIT.

For decades, synthetic biologists have been developing gene circuits that can be transferred into cells for applications such as reprogramming a stem cell into a neuron or generating a protein that could help treat a disease such as fragile X syndrome.

These gene circuits are typically delivered into cells by carriers such as nonpathogenic viruses. However, it has been difficult to ensure that these cells end up producing the correct amount of the protein encoded by the synthetic gene.

To overcome that obstacle, MIT engineers have designed a new control mechanism that allows them to establish a desired protein level, or set point, for any gene circuit. This approach also allows them to edit the set point after the circuit is delivered.

“This is a really stable and multifunctional tool. The tool is very modular, so there are a lot of transgenes you could control with this system,” says Katie Galloway, an assistant professor in Chemical Engineering at MIT and the senior author of the new study.

Using this strategy, the researchers showed that they could induce cells to generate consistent levels of target proteins. In one application that they demonstrated, they converted mouse embryonic fibroblasts to motor neurons by delivering high levels of a gene that promotes that conversion.

MIT graduate student Sneha Kabaria is the lead author of the paper, which appears today in Nature Biotechnology. Other authors include Yunbeen Bae ’24; MIT graduate students Mary Ehmann, Brittany Lende-Dorn, Emma Peterman, and Kasey Love; Adam Beitz PhD ’25; and former MIT postdoc Deon Ploessl.

Dialing up gene expression

Synthetic gene circuits are engineered to include not only the gene of interest, but also a promoter region. At this site, transcription factors and other regulators can bind, turning on the expression of the synthetic gene.

However, it’s not always possible to get all of the cells in a population to express the desired gene at a uniform level. One reason for that is that some cells may take up just one copy of the circuit, while others receive many more. Additionally, cells have natural variation in how much protein they produce.

That has made reprogramming cells challenging because it’s difficult to ensure that every cell in a population of skin cells, for example, will produce enough of the necessary transcription factors to successfully transition into a new cell identity, such as a neuron or induced pluripotent stem cell.

In the new paper, the researchers devised a way to control gene expression levels by changing the distance between the synthetic gene and its promoter. They found that when there was a longer DNA “spacer” between the promoter region and the gene, the gene would be expressed at a lower level. That extra distance, they showed, makes it less likely that transcription factors bound to the promoter will effectively turn on gene transcription.

Then, to create set points that could be edited, the researchers incorporated sites within the spacer that can be excised by an enzyme called Cre recombinase. As parts of the spacer are cut out, it helps bring the transcription factors closer to the gene of interest, which turns up gene expression.

The researchers showed they could create spacers with multiple excision points, each targeted by different recombinases. This allowed them to create a system called DIAL, that they could use to establish “high,” “med,” “low” and “off” set points for gene expression.

After the DNA segment carrying the gene and its promoter is delivered into cells, recombinases can be added to the cells, allowing the set point to be edited at any time.

The researchers demonstrated their system in mouse and human cells by delivering the gene for different fluorescent proteins and functional genes, and showed that they could get uniform expression across the a population of cells at the target level.

“We achieved uniform and stable control. This is very exciting for us because lack of uniform, stable control has been one of the things that's been limiting our ability to build reliable systems in synthetic biology. When there are too many variables that affect your system, and then you add in normal biological variation, it’s very hard to build stable systems,” Galloway says.

Reprogramming cells

To demonstrate potential applications of the DIAL system, the researchers then used it to deliver different levels of the gene HRas^G12V to mouse embryonic fibroblasts. This HRas variant has previously been shown to increase the rate of conversion of fibroblasts to neurons. The MIT team found that in cells that received a higher dose of the gene, a larger percentage of them were able to successfully transform into neurons.

Using this system, researchers now hope to perform more systematic studies of different transcription factors that can induce cells to transition to different cell types. Such studies could reveal how different levels of those factors affect the success rate, and whether changing the transcription factors levels might alter the cell type that is generated.

In ongoing work, the researchers have shown that DIAL can be combined with a system they previously developed, known as ComMAND, that uses a feedforward loop to help prevent cells from overexpressing a therapeutic gene.

Using these systems together, it could be possible to tailor gene therapies to produce specific, consistent protein levels in the target cells of individual patients, the researchers say.

“This is something we’re excited about because both DIAL and ComMAND are highly modular, so you could not only have a well-controlled gene therapy that’s somewhat general for a population, but you could, in theory, tailor it for any given person or any given cell type,” Galloway says.

The research was funded, in part, by the National Institute of General Medical Sciences, the National Science Foundation, and the Institute for Collaborative Biotechnologies.

Understanding how interactions between the central nervous system and the immune system contribute to problems of aging, including Alzheimer’s disease, Parkinson’s disease, arthritis, and more, can generate new leads for therapeutic development, speakers said at MIT’s symposium “The Neuro-Immune Axis and the Aging Brain” on Sept 18.

“The past decade has brought rapid progress in our understanding of how adaptive and innate immune systems impact the pathogenesis of neurodegenerative disorders,” said Picower Professor Li-Huei Tsai, director of The Picower Institute for Learning and Memory and MIT’s Aging Brain Initiative (ABI), in her introduction to the event, which more than 450 people registered to attend. “Together, today’s speakers will trace how the neuro-immune axis shapes brain health and disease … Their work converges on the promise of immunology-informed therapies to slow or prevent neurodegeneration and age-related cognitive decline.”

For instance, keynote speaker Michal Schwartz of the Weizmann Institute in Israel described her decades of pioneering work to understand the neuro-immune “ecosystem.” Immune cells, she said, help the brain heal, and support many of its functions, including its “plasticity,” the ability it has to adapt to and incorporate new information. But Schwartz’s lab also found that an immune signaling cascade can arise with aging that undermines cognitive function. She has leveraged that insight to investigate and develop corrective immunotherapies that improve the brain’s immune response to Alzheimer’s both by rejuvenating the brain’s microglia immune cells and bringing in the help of peripheral immune cells called macrophages. Schwartz has brought the potential therapy to market as the chief science officer of ImmunoBrain, a company testing it in a clinical trial.

In her presentation, Tsai noted recent work from her lab and that of computer science professor and fellow ABI member Manolis Kellis showing that many of the genes associated with Alzheimer’s disease are most strongly expressed in microglia, giving it an expression profile more similar to autoimmune disorders than to many psychiatric ones (where expression of disease-associated genes typically is highest in neurons). The study showed that microglia become “exhausted” over the course of disease progression, losing their cellular identity and becoming harmfully inflammatory.

“Genetic risk, epigenomic instability, and microglia exhaustion really play a central role in Alzheimer’s disease,” Tsai said, adding that her lab is now also looking into how immune T cells, recruited by microglia, may also contribute to Alzheimer’s disease progression.

The body and the brain

The neuro-immune “axis” connects not only the nervous and immune systems, but also extends between the whole body and the brain, with numerous implications for aging. Several speakers focused on the key conduit: the vagus nerve, which runs from the brain to the body’s major organs.

For instance, Sara Prescott, an investigator in the Picower Institute and an MIT assistant professor of biology, presented evidence her lab is amassing that the brain’s communication via vagus nerve terminals in the body’s airways is crucial for managing the body’s defense of respiratory tissues. Given that we inhale about 20,000 times a day, our airways are exposed to many environmental challenges, Prescott noted, and her lab and others are finding that the nervous system interacts directly with immune pathways to mount physiological responses. But vagal reflexes decline in aging, she noted, increasing susceptibility to infection, and so her lab is now working in mouse models to study airway-to-brain neurons throughout the lifespan to better understand how they change with aging.

In his talk, Caltech Professor Sarkis Mazmanian focused on work in his lab linking the gut microbiome to Parkinson’s disease (PD), for instance by promoting alpha-synuclein protein pathology and motor problems in mouse models. His lab hypothesizes that the microbiome can nucleate alpha-synuclein in the gut via a bacterial amyloid protein that may subsequently promote pathology in the brain, potentially via the vagus nerve. Based on its studies, the lab has developed two interventions. One is giving alpha-synuclein overexpressing mice a high-fiber diet to increase short-chain fatty acids in their gut, which actually modulates the activity of microglia in the brain. The high-fiber diet helps relieve motor dysfunction, corrects microglia activity, and reduces protein pathology, he showed. Another is a drug to disrupt the bacterial amyloid in the gut. It prevents alpha synuclein formation in the mouse brain and ameliorates PD-like symptoms. These results are pending publication.

Meanwhile, Kevin Tracey, professor at Hofstra University and Northwell Health, took listeners on a journey up and down the vagus nerve to the spleen, describing how impulses in the nerve regulate immune system emissions of signaling molecules, or “cytokines.” Too great a surge can become harmful, for instance causing the autoimmune disorder rheumatoid arthritis. Tracey described how a newly U.S. Food and Drug Administration-approved pill-sized neck implant to stimulate the vagus nerve helps patients with severe forms of the disease without suppressing their immune system.

The brain’s border

Other speakers discussed opportunities for understanding neuro-immune interactions in aging and disease at the “borders” where the brain’s and body’s immune system meet. These areas include the meninges that surround the brain, the choroid plexus (proximate to the ventricles, or open spaces, within the brain), and the interface between brain cells and the circulatory system.

For instance, taking a cue from studies showing that circadian disruptions are a risk factor for Alzheimer’s disease, Harvard Medical School Professor Beth Stevens of Boston Children’s Hospital described new research in her lab that examined how brain immune cells may function differently around the day-night cycle. The project, led by newly minted PhD Helena Barr, found that “border-associated macrophages” — long-lived immune cells residing in the brain’s borders — exhibited circadian rhythms in gene expression and function. Stevens described how these cells are tuned by the circadian clock to “eat” more during the rest phase, a process that may help remove material draining from the brain, including Alzheimer’s disease-associated peptides such as amyloid-beta. So, Stevens hypothesizes, circadian disruptions, for example due to aging or night-shift work, may contribute to disease onset by disrupting the delicate balance in immune-mediated “clean-up” of the brain and its borders.

Following Stevens at the podium, Washington University Professor Marco Colonna traced how various kinds of macrophages, including border macrophages and microglia, develop from the embryonic stage. He described the different gene-expression programs that guide their differentiation into one type or another. One gene he highlighted, for instance, is necessary for border macrophages along the brain’s vasculature to help regulate the waste-clearing cerebrospinal fluid (CSF) flow that Stevens also discussed. Knocking out the gene also impairs blood flow. Importantly, his lab has found that versions of the gene may be somewhat protective against Alzheimer’s, and that regulating expression of the gene could be a therapeutic strategy.

Colonna’s WashU colleague Jonathan Kipnis (a former student of Schwartz) also discussed macrophages that are associated with the particular border between brain tissue and the plumbing alongside the vasculature that carries CSF. The macrophages, his lab showed in 2022, actively govern the flow of CSF. He showed that removing the macrophages let Alzheimer’s proteins accumulate in mice. His lab is continuing to investigate ways in which these specific border macrophages may play roles in disease. He’s also looking in separate studies of how the skull’s brain marrow contributes to the population of immune cells in the brain and may play a role in neurodegeneration.

For all the talk of distant organs and the brain’s borders, neurons themselves were never far from the discussion. Harvard Medical School Professor Isaac Chiu gave them their direct due in a talk focusing on how they participate in their own immune defense, for instance by directly sensing pathogens and giving off inflammation signals upon cell death. He discussed a key molecule in that latter process, which is expressed among neurons all over the brain.

Whether they were looking within the brain, at its border, or throughout the body, speakers showed that age-related nervous system diseases are not only better understood but also possibly better treated by accounting not only for the nerve cells, but their immune system partners. 

The MIT Schwarzman College of Computing and the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) recently celebrated the launch of the MIT–MBZUAI Collaborative Research Program, a new effort to strengthen the building blocks of artificial intelligence and accelerate its use in pressing scientific and societal challenges.

Under the five-year agreement, faculty, students, and research staff from both institutions will collaborate on fundamental research projects to advance the technological foundations of AI and its applications in three core areas: scientific discovery, human thriving, and the health of the planet.

“Artificial intelligence is transforming nearly every aspect of human endeavor. MIT’s leadership in AI is greatly enriched through collaborations with leading academic institutions in the U.S. and around the world,” says Dan Huttenlocher, dean of the MIT Schwarzman College of Computing and the Henry Ellis Warren Professor of Electrical Engineering and Computer Science. “Our collaboration with MBZUAI reflects a shared commitment to advancing AI in ways that are responsible, inclusive, and globally impactful. Together, we can explore new horizons in AI and bring broad benefits to society.”

“This agreement will unite the efforts of researchers at two world-class institutions to advance frontier AI research across scientific discovery, human thriving, and the health of the planet. By combining MBZUAI’s focus on foundational models and real-world deployment with MIT’s depth in computing and interdisciplinary innovation, we are creating a transcontinental bridge for discovery. Together, we will not only expand the boundaries of AI science, but also ensure that these breakthroughs are pursued responsibly and applied where they matter most — improving human health, enabling intelligent robotics, and driving sustainable AI at scale,” says Eric Xing, president and university professor at MBZUAI.

Each institution has appointed an academic director to oversee the program on its campus. At MIT, Philip Isola, the Class of 1948 Career Development Professor in the Department of Electrical Engineering and Computer Science, will serve as program lead. At MBZUAI, Le Song, professor of machine learning, will take on the role.

Supported by MBZUAI — the first university dedicated entirely to advancing science through AI, and based in Abu Dhabi, U.A.E. — the collaboration will fund a number of joint research projects per year. The findings will be openly publishable, and each project will be led by a principal investigator from MIT and one from MBZUAI, with project selections made by a steering committee composed of representatives from both institutions.

Ammonia is one of the most widely produced chemicals in the world, used mostly as fertilizer, but also for the production of some plastics, textiles, and other applications. Its production, through processes that require high heat and pressure, accounts for up to 20 percent of all the greenhouse gases from the entire chemical industry, so efforts have been underway worldwide to find ways to reduce those emissions.

Now, researchers at MIT have come up with a clever way of combining two different methods of producing the compound that minimizes waste products, that, when combined with some other simple upgrades, could reduce the greenhouse emissions from production by as much as 63 percent, compared to the leading “low-emissions” approach being used today.

The new approach is described in the journal Energy & Fuels, in a paper by MIT Energy Initiative (MITEI) Director William H. Green, graduate student Sayandeep Biswas, MITEI Director of Research Randall Field, and two others.

“Ammonia has the most carbon dioxide emissions of any kind of chemical,” says Green, who is the Hoyt C. Hottel Professor in Chemical Engineering. “It’s a very important chemical,” he says, because its use as a fertilizer is crucial to being able to feed the world’s population.

Until late in the 19th century, the most widely used source of nitrogen fertilizer was mined deposits of bat or bird guano, mostly from Chile, but that source was beginning to run out, and there were predictions that the world would soon be running short of food to sustain the population. But then a new chemical process, called the Haber-Bosch process after its inventors, made it possible to make ammonia out of nitrogen from the air and hydrogen, which was mostly derived from methane. But both the burning of fossil fuels to provide the needed heat and the use of methane to make the hydrogen led to massive climate-warming emissions from the process.

To address this, two newer variations of ammonia production have been developed: so-called “blue ammonia,” where the greenhouse gases are captured right at the factory and then sequestered deep underground, and “green ammonia,” produced by a different chemical pathway, using electricity instead of fossil fuels to hydrolyze water to make hydrogen.

Blue ammonia is already beginning to be used, with a few plants operating now in Louisiana, Green says, and the ammonia mostly being shipped to Japan, “so that’s already kind of commercial.” Other parts of the world are starting to use green ammonia, especially in places that have lots of hydropower, solar, or wind to provide inexpensive electricity, including a giant plant now under construction in Saudi Arabia.

But in most places, both blue and green ammonia are still more expensive than the traditional fossil-fuel-based version, so many teams around the world have been working on ways to cut these costs as much as possible so that the difference is small enough to be made up through tax subsidies or other incentives.

The problem is growing, because as the population grows, and as wealth increases, there will be ever-increasing demands for nitrogen fertilizer. At the same time, ammonia is a promising substitute fuel to power hard-to-decarbonize transportation such as cargo ships and heavy trucks, which could lead to even greater needs for the chemical.

“It definitely works” as a transportation fuel, by powering fuel cells that have been demonstrated for use by everything from drones to barges and tugboats and trucks, Green says. “People think that the most likely market of that type would be for shipping,” he says, “because the downside of ammonia is it’s toxic and it’s smelly, and that makes it slightly dangerous to handle and to ship around.” So its best uses may be where it’s used in high volume and in relatively remote locations, like the high seas. In fact, the International Maritime Organization will soon be voting on new rules that might give a strong boost to the ammonia alternative for shipping.

The key to the new proposed system is to combine the two existing approaches in one facility, with a blue ammonia factory next to a green ammonia factory. The process of generating hydrogen for the green ammonia plant leaves a lot of leftover oxygen that just gets vented to the air. Blue ammonia, on the other hand, uses a process called autothermal reforming that requires a source of pure oxygen, so if there’s a green ammonia plant next door, it can use that excess oxygen.

“Putting them next to each other turns out to have significant economic value,” Green says. This synergy could help hybrid “blue-green ammonia” facilities serve as an important bridge toward a future where eventually green ammonia, the cleanest version, could finally dominate. But that future is likely decades away, Green says, so having the combined plants could be an important step along the way.

“It might be a really long time before [green ammonia] is actually attractive” economically, he says. “Right now, it’s nowhere close, except in very special situations.” But the combined plants “could be a really appealing concept, and maybe a good way to start the industry,” because so far only small, standalone demonstration plants of the green process are being built.

“If green or blue ammonia is going to become the new way of making ammonia, you need to find ways to make it relatively affordable in a lot of countries, with whatever resources they’ve got,” he says. This new proposed combination, he says, “looks like a really good idea that can help push things along. Ultimately, there’s got to be a lot of green ammonia plants in a lot of places,” and starting out with the combined plants, which could be more affordable now, could help to make that happen. The team has filed for a patent on the process.

Although the team did a detailed study of both the technology and the economics that show the system has great promise, Green points out that “no one has ever built one. We did the analysis, it looks good, but surely when people build the first one, they’ll find funny little things that need some attention,” such as details of how to start up or shut down the process. “I would say there’s plenty of additional work to do to make it a real industry.” But the results of this study, which shows the costs to be much more affordable than existing blue or green plants in isolation, “definitely encourages the possibility of people making the big investments that would be needed to really make this industry feasible.”

This proposed integration of the two methods “improves efficiency, reduces greenhouse gas emissions, and lowers overall cost,” says Kevin van Geem, a professor in the Center for Sustainable Chemistry at Ghent University, who was not associated with this research. “The analysis is rigorous, with validated process models, transparent assumptions, and comparisons to literature benchmarks. By combining techno-economic analysis with emissions accounting, the work provides a credible and balanced view of the trade-offs.”

He adds that, “given the scale of global ammonia production, such a reduction could have a highly impactful effect on decarbonizing one of the most emissions-intensive chemical industries.”

The research team also included MIT postdoc Angiras Menon and MITEI research lead Guiyan Zang. The work was supported by IHI Japan through the MIT Energy Initiative and the Martin Family Society of Fellows for Sustainability. 

Chatbots like ChatGPT and Claude have experienced a meteoric rise in usage over the past three years because they can help you with a wide range of tasks. Whether you’re writing Shakespearean sonnets, debugging code, or need an answer to an obscure trivia question, artificial intelligence systems seem to have you covered. The source of this versatility? Billions, or even trillions, of textual data points across the internet.

Those data aren’t enough to teach a robot to be a helpful household or factory assistant, though. To understand how to handle, stack, and place various arrangements of objects across diverse environments, robots need demonstrations. You can think of robot training data as a collection of how-to videos that walk the systems through each motion of a task. Collecting these demonstrations on real robots is time-consuming and not perfectly repeatable, so engineers have created training data by generating simulations with AI (which don’t often reflect real-world physics), or tediously handcrafting each digital environment from scratch.

Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Toyota Research Institute may have found a way to create the diverse, realistic training grounds robots need. Their “steerable scene generation” approach creates digital scenes of things like kitchens, living rooms, and restaurants that engineers can use to simulate lots of real-world interactions and scenarios. Trained on over 44 million 3D rooms filled with models of objects such as tables and plates, the tool places existing assets in new scenes, then refines each one into a physically accurate, lifelike environment.

Steerable scene generation creates these 3D worlds by “steering” a diffusion model — an AI system that generates a visual from random noise — toward a scene you’d find in everyday life. The researchers used this generative system to “in-paint” an environment, filling in particular elements throughout the scene. You can imagine a blank canvas suddenly turning into a kitchen scattered with 3D objects, which are gradually rearranged into a scene that imitates real-world physics. For example, the system ensures that a fork doesn’t pass through a bowl on a table — a common glitch in 3D graphics known as “clipping,” where models overlap or intersect.

How exactly steerable scene generation guides its creation toward realism, however, depends on the strategy you choose. Its main strategy is “Monte Carlo tree search” (MCTS), where the model creates a series of alternative scenes, filling them out in different ways toward a particular objective (like making a scene more physically realistic, or including as many edible items as possible). It’s used by the AI program AlphaGo to beat human opponents in Go (a game similar to chess), as the system considers potential sequences of moves before choosing the most advantageous one.

“We are the first to apply MCTS to scene generation by framing the scene generation task as a sequential decision-making process,” says MIT Department of Electrical Engineering and Computer Science (EECS) PhD student Nicholas Pfaff, who is a CSAIL researcher and a lead author on a paper presenting the work. “We keep building on top of partial scenes to produce better or more desired scenes over time. As a result, MCTS creates scenes that are more complex than what the diffusion model was trained on.”

In one particularly telling experiment, MCTS added the maximum number of objects to a simple restaurant scene. It featured as many as 34 items on a table, including massive stacks of dim sum dishes, after training on scenes with only 17 objects on average.

Steerable scene generation also allows you to generate diverse training scenarios via reinforcement learning — essentially, teaching a diffusion model to fulfill an objective by trial-and-error. After you train on the initial data, your system undergoes a second training stage, where you outline a reward (basically, a desired outcome with a score indicating how close you are to that goal). The model automatically learns to create scenes with higher scores, often producing scenarios that are quite different from those it was trained on.

Users can also prompt the system directly by typing in specific visual descriptions (like “a kitchen with four apples and a bowl on the table”). Then, steerable scene generation can bring your requests to life with precision. For example, the tool accurately followed users’ prompts at rates of 98 percent when building scenes of pantry shelves, and 86 percent for messy breakfast tables. Both marks are at least a 10 percent improvement over comparable methods like “MiDiffusion” and “DiffuScene.”

The system can also complete specific scenes via prompting or light directions (like “come up with a different scene arrangement using the same objects”). You could ask it to place apples on several plates on a kitchen table, for instance, or put board games and books on a shelf. It’s essentially “filling in the blank” by slotting items in empty spaces, but preserving the rest of a scene.

According to the researchers, the strength of their project lies in its ability to create many scenes that roboticists can actually use. “A key insight from our findings is that it’s OK for the scenes we pre-trained on to not exactly resemble the scenes that we actually want,” says Pfaff. “Using our steering methods, we can move beyond that broad distribution and sample from a ‘better’ one. In other words, generating the diverse, realistic, and task-aligned scenes that we actually want to train our robots in.”

Such vast scenes became the testing grounds where they could record a virtual robot interacting with different items. The machine carefully placed forks and knives into a cutlery holder, for instance, and rearranged bread onto plates in various 3D settings. Each simulation appeared fluid and realistic, resembling the real-world, adaptable robots steerable scene generation could help train, one day.

While the system could be an encouraging path forward in generating lots of diverse training data for robots, the researchers say their work is more of a proof of concept. In the future, they’d like to use generative AI to create entirely new objects and scenes, instead of using a fixed library of assets. They also plan to incorporate articulated objects that the robot could open or twist (like cabinets or jars filled with food) to make the scenes even more interactive.

To make their virtual environments even more realistic, Pfaff and his colleagues may incorporate real-world objects by using a library of objects and scenes pulled from images on the internet and using their previous work on “Scalable Real2Sim.” By expanding how diverse and lifelike AI-constructed robot testing grounds can be, the team hopes to build a community of users that’ll create lots of data, which could then be used as a massive dataset to teach dexterous robots different skills.

“Today, creating realistic scenes for simulation can be quite a challenging endeavor; procedural generation can readily produce a large number of scenes, but they likely won’t be representative of the environments the robot would encounter in the real world. Manually creating bespoke scenes is both time-consuming and expensive,” says Jeremy Binagia, an applied scientist at Amazon Robotics who wasn’t involved in the paper. “Steerable scene generation offers a better approach: train a generative model on a large collection of pre-existing scenes and adapt it (using a strategy such as reinforcement learning) to specific downstream applications. Compared to previous works that leverage an off-the-shelf vision-language model or focus just on arranging objects in a 2D grid, this approach guarantees physical feasibility and considers full 3D translation and rotation, enabling the generation of much more interesting scenes.”

“Steerable scene generation with post training and inference-time search provides a novel and efficient framework for automating scene generation at scale,” says Toyota Research Institute roboticist Rick Cory SM ’08, PhD ’10, who also wasn’t involved in the paper. “Moreover, it can generate ‘never-before-seen’ scenes that are deemed important for downstream tasks. In the future, combining this framework with vast internet data could unlock an important milestone towards efficient training of robots for deployment in the real world.”

Pfaff wrote the paper with senior author Russ Tedrake, the Toyota Professor of Electrical Engineering and Computer Science, Aeronautics and Astronautics, and Mechanical Engineering at MIT; a senior vice president of large behavior models at the Toyota Research Institute; and CSAIL principal investigator. Other authors were Toyota Research Institute robotics researcher Hongkai Dai SM ’12, PhD ’16; team lead and Senior Research Scientist Sergey Zakharov; and Carnegie Mellon University PhD student Shun Iwase. Their work was supported, in part, by Amazon and the Toyota Research Institute. The researchers presented their work at the Conference on Robot Learning (CoRL) in September.

Every time you check the time on your phone, make an online transaction, or use a navigation app, you are depending on the precision of atomic clocks.

An atomic clock keeps time by relying on the “ticks” of atoms as they naturally oscillate at rock-steady frequencies. Today’s atomic clocks operate by tracking cesium atoms, which tick over 10 billion times per second. Each of those ticks is precisely tracked using lasers that oscillate in sync, at microwave frequencies.

Scientists are developing next-generation atomic clocks that rely on even faster-ticking atoms such as ytterbium, which can be tracked with lasers at higher, optical frequencies. If they can be kept stable, optical atomic clocks could track even finer intervals of time, up to 100 trillion times per second.

Now, MIT physicists have found a way to improve the stability of optical atomic clocks, by reducing “quantum noise” — a fundamental measurement limitation due to the effects of quantum mechanics, which obscures the atoms’ pure oscillations. In addition, the team discovered that an effect of a clock’s laser on the atoms, previously considered irrelevant, can be used to further stabilize the laser.

The researchers developed a method to harness a laser-induced “global phase” in ytterbium atoms, and have boosted this effect with a quantum-amplification technique. The new approach doubles the precision of an optical atomic clock, enabling it to discern twice as many ticks per second compared to the same setup without the new method. What’s more, they anticipate that the precision of the method should increase steadily with the number of atoms in an atomic clock.

The researchers detail the method, which they call global phase spectroscopy, in a study appearing today in the journal Nature. They envision that the clock-stabilizing technique could one day enable portable optical atomic clocks that can be transported to various locations to measure all manner of phenomena.

“With these clocks, people are trying to detect dark matter and dark energy, and test whether there really are just four fundamental forces, and even to see if these clocks can predict earthquakes,” says study author Vladan Vuletić, the Lester Wolfe Professor of Physics at MIT. “We think our method can help make these clocks transportable and deployable to where they’re needed.”

The paper’s co-authors are Leon Zaporski, Qi Liu, Gustavo Velez, Matthew Radzihovsky, Zeyang Li, Simone Colombo, and Edwin Pedrozo-Peñafiel, who are members of the MIT-Harvard Center for Ultracold Atoms and the MIT Research Laboratory of Electronics.

Ticking time

In 2020, Vuletić and his colleagues demonstrated that an atomic clock could be made more precise by quantumly entangling the clock’s atoms. Quantum entanglement is a phenomenon by which particles can be made to behave in a collective, highly correlated manner. When atoms are quantumly entangled, they redistribute any noise, or uncertainty in measuring the atoms’ oscillations, in a way that reveals a clearer, more measurable “tick.”

In their previous work, the team induced quantum entanglement among several hundred ytterbium atoms that they first cooled and trapped in a cavity formed by two curved mirrors. They sent a laser into the cavity, which bounced thousands of times between the mirrors, interacting with the atoms and causing the ensemble to entangle. They were able to show that quantum entanglement could improve the precision of existing atomic clocks by essentially reducing the noise, or uncertainty between the laser’s and atoms’ tick rates.

At the time, however, they were limited by the ticking instability of the clock’s laser. In 2022, the same team derived a way to further amplify the difference in laser versus atom tick rates with “time reversal” — a trick that relies on entangling and de-entangling the atoms to boost the signal acquired in between.

However, in that work the team was still using traditional microwaves, which oscillate at much lower frequencies than the optical frequency standards ytterbium atoms can provide. It was as if they had painstakingly lifted a film of dust off a painting, only to then photograph it with a low-resolution camera.

“When you have atoms that tick 100 trillion times per second, that’s 10,000 times faster than the frequency of microwaves,” Vuletić says. “We didn’t know at the time how to apply these methods to higher-frequency optical clocks that are much harder to keep stable.”

About phase

In their new study, the team has found a way to apply their previously developed approach of time reversal to optical atomic clocks. They then sent in a laser that oscillates near the optical frequency of the entangled atoms.

“The laser ultimately inherits the ticking of the atoms,” says first author Zaporski. “But in order for this inheritance to hold for a long time, the laser has to be quite stable.”

The researchers found they were able to improve the stability of an optical atomic clock by taking advantage of a phenomenon that scientists had assumed was inconsequential to the operation. They realized that when light is sent through entangled atoms, the interaction can cause the atoms to jump up in energy, then settle back down into their original energy state and still carry the memory about their round trip.

“One might think we’ve done nothing,” Vuletić says. “You get this global phase of the atoms, which is usually considered irrelevant. But this global phase contains information about the laser frequency.”

In other words, they realized that the laser was inducing a measurable change in the atoms, despite bringing them back to the original energy state, and that the magnitude of this change depends on the laser’s frequency.

“Ultimately, we are looking for the difference of laser frequency and the atomic transition frequency,” explains co-author Liu. “When that difference is small, it gets drowned by quantum noise. Our method amplifies this difference above this quantum noise.”

In their experiments, the team applied this new approach and found that through entanglement they were able to double the precision of their optical atomic clock.

“We saw that we can now resolve nearly twice as small a difference in the optical frequency or, the clock ticking frequency, without running into the quantum noise limit,” Zaporski says. “Although it’s a hard problem in general to run atomic clocks, the technical benefits of our method it will make it easier, and we think this can enable stable, transportable atomic clocks.”

This research was supported, in part, by the U.S. Office of Naval Research, the National Science Foundation, the U.S. Defense Advanced Research Projects Agency, the U.S. Department of Energy, the U.S. Office of Science, the National Quantum Information Science Research Centers, and the Quantum Systems Accelerator.

For decades, it’s been known that subtle chemical patterns exist in metal alloys, but researchers thought they were too minor to matter — or that they got erased during manufacturing. However, recent studies have shown that in laboratory settings, these patterns can change a metal’s properties, including its mechanical strength, durability, heat capacity, radiation tolerance, and more.

Now, researchers at MIT have found that these chemical patterns also exist in conventionally manufactured metals. The surprising finding revealed a new physical phenomenon that explains the persistent patterns.

In a paper published in Nature Communications today, the researchers describe how they tracked the patterns and discovered the physics that explains them. The authors also developed a simple model to predict chemical patterns in metals, and they show how engineers could use the model to tune the effect of such patterns on metallic properties, for use in aerospace, semiconductors, nuclear reactors, and more.

“The conclusion is: You can never completely randomize the atoms in a metal. It doesn’t matter how you process it,” says Rodrigo Freitas, the TDK Assistant Professor in the Department of Materials Science and Engineering. “This is the first paper showing these non-equilibrium states that are retained in the metal. Right now, this chemical order is not something we’re controlling for or paying attention to when we manufacture metals.”

For Freitas, an early-career researcher, the findings offer vindication for exploring a crowded field that he says few believed would lead to unique or broadly impactful results. He credits the U.S. Air Force Office of Scientific Research, which supported the work through their Young Investigator Program. He also credits the collaborative effort that enabled the paper, which features three MIT PhD students as co-first authors: Mahmudul Islam, Yifan Cao, and Killian Sheriff.

“There was the question of whether I should even be tackling this specific problem because people have been working on it for a long time,” Freitas says. “But the more I learned about it, the more I saw researchers were thinking about this in idealized laboratory scenarios. We wanted to perform simulations that were as realistic as possible to reproduce these manufacturing processes with high fidelity. My favorite part of this project is how non-intuitive the findings are. The fact that you cannot completely mix something together, people didn’t see that coming.”

From surprises to theories

Freitas’ research team began with a practical question: How fast do chemical elements mix during metal processing? Conventional wisdom held that there’s a point where the chemical composition of metals becomes completely uniform from mixing during manufacturing. By finding that point, the researchers thought they could develop a simple way to design alloys with different levels of atomic order, also known as short-range order.

The researchers used machine-learning techniques to track millions of atoms as they moved and rearranged themselves under conditions that mimicked metal processing.

“The first thing we did was to deform a piece of metal,” Freitas explains. “That’s a common step during manufacturing: You roll the metal and deform it and heat it up again and deform it a little more, so it develops the structure you want. We did that and we tracked chemical order. The thought was as you deform the material, its chemical bonds are broken and that randomizes the system. These violent manufacturing processes essentially shuffle the atoms.”

The researchers hit a snag during the mixing process: The alloys never reached a fully random state. That was a surprise, because no known physical mechanism could explain the result.

“It pointed to a new piece of physics in metals,” the researchers write in the paper. “It was one of those cases where applied research led to a fundamental discovery.”

To uncover the new physics, the researchers developed computational tools, including high-fidelity machine-learning models, to capture atomic interactions, along with new statistical methods that quantify how chemical order changes over time. They then applied these tools in large-scale molecular dynamics simulations to track how atoms rearrange during processing.

The researchers found some standard chemical arrangements in their processed metals, but at higher temperatures than would normally be expected. Even more surprisingly, they found completely new chemical patterns never seen outside of manufacturing processes. This was the first time such patterns were observed. The researchers referred to the patterns as “far-from-equilibrium states.”

The researchers also built a simple model that reproduced key features of the simulations. The model explains how the chemical patterns arise from defects known as dislocations, which are like three-dimensional scribbles within a metal. As the metal is deformed, those scribbles warp, shuffling nearby atoms along the way. Previously, researchers believed that shuffling completely erased order in the metals, but they found that dislocations favor some atomic swaps over others, resulting not in randomness but in subtle patterns that explain their findings.

“These defects have chemical preferences that guide how they move,” Freitas says. “They look for low energy pathways, so given a choice between breaking chemical bonds, they tend to break the weakest bonds, and it’s not completely random. This is very exciting because it’s a non-equilibrium state: It’s not something you’d see naturally occurring in materials. It’s the same way our bodies live in non-equilibrium. The temperature outside is always hotter or colder than our bodies, and we’re maintaining that steady state equilibrium to stay alive. That’s why these states exist in metal: the balance between an internal push toward disorder plus this ordering tendency of breaking certain bonds that are always weaker than others.”

Applying a new theory

The researchers are now exploring how these chemical patterns develop across a wide range of manufacturing conditions. The result is a map that links various metal processing steps to different chemical patterns in metal.

To date, this chemical order and the properties they tune have been largely considered an academic subject. With this map, the researchers hope engineers can begin thinking of these patterns as levers in design that can be pulled during production to get new properties.

“Researchers have been looking at the ways these atomic arrangements change metallic properties — a big one is catalysis,” Freitas says of the process that drives chemical reactions. “Electrochemistry happens at the surface of the metal, and it’s very sensitive to local atomic arrangements. And there have been other properties that you wouldn't think would be influenced by these factors. Radiation damage is another big one. That affects these materials’ performance in nuclear reactors.”

Researchers have already told Freitas the paper could help explain other surprise findings about metallic properties, and he’s excited for the field to move from fundamental research into chemical order to more applied work.

“You can think of areas where you need very optimized alloys like aerospace,” Freitas says. “They care about very specific compositions. Advanced manufacturing now makes it possible to combine metals that normally wouldn’t mix through deformation. Understanding how atoms actually shuffle and mix in those processes is crucial, because it’s the key to gaining strength while still keeping the low density. So, this could be a huge deal for them.”

This work was supported, in part, by the U.S. Air Force Office of Scientific Research, MathWorks, and the MIT-Portugal Program.

One of the newest weapons that scientists have developed against cancer is a type of engineered immune cell known as CAR-NK (natural killer) cells. Similar to CAR-T cells, these cells can be programmed to attack cancer cells.

MIT and Harvard Medical School researchers have now come up with a new way to engineer CAR-NK cells that makes them much less likely to be rejected by the patient’s immune system, which is a common drawback of this type of treatment.

The new advance may also make it easier to develop “off-the-shelf” CAR-NK cells that could be given to patients as soon as they are diagnosed. Traditional approaches to engineering CAR-NK or CAR-T cells usually take several weeks.

“This enables us to do one-step engineering of CAR-NK cells that can avoid rejection by host T cells and other immune cells. And, they kill cancer cells better and they’re safer,” says Jianzhu Chen, an MIT professor of biology, a member of the Koch Institute for Integrative Cancer Research,and one of the senior authors of the study.

In a study of mice with humanized immune systems, the researchers showed that these CAR-NK cells could destroy most cancer cells while evading the host immune system.

Rizwan Romee, an associate professor of medicine at Harvard Medical School and Dana-Farber Cancer Institute, is also a senior author of the paper, which appears today in Nature Communications. The paper’s lead author is Fuguo Liu, a postdoc at the Koch Institute and a research fellow at Dana-Farber.

Evading the immune system

NK cells are a critical part of the body’s natural immune defenses, and their primary responsibility is to locate and kill cancer cells and virus-infected cells. One of their cell-killing strategies, also used by T cells, is a process called degranulation. Through this process, immune cells release a protein called perforin, which can poke holes in another cell to induce cell death.

To create CAR-NK cells to treat cancer patients, doctors first take a blood sample from the patient. NK cells are isolated from the sample and engineered to express a protein called a chimeric antigen receptor (CAR), which can be designed to target specific proteins found on cancer cells.

Then, the cells spend several weeks proliferating until there are enough to transfuse back into the patient. A similar approach is also used to create CAR-T cells. Several CAR-T cell therapies have been approved to treat blood cancers such as lymphoma and leukemia, but CAR-NK treatments are still in clinical trials.

Because it takes so long to grow a population of engineered cells that can be infused into the patient, and those cells may not be as viable as cells that came from a healthy person, researchers are exploring an alternative approach: using NK cells from a healthy donor.

Such cells could be grown in large quantities and would be ready whenever they were needed. However, the drawback to these cells is that the recipient’s immune system may see them as foreign and attack them before they can start killing cancer cells.

In the new study, the MIT team set out to find a way to help NK cells “hide” from a patient’s immune system. Through studies of immune cell interactions, they showed that NK cells could evade a host T-cell response if they did not carry surface proteins called HLA class 1 proteins. These proteins, usually expressed on NK cell surfaces, can trigger T cells to attack if the immune system doesn’t recognize them as “self.”

To take advantage of this, the researchers engineered the cells to express a sequence of siRNA (short interfering RNA) that interferes with the genes for HLA class 1. They also delivered the CAR gene, as well as the gene for either PD-L1 or single-chain HLA-E (SCE). PD-L1 and SCE are proteins that make NK cells more effective by turning up genes that are involved in killing cancer cells.

All of these genes can be carried on a single piece of DNA, known as a construct, making it simple to transform donor NK cells into immune-evasive CAR-NK cells. The researchers used this construct to create CAR-NK cells targeting a protein called CD-19, which is often found on cancerous B cells in lymphoma patients.

NK cells unleashed

The researchers tested these CAR-NK cells in mice with a human-like immune system. These mice were also injected with lymphoma cells.

Mice that received CAR-NK cells with the new construct maintained the NK cell population for at least three weeks, and the NK cells were able to nearly eliminate cancer in those mice. In mice that received either NK cells with no genetic modifications or NK cells with only the CAR gene, the host immune cells attacked the donor NK cells. In these mice, the NK cells died out within two weeks, and the cancer spread unchecked.

The researchers also found that these engineered CAR-NK cells were much less likely to induce cytokine release syndrome — a common side effect of immunotherapy treatments, which can cause life-threatening complications.

Because of CAR-NK cells’ potentially better safety profile, Chen anticipates that they could eventually be used in place of CAR-T cells. For any CAR-NK cells that are now in development to target lymphoma or other types of cancer, it should be possible to adapt them by adding the construct developed in this study, he says.

The researchers now hope to run a clinical trial of this approach, working with colleagues at Dana-Farber. They are also working with a local biotech company to test CAR-NK cells to treat lupus, an autoimmune disorder that causes the immune system to attack healthy tissues and organs.

The research was funded, in part, by Skyline Therapeutics, the Koch Institute Frontier Research Program through the Kathy and Curt Marble Cancer Research Fund and the Elisa Rah (2004, 2006) Memorial Fund, the Claudia Adams Barr Foundation, and the Koch Institute Support (core) Grant from the National Cancer Institute.

Tokamaks are machines that are meant to hold and harness the power of the sun. These fusion machines use powerful magnets to contain a plasma hotter than the sun’s core and push the plasma’s atoms to fuse and release energy. If tokamaks can operate safely and efficiently, the machines could one day provide clean and limitless fusion energy.

Today, there are a number of experimental tokamaks in operation around the world, with more underway. Most are small-scale research machines built to investigate how the devices can spin up plasma and harness its energy. One of the challenges that tokamaks face is how to safely and reliably turn off a plasma current that is circulating at speeds of up to 100 kilometers per second, at temperatures of over 100 million degrees Celsius.

Such “rampdowns” are necessary when a plasma becomes unstable. To prevent the plasma from further disrupting and potentially damaging the device’s interior, operators ramp down the plasma current. But occasionally the rampdown itself can destabilize the plasma. In some machines, rampdowns have caused scrapes and scarring to the tokamak’s interior — minor damage that still requires considerable time and resources to repair.

Now, scientists at MIT have developed a method to predict how plasma in a tokamak will behave during a rampdown. The team combined machine-learning tools with a physics-based model of plasma dynamics to simulate a plasma’s behavior and any instabilities that may arise as the plasma is ramped down and turned off. The researchers trained and tested the new model on plasma data from an experimental tokamak in Switzerland. They found the method quickly learned how plasma would evolve as it was tuned down in different ways. What’s more, the method achieved a high level of accuracy using a relatively small amount of data. This training efficiency is promising, given that each experimental run of a tokamak is expensive and quality data is limited as a result.

The new model, which the team highlights this week in an open-access Nature Communications paper, could improve the safety and reliability of future fusion power plants.

“For fusion to be a useful energy source it’s going to have to be reliable,” says lead author Allen Wang, a graduate student in aeronautics and astronautics and a member of the Disruption Group at MIT’s Plasma Science and Fusion Center (PSFC). “To be reliable, we need to get good at managing our plasmas.”

The study’s MIT co-authors include PSFC Principal Research Scientist and Disruptions Group leader Cristina Rea, and members of the Laboratory for Information and Decision Systems (LIDS) Oswin So, Charles Dawson, and Professor Chuchu Fan, along with Mark (Dan) Boyer of Commonwealth Fusion Systems and collaborators from the Swiss Plasma Center in Switzerland.

“A delicate balance”

Tokamaks are experimental fusion devices that were first built in the Soviet Union in the 1950s. The device gets its name from a Russian acronym that translates to a “toroidal chamber with magnetic coils.” Just as its name describes, a tokamak is toroidal, or donut-shaped, and uses powerful magnets to contain and spin up a gas to temperatures and energies high enough that atoms in the resulting plasma can fuse and release energy.

Today, tokamak experiments are relatively low-energy in scale, with few approaching the size and output needed to generate safe, reliable, usable energy. Disruptions in experimental, low-energy tokamaks are generally not an issue. But as fusion machines scale up to grid-scale dimensions, controlling much higher-energy plasmas at all phases will be paramount to maintaining a machine’s safe and efficient operation.

“Uncontrolled plasma terminations, even during rampdown, can generate intense heat fluxes damaging the internal walls,” Wang notes. “Quite often, especially with the high-performance plasmas, rampdowns actually can push the plasma closer to some instability limits. So, it’s a delicate balance. And there’s a lot of focus now on how to manage instabilities so that we can routinely and reliably take these plasmas and safely power them down. And there are relatively few studies done on how to do that well.”

Bringing down the pulse

Wang and his colleagues developed a model to predict how a plasma will behave during tokamak rampdown. While they could have simply applied machine-learning tools such as a neural network to learn signs of instabilities in plasma data, “you would need an ungodly amount of data” for such tools to discern the very subtle and ephemeral changes in extremely high-temperature, high-energy plasmas, Wang says.

Instead, the researchers paired a neural network with an existing model that simulates plasma dynamics according to the fundamental rules of physics. With this combination of machine learning and a physics-based plasma simulation, the team found that only a couple hundred pulses at low performance, and a small handful of pulses at high performance, were sufficient to train and validate the new model.

The data they used for the new study came from the TCV, the Swiss “variable configuration tokamak” operated by the Swiss Plasma Center at EPFL (the Swiss Federal Institute of Technology Lausanne). The TCV is a small experimental fusion experimental device that is used for research purposes, often as test bed for next-generation device solutions. Wang used the data from several hundred TCV plasma pulses that included properties of the plasma such as its temperature and energies during each pulse’s ramp-up, run, and ramp-down. He trained the new model on this data, then tested it and found it was able to accurately predict the plasma’s evolution given the initial conditions of a particular tokamak run.

The researchers also developed an algorithm to translate the model’s predictions into practical “trajectories,” or plasma-managing instructions that a tokamak controller can automatically carry out to for instance adjust the magnets or temperature maintain the plasma’s stability. They implemented the algorithm on several TCV runs and found that it produced trajectories that safely ramped down a plasma pulse, in some cases faster and without disruptions compared to runs without the new method.

“At some point the plasma will always go away, but we call it a disruption when the plasma goes away at high energy. Here, we ramped the energy down to nothing,” Wang notes. “We did it a number of times. And we did things much better across the board. So, we had statistical confidence that we made things better.”

The work was supported in part by Commonwealth Fusion Systems (CFS), an MIT spinout that intends to build the world’s first compact, grid-scale fusion power plant. The company is developing a demo tokamak, SPARC, designed to produce net-energy plasma, meaning that it should generate more energy than it takes to heat up the plasma. Wang and his colleagues are working with CFS on ways that the new prediction model and tools like it can better predict plasma behavior and prevent costly disruptions to enable safe and reliable fusion power.

“We’re trying to tackle the science questions to make fusion routinely useful,” Wang says. “What we’ve done here is the start of what is still a long journey. But I think we’ve made some nice progress.”

Additional support for the research came from the framework of the EUROfusion Consortium, via the Euratom Research and Training Program and funded by the Swiss State Secretariat for Education, Research, and Innovation.

MIT engineers have developed a printable aluminum alloy that can withstand high temperatures and is five times stronger than traditionally manufactured aluminum.

The new printable metal is made from a mix of aluminum and other elements that the team identified using a combination of simulations and machine learning, which significantly pruned the number of possible combinations of materials to search through. While traditional methods would require simulating over 1 million possible combinations of materials, the team’s new machine learning-based approach needed only to evaluate 40 possible compositions before identifying an ideal mix for a high-strength, printable aluminum alloy.

When they printed the alloy and tested the resulting material, the team confirmed that, as predicted, the aluminum alloy was as strong as the strongest aluminum alloys that are manufactured today using traditional casting methods.

The researchers envision that the new printable aluminum could be made into stronger, more lightweight and temperature-resistant products, such as fan blades in jet engines. Fan blades are traditionally cast from titanium — a material that is more than 50 percent heavier and up to 10 times costlier than aluminum — or made from advanced composites.

“If we can use lighter, high-strength material, this would save a considerable amount of energy for the transportation industry,” says Mohadeseh Taheri-Mousavi, who led the work as a postdoc at MIT and is now an assistant professor at Carnegie Mellon University.

“Because 3D printing can produce complex geometries, save material, and enable unique designs, we see this printable alloy as something that could also be used in advanced vacuum pumps, high-end automobiles, and cooling devices for data centers,” adds John Hart, the Class of 1922 Professor and head of the Department of Mechanical Engineering at MIT.

Hart and Taheri-Mousavi provide details on the new printable aluminum design in a paper published in the journal Advanced Materials. The paper’s MIT co-authors include Michael Xu, Clay Houser, Shaolou Wei, James LeBeau, and Greg Olson, along with Florian Hengsbach and Mirko Schaper of Paderborn University in Germany, and Zhaoxuan Ge and Benjamin Glaser of Carnegie Mellon University.

Micro-sizing

The new work grew out of an MIT class that Taheri-Mousavi took in 2020, which was taught by Greg Olson, professor of the practice in the Department of Materials Science and Engineering. As part of the class, students learned to use computational simulations to design high-performance alloys. Alloys are materials that are made from a mix of different elements, the combination of which imparts exceptional strength and other unique properties to the material as a whole.

Olson challenged the class to design an aluminum alloy that would be stronger than the strongest printable aluminum alloy designed to date. As with most materials, the strength of aluminum depends in large part on its microstructure: The smaller and more densely packed its microscopic constituents, or “precipitates,” the stronger the alloy would be.

With this in mind, the class used computer simulations to methodically combine aluminum with various types and concentrations of elements, to simulate and predict the resulting alloy’s strength. However, the exercise failed to produce a stronger result. At the end of the class, Taheri-Mousavi wondered: Could machine learning do better?

“At some point, there are a lot of things that contribute nonlinearly to a material’s properties, and you are lost,” Taheri-Mousavi says. “With machine-learning tools, they can point you to where you need to focus, and tell you for example, these two elements are controlling this feature. It lets you explore the design space more efficiently.”

Layer by layer

In the new study, Taheri-Mousavi continued where Olson’s class left off, this time looking to identify a stronger recipe for aluminum alloy. This time, she used machine-learning techniques designed to efficiently comb through data such as the properties of elements, to identify key connections and correlations that should lead to a more desirable outcome or product.

She found that, using just 40 compositions mixing aluminum with different elements, their machine-learning approach quickly homed in on a recipe for an aluminum alloy with higher volume fraction of small precipitates, and therefore higher strength, than what the previous studies identified. The alloy’s strength was even higher than what they could identify after simulating over 1 million possibilities without using machine learning.

To physically produce this new strong, small-precipitate alloy, the team realized 3D printing would be the way to go instead of traditional metal casting, in which molten liquid aluminum is poured into a mold and is left to cool and harden. The longer this cooling time is, the more likely the individual precipitate is to grow.

The researchers showed that 3D printing, broadly also known as additive manufacturing, can be a faster way to cool and solidify the aluminum alloy. Specifically, they considered laser bed powder fusion (LBPF) — a technique by which a powder is deposited, layer by layer, on a surface in a desired pattern and then quickly melted by a laser that traces over the pattern. The melted pattern is thin enough that it solidfies quickly before another layer is deposited and similarly “printed.” The team found that LBPF’s inherently rapid cooling and solidification enabled the small-precipitate, high-strength aluminum alloy that their machine learning method predicted.

“Sometimes we have to think about how to get a material to be compatible with 3D printing,” says study co-author John Hart. “Here, 3D printing opens a new door because of the unique characteristics of the process — particularly, the fast cooling rate. Very rapid freezing of the alloy after it’s melted by the laser creates this special set of properties.”

Putting their idea into practice, the researchers ordered a formulation of printable powder, based on their new aluminum alloy recipe. They sent the powder — a mix of aluminum and five other elements — to collaborators in Germany, who printed small samples of the alloy using their in-house LPBF system. The samples were then sent to MIT where the team ran multiple tests to measure the alloy’s strength and image the samples’ microstructure.

Their results confirmed the predictions made by their initial machine learning search: The printed alloy was five times stronger than a casted counterpart and 50 percent stronger than alloys designed using conventional simulations without machine learning. The new alloy’s microstructure also consisted of a higher volume fraction of small precipitates, and was stable at high temperatures of up to 400 degrees Celsius — a very high temperature for aluminum alloys.

The researchers are applying similar machine-learning techniques to further optimize other properties of the alloy.

“Our methodology opens new doors for anyone who wants to do 3D printing alloy design,” Taheri-Mousavi says. “My dream is that one day, passengers looking out their airplane window will see fan blades of engines made from our aluminum alloys.”

This work was carried out, in part, using MIT.nano’s characterization facilities.

In a world full of competing sounds, we often have to filter out a lot of noise to hear what’s most important. This critical skill may come more easily for people with musical training, according to scientists at MIT’s McGovern Institute for Brain Research, who used brain imaging to follow what happens when people try to focus their attention on certain sounds.

When Cassia Low Manting, a recent MIT postdoc working in the labs of MIT Professor and McGovern Institute PI John Gabrieli and former McGovern Institute PI Dimitrios Pantazis, asked people to focus on a particular melody while another melody played at the same time, individuals with musical backgrounds were, unsurprisingly, better able to follow the target tune. An analysis of study participants’ brain activity suggests this advantage arises because musical training sharpens neural mechanisms that amplify the sounds they want to listen to while turning down distractions. 

“People can hear, understand, and prioritize multiple sounds around them that flow on a moment-to-moment basis,” explains Gabrieli, who is the Grover Hermann Professor of Health Sciences and Technology at MIT. “This study reveals the specific brain mechanisms that successfully process simultaneous sounds on a moment-to-moment basis and promote attention to the most important sounds. It also shows how musical training alters that processing in the mind and brain, offering insight into how experience shapes the way we listen and pay attention.”

The research team, which also included senior author Daniel Lundqvist at the Karolinska Institute in Sweden, reported their open-access findings Sept. 17 in the journal Science Advances. Manting, who is now at the Karolinska Institute, notes that the research is part of an ongoing collaboration between the two institutions.

Overcoming challenges

Participants in the study had vastly difference backgrounds when it came to music. Some were professional musicians with deep training and experience, while others struggled to differentiate between the two tunes they were played, despite each one’s distinct pitch. This disparity allowed the researchers to explore how the brain’s capacity for attention might change with experience. “Musicians are very fun to study because their brains have been morphed in ways based on their training,” Manting says. “It’s a nice model to study these training effects.”

Still, the researchers had significant challenges to overcome. It has been hard to study how the brain manages auditory attention, because when researchers use neuroimaging to monitor brain activity, they see the brain’s response to all sounds: those that the listener cares most about, as well as those the listener is trying to ignore. It is usually difficult to figure out which brain signals were triggered by which sounds.

Manting and her colleagues overcame this challenge with a method called frequency tagging. Rather than playing the melodies in their experiments at a constant volume, the volume of each melody oscillated, rising and falling with a particular frequency. Each melody had its own frequency, creating detectable patterns in the brain signals that responded to it. “When you play these two sounds simultaneously to the subject and you record the brain signal, you can say, this 39-Hertz activity corresponds to the lower-pitch sound and the 43-Hertz activity corresponds specifically to the higher-pitch sound,” Manting explains. “It is very clean and very clear.”

When they paired frequency tagging with magnetoencephalography, a noninvasive method of monitoring brain activity, the team was able to track how their study participants’ brains responded to each of two melodies during their experiments. While the two tunes played, subjects were instructed to follow either the higher-pitched or the lower-pitched melody. When the music stopped, they were asked about the final notes of the target tune: did they rise or did they fall? The researchers could make this task harder by making the two tunes closer together in pitch, as well as by altering the timing of the notes.

Manting used a survey that asked about musical experience to score each participant’s musicality, and this measure had an obvious effect on task performance: The more musical a person was, the more successful they were at following the tune they had been asked to track.

To look for differences in brain activity that might explain this, the research team developed a new machine-learning approach to analyze their data. They used it to tease apart what was happening in the brain as participants focused on the target tune — even, in some cases, when the notes of the distracting tune played at the exact same time.

Top-down versus bottom-up attention

What they found was a clear separation of brain activity associated with two kinds of attention, known as top-down and bottom-up attention. Manting explains that top-down attention is goal-oriented, involving a conscious focus — the kind of attention listeners called on as they followed the target tune. Bottom-up attention, on the other hand, is triggered by the nature of the sound itself. A fire alarm would be expected to trigger this kind of attention, both with its volume and its suddenness. The distracting tune in the team’s experiments triggered activity associated with bottom-up attention — but more so in some people than in others.

“The more musical someone is, the better they are at focusing their top-down selective attention, and the less the effect of bottom-up attention is,” Manting explains.

Manting expects that musicians use their heightened capacity for top-down attention in other situations, as well. For example, they might be better than others at following a conversation in a room filled with background chatter. “I would put my bet on it that there is a high chance that they will be great at zooming into sounds,” she says.

She wonders, however, if one kind of distraction might actually be harder for a musician to filter out: the sound of their own instrument. Manting herself plays both the piano and the Chinese harp, and she says hearing those instruments is “like someone calling my name.” It’s one of many questions about how musical training affects cognition that she plans to explore in her future work.

MIT chemists have designed a new type of fluorescent molecule that they hope could be used for applications such as generating clearer images of tumors.

The new dye is based on a borenium ion — a positively charged form of boron that can emit light in the red to near-infrared range. Until recently, these ions have been too unstable to be used for imaging or other biomedical applications.

In a study appearing today in Nature Chemistry, the researchers showed that they could stabilize borenium ions by attaching them to a ligand. This approach allowed them to create borenium-containing films, powders, and crystals, all of which emit and absorb light in the red and near-infrared range.

That is important because near-IR light is easier to see when imaging structures deep within tissues, which could allow for clearer images of tumors and other structures in the body.

“One of the reasons why we focus on red to near-IR is because those types of dyes penetrate the body and tissue much better than light in the UV and visible range. Stability and brightness of those red dyes are the challenges that we tried to overcome in this study,” says Robert Gilliard, the Novartis Professor of Chemistry at MIT and the senior author of the study.

MIT research scientist Chun-Lin Deng is the lead author of the paper. Other authors include Bi Youan (Eric) Tra PhD ’25, former visiting graduate student Xibao Zhang, and graduate student Chonghe Zhang.

Stabilized borenium

Most fluorescent imaging relies on dyes that emit blue or green light. Those imaging agents work well in cells, but they are not as useful in tissue because low levels of blue and green fluorescence produced by the body interfere with the signal. Blue and green light also scatters in tissue, limiting how deeply it can penetrate.

Imaging agents that emit red fluorescence can produce clearer images, but most red dyes are inherently unstable and don’t produce a bright signal, because of their low quantum yields (the ratio of fluorescent photons emitted per photon of light is absorbed). For many red dyes, the quantum yield is only about 1 percent.

Among the molecules that can emit near-infrared light are borenium cations —positively charged ions containing an atom of boron attached to three other atoms.

When these molecules were first discovered in the mid-1980s, they were considered “laboratory curiosities,” Gilliard says. These molecules were so unstable that they had to be handled in a sealed container called a glovebox to protect them from exposure to air, which can lead them to break down.

Later, chemists realized they could make these ions more stable by attaching them to molecules called ligands. Working with these more stable ions, Gillliard’s lab discovered in 2019 that they had some unusual properties: Namely, they could respond to changes in temperature by emitting different colors of light.

However, at that point, “there was a substantial problem in that they were still too reactive to be handled in open air,” Gilliard says.

His lab began working on new ways to further stabilize them using ligands known as carbodicarbenes (CDCs), which they reported in a 2022 study. Due to this stabilization, the compounds can now be studied and handled without using a glovebox. They are also resistant to being broken down by light, unlike many previous borenium-based compounds.

In the new study, Gilliard began experimenting with the anions (negatively charged ions) that are a part of the CDC-borenium compounds. Interactions between these anions and the borenium cation generate a phenomenon known as exciton coupling, the researchers discovered. This coupling, they found, shifted the molecules’ emission and absorption properties toward the infrared end of the color spectrum. These molecules also generated a high quantum yield, allowing them to shine more brightly.

“Not only are we in the correct region, but the efficiency of the molecules is also very suitable,” Gilliard says. “We’re up to percentages in the thirties for the quantum yields in the red region, which is considered to be high for that region of the electromagnetic spectrum.”

Potential applications

The researchers also showed that they could convert their borenium-containing compounds into several different states, including solid crystals, films, powders, and colloidal suspensions.

For biomedical imaging, Gilliard envisions that these borenium-containing materials could be encapsulated in polymers, allowing them to be injected into the body to use as an imaging dye. As a first step, his lab plans to work with researchers in the chemistry department at MIT and at the Broad Institute of MIT and Harvard to explore the potential of imaging these materials within cells.

Because of their temperature responsiveness, these materials could also be deployed as temperature sensors, for example, to monitor whether drugs or vaccines have been exposed to temperatures that are too high or low during shipping.

“For any type of application where temperature tracking is important, these types of ‘molecular thermometers’ can be very useful,” Gilliard says.

If incorporated into thin films, these molecules could also be useful as organic light-emitting diodes (OLEDs), particularly in new types of materials such as flexible screens, Gilliard says.

“The very high quantum yields achieved in the near-IR, combined with the excellent environmental stability, make this class of compounds extremely interesting for biological applications,” says Frieder Jaekle, a professor of chemistry at Rutgers University, who was not involved in the study. “Besides the obvious utility in bioimaging, the strong and tunable near-IR emission also makes these new fluorophores very appealing as smart materials for anticounterfeiting, sensors, switches, and advanced optoelectronic devices.”

In addition to exploring possible applications for these dyes, the researchers are now working on extending their color emission further into the near-infrared region, which they hope to achieve by incorporating additional boron atoms. Those extra boron atoms could make the molecules less stable, so the researchers are also working on new types of carbodicarbenes to help stabilize them.

The research was funded by the Arnold and Mabel Beckman Foundation and the National Institutes of Health.

For patients with inflammatory bowel disease, antibiotics can be a double-edged sword. The broad-spectrum drugs often prescribed for gut flare-ups can kill helpful microbes alongside harmful ones, sometimes worsening symptoms over time. When fighting gut inflammation, you don’t always want to bring a sledgehammer to a knife fight.

Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and McMaster University have identified a new compound that takes a more targeted approach. The molecule, called enterololin, suppresses a group of bacteria linked to Crohn’s disease flare-ups while leaving the rest of the microbiome largely intact. Using a generative AI model, the team mapped how the compound works, a process that usually takes years but was accelerated here to just months.

“This discovery speaks to a central challenge in antibiotic development,” says Jon Stokes, senior author of a new paper on the work, assistant professor of biochemistry and biomedical sciences at McMaster, and research affiliate at MIT’s Abdul Latif Jameel Clinic for Machine Learning in Health. “The problem isn’t finding molecules that kill bacteria in a dish — we’ve been able to do that for a long time. A major hurdle is figuring out what those molecules actually do inside bacteria. Without that detailed understanding, you can’t develop these early-stage antibiotics into safe and effective therapies for patients.”

Enterololin is a stride toward precision antibiotics: treatments designed to knock out only the bacteria causing trouble. In mouse models of Crohn’s-like inflammation, the drug zeroed in on Escherichia coli, a gut-dwelling bacterium that can worsen flares, while leaving most other microbial residents untouched. Mice given enterololin recovered faster and maintained a healthier microbiome than those treated with vancomycin, a common antibiotic.

Pinning down a drug’s mechanism of action, the molecular target it binds inside bacterial cells, normally requires years of painstaking experiments. Stokes’ lab discovered enterololin using a high-throughput screening approach, but determining its target would have been the bottleneck. Here, the team turned to DiffDock, a generative AI model developed at CSAIL by MIT PhD student Gabriele Corso and MIT Professor Regina Barzilay.

DiffDock was designed to predict how small molecules fit into the binding pockets of proteins, a notoriously difficult problem in structural biology. Traditional docking algorithms search through possible orientations using scoring rules, often producing noisy results. DiffDock instead frames docking as a probabilistic reasoning problem: a diffusion model iteratively refines guesses until it converges on the most likely binding mode.

“In just a couple of minutes, the model predicted that enterololin binds to a protein complex called LolCDE, which is essential for transporting lipoproteins in certain bacteria,” says Barzilay, who also co-leads the Jameel Clinic. “That was a very concrete lead — one that could guide experiments, rather than replace them.”

Stokes’ group then put that prediction to the test. Using DiffDock predictions as an experimental GPS, they first evolved enterololin-resistant mutants of E. coli in the lab, which revealed that changes in the mutant’s DNA mapped to lolCDE, precisely where DiffDock had predicted enterololin to bind. They also performed RNA sequencing to see which bacterial genes switched on or off when exposed to the drug, as well as used CRISPR to selectively knock down expression of the expected target. These laboratory experiments all revealed disruptions in pathways tied to lipoprotein transport, exactly what DiffDock had predicted.

“When you see the computational model and the wet-lab data pointing to the same mechanism, that’s when you start to believe you’ve figured something out,” says Stokes.

For Barzilay, the project highlights a shift in how AI is used in the life sciences. “A lot of AI use in drug discovery has been about searching chemical space, identifying new molecules that might be active,” she says. “What we’re showing here is that AI can also provide mechanistic explanations, which are critical for moving a molecule through the development pipeline.”

That distinction matters because mechanism-of-action studies are often a major rate-limiting step in drug development. Traditional approaches can take 18 months to two years, or more, and cost millions of dollars. In this case, the MIT–McMaster team cut the timeline to about six months, at a fraction of the cost.

Enterololin is still in the early stages of development, but translation is already underway. Stokes’ spinout company, Stoked Bio, has licensed the compound and is optimizing its properties for potential human use. Early work is also exploring derivatives of the molecule against other resistant pathogens, such as Klebsiella pneumoniae. If all goes well, clinical trials could begin within the next few years.

The researchers also see broader implications. Narrow-spectrum antibiotics have long been sought as a way to treat infections without collateral damage to the microbiome, but they have been difficult to discover and validate. AI tools like DiffDock could make that process more practical, rapidly enabling a new generation of targeted antimicrobials.

For patients with Crohn’s and other inflammatory bowel conditions, the prospect of a drug that reduces symptoms without destabilizing the microbiome could mean a meaningful improvement in quality of life. And in the bigger picture, precision antibiotics may help tackle the growing threat of antimicrobial resistance.

“What excites me is not just this compound, but the idea that we can start thinking about the mechanism of action elucidation as something we can do more quickly, with the right combination of AI, human intuition, and laboratory experiments,” says Stokes. “That has the potential to change how we approach drug discovery for many diseases, not just Crohn’s.”

“One of the greatest challenges to our health is the increase of antimicrobial-resistant bacteria that evade even our best antibiotics,” adds Yves Brun, professor at the University of Montreal and distinguished professor emeritus at Indiana University Bloomington, who wasn’t involved in the paper. “AI is becoming an important tool in our fight against these bacteria. This study uses a powerful and elegant combination of AI methods to determine the mechanism of action of a new antibiotic candidate, an important step in its potential development as a therapeutic.”

Corso, Barzilay, and Stokes wrote the paper with McMaster researchers Denise B. Catacutan, Vian Tran, Jeremie Alexander, Yeganeh Yousefi, Megan Tu, Stewart McLellan, and Dominique Tertigas, and professors ​​Jakob Magolan, Michael Surette, Eric Brown, and Brian Coombes. Their research was supported, in part, by the Weston Family Foundation; the David Braley Centre for Antibiotic Discovery; the Canadian Institutes of Health Research; the Natural Sciences and Engineering Research Council of Canada; M. and M. Heersink; Canadian Institutes for Health Research; Ontario Graduate Scholarship Award; the Jameel Clinic; and the U.S. Defense Threat Reduction Agency Discovery of Medical Countermeasures Against New and Emerging Threats program.

The researchers posted sequencing data in public repositories and released the DiffDock-L code openly on GitHub.

At the heart of all lithium-ion batteries is a simple reaction: Lithium ions dissolved in an electrolyte solution “intercalate” or insert themselves into a solid electrode during battery discharge. When they de-intercalate and return to the electrolyte, the battery charges.

This process happens thousands of times throughout the life of a battery. The amount of power that the battery can generate, and how quickly it can charge, depend on how fast this reaction happens. However, little is known about the exact mechanism of this reaction, or the factors that control its rate.

In a new study, MIT researchers have measured lithium intercalation rates in a variety of different battery materials and used that data to develop a new model of how the reaction is controlled. Their model suggests that lithium intercalation is governed by a process known as coupled ion-electron transfer, in which an electron is transferred to the electrode along with a lithium ion.

Insights gleaned from this model could guide the design of more powerful and faster charging lithium-ion batteries, the researchers say.

“What we hope is enabled by this work is to get the reactions to be faster and more controlled, which can speed up charging and discharging,” says Martin Bazant, the Chevron Professor of Chemical Engineering and a professor of mathematics at MIT.

The new model may also help scientists understand why tweaking electrodes and electrolytes in certain ways leads to increased energy, power, and battery life — a process that has mainly been done by trial and error.

“This is one of these papers where now we began to unify the observations of reaction rates that we see with different materials and interfaces, in one theory of coupled electron and ion transfer for intercalation, building up previous work on reaction rates,” says Yang Shao-Horn, the J.R. East Professor of Engineering at MIT and a professor of mechanical engineering, materials science and engineering, and chemistry.

Shao-Horn and Bazant are the senior authors of the paper, which appears today in Science. The paper’s lead authors are Yirui Zhang PhD ’22, who is now an assistant professor at Rice University; Dimitrios Fraggedakis PhD ’21, who is now an assistant professor at Princeton University; Tao Gao, a former MIT postdoc who is now an assistant professor at the University of Utah; and MIT graduate student Shakul Pathak.

Modeling lithium flow

For many decades, scientists have hypothesized that the rate of lithium intercalation at a lithium-ion battery electrode is determined by how quickly lithium ions can diffuse from the electrolyte into the electrode. This reaction, they believed, was governed by a model known as the Butler-Volmer equation, originally developed almost a century ago to describe the rate of charge transfer during an electrochemical reaction.

However, when researchers have tried to measure lithium intercalation rates, the measurements they obtained were not always consistent with the rates predicted by the Butler-Volmer equation. Furthermore, obtaining consistent measurements across labs has been difficult, with different research teams reporting measurements for the same reaction that varied by a factor of up to 1 billion.

In the new study, the MIT team measured lithium intercalation rates using an electrochemical technique that involves applying repeated, short bursts of voltage to an electrode. They generated these measurements for more than 50 combinations of electrolytes and electrodes, including lithium nickel manganese cobalt oxide, which is commonly used in electric vehicle batteries, and lithium cobalt oxide, which is found in the batteries that power most cell phones, laptops, and other portable electronics.

For these materials, the measured rates are much lower than has previously been reported, and they do not correspond to what would be predicted by the traditional Butler-Volmer model.

The researchers used the data to come up with an alternative theory of how lithium intercalation occurs at the surface of an electrode. This theory is based on the assumption that in order for a lithium ion to enter an electrode, an electron from the electrolyte solution must be transferred to the electrode at the same time.

“The electrochemical step is not lithium insertion, which you might think is the main thing, but it’s actually electron transfer to reduce the solid material that is hosting the lithium,” Bazant says. “Lithium is intercalated at the same time that the electron is transferred, and they facilitate one another.”

This coupled-electron ion transfer (CIET) lowers the energy barrier that must be overcome for the intercalation reaction to occur, making it more likely to happen. The mathematical framework of CIET allowed the researchers to make reaction rate predictions, which were validated by their experiments and substantially different from those made by the Butler-Volmer model.

Faster charging

In this study, the researchers also showed that they could tune intercalation rates by changing the composition of the electrolyte. For example, swapping in different anions can lower the amount of energy needed to transfer the lithium and electron, making the process more efficient.

“Tuning the intercalation kinetics by changing electrolytes offers great opportunities to enhance the reaction rates, alter electrode designs, and therefore enhance the battery power and energy,” Shao-Horn says.

Shao-Horn’s lab and their collaborators have been using automated experiments to make and test thousands of different electrolytes, which are used to develop machine-learning models to predict electrolytes with enhanced functions.

The findings could also help researchers to design batteries that would charge faster, by speeding up the lithium intercalation reaction. Another goal is reducing the side reactions that can cause battery degradation when electrons are picked off the electrode and dissolve into the electrolyte.

“If you want to do that rationally, not just by trial and error, you need some kind of theoretical framework to know what are the important material parameters that you can play with,” Bazant says. “That’s what this paper tries to provide.”

The research was funded by Shell International Exploration and Production and the Toyota Research Institute through the D3BATT Center for Data-Driven Design of Rechargeable Batteries.

Designing a complex electronic device like a delivery drone involves juggling many choices, such as selecting motors and batteries that minimize cost while maximizing the payload the drone can carry or the distance it can travel.

Unraveling that conundrum is no easy task, but what happens if the designers don’t know the exact specifications of each battery and motor? On top of that, the real-world performance of these components will likely be affected by unpredictable factors, like changing weather along the drone’s route.

MIT researchers developed a new framework that helps engineers design complex systems in a way that explicitly accounts for such uncertainty. The framework allows them to model the performance tradeoffs of a device with many interconnected parts, each of which could behave in unpredictable ways.

Their technique captures the likelihood of many outcomes and tradeoffs, giving designers more information than many existing approaches which, at most, can usually only model best-case and worst-case scenarios.

Ultimately, this framework could help engineers develop complex systems like autonomous vehicles, commercial aircraft, or even regional transportation networks that are more robust and reliable in the face of real-world unpredictability.

“In practice, the components in a device never behave exactly like you think they will. If someone has a sensor whose performance is uncertain, and an algorithm that is uncertain, and the design of a robot that is also uncertain, now they have a way to mix all these uncertainties together so they can come up with a better design,” says Gioele Zardini, the Rudge and Nancy Allen Assistant Professor of Civil and Environmental Engineering at MIT, a principal investigator in the Laboratory for Information and Decision Systems (LIDS), an affiliate faculty with the Institute for Data, Systems, and Society (IDSS), and senior author of a paper on this framework.

Zardini is joined on the paper by lead author Yujun Huang, an MIT graduate student; and Marius Furter, a graduate student at the University of Zurich. The research will be presented at the IEEE Conference on Decision and Control.

Considering uncertainty

The Zardini Group studies co-design, a method for designing systems made of many interconnected components, from robots to regional transportation networks.

The co-design language breaks a complex problem into a series of boxes, each representing one component, that can be combined in different ways to maximize outcomes or minimize costs. This allows engineers to solve complex problems in a feasible amount of time.

In prior work, the researchers modeled each co-design component without considering uncertainty. For instance, the performance of each sensor the designers could choose for a drone was fixed.

But engineers often don’t know the exact performance specifications of each sensor, and even if they do, it is unlikely the senor will perfectly follow its spec sheet. At the same time, they don’t know how each sensor will behave once integrated into a complex device, or how performance will be affected by unpredictable factors like weather.

“With our method, even if you are unsure what the specifications of your sensor will be, you can still design the robot to maximize the outcome you care about,” says Furter.

To accomplish this, the researchers incorporated this notion of uncertainty into an existing framework based on category theory.

Using some mathematical tricks, they simplified the problem into a more general structure. This allows them to use the tools of category theory to solve co-design problems in a way that considers a range of uncertain outcomes.

By reformulating the problem, the researchers can capture how multiple design choices affect one another even when their individual performance is uncertain.

This approach is also simpler than many existing tools that typically require extensive domain expertise. With their plug-and-play system, one can rearrange the components in the system without violating any mathematical constraints.

And because no specific domain expertise is required, the framework could be used by a multidisciplinary team where each member designs one component of a larger system.

“Designing an entire UAV isn’t feasible for just one person, but designing a component of a UAV is. By providing the framework for how these components work together in a way that considers uncertainty, we’ve made it easier for people to evaluate the performance of the entire UAV system,” Huang says.

More detailed information

The researchers used this new approach to choose perception systems and batteries for a drone that would maximize its payload while minimizing its lifetime cost and weight.

While each perception system may offer a different detection accuracy under varying weather conditions, the designer doesn’t know exactly how its performance will fluctuate. This new system allows the designer to take these uncertainties into consideration when thinking about the drone’s overall performance.

And unlike other approaches, their framework reveals distinct advantages of each battery technology.

For instance, their results show that at lower payloads, nickel-metal hydride batteries provide the lowest expected lifetime cost. This insight would be impossible to fully capture without accounting for uncertainty, Zardini says.

While another method might only be able to show the best-case and worst-case performance scenarios of lithium polymer batteries, their framework gives the user more detailed information.

For example, it shows that if the drone’s payload is 1,750 grams, there is a 12.8 percent chance the battery design would be infeasible.

“Our system provides the tradeoffs, and then the user can reason about the design,” he adds.

In the future, the researchers want to improve the computational efficiency of their problem-solving algorithms. They also want to extend this approach to situations where a system is designed by multiple parties that are collaborative and competitive, like a transportation network in which rail companies operate using the same infrastructure.

“As the complexity of systems grow, and involves more disparate components, we need a formal framework in which to design these systems. This paper presents a way to compose large systems from modular components, understand design trade-offs, and importantly do so with a notion of uncertainty. This creates an opportunity to formalize the design of large-scale systems with learning-enabled components,” says Aaron Ames, the Bren Professor of Mechanical and Civil Engineering, Control and Dynamical Systems, and Aerospace at Caltech, who was not involved with this research. 

Concrete already builds our world, and now it’s one step closer to powering it, too. Made by combining cement, water, ultra-fine carbon black (with nanoscale particles), and electrolytes, electron-conducting carbon concrete (ec³, pronounced “e-c-cubed”) creates a conductive “nanonetwork” inside concrete that could enable everyday structures like walls, sidewalks, and bridges to store and release electrical energy. In other words, the concrete around us could one day double as giant “batteries.”

As MIT researchers report in a new PNAS paper, optimized electrolytes and manufacturing processes have increased the energy storage capacity of the latest ec³ supercapacitors by an order of magnitude. In 2023, storing enough energy to meet the daily needs of the average home would have required about 45 cubic meters of ec³, roughly the amount of concrete used in a typical basement. Now, with the improved electrolyte, that same task can be achieved with about 5 cubic meters, the volume of a typical basement wall.

“A key to the sustainability of concrete is the development of ‘multifunctional concrete,’ which integrates functionalities like this energy storage, self-healing, and carbon sequestration. Concrete is already the world’s most-used construction material, so why not take advantage of that scale to create other benefits?” asks Admir Masic, lead author of the new study, MIT Electron-Conducting Carbon-Cement-Based Materials Hub (EC³ Hub) co-director, and associate professor of civil and environmental engineering (CEE) at MIT.

The improved energy density was made possible by a deeper understanding of how the nanocarbon black network inside ec³ functions and interacts with electrolytes. Using focused ion beams for the sequential removal of thin layers of the ec³ material, followed by high-resolution imaging of each slice with a scanning electron microscope (a technique called FIB-SEM tomography), the team across the EC³ Hub and MIT Concrete Sustainability Hub was able to reconstruct the conductive nanonetwork at the highest resolution yet. This approach allowed the team to discover that the network is essentially a fractal-like “web” that surrounds ec³ pores, which is what allows the electrolyte to infiltrate and for current to flow through the system. 

“Understanding how these materials ‘assemble’ themselves at the nanoscale is key to achieving these new functionalities,” adds Masic.

Equipped with their new understanding of the nanonetwork, the team experimented with different electrolytes and their concentrations to see how they impacted energy storage density. As Damian Stefaniuk, first author and EC³ Hub research scientist, highlights, “we found that there is a wide range of electrolytes that could be viable candidates for ec³. This even includes seawater, which could make this a good material for use in coastal and marine applications, perhaps as support structures for offshore wind farms.”

At the same time, the team streamlined the way they added electrolytes to the mix. Rather than curing ec³ electrodes and then soaking them in electrolyte, they added the electrolyte directly into the mixing water. Since electrolyte penetration was no longer a limitation, the team could cast thicker electrodes that stored more energy.

The team achieved the greatest performance when they switched to organic electrolytes, especially those that combined quaternary ammonium salts — found in everyday products like disinfectants — with acetonitrile, a clear, conductive liquid often used in industry. A cubic meter of this version of ec³ — about the size of a refrigerator — can store over 2 kilowatt-hours of energy. That’s about enough to power an actual refrigerator for a day.

While batteries maintain a higher energy density, ec³ can in principle be incorporated directly into a wide range of architectural elements — from slabs and walls to domes and vaults — and last as long as the structure itself.

“The Ancient Romans made great advances in concrete construction. Massive structures like the Pantheon stand to this day without reinforcement. If we keep up their spirit of combining material science with architectural vision, we could be at the brink of a new architectural revolution with multifunctional concretes like ec³,” proposes Masic.

Taking inspiration from Roman architecture, the team built a miniature ec³ arch to show how structural form and energy storage can work together. Operating at 9 volts, the arch supported its own weight and additional load while powering an LED light.

However, something unique happened when the load on the arch increased: the light flickered. This is likely due to the way stress impacts electrical contacts or the distribution of charges. “There may be a kind of self-monitoring capacity here. If we think of an ec³ arch at architectural scale, its output may fluctuate when it’s impacted by a stressor like high winds. We may be able to use this as a signal of when and to what extent a structure is stressed, or monitor its overall health in real time,” envisions Masic.

The latest developments in ec³ technology bring it a step closer to real-world scalability. It’s already been used to heat sidewalk slabs in Sapporo, Japan, due to its thermally conductive properties, representing a potential alternative to salting. “With these higher energy densities and demonstrated value across a broader application space, we now have a powerful and flexible tool that can help us address a wide range of persistent energy challenges,” explains Stefaniuk. “One of our biggest motivations was to help enable the renewable energy transition. Solar power, for example, has come a long way in terms of efficiency. However, it can only generate power when there’s enough sunlight. So, the question becomes: How do you meet your energy needs at night, or on cloudy days?”

Franz-Josef Ulm, EC³ Hub co-director and CEE professor, continues the thread: “The answer is that you need a way to store and release energy. This has usually meant a battery, which often relies on scarce or harmful materials. We believe that ec³ is a viable substitute, letting our buildings and infrastructure meet our energy storage needs.” The team is working toward applications like parking spaces and roads that could charge electric vehicles, as well as homes that can operate fully off the grid.

“What excites us most is that we’ve taken a material as ancient as concrete and shown that it can do something entirely new,” says James Weaver, a co-author on the paper who is an associate professor of design technology and materials science and engineering at Cornell University, as well as a former EC³ Hub researcher. “By combining modern nanoscience with an ancient building block of civilization, we’re opening a door to infrastructure that doesn’t just support our lives, it powers them.”

Palladium is one of the keys to jump-starting a hydrogen-based energy economy. The silvery metal is a natural gatekeeper against every gas except hydrogen, which it readily lets through. For its exceptional selectivity, palladium is considered one of the most effective materials at filtering gas mixtures to produce pure hydrogen.

Today, palladium-based membranes are used at commercial scale to provide pure hydrogen for semiconductor manufacturing, food processing, and fertilizer production, among other applications in which the membranes operate at modest temperatures. If palladium membranes get much hotter than around 800 kelvins, they can break down.

Now, MIT engineers have developed a new palladium membrane that remains resilient at much higher temperatures. Rather than being made as a continuous film, as most membranes are, the new design is made from palladium that is deposited as “plugs” into the pores of an underlying supporting material. At high temperatures, the snug-fitting plugs remain stable and continue separating out hydrogen, rather than degrading as a surface film would.

The thermally stable design opens opportunities for membranes to be used in hydrogen-fuel-generating technologies such as compact steam methane reforming and ammonia cracking — technologies that are designed to operate at much higher temperatures to produce hydrogen for zero-carbon-emitting fuel and electricity.

“With further work on scaling and validating performance under realistic industrial feeds, the design could represent a promising route toward practical membranes for high-temperature hydrogen production,” says Lohyun Kim PhD ’24, a former graduate student in MIT’s Department of Mechanical Engineering.

Kim and his colleagues report details of the new membrane in a study appearing today in the journal Advanced Functional Materials. The study’s co-authors are Randall Field, director of research at the MIT Energy Initiative (MITEI); former MIT chemical engineering graduate student Chun Man Chow PhD ’23; Rohit Karnik, the Jameel Professor in the Department of Mechanical Engineering at MIT and the director of the Abdul Latif Jameel Water and Food Systems Lab (J-WAFS); and Aaron Persad, a former MIT research scientist in mechanical engineering who is now an assistant professor at the University of Maryland Eastern Shore.

Compact future

The team’s new design came out of a MITEI project related to fusion energy. Future fusion power plants, such as the one MIT spinout Commonwealth Fusion Systems is designing, will involve circulating hydrogen isotopes of deuterium and tritium at extremely high temperatures to produce energy from the isotopes’ fusing. The reactions inevitably produce other gases that will have to be separated, and the hydrogen isotopes will be recirculated into the main reactor for further fusion.

Similar issues arise in a number of other processes for producing hydrogen, where gases must be separated and recirculated back into a reactor. Concepts for such recirculating systems would require first cooling down the gas before it can pass through hydrogen-separating membranes — an expensive and energy-intensive step that would involve additional machinery and hardware.

“One of the questions we were thinking about is: Can we develop membranes which could be as close to the reactor as possible, and operate at higher temperatures, so we don’t have to pull out the gas and cool it down first?” Karnik says. “It would enable more energy-efficient, and therefore cheaper and compact, fusion systems.”

The researchers looked for ways to improve the temperature resistance of palladium membranes. Palladium is the most effective metal used today to separate hydrogen from a variety of gas mixtures. It naturally attracts hydrogen molecules (H₂) to its surface, where the metal’s electrons interact with and weaken the molecule’s bonds, causing H₂ to temporarily break apart into its respective atoms. The individual atoms then diffuse through the metal and join back up on the other side as pure hydrogen.

Palladium is highly effective at permeating hydrogen, and only hydrogen, from streams of various gases. But conventional membranes typically can operate at temperatures of up to 800 kelvins before the film starts to form holes or clumps up into droplets, allowing other gases to flow through.

Plugging in

Karnik, Kim and their colleagues took a different design approach. They observed that at high temperatures, palladium will start to shrink up. In engineering terms, the material is acting to reduce surface energy. To do this, palladium, and most other materials and even water, will pull apart and form droplets with the smallest surface energy. The lower the surface energy, the more stable the material can be against further heating.

This gave the team an idea: If a supporting material’s pores could be “plugged” with deposits of palladium — essentially already forming a droplet with the lowest surface energy — the tight quarters might substantially increase palladium’s heat tolerance while preserving the membrane’s selectivity for hydrogen.

To test this idea, they fabricated small chip-sized samples of membrane using a porous silica supporting layer (each pore measuring about half a micron wide), onto which they deposited a very thin layer of palladium. They applied techniques to essentially grow the palladium into the pores, and polished down the surface to remove the palladium layer and leave palladium only inside the pores.

They then placed samples in a custom-built apparatus in which they flowed hydrogen-containing gas of various mixtures and temperatures to test its separation performance. The membranes remained stable and continued to separate hydrogen from other gases even after experiencing temperatures of up to 1,000 kelvins for over 100 hours — a significant improvement over conventional film-based membranes.

“The use of palladium film membranes are generally limited to below around 800 kelvins, at which point they degrade,” Kim says. “Our plug design therefore extends palladium’s effective heat resilience by roughly at least 200 kelvins and maintains integrity far longer under extreme conditions.”

These conditions are within the range of hydrogen-generating technologies such as steam methane reforming and ammonia cracking.

Steam methane reforming is an established process that has required complex, energy-intensive systems to preprocess methane to a form where pure hydrogen can be extracted. Such preprocessing steps could be replaced with a compact “membrane reactor,” through which a methane gas would directly flow, and the membrane inside would filter out pure hydrogen. Such reactors would significantly cut down the size, complexity, and cost of producing hydrogen from steam methane reforming, and Kim estimates a membrane would have to work reliably in temperatures of up to nearly 1,000 kelvins. The team’s new membrane could work well within such conditions.

Ammonia cracking is another way to produce hydrogen, by “cracking” or breaking apart ammonia. As ammonia is very stable in liquid form, scientists envision that it could be used as a carrier for hydrogen and be safely transported to a hydrogen fuel station, where ammonia could be fed into a membrane reactor that again pulls out hydrogen and pumps it directly into a fuel cell vehicle. Ammonia cracking is still largely in pilot and demonstration stages, and Kim says any membrane in an ammonia cracking reactor would likely operate at temperatures of around 800 kelvins — within the range of the group’s new plug-based design.

Karnik emphasizes that their results are just a start. Adopting the membrane into working reactors will require further development and testing to ensure it remains reliable over much longer periods of time.

“We showed that instead of making a film, if you make discretized nanostructures you can get much more thermally stable membranes,” Karnik says. “It provides a pathway for designing membranes for extreme temperatures, with the added possibility of using smaller amounts of expensive palladium, toward making hydrogen production more efficient and affordable. There is potential there.”

This work was supported by Eni S.p.A. via the MIT Energy Initiative.

This work utilized facilities at the MIT Materials Research Laboratory (MRL), the MIT Laboratory for Manufacturing and Productivity (LMP), and MIT.nano.

A diet rich in the amino acid cysteine may have rejuvenating effects in the small intestine, according to a new study from MIT. This amino acid, the researchers discovered, can turn on an immune signaling pathway that helps stem cells to regrow new intestinal tissue.

This enhanced regeneration may help to heal injuries from radiation, which often occur in patients undergoing radiation therapy for cancer. The research was conducted in mice, but if future research shows similar results in humans, then delivering elevated quantities of cysteine, through diet or supplements, could offer a new strategy to help damaged tissue heal faster, the researchers say.

“The study suggests that if we give these patients a cysteine-rich diet or cysteine supplementation, perhaps we can dampen some of the chemotherapy or radiation-induced injury,” says Omer Yilmaz, director of the MIT Stem Cell Initiative, an associate professor of biology at MIT, and a member of MIT’s Koch Institute for Integrative Cancer Research. “The beauty here is we’re not using a synthetic molecule; we’re exploiting a natural dietary compound.”

While previous research has shown that certain types of diets, including low-calorie diets, can enhance intestinal stem cell activity, the new study is the first to identify a single nutrient that can help intestinal cells to regenerate.

Yilmaz is the senior author of the study, which appears today in Nature. Koch Institute postdoc Fangtao Chi is the paper’s lead author.

Boosting regeneration

It is well-established that diet can affect overall health: High-fat diets can lead to obesity, diabetes, and other health problems, while low-calorie diets have been shown to extend lifespans in many species. In recent years, Yilmaz’s lab has investigated how different types of diets influence stem cell regeneration, and found that high-fat diets, as well as short periods of fasting, can enhance stem cell activity in different ways.

“We know that macro diets such as high-sugar diets, high-fat diets, and low-calorie diets have a clear impact on health. But at the granular level, we know much less about how individual nutrients impact stem cell fate decisions, as well as tissue function and overall tissue health,” Yilmaz says.

In their new study, the researchers began by feeding mice a diet high in one of 20 different amino acids, the building blocks of proteins. For each group, they measured how the diet affected intestinal stem cell regeneration. Among these amino acids, cysteine had the most dramatic effects on stem cells and progenitor cells (immature cells that differentiate into adult intestinal cells).

Further studies revealed that cysteine initiates a chain of events leading to the activation of a population of immune cells called CD8 T cells. When cells in the lining of the intestine absorb cysteine from digested food, they convert it into CoA, a cofactor that is released into the mucosal lining of the intestine. There, CD8 T cells absorb CoA, which stimulates them to begin proliferating and producing a cytokine called IL-22.

IL-22 is an important player in the regulation of intestinal stem cell regeneration, but until now, it wasn’t known that CD8 T cells can produce it to boost intestinal stem cells. Once activated, those IL-22-releasing T cells are primed to help combat any kind of injury that could occur within the intestinal lining.

“What’s really exciting here is that feeding mice a cysteine-rich diet leads to the expansion of an immune cell population that we typically don’t associate with IL-22 production and the regulation of intestinal stemness,” Yilmaz says. “What happens in a cysteine-rich diet is that the pool of cells that make IL-22 increases, particularly the CD8 T-cell fraction.”

These T cells tend to congregate within the lining of the intestine, so they are already in position when needed. The researchers found that the stimulation of CD8 T cells occurred primarily in the small intestine, not in any other part of the digestive tract, which they believe is because most of the protein that we consume is absorbed by the small intestine.

Healing the intestine

In this study, the researchers showed that regeneration stimulated by a cysteine-rich diet could help to repair radiation damage to the intestinal lining. Also, in work that has not been published yet, they showed that a high-cysteine diet had a regenerative effect following treatment with a chemotherapy drug called 5-fluorouracil. This drug, which is used to treat colon and pancreatic cancers, can also damage the intestinal lining.

Cysteine is found in many high-protein foods, including meat, dairy products, legumes, and nuts. The body can also synthesize its own cysteine, by converting the amino acid methionine to cysteine — a process that takes place in the liver. However, cysteine produced in the liver is distributed through the entire body and doesn’t lead to a buildup in the small intestine the way that consuming cysteine in the diet does.

“With our high-cysteine diet, the gut is the first place that sees a high amount of cysteine,” Chi says.

Cysteine has been previously shown to have antioxidant effects, which are also beneficial, but this study is the first to demonstrate its effect on intestinal stem cell regeneration. The researchers now hope to study whether it may also help other types of stem cells regenerate new tissues. In one ongoing study, they are investigating whether cysteine might stimulate hair follicle regeneration.

They also plan to further investigate some of the other amino acids that appear to influence stem cell regeneration.

“I think we’re going to uncover multiple new mechanisms for how these amino acids regulate cell fate decisions and gut health in the small intestine and colon,” Yilmaz says.

The research was funded, in part, by the National Institutes of Health, the V Foundation, the Koch Institute Frontier Research Program via the Kathy and Curt Marble Cancer Research Fund, the Bridge Project — a partnership between the Koch Institute for Integrative Cancer Research at MIT and the Dana-Farber/Harvard Cancer Center, the American Federation for Aging Research, the MIT Stem Cell Initiative, and the Koch Institute Support (core) Grant from the National Cancer Institute.

Say a local concert venue wants to engage its community by giving social media followers an easy way to share and comment on new music from emerging artists. Rather than working within the constraints of existing social platforms, the venue might want to create its own social app with the functionality that would be best for its community. But building a new social app from scratch involves many complicated programming steps, and even if the venue can create a customized app, the organization’s followers may be unwilling to join the new platform because it could mean leaving their connections and data behind.

Now, researchers from MIT have launched a framework called Graffiti that makes building personalized social applications easier, while allowing users to migrate between multiple applications without losing their friends or data.

“We want to empower people to have control over their own designs rather than having them dictated from the top down,” says electrical engineering and computer science graduate student Theia Henderson.

Henderson and her colleagues designed Graffiti with a flexible structure so individuals have the freedom to create a variety of customized applications, from messenger apps like WhatsApp to microblogging platforms like X to location-based social networking sites like Nextdoor, all using only front-end development tools like HTML.

The protocol ensures all applications can interoperate, so content posted on one application can appear on any other application, even those with disparate designs or functionality. Importantly, Graffiti users retain control of their data, which is stored on a decentralized infrastructure rather than being held by a specific application.

While the pros and cons of implementing Graffiti at scale remain to be fully explored, the researchers hope this new approach can someday lead to healthier online interactions.

“We’ve shown that you can have a rich social ecosystem where everyone owns their own data and can use whatever applications they want to interact with whoever they want in whatever way they want. And they can have their own experiences without losing connection with the people they want to stay connected with,” says David Karger, professor of EECS and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL).

Henderson, the lead author, and Karger are joined by MIT Research Scientist David D. Clark on a paper about Graffiti, which will be presented at the ACM Symposium on User Interface Software and Technology.

Personalized, integrated applications

With Graffiti, the researchers had two main goals: to lower the barrier to creating personalized social applications and to enable those personalized applications to interoperate without requiring permission from developers.

To make the design process easier, they built a collective back-end infrastructure that all applications access to store and share content. This means developers don’t need to write any complex server code. Instead, designing a Graffiti application is more like making a website using popular tools like Vue.

Developers can also easily introduce new features and new types of content, giving them more freedom and fostering creativity.

“Graffiti is so straightforward that we used it as the infrastructure for the intro to web design class I teach, and students were able to write the front-end very easily to come up with all sorts of applications,” Karger says.

The open, interoperable nature of Graffiti means no one entity has the power to set a moderation policy for the entire platform. Instead, multiple competing and contradictory moderation services can operate, and people can choose the ones they like. 

Graffiti uses the idea of “total reification,” where every action taken in Graffiti, such as liking, sharing, or blocking a post, is represented and stored as its own piece of data. A user can configure their social application to interpret or ignore those data using its own rules.

For instance, if an application is designed so a certain user is a moderator, posts blocked by that user won’t appear in the application. But for an application with different rules where that person isn’t considered a moderator, other users might just see a warning or no flag at all.

“Theia’s system lets each person pick their own moderators, avoiding the one-sized-fits-all approach to moderation taken by the major social platforms,” Karger says.

But at the same time, having no central moderator means there is no one to remove content from the platform that might be offensive or illegal.

“We need to do more research to understand if that is going to provide real, damaging consequences or if the kind of personal moderation we created can provide the protections people need,” he adds.

Empowering social media users

The researchers also had to overcome a problem known as context collapse, which conflicts with their goal of interoperation.

For instance, context collapse would occur if a person’s Tinder profile appeared on LinkedIn, or if a post intended for one group, like close friends, would create conflict with another group, such as family members. Context collapse can lead to anxiety and have social repercussions for the user and their different communities.

“We realize that interoperability can sometimes be a bad thing. People have boundaries between different social contexts, and we didn’t want to violate those,” Henderson says.

To avoid context collapse, the researchers designed Graffiti so all content is organized into distinct channels. Channels are flexible and can represent a variety of contexts, such as people, applications, locations, etc.

If a user’s post appears in an application channel but not their personal channel, others using that application will see the post, but those who only follow this user will not.

“Individuals should have the power to choose the audience for whatever they want to say,” Karger adds.

The researchers created multiple Graffiti applications to showcase personalization and interoperability, including a community-specific application for a local concert venue, a text-centric microblogging platform patterned off X, a Wikipedia-like application that enables collective editing, and a real-time messaging app with multiple moderation schemes patterned off WhatsApp and Slack.

“It also leaves room to create so many social applications people haven’t thought of yet. I’m really excited to see what people come up with when they are given full creative freedom,” Henderson says.

In the future, she and her colleagues want to explore additional social applications they could build with Graffiti. They also intend to incorporate tools like graphical editors to simplify the design process. In addition, they want to strengthen Graffiti’s security and privacy.

And while there is still a long way to go before Graffiti could be implemented at scale, the researchers are currently running a user study as they explore the potential positive and negative impacts the system could have on the social media landscape. 

“You still had to prove yourself.”

“Every cloud has a blue lining!”

Which of those sentences are you most likely to remember a few minutes from now? If you guessed the second, you’re probably correct.

According to a new study from MIT cognitive scientists, sentences that stick in your mind longer are those that have distinctive meanings, making them stand out from sentences you’ve previously seen. They found that meaning, not any other trait, is the most important feature when it comes to memorability.

“One might have thought that when you remember sentences, maybe it’s all about the visual features of the sentence, but we found that that was not the case. A big contribution of this paper is pinning down that it is the meaning-related space that makes sentences memorable,” says Greta Tuckute PhD ’25, who is now a research fellow at Harvard University’s Kempner Institute.

The findings support the hypothesis that sentences with distinctive meanings — like “Does olive oil work for tanning?” — are stored in brain space that is not cluttered with sentences that mean almost the same thing. Sentences with similar meanings end up densely packed together and are therefore more difficult to recognize confidently later on, the researchers believe.

“When you encode sentences that have a similar meaning, there’s feature overlap in that space. Therefore, a particular sentence you’ve encoded is not linked to a unique set of features, but rather to a whole bunch of features that may overlap with other sentences,” says Evelina Fedorenko, an MIT associate professor of brain and cognitive sciences (BCS), a member of MIT’s McGovern Institute for Brain Research, and the senior author of the study.

Tuckute and Thomas Clark, an MIT graduate student, are the lead authors of the paper, which appears in the Journal of Memory and Language. MIT graduate student Bryan Medina is also an author.

Distinctive sentences

What makes certain things more memorable than others is a longstanding question in cognitive science and neuroscience. In a 2011 study, Aude Oliva, now a senior research scientist at MIT and MIT director of the MIT-IBM Watson AI Lab, showed that not all items are created equal: Some types of images are much easier to remember than others, and people are remarkably consistent in what images they remember best.

In that study, Oliva and her colleagues found that, in general, images with people in them are the most memorable, followed by images of human-scale space and close-ups of objects. Least memorable are natural landscapes.

As a follow-up to that study, Fedorenko and Oliva, along with Ted Gibson, another faculty member in BCS, teamed up to determine if words also vary in their memorability. In a study published earlier this year, co-led by Tuckute and Kyle Mahowald, a former PhD student in BCS, the researchers found that the most memorable words are those that have the most distinctive meanings.

Words are categorized as being more distinctive if they have a single meaning, and few or no synonyms — for example, words like “pineapple” or “avalanche” which were found to be very memorable. On the other hand, words that can have multiple meanings, such as “light,” or words that have many synonyms, like “happy,” were more difficult for people to recognize accurately.

In the new study, the researchers expanded their scope to analyze the memorability of sentences. Just like words, some sentences have very distinctive meanings, while others communicate similar information in slightly different ways.

To do the study, the researchers assembled a collection of 2,500 sentences drawn from publicly available databases that compile text from novels, news articles, movie dialogues, and other sources. Each sentence that they chose contained exactly six words.

The researchers then presented a random selection of about 1,000 of these sentences to each study participant, including repeats of some sentences. Each of the 500 participants in the study was asked to press a button when they saw a sentence that they remembered seeing earlier.

The most memorable sentences — the ones where participants accurately and quickly indicated that they had seen them before — included strings such as “Homer Simpson is hungry, very hungry,” and “These mosquitoes are — well, guinea pigs.”

Those memorable sentences overlapped significantly with sentences that were determined as having distinctive meanings as estimated through the high-dimensional vector space of a large language model (LLM) known as Sentence BERT. That model is able to generate sentence-level representations of sentences, which can be used for tasks like judging meaning similarity between sentences. This model provided researchers with a distinctness score for each sentence based on its semantic similarity to other sentences.

The researchers also evaluated the sentences using a model that predicts memorability based on the average memorability of the individual words in the sentence. This model performed fairly well at predicting overall sentence memorability, but not as well as Sentence BERT. This suggests that the meaning of a sentence as a whole — above and beyond the contributions from individual words — determines how memorable it will be, the researchers say.

Noisy memories

While cognitive scientists have long hypothesized that the brain’s memory banks have a limited capacity, the findings of the new study support an alternative hypothesis that would help to explain how the brain can continue forming new memories without losing old ones.

This alternative, known as the noisy representation hypothesis, says that when the brain encodes a new memory, be it an image, a word, or a sentence, it is represented in a noisy way — that is, this representation is not identical to the stimulus, and some information is lost. For example, for an image, you may not encode the exact viewing angle at which an object is shown, and for a sentence, you may not remember the exact construction used.

Under this theory, a new sentence would be encoded in a similar part of the memory space as sentences that carry a similar meanings, whether they were encountered recently or sometime across a lifetime of language experience. This jumbling of similar meanings together increases the amount of noise and can make it much harder, later on, to remember the exact sentence you have seen before.

“The representation is gradually going to accumulate some noise. As a result, when you see an image or a sentence for a second time, your accuracy at judging whether you’ve seen it before will be affected, and it’ll be less than 100 percent in most cases,” Clark says.

However, if a sentence has a unique meaning that is encoded in a less densely crowded space, it will be easier to pick out later on.

“Your memory may still be noisy, but your ability to make judgments based on the representations is less affected by that noise because the representation is so distinctive to begin with,” Clark says.

The researchers now plan to study whether other features of sentences, such as more vivid and descriptive language, might also contribute to making them more memorable, and how the language system may interact with the hippocampal memory structures during the encoding and retrieval of memories.

The research was funded, in part, by the National Institutes of Health, the McGovern Institute, the Department of Brain and Cognitive Sciences, the Simons Center for the Social Brain, and the MIT Quest for Intelligence.

The Center for International Studies (CIS) empowers students, faculty, and scholars to bring MIT’s interdisciplinary style of research and scholarship to address complex global challenges. 

In this Q&A, Mihaela Papa, the center's director of research and a principal research scientist at MIT, describes her role and her founding of the BRICS Lab, which studies how the BRICS group of major powers and emerging markets — comprising Brazil, Russia, India, China and South Africa, along with new members and partners — seeks to reform global policymaking. She also discusses the ongoing mission of CIS to tackle the world's most complex challenges in new and creative ways.

Q: What is your role at CIS, and some of your key accomplishments since joining the center just over a year ago?

A: I serve as director of research and principal research scientist at CIS, a role that bridges management and scholarship. I oversee grant and fellowship programs, spearhead new research initiatives, build research communities across our center's area programs and MIT schools, and mentor the next generation of scholars. My academic expertise is in international relations, and I publish on global governance and sustainable development, particularly through my new BRICS Lab. 

This past year, I focused on building collaborative platforms that highlight CIS’ role as an interdisciplinary hub and expand its research reach. With Evan Lieberman, the director of CIS, I launched the CIS Global Research and Policy Seminar series to address current challenges in global development and governance, foster cross-disciplinary dialogue, and connect theoretical insights to policy solutions. We also convened a Climate Adaptation Workshop, which examined promising strategies for financing adaptation and advancing policy innovation. We documented the outcomes in a workshop report that outlines a broader research agenda contributing to MIT’s larger climate mission.

In parallel, I have been reviewing CIS’ grant-making programs to improve how we serve our community, while also supporting regional initiatives such as research planning related to Ukraine. Together with the center's MIT-Brazil faculty director Brad Olsen, I secured a MITHIC [MIT Human Insight Collaboration] Connectivity grant to build an MIT Amazonia research community that connects MIT scholars with regional partners and strengthens collaboration across the Amazon. Finally, I launched the BRICS Lab to analyze transformations in global governance and have ongoing research on BRICS and food security and data centers in BRICS. 

Q: Tell us more about the BRICS Lab.

A: The BRICS countries comprise the majority of the world’s population and an expanding share of the global economy. [Originally comprising Brazil, Russia, India, and China, BRICS currently includes 10 members (Brazil, Russia, India, China, South Africa, Egypt, Ethiopia, Indonesia, Iran and the United Arab Emirates) and 10 partner countries.] As a group, they carry the collective weight to shape international rules, influence global markets, and redefine norms — yet the question remains: Will they use this power effectively? The BRICS Lab explores the implications of the bloc’s rise for international cooperation and its role in reshaping global politics. Our work focuses on three areas: the design and strategic use of informal groups like BRICS in world affairs; the coalition’s potential to address major challenges such as food security, climate change, and artificial intelligence; and the implications of U.S. policy toward BRICS for the future of multilateralism.

Q: What are the center’s biggest research priorities right now?

A: Our center was founded in response to rising geopolitical tensions and the urgent need for policy rooted in rigorous, evidence-based research. Since then, we have grown into a hub that combines interdisciplinary scholarship and actively engages with policymakers and the public. Today, as in our early years, the center brings together exceptional researchers with the ambition to address the world’s most pressing challenges in new and creative ways.

Our core focus spans security, development, and human dignity. Security studies have been a priority for the center, and our new nuclear security programming advances this work while training the next generation of scholars in this critical field. On the development front, our work has explored how societies manage diverse populations, navigate international migration, as well as engage with human rights and the changing patterns of regime dynamics.

We are pursuing new research in three areas. First, on climate change, we seek to understand how societies confront environmental risks and harms, from insurance to water and food security in the international context. Second, we examine shifting patterns of global governance as rising powers set new agendas and take on greater responsibilities in the international system. Finally, we are initiating research on the impact of AI — how it reshapes governance across international relations, what is the role of AI corporations, and how AI-related risks can be managed.

As we approach our 75th anniversary in 2026, we are excited to bring researchers together to spark bold ideas that open new possibilities for the future.

A Saab 340 aircraft recently became a permanent fixture of the fleet at the MIT Lincoln Laboratory Flight Test Facility, which supports R&D programs across the lab. 

Over the past five years, the facility leased and operated the twin-engine turboprop, once commercially used for the regional transport of passengers and cargo. During this time, staff modified the aircraft with a suite of radar, sensing, and communications capabilities. Transitioning the aircraft from a leased to a government-owned asset retains the aircraft's capabilities for present and future R&D in support of national security and reduces costs for Lincoln Laboratory sponsors. 

With the acquisition of the Saab, the Flight Test Facility currently maintains five government-owned aircraft — including three Gulfstream IVs and a Cessna 206 — as well as a leased Twin Otter, all housed on Hanscom Air Force Base, just over a mile from the laboratory's main campus.

"Of all our aircraft, the Saab is the most multi-mission-capable," says David Culbertson, manager of the Flight Test Facility. "It's highly versatile and adaptable, like a Swiss Army knife. Researchers from across the laboratory have conducted flight tests on the Saab to develop all kinds of technologies for national security."

For example, the Saab was modified to host the Airborne Radar Testbed (ARTB), a high-performance radar system based on a computer-controlled array of antennas that can be electronically steered (instead of physically moved) in different directions. With the ARTB, researchers have matured innovative radio-frequency technology; prototyped advanced system concepts; and demonstrated concepts of operation for intelligence, surveillance, and reconnaissance (ISR) missions. With its open-architecture design and compliance with open standards, the ARTB can easily be reconfigured to suit specific R&D needs.

"The Saab has enabled us to rapidly prototype and mature the complex system-of-systems solutions needed to realize critical warfighter capabilities," says Ramu Bhagavatula, an assistant leader of the laboratory's Embedded and Open Systems Group. "Recently, the Saab participated in a major national exercise as a surrogate multi-INT [intelligence] ISR platform. We demonstrated machine-to-machine cueing of our multi-INT payload to automatically recognize targets designated by an operational U.S. Air Force platform. The Saab's flexibility was key to integrating diverse technologies to develop this important capability."

In anticipation of the expiration of the Saab's lease, the Flight Test Facility and Financial Services Department conducted an extensive analysis of alternatives. Comparing the operational effectiveness, suitability, and life-cycle cost of various options, this analysis determined that the optimal solution for the laboratory and the government was to purchase the aircraft.

"Having the Saab in our permanent inventory allows research groups from across the laboratory to continuously leverage each other's test beds and expertise," says Linda McCabe, a project manager in the laboratory's Communication Networks and Analysis Group. "In addition, we can invest in long-term infrastructure updates that will benefit a wide range of users. For instance, my group helped obtain authorizations from various agencies to equip the Saab with Link 16, a secure communications network used by NATO and its allies to share tactical information."

The Saab acquisition is part of a larger recapitalization effort at the Flight Test Facility to support emerging technology development for years to come. This 10-year effort, slated for completion in 2026, is retiring aging, obsolete aircraft and replacing them with newer platforms that will be more cost-effective to maintain, easier to integrate rapidly prototyped systems into, and able to operate under expanded flight envelopes (the performance limits within which an aircraft can safely fly, defined by parameters such as speed, altitude, and maneuverability).

The successful test of SpaceX’s Starship launch vehicle, following a series of engineering challenges and failed launches, has reignited excitement over the possibilities this massive rocket may unlock for humanity’s greatest ambitions in space. The largest rocket ever built, Starship and its 33-engine “super heavy” booster completed a full launch into Earth orbit on Aug. 26, deployed eight test prototype satellites, and survived reentry for a simulated landing before coming down, mostly intact, in the Indian Ocean. The 400-foot rocket is designed to carry up to 150 tons of cargo to low Earth orbit, dramatically increasing potential payload volume from rockets currently in operation. In addition to the planned Artemis III mission to the lunar surface and proposed missions to Mars in the near future, Starship also poses an opportunity for large-scale scientific missions throughout the solar system.

The National Academy of Sciences Planetary Science Decadal Survey published a recommendation in 2022 outlining exploration of Uranus as its highest-priority flagship mission. This proposed mission was envisioned for the 2030s, assuming use of a Falcon Heavy expendable rocket and anticipating arrival at the planet before 2050. Earlier this summer, a paper from researchers in MIT’s Engineering Systems Lab found that Starship may enable this flagship mission to Uranus in half the flight time. 

In this 3Q, Chloe Gentgen, a PhD student in aeronautics and astronautics and co-author on the recent study, describes the significance of Uranus as a flagship mission and what the current trajectory of Starship means for scientific exploration.

Q: Why has Uranus been identified as the highest-priority flagship mission? 

A: Uranus is one of the most intriguing and least-explored planets in our solar system. The planet is tilted on its side, is extremely cold, presents a highly dynamic atmosphere with fast winds, and has an unusual and complex magnetic field. A few of Uranus’ many moons could be ocean worlds, making them potential candidates in the search for life in the solar system. The ice giants Uranus and Neptune also represent the closest match to most of the exoplanets discovered. A mission to Uranus would therefore radically transform our understanding of ice giants, the solar system, and exoplanets. 

What we know about Uranus largely dates back to Voyager 2’s brief flyby nearly 40 years ago. No spacecraft has visited Uranus or Neptune since, making them the only planets yet to have a dedicated orbital mission. One of the main obstacles has been the sheer distance. Uranus is 19 times farther from the sun than the Earth is, and nearly twice as far as Saturn. Reaching it requires a heavy-lift launch vehicle and trajectories involving gravity assists from other planets. 

Today, such heavy-lift launch vehicles are available, and trajectories have been identified for launch windows throughout the 2030s, which resulted in selecting a Uranus mission as the highest priority flagship in the 2022 decadal survey. The proposed concept, called Uranus Orbiter and Probe (UOP), would release a probe into the planet’s atmosphere and then embark on a multiyear tour of the system to study the planet’s interior, atmosphere, magnetosphere, rings, and moons. 

Q: How do you envision your work on the Starship launch vehicle being deployed for further development?

A: Our study assessed the feasibility and potential benefits of launching a mission to Uranus with a Starship refueled in Earth’s orbit, instead of a Falcon Heavy (another SpaceX launch vehicle, currently operational). The Uranus decadal study showed that launching on a Falcon Heavy Expendable results in a cruise time of at least 13 years. Long cruise times present challenges, such as loss of team expertise and a higher operational budget. With the mission not yet underway, we saw an opportunity to evaluate launch vehicles currently in development, particularly Starship. 

When refueled in orbit, Starship could launch a spacecraft directly to Uranus, without detours by other planets for gravity-assist maneuvers. The proposed spacecraft could then arrive at Uranus in just over six years, less than half the time currently envisioned. These high-energy trajectories require significant deceleration at Uranus to capture in orbit. If the spacecraft slows down propulsively, the burn would require 5 km/s of delta v (which quantifies the energy needed for the maneuver), much higher than is typically performed by spacecraft, which might result in a very complex design. A more conservative approach, assuming a maximum burn of 2 km/s at Uranus, would result in a cruise time of 8.5 years. 

An alternative to propulsive orbit insertion at Uranus is aerocapture, where the spacecraft, enclosed in a thermally protective aeroshell, dips into the planet’s atmosphere and uses aerodynamic drag to decelerate. We examined whether Starship itself could perform aerocapture, rather than being separated from the spacecraft shortly after launch. Starship is already designed to withstand atmospheric entry at Earth and Mars, and thus already has a thermal protection system that could, potentially, be modified for aerocapture at Uranus. While bringing a Starship vehicle all the way to Uranus presents significant challenges, our analysis showed that aerocapture with Starship would produce deceleration and heating loads similar to those of other Uranus aerocapture concepts and would enable a cruise time of six years.

In addition to launching the proposed spacecraft on a faster trajectory that would reach Uranus sooner, Starship’s capabilities could also be leveraged to deploy larger masses to Uranus, enabling an enhanced mission with additional instruments or probes.

Q: What does the recent successful test of Starship tell us about the viability and timeline for a potential mission to the outer solar system?

A: The latest Starship launch marked an important milestone for the company after three failed launches in recent months, renewing optimism about the rocket’s future capabilities. Looking ahead, the program will need to demonstrate on-orbit refueling, a capability central to both SpaceX’s long-term vision of deep-space exploration and this proposed mission.

Launch vehicle selection for flagship missions typically occurs approximately two years after the official mission formulation process begins, which has not yet commenced for the Uranus mission. As such, Starship still has a few more years to demonstrate its on-orbit refueling architecture before a decision has to be made.

Overall, Starship is still under development, and significant uncertainty remains about its performance, timelines, and costs. Even so, our initial findings paint a promising picture of the benefits that could be realized by using Starship for a flagship mission to Uranus.

The following article is adapted from a joint press release issued today by MIT and the Giant Magellan Telescope.

MIT is lending its support to the Giant Magellan Telescope, joining the international consortium to advance the $2.6 billion observatory in Chile. The Institute’s participation, enabled by a transformational gift from philanthropists Phillip (Terry) Ragon ’72 and Susan Ragon, adds to the momentum to construct the Giant Magellan Telescope, whose 25.4-meter aperture will have five times the light-collecting area and up to 200 times the power of existing observatories.

“As philanthropists, Terry and Susan have an unerring instinct for finding the big levers: those interventions that truly transform the scientific landscape,” says MIT President Sally Kornbluth. “We saw this with their founding of the Ragon Institute, which pursues daring approaches to harnessing the immune system to prevent and cure human diseases. With today’s landmark gift, the Ragons enable an equally lofty mission to better understand the universe — and we could not be more grateful for their visionary support."

MIT will be the 16th member of the international consortium advancing the Giant Magellan Telescope and the 10th participant based in the United States. Together, the consortium has invested $1 billion in the observatory — the largest-ever private investment in ground-based astronomy. The Giant Magellan Telescope is already 40 percent under construction, with major components being designed and manufactured across 36 U.S. states.

“MIT is honored to join the consortium and participate in this exceptional scientific endeavor,” says Ian A. Waitz, MIT’s vice president for research. “The Giant Magellan Telescope will bring tremendous new capabilities to MIT astronomy and to U.S. leadership in fundamental science. The construction of this uniquely powerful telescope represents a vital private and public investment in scientific excellence for decades to come.”

MIT brings to the consortium powerful scientific capabilities and a legacy of astronomical excellence. MIT’s departments of Physics and of Earth, Atmospheric and Planetary Sciences, and the MIT Kavli Institute for Astrophysics and Space Research, are internationally recognized for research in exoplanets, cosmology, and environments of extreme gravity, such as black holes and compact binary stars. MIT’s involvement will strengthen the Giant Magellan Telescope’s unique capabilities in high-resolution spectroscopy, adaptive optics, and the search for life beyond Earth. It also deepens a long-standing scientific relationship: MIT is already a partner in the existing twin Magellan Telescopes at Las Campanas Observatory in Chile — one of the most scientifically valuable observing sites on Earth, and the same site where the Giant Magellan Telescope is now under construction.

“Since Galileo’s first spyglass, the world’s largest telescope has doubled in aperture every 40 to 50 years,” says Robert A. Simcoe, director of the MIT Kavli Institute and the Francis L. Friedman Professor of Physics. “Each generation’s leading instruments have resolved important scientific questions of the day and then surprised their builders with new discoveries not yet even imagined, helping humans understand our place in the universe. Together with the Giant Magellan Telescope, MIT is helping to realize our generation’s contribution to this lineage, consistent with our mission to advance the frontier of fundamental science by undertaking the most audacious and advanced engineering challenges.”

Contributing to the national strategy

MIT’s support comes at a pivotal time for the observatory. In June 2025, the National Science Foundation (NSF) advanced the Giant Magellan Telescope into its Final Design Phase, one of the final steps before it becomes eligible for federal construction funding. To demonstrate readiness and a strong commitment to U.S. leadership, the consortium offered to privately fund this phase, which is traditionally supported by the NSF.

MIT’s investment is an integral part of the national strategy to secure U.S. access to the next generation of research facilities known as “extremely large telescopes.” The Giant Magellan Telescope is a core partner in the U.S. Extremely Large Telescope Program, the nation’s top priority in astronomy. The National Academies’ Astro2020 Decadal Survey called the program “absolutely essential if the United States is to maintain a position as a leader in ground-based astronomy.” This long-term strategy also includes the recently commissioned Vera C. Rubin Observatory in Chile. Rubin is scanning the sky to detect rare, fast-changing cosmic events, while the Giant Magellan Telescope will provide the sensitivity, resolution, and spectroscopic instruments needed to study them in detail. Together, these Southern Hemisphere observatories will give U.S. scientists the tools they need to lead 21st-century astrophysics.

“Without direct access to the Giant Magellan Telescope, the U.S. risks falling behind in fundamental astronomy, as Rubin’s most transformational discoveries will be utilized by other nations with access to their own ‘extremely large telescopes’ under development,” says Walter Massey, board chair of the Giant Magellan Telescope.

MIT’s participation brings the United States a step closer to completing the promise of this powerful new observatory on a globally competitive timeline. With federal construction funding, it is expected that the observatory could reach 90 percent completion in less than two years and become operational by the 2030s.

“MIT brings critical expertise and momentum at a time when global leadership in astronomy hangs in the balance,” says Robert Shelton, president of the Giant Magellan Telescope. “With MIT, we are not just adding a partner; we are accelerating a shared vision for the future and reinforcing the United States’ position at the forefront of science.”

Other members of the Giant Magellan Telescope consortium include the University of Arizona, Carnegie Institution for Science, The University of Texas at Austin, Korea Astronomy and Space Science Institute, University of Chicago, São Paulo Research Foundation (FAPESP), Texas A&M University, Northwestern University, Harvard University, Astronomy Australia Ltd., Australian National University, Smithsonian Institution, Weizmann Institute of Science, Academia Sinica Institute of Astronomy and Astrophysics, and Arizona State University.

A boon for astrophysics research and education

Access to the world’s best optical telescopes is a critical resource for MIT researchers. More than 150 individual science programs at MIT have relied on major astronomical observatories in the past three years, engaging faculty, researchers, and students in investigations into the marvels of the universe. Recent research projects have included chemical studies of the universe’s oldest stars, led by Professor Anna Frebel; spectroscopy of stars shredded by dormant black holes, led by Professor Erin Kara; and measurements of a white dwarf teetering on the precipice of a black hole, led by Professor Kevin Burdge. 

“Over many decades, researchers at the MIT Kavli Institute have used unparalleled instruments to discover previously undetected cosmic phenomena from both ground-based observations and spaceflight missions,” says Nergis Mavalvala, dean of the MIT School of Science and the Curtis (1963) and Kathleen Marble Professor of Astrophysics. “I have no doubt our brilliant colleagues will carry on that tradition with the Giant Magellan Telescope, and I can’t wait to see what they will discover next.”

The Giant Magellan Telescope will also provide a platform for advanced R&D in remote sensing, creating opportunities to build custom infrared and optical spectrometers and high-speed imagers to further study our universe.

“One cannot have a leading physics program without a leading astrophysics program. Access to time on the Giant Magellan Telescope will ensure that future generations of MIT researchers will continue to work at the forefront of astrophysical discovery for decades to come,” says Deepto Chakrabarty, head of the MIT Department of Physics, the William A. M. Burden Professor in Astrophysics, and principal investigator at the MIT Kavli Institute. “Our institutional access will help attract and retain top researchers in astrophysics, planetary science, and advanced optics, and will give our PhD students and postdocs unrivaled educational opportunities.”

In part 2 of our two-part series on generative artificial intelligence’s environmental impacts, MIT News explores some of the ways experts are working to reduce the technology’s carbon footprint.

The energy demands of generative AI are expected to continue increasing dramatically over the next decade.

For instance, an April 2025 report from the International Energy Agency predicts that the global electricity demand from data centers, which house the computing infrastructure to train and deploy AI models, will more than double by 2030, to around 945 terawatt-hours. While not all operations performed in a data center are AI-related, this total amount is slightly more than the energy consumption of Japan.

Moreover, an August 2025 analysis from Goldman Sachs Research forecasts that about 60 percent of the increasing electricity demands from data centers will be met by burning fossil fuels, increasing global carbon emissions by about 220 million tons. In comparison, driving a gas-powered car for 5,000 miles produces about 1 ton of carbon dioxide.

These statistics are staggering, but at the same time, scientists and engineers at MIT and around the world are studying innovations and interventions to mitigate AI’s ballooning carbon footprint, from boosting the efficiency of algorithms to rethinking the design of data centers.

Considering carbon emissions

Talk of reducing generative AI’s carbon footprint is typically centered on “operational carbon” — the emissions used by the powerful processors, known as GPUs, inside a data center. It often ignores “embodied carbon,” which are emissions created by building the data center in the first place, says Vijay Gadepally, senior scientist at MIT Lincoln Laboratory, who leads research projects in the Lincoln Laboratory Supercomputing Center.

Constructing and retrofitting a data center, built from tons of steel and concrete and filled with air conditioning units, computing hardware, and miles of cable, consumes a huge amount of carbon. In fact, the environmental impact of building data centers is one reason companies like Meta and Google are exploring more sustainable building materials. (Cost is another factor.)

Plus, data centers are enormous buildings — the world’s largest, the China Telecomm-Inner Mongolia Information Park, engulfs roughly 10 million square feet — with about 10 to 50 times the energy density of a normal office building, Gadepally adds. 

“The operational side is only part of the story. Some things we are working on to reduce operational emissions may lend themselves to reducing embodied carbon, too, but we need to do more on that front in the future,” he says.

Reducing operational carbon emissions

When it comes to reducing operational carbon emissions of AI data centers, there are many parallels with home energy-saving measures. For one, we can simply turn down the lights.

“Even if you have the worst lightbulbs in your house from an efficiency standpoint, turning them off or dimming them will always use less energy than leaving them running at full blast,” Gadepally says.

In the same fashion, research from the Supercomputing Center has shown that “turning down” the GPUs in a data center so they consume about three-tenths the energy has minimal impacts on the performance of AI models, while also making the hardware easier to cool.

Another strategy is to use less energy-intensive computing hardware.

Demanding generative AI workloads, such as training new reasoning models like GPT-5, usually need many GPUs working simultaneously. The Goldman Sachs analysis estimates that a state-of-the-art system could soon have as many as 576 connected GPUs operating at once.

But engineers can sometimes achieve similar results by reducing the precision of computing hardware, perhaps by switching to less powerful processors that have been tuned to handle a specific AI workload.

There are also measures that boost the efficiency of training power-hungry deep-learning models before they are deployed.

Gadepally’s group found that about half the electricity used for training an AI model is spent to get the last 2 or 3 percentage points in accuracy. Stopping the training process early can save a lot of that energy.

“There might be cases where 70 percent accuracy is good enough for one particular application, like a recommender system for e-commerce,” he says.

Researchers can also take advantage of efficiency-boosting measures.

For instance, a postdoc in the Supercomputing Center realized the group might run a thousand simulations during the training process to pick the two or three best AI models for their project.

By building a tool that allowed them to avoid about 80 percent of those wasted computing cycles, they dramatically reduced the energy demands of training with no reduction in model accuracy, Gadepally says.

Leveraging efficiency improvements

Constant innovation in computing hardware, such as denser arrays of transistors on semiconductor chips, is still enabling dramatic improvements in the energy efficiency of AI models.

Even though energy efficiency improvements have been slowing for most chips since about 2005, the amount of computation that GPUs can do per joule of energy has been improving by 50 to 60 percent each year, says Neil Thompson, director of the FutureTech Research Project at MIT’s Computer Science and Artificial Intelligence Laboratory and a principal investigator at MIT’s Initiative on the Digital Economy.

“The still-ongoing ‘Moore’s Law’ trend of getting more and more transistors on chip still matters for a lot of these AI systems, since running operations in parallel is still very valuable for improving efficiency,” says Thomspon.

Even more significant, his group’s research indicates that efficiency gains from new model architectures that can solve complex problems faster, consuming less energy to achieve the same or better results, is doubling every eight or nine months.

Thompson coined the term “negaflop” to describe this effect. The same way a “negawatt” represents electricity saved due to energy-saving measures, a “negaflop” is a computing operation that doesn’t need to be performed due to algorithmic improvements.

These could be things like “pruning” away unnecessary components of a neural network or employing compression techniques that enable users to do more with less computation.

“If you need to use a really powerful model today to complete your task, in just a few years, you might be able to use a significantly smaller model to do the same thing, which would carry much less environmental burden. Making these models more efficient is the single-most important thing you can do to reduce the environmental costs of AI,” Thompson says.

Maximizing energy savings

While reducing the overall energy use of AI algorithms and computing hardware will cut greenhouse gas emissions, not all energy is the same, Gadepally adds.

“The amount of carbon emissions in 1 kilowatt hour varies quite significantly, even just during the day, as well as over the month and year,” he says.

Engineers can take advantage of these variations by leveraging the flexibility of AI workloads and data center operations to maximize emissions reductions. For instance, some generative AI workloads don’t need to be performed in their entirety at the same time.

Splitting computing operations so some are performed later, when more of the electricity fed into the grid is from renewable sources like solar and wind, can go a long way toward reducing a data center’s carbon footprint, says Deepjyoti Deka, a research scientist in the MIT Energy Initiative.

Deka and his team are also studying “smarter” data centers where the AI workloads of multiple companies using the same computing equipment are flexibly adjusted to improve energy efficiency.

“By looking at the system as a whole, our hope is to minimize energy use as well as dependence on fossil fuels, while still maintaining reliability standards for AI companies and users,” Deka says.

He and others at MITEI are building a flexibility model of a data center that considers the differing energy demands of training a deep-learning model versus deploying that model. Their hope is to uncover the best strategies for scheduling and streamlining computing operations to improve energy efficiency.

The researchers are also exploring the use of long-duration energy storage units at data centers, which store excess energy for times when it is needed.

With these systems in place, a data center could use stored energy that was generated by renewable sources during a high-demand period, or avoid the use of diesel backup generators if there are fluctuations in the grid.

“Long-duration energy storage could be a game-changer here because we can design operations that really change the emission mix of the system to rely more on renewable energy,” Deka says.

In addition, researchers at MIT and Princeton University are developing a software tool for investment planning in the power sector, called GenX, which could be used to help companies determine the ideal place to locate a data center to minimize environmental impacts and costs.

Location can have a big impact on reducing a data center’s carbon footprint. For instance, Meta operates a data center in Lulea, a city on the coast of northern Sweden where cooler temperatures reduce the amount of electricity needed to cool computing hardware.

Thinking farther outside the box (way farther), some governments are even exploring the construction of data centers on the moon where they could potentially be operated with nearly all renewable energy.

AI-based solutions

Currently, the expansion of renewable energy generation here on Earth isn’t keeping pace with the rapid growth of AI, which is one major roadblock to reducing its carbon footprint, says Jennifer Turliuk MBA ’25, a short-term lecturer, former Sloan Fellow, and former practice leader of climate and energy AI at the Martin Trust Center for MIT Entrepreneurship.

The local, state, and federal review processes required for a new renewable energy projects can take years.

Researchers at MIT and elsewhere are exploring the use of AI to speed up the process of connecting new renewable energy systems to the power grid.

For instance, a generative AI model could streamline interconnection studies that determine how a new project will impact the power grid, a step that often takes years to complete.

And when it comes to accelerating the development and implementation of clean energy technologies, AI could play a major role.

“Machine learning is great for tackling complex situations, and the electrical grid is said to be one of the largest and most complex machines in the world,” Turliuk adds.

For instance, AI could help optimize the prediction of solar and wind energy generation or identify ideal locations for new facilities.

It could also be used to perform predictive maintenance and fault detection for solar panels or other green energy infrastructure, or to monitor the capacity of transmission wires to maximize efficiency.

By helping researchers gather and analyze huge amounts of data, AI could also inform targeted policy interventions aimed at getting the biggest “bang for the buck” from areas such as renewable energy, Turliuk says.

To help policymakers, scientists, and enterprises consider the multifaceted costs and benefits of AI systems, she and her collaborators developed the Net Climate Impact Score.

The score is a framework that can be used to help determine the net climate impact of AI projects, considering emissions and other environmental costs along with potential environmental benefits in the future.

At the end of the day, the most effective solutions will likely result from collaborations among companies, regulators, and researchers, with academia leading the way, Turliuk adds.

“Every day counts. We are on a path where the effects of climate change won’t be fully known until it is too late to do anything about it. This is a once-in-a-lifetime opportunity to innovate and make AI systems less carbon-intense,” she says.

A team of MIT geochemists has unearthed new evidence in very old rocks suggesting that some of the first animals on Earth were likely ancestors of the modern sea sponge.

In a study appearing today in the Proceedings of the National Academy of Sciences, the researchers report that they have identified “chemical fossils” that may have been left by ancient sponges in rocks that are more than 541 million years old. A chemical fossil is a remnant of a biomolecule that originated from a living organism that has since been buried, transformed, and preserved in sediment, sometimes for hundreds of millions of years.

The newly identified chemical fossils are special types of steranes, which are the geologically stable form of sterols, such as cholesterol, that are found in the cell membranes of complex organisms. The researchers traced these special steranes to a class of sea sponges known as demosponges. Today, demosponges come in a huge variety of sizes and colors, and live throughout the oceans as soft and squishy filter feeders. Their ancient counterparts may have shared similar characteristics.

“We don’t know exactly what these organisms would have looked like back then, but they absolutely would have lived in the ocean, they would have been soft-bodied, and we presume they didn’t have a silica skeleton,” says Roger Summons, the Schlumberger Professor of Geobiology Emeritus in MIT’s Department of Earth, Atmospheric and Planetary Sciences (EAPS).

The group’s discovery of sponge-specific chemical fossils offers strong evidence that the ancestors of demosponges were among the first animals to evolve, and that they likely did so much earlier than the rest of Earth’s major animal groups.

The study’s authors, including Summons, are lead author and former MIT EAPS Crosby Postdoctoral Fellow Lubna Shawar, who is now a research scientist at Caltech, along with Gordon Love from the University of California at Riverside, Benjamin Uveges of Cornell University, Alex Zumberge of GeoMark Research in Houston, Paco Cárdenas of Uppsala University in Sweden, and José-Luis Giner of the State University of New York College of Environmental Science and Forestry.

Sponges on steroids

The new study builds on findings that the group first reported in 2009. In that study, the team identified the first chemical fossils that appeared to derive from ancient sponges. They analyzed rock samples from an outcrop in Oman and found a surprising abundance of steranes that they determined were the preserved remnants of 30-carbon (C₃₀) sterols — a rare form of steroid that they showed was likely derived from ancient sea sponges.

The steranes were found in rocks that were very old and formed during the Ediacaran Period — which spans from roughly 541 million to about 635 million years ago. This period took place just before the Cambrian, when the Earth experienced a sudden and global explosion of complex multicellular life. The team’s discovery suggested that ancient sponges appeared much earlier than most multicellular life, and were possibly one of Earth’s first animals.

However, soon after these findings were released, alternative hypotheses swirled to explain the C₃₀ steranes’ origins, including that the chemicals could have been generated by other groups of organisms or by nonliving geological processes.

The team says the new study reinforces their earlier hypothesis that ancient sponges left behind this special chemical record, as they have identified a new chemical fossil in the same Precambrian rocks that is almost certainly biological in origin.

Building evidence

Just as in their previous work, the researchers looked for chemical fossils in rocks that date back to the Ediacaran Period. They acquired samples from drill cores and outcrops in Oman, western India, and Siberia, and analyzed the rocks for signatures of steranes, the geologically stable form of sterols found in all eukaryotes (plants, animals, and any organism with a nucleus and membrane-bound organelles).

“You’re not a eukaryote if you don’t have sterols or comparable membrane lipids,” Summons says.

A sterol’s core structure consists of four fused carbon rings. Additional carbon side chain and chemical add-ons can attach to and extend a sterol’s structure, depending on what an organism’s particular genes can produce. In humans, for instance, the sterol cholesterol contains 27 carbon atoms, while the sterols in plants generally have 29 carbon atoms.

“It’s very unusual to find a sterol with 30 carbons,” Shawar says.

The chemical fossil the researchers identified in 2009 was a 30-carbon sterol. What’s more, the team determined that the compound could be synthesized because of the presence of a distinctive enzyme which is encoded by a gene that is common to demosponges.

In their new study, the team focused on the chemistry of these compounds and realized the same sponge-derived gene could produce an even rarer sterol, with 31 carbon atoms (C₃₁). When they analyzed their rock samples for C₃₁ steranes, they found it in surprising abundance, along with the aforementioned C₃₀ steranes.

“These special steranes were there all along,” Shawar says. “It took asking the right questions to seek them out and to really understand their meaning and from where they come.”

The researchers also obtained samples of modern-day demosponges and analyzed them for C₃₁ sterols. They found that, indeed, the sterols — biological precursors of the C₃₁ steranes found in rocks — are present in some species of contemporary demosponges. Going a step further, they chemically synthesized eight different C₃₁ sterols in the lab as reference standards to verify their chemical structures. Then, they processed the molecules in ways that simulate how the sterols would change when deposited, buried, and pressurized over hundreds of millions of years. They found that the products of only two such sterols were an exact match with the form of C₃₁ sterols that they found in ancient rock samples. The presence of two and the absence of the other six demonstrates that these compounds were not produced by a random nonbiological process.

The findings, reinforced by multiple lines of inquiry, strongly support the idea that the steranes that were found in ancient rocks were indeed produced by living organisms, rather than through geological processes. What’s more, those organisms were likely the ancestors of demosponges, which to this day have retained the ability to produce the same series of compounds.

“It’s a combination of what’s in the rock, what’s in the sponge, and what you can make in a chemistry laboratory,” Summons says. “You’ve got three supportive, mutually agreeing lines of evidence, pointing to these sponges being among the earliest animals on Earth.”

“In this study we show how to authenticate a biomarker, verifying that a signal truly comes from life rather than contamination or non-biological chemistry,” Shawar adds.

Now that the team has shown C₃₀ and C₃₁ sterols are reliable signals of ancient sponges, they plan to look for the chemical fossils in ancient rocks from other regions of the world. They can only tell from the rocks they’ve sampled so far that the sediments, and the sponges, formed some time during the Ediacaran Period. With more samples, they will have a chance to narrow in on when some of the first animals took form.

This research was supported, in part, by the MIT Crosby Fund, the Distinguished Postdoctoral Fellowship program, the Simons Foundation Collaboration on the Origins of Life, and the NASA Exobiology Program. 

The brain divides vision between its two hemispheres — what’s on your left is processed by your right hemisphere, and vice versa — but your experience with every bike or bird that you see zipping by is seamless. A new study by neuroscientists at The Picower Institute for Learning and Memory at MIT reveals how the brain handles the transition.

“It’s surprising to some people to hear that there’s some independence between the hemispheres, because that doesn’t really correspond to how we perceive reality,” says Earl K. Miller, Picower Professor in the Picower Institute and MIT’s Department of Brain and Cognitive Sciences. “In our consciousness, everything seems to be unified.”

There are advantages to separately processing vision on either side of the brain, including the ability to keep track of more things at once, Miller and other researchers have found, but neuroscientists have been eager to fully understand how perception ultimately appears so unified in the end.

Led by Picower Fellow Matthew Broschard and Research Scientist Jefferson Roy, the research team measured neural activity in the brains of animals as they tracked objects crossing their field of view. The results reveal that different frequencies of brain waves encoded and then transferred information from one hemisphere to the other in advance of the crossing, and then held on to the object representation in both hemispheres until after the crossing was complete. The process is analogous to how relay racers hand off a baton, how a child swings from one monkey bar to the next, and how cellphone towers hand off a call from one to the next as a train passenger travels through their area. In all cases, both towers or hands actively hold what’s being transferred until the handoff is confirmed.

Witnessing the handoff

To conduct the study, published Sept. 19 in the Journal of Neuroscience, the researchers measured both the electrical spiking of individual neurons and the various frequencies of brain waves that emerge from the coordinated activity of many neurons. They studied the dorsal and ventrolateral prefrontal cortex in both hemispheres, brain areas associated with executive brain functions.

The power fluctuations of the wave frequencies in each hemisphere told the researchers a clear story about how the subject’s brains transferred information from the “sending” to the “receiving” hemisphere whenever a target object crossed the middle of their field of view. In the experiments, the target was accompanied by a distractor object on the opposite side of the screen to confirm that the subjects were consciously paying attention to the target object’s motion, and not just indiscriminately glancing at whatever happened to pop up on to the screen.

The highest-frequency “gamma” waves, which encode sensory information, peaked in both hemispheres when the subjects first looked at the screen and again when the two objects appeared. When a color change signaled which object was the target to track, the gamma increase was only evident in the “sending” hemisphere (on the opposite side as the target object), as expected. Meanwhile, the power of somewhat lower-frequency “beta” waves, which regulate when gamma waves are active, varied inversely with the gamma waves. These sensory encoding dynamics were stronger in the ventrolateral locations compared to the dorsolateral ones.

Meanwhile, two distinct bands of lower-frequency waves showed greater power in the dorsolateral locations at key moments related to achieving the handoff. About a quarter of a second before a target object crossed the middle of the field of view, “alpha” waves ramped up in both hemispheres and then peaked just after the object crossed. Meanwhile, “theta” band waves peaked after the crossing was complete, only in the “receiving” hemisphere (opposite from the target’s new position).

Accompanying the pattern of wave peaks, neuron spiking data showed how the brain’s representation of the target’s location traveled. Using decoder software, which interprets what information the spikes represent, the researchers could see the target representation emerge in the sending hemisphere’s ventrolateral location when it was first cued by the color change. Then they could see that as the target neared the middle of the field of view, the receiving hemisphere joined the sending hemisphere in representing the object, so that they both encoded the information during the transfer.

Doing the wave

Taken together, the results showed that after the sending hemisphere initially encoded the target with a ventrolateral interplay of beta and gamma waves, a dorsolateral ramp up of alpha waves caused the receiving hemisphere to anticipate the handoff by mirroring the sending hemisphere’s encoding of the target information. Alpha peaked just after the target crossed the middle of the field of view, and when the handoff was complete, theta peaked in the receiving hemisphere as if to say, “I got it.”

And in trials where the target never crossed the middle of the field of view, these handoff dynamics were not apparent in the measurements.

The study shows that the brain is not simply tracking objects in one hemisphere and then just picking them up anew when they enter the field of view of the other hemisphere.

“These results suggest there are active mechanisms that transfer information between cerebral hemispheres,” the authors wrote. “The brain seems to anticipate the transfer and acknowledge its completion.”

But they also note, based on other studies, that the system of interhemispheric coordination can sometimes appear to break down in certain neurological conditions including schizophrenia, autism, depression, dyslexia, and multiple sclerosis. The new study may lend insight into the specific dynamics needed for it to succeed.

In addition to Broschard, Roy, and Miller, the paper’s other authors are Scott Brincat and Meredith Mahnke.

Funding for the study came from the Office of Naval Research, the National Eye Institute of the National Institutes of Health, The Freedom Together Foundation, and The Picower Institute for Learning and Memory.

Gleevec, a cancer drug first approved for sale in 2001, has dramatically changed the lives of people with chronic myeloid leukemia. This form of cancer was once regarded as very difficult to combat, but survival rates of patients who respond to Gleevec now resemble that of the population at large.

Gleevec is also a medicine developed with the help of federally funded research. That support helped scientists better understand how to create drugs targeting the BCR-ABL oncoprotein, the cancer-causing protein behind chronic myeloid leukemia.

A new study co-authored by MIT researchers quantifies how many such examples of drug development exist. The current administration is proposing a nearly 40 percent budget reduction to the National Institutes of Health (NIH), which sponsors a significant portion of biomedical research. The study finds that over 50 percent of small-molecule drug patents this century cite at least one piece of NIH-backed research that would likely be vulnerable to that potential level of funding change.

“What we found was quite striking,” says MIT economist Danielle Li, co-author of a newly published paper outlining the study’s results. “More than half of the drugs approved by the FDA since 2000 are connected to NIH research that would likely have been cut under a 40 percent budget reduction.”

Or, as the researchers write in the paper: “We found extensive connections between medical advances and research that was funded by grants that would have been cut if the NIH budget was sharply reduced.”

The paper, “What if NIH funding had been 40% smaller?” is published today as a Policy Article in the journal Science. The authors are Pierre Azoulay, the China Program Professor of International Management at the MIT Sloan School of Management; Matthew Clancy, an economist with the group Open Philanthropy; Li, the David Sarnoff Professor of Management of Technology at MIT Sloan; and Bhaven N. Sampat, an economist at Johns Hopkins University. (Biomedical researchers at both MIT and Johns Hopkins could be affected by adjustments to NIH funding.)

To conduct the study, the researchers leveraged the fact that the NIH uses priority lists to determine which projects get funded. That makes it possible to discern which projects were in the lower 40 percent of NIH-backed projects, priority-wise, for a given time period. The researchers call these “at-risk” pieces of research. Applying these data from 1980 through 2007, the scholars examined the patents of the new molecular entities — drugs with a new active ingredient — approved by the U.S. Food and Drug Administration since 2000. There is typically a time interval between academic research and subsequent related drug development.

The study focuses on small-molecule drugs — compact organic compounds, often taken orally as medicine — whereas NIH funding supports a wider range of advancements in medicine generally. Based on how many of these FDA-approved small-molecule medicines were linked to at-risk research from the prior period, the researchers estimated what kinds of consequences a 40 percent cut in funding would have generated going forward.

The study distinguishes between two types of links new drugs have to NIH funding. Some drug patents have what the researchers call “direct” links to new NIH-backed projects that generated new findings relevant to development of those particular drugs. Other patents have “indirect “ links to the NIH, when they cite prior NIH-funded studies that contributed to the overall body of knowledge used in drug development.

The analysis finds that 40 of the FDA-approved medications have direct links to new NIH-supported studies cited in the patents — or 7.1 percent. Of these, 14 patents cite at-risk pieces of NIH research.

When it comes to indirect links, of the 557 drugs approved by the FDA from 2000 to 2023, the study found that 59.4 percent have a patent citing at least one NIH-supported research publication. And, 51.4 percent cite at least one NIH-funded study from the at-risk category of projects. 

“The indirect connection is where we see the real breadth of NIH's impact,” Li says. “What the NIH does is fund research that forms the scientific foundation upon which companies and other drug developers build.”

As the researchers emphasize in the paper, there are many nuances involved in the study. A single citation of an NIH-funded study could appear in a patent for a variety of reasons, and does not necessarily mean “that the drug in question could never have been developed in its absence,” as they write in the paper. To reckon with this, the study also analyzes how many patents had at least 25 percent of their citations fall in the category of at-risk NIH-backed research. By this metric, they found that 65 of the 557 FDA-approved drugs, or 11.7 percent, met the threshold.

On the other hand, as the researchers state in the paper, it is possible the study “understates the extent to which medical advances are connected to NIH research.” For one thing, as the study’s endpoint for examining NIH data is 2007, there could have been more recent pieces of research informing medications that have already received FDA approval. The study does not quantify “second-order connections,” in which NIH-supported findings may have led to additional research that directly led to drug development. Again, NIH funding also supports a broad range of studies beyond the type examined in the current paper.

It is also likely, the scholars suggest, that NIH cuts would curtail the careers of many promising scientists, and in so doing slowdown medical progress. For a variety of these reasons, in addition to the core data itself, the scholars say the study indicates how broadly NIH-backed research has helped advance medicine.

“The worry is that these kinds of deep cuts to the NIH risk that foundation and therefore endanger the development of medicines that might be used to treat us, or our kids and grandkids, 20 years from now,” Li says.

Azoulay and Sampat have received past NIH funding. They also serve on an NIH working group about the empirical analysis of the scientific enterprise.

Machine-learning models can speed up the discovery of new materials by making predictions and suggesting experiments. But most models today only consider a few specific types of data or variables. Compare that with human scientists, who work in a collaborative environment and consider experimental results, the broader scientific literature, imaging and structural analysis, personal experience or intuition, and input from colleagues and peer reviewers.

Now, MIT researchers have developed a method for optimizing materials recipes and planning experiments that incorporates information from diverse sources like insights from the literature, chemical compositions, microstructural images, and more. The approach is part of a new platform, named Copilot for Real-world Experimental Scientists (CRESt), that also uses robotic equipment for high-throughput materials testing, the results of which are fed back into large multimodal models to further optimize materials recipes.

Human researchers can converse with the system in natural language, with no coding required, and the system makes its own observations and hypotheses along the way. Cameras and visual language models also allow the system to monitor experiments, detect issues, and suggest corrections.

“In the field of AI for science, the key is designing new experiments,” says Ju Li, School of Engineering Carl Richard Soderberg Professor of Power Engineering. “We use multimodal feedback — for example information from previous literature on how palladium behaved in fuel cells at this temperature, and human feedback — to complement experimental data and design new experiments. We also use robots to synthesize and characterize the material’s structure and to test performance.”

The system is described in a paper published in Nature. The researchers used CRESt to explore more than 900 chemistries and conduct 3,500 electrochemical tests, leading to the discovery of a catalyst material that delivered record power density in a fuel cell that runs on formate salt to produce electricity.

Joining Li on the paper as first authors are PhD student Zhen Zhang, Zhichu Ren PhD ’24, PhD student Chia-Wei Hsu, and postdoc Weibin Chen. Their coauthors are MIT Assistant Professor Iwnetim Abate; Associate Professor Pulkit Agrawal; JR East Professor of Engineering Yang Shao-Horn; MIT.nano researcher Aubrey Penn; Zhang-Wei Hong PhD ’25, Hongbin Xu PhD ’25; Daniel Zheng PhD ’25; MIT graduate students Shuhan Miao and Hugh Smith; MIT postdocs Yimeng Huang, Weiyin Chen, Yungsheng Tian, Yifan Gao, and Yaoshen Niu; former MIT postdoc Sipei Li; and collaborators including Chi-Feng Lee, Yu-Cheng Shao, Hsiao-Tsu Wang, and Ying-Rui Lu.

A smarter system

Materials science experiments can be time-consuming and expensive. They require researchers to carefully design workflows, make new material, and run a series of tests and analysis to understand what happened. Those results are then used to decide how to improve the material.

To improve the process, some researchers have turned to a machine-learning strategy known as active learning to make efficient use of previous experimental data points and explore or exploit those data. When paired with a statistical technique known as Bayesian optimization (BO), active learning has helped researchers identify new materials for things like batteries and advanced semiconductors.

“Bayesian optimization is like Netflix recommending the next movie to watch based on your viewing history, except instead it recommends the next experiment to do,” Li explains. “But basic Bayesian optimization is too simplistic. It uses a boxed-in design space, so if I say I’m going to use platinum, palladium, and iron, it only changes the ratio of those elements in this small space. But real materials have a lot more dependencies, and BO often gets lost.”

Most active learning approaches also rely on single data streams that don’t capture everything that goes on in an experiment. To equip computational systems with more human-like knowledge, while still taking advantage of the speed and control of automated systems, Li and his collaborators built CRESt.

CRESt’s robotic equipment includes a liquid-handling robot, a carbothermal shock system to rapidly synthesize materials, an automated electrochemical workstation for testing, characterization equipment including automated electron microscopy and optical microscopy, and auxiliary devices such as pumps and gas valves, which can also be remotely controlled.  Many processing parameters can also be tuned.

With the user interface, researchers can chat with CRESt and tell it to use active learning to find promising materials recipes for different projects. CRESt can include up to 20 precursor molecules and substrates into its recipe. To guide material designs, CRESt’s models search through scientific papers for descriptions of elements or precursor molecules that might be useful. When human researchers tell CRESt to pursue new recipes, it kicks off a robotic symphony of sample preparation, characterization, and testing. The researcher can also ask CRESt to perform image analysis from scanning electron microscopy imaging, X-ray diffraction, and other sources.

Information from those processes is used to train the active learning models, which use both literature knowledge and current experimental results to suggest further experiments and accelerate materials discovery.

“For each recipe we use previous literature text or databases, and it creates these huge representations of every recipe based on the previous knowledge base before even doing the experiment,” says Li. “We perform principal component analysis in this knowledge embedding space to get a reduced search space that captures most of the performance variability. Then we use Bayesian optimization in this reduced space to design the new experiment. After the new experiment, we feed newly acquired multimodal experimental data and human feedback into a large language model to augment the knowledgebase and redefine the reduced search space, which gives us a big boost in active learning efficiency.”

Materials science experiments can also face reproducibility challenges. To address the problem, CRESt monitors its experiments with cameras, looking for potential problems and suggesting solutions via text and voice to human researchers.

The researchers used CRESt to develop an electrode material for an advanced type of high-density fuel cell known as a direct formate fuel cell. After exploring more than 900 chemistries over three months, CRESt discovered a catalyst material made from eight elements that achieved a 9.3-fold improvement in power density per dollar over pure palladium, an expensive precious metal. In further tests, CRESTs material was used to deliver a record power density to a working direct formate fuel cell even though the cell contained just one-fourth of the precious metals of previous devices.

The results show the potential for CRESt to find solutions to real-world energy problems that have plagued the materials science and engineering community for decades.

“A significant challenge for fuel-cell catalysts is the use of precious metal,” says Zhang. “For fuel cells, researchers have used various precious metals like palladium and platinum. We used a multielement catalyst that also incorporates many other cheap elements to create the optimal coordination environment for catalytic activity and resistance to poisoning species such as carbon monoxide and adsorbed hydrogen atom. People have been searching low-cost options for many years. This system greatly accelerated our search for these catalysts.”

A helpful assistant

Early on, poor reproducibility emerged as a major problem that limited the researchers’ ability to perform their new active learning technique on experimental datasets. Material properties can be influenced by the way the precursors are mixed and processed, and any number of problems can subtly alter experimental conditions, requiring careful inspection to correct.

To partially automate the process, the researchers coupled computer vision and vision language models with domain knowledge from the scientific literature, which allowed the system to hypothesize sources of irreproducibility and propose solutions. For example, the models can notice when there’s a millimeter-sized deviation in a sample’s shape or when a pipette moves something out of place. The researchers incorporated some of the model’s suggestions, leading to improved consistency, suggesting the models already make good experimental assistants.

The researchers noted that humans still performed most of the debugging in their experiments.

“CREST is an assistant, not a replacement, for human researchers,” Li says. “Human researchers are still indispensable. In fact, we use natural language so the system can explain what it is doing and present observations and hypotheses. But this is a step toward more flexible, self-driving labs.”

Mucus is more than just a sticky substance: It contains a wealth of powerful molecules called mucins that help to tame microbes and prevent infection. In a new study, MIT researchers have identified mucins that defend against Salmonella and other bacteria that cause diarrhea.

The researchers now hope to mimic this defense system to create synthetic mucins that could help prevent or treat illness in soldiers or other people at risk of exposure to Salmonella. It could also help prevent “traveler’s diarrhea,” a gastrointestinal infection caused by consuming contaminated food or water.

Mucins are bottlebrush-shaped polymers made of complex sugar molecules known as glycans, which are tethered to a peptide backbone. In this study, the researchers discovered that a mucin called MUC2 turns off genes that Salmonella uses to enter and infect host cells.

“By using and reformatting this motif from the natural innate immune system, we hope to develop strategies to preventing diarrhea before it even starts. This approach could provide a low-cost solution to a major global health challenge that costs billions annually in lost productivity, health care expenses, and human suffering,” says Katharina Ribbeck, the Andrew and Erna Viterbi Professor of Biological Engineering at MIT and the senior author of the study.

MIT Research Scientist Kelsey Wheeler PhD ’21 and Michaela Gold PhD ’22 are the lead authors of the paper, which appeared Tuesday in the journal Cell Reports.

Blocking infection

Mucus lines much of the body, providing a physical barrier to infection, but that’s not all it does. Over the past decade, Ribbeck has identified mucins that can help to disarm Vibrio cholerae, as well as Pseudomonas aeruginosa, which can infect the lungs and other organs, and the yeast Candida albicans.

In the new study, the researchers wanted to explore how mucins from the digestive tract might interact with Salmonella enterica, a foodborne pathogen that can cause illness after consuming raw or undercooked food, or contaminated water.

To infect host cells, Salmonella must produce proteins that are part of the type 3 secretion system (T3SS), which helps bacteria form needle-like complexes that transfer bacterial proteins directly into host cells. These proteins are all encoded on a segment of DNA called Salmonella pathogenicity island 1 (SPI-1).

The researchers found that when they exposed Salmonella to a mucin called MUC2, which is found in the intestines, the bacteria stopped producing the proteins encoded by SPI-1, and they were no longer able to infect cells.

Further studies revealed that MUC2 achieves this by turning off a regulatory bacterial protein known as HilD. When this protein is blocked by mucins, it can no longer activate the T3SS genes.

Using computational simulations, the researchers showed that certain monosaccharides found in glycans, including GlcNAc and GalNAc, can attach to a specific binding site of the HilD protein. However, their studies showed that these monosaccharides can’t turn off HilD on their own — the shutoff only occurs when the glycans are tethered to the peptide backbone of the mucin.

The researchers also discovered that a similar mucin called MUC5AC, which is found in the stomach, can block HilD. And, both MUC2 and MUC5AC can turn off virulence genes in other foodborne pathogens that also use HilD as a gene regulator.

Mucins as medicine

Ribbeck and her students now plan to explore ways to use synthetic versions of these mucins to help boost the body’s natural defenses and protect the GI tract from Salmonella and other infections.

Studies from other labs have shown that in mice, Salmonella tends to infect portions of the GI tract that have a thin mucus barrier, or no barrier at all.

“Part of Salmonella’s evasion strategy for this host defense is to find locations where mucus is absent and then infect there. So, one could imagine a strategy where we try to bolster mucus barriers to protect those areas with limited mucin,” Wheeler says.

One way to deploy synthetic mucins could be to add them to oral rehydration salts — mixtures of electrolytes that are dissolved in water and used to treat dehydration caused by diarrhea and other gastrointestinal illnesses.

Another potential application for synthetic mucins would be to incorporate them into a chewable tablet that could be consumed before traveling to areas where Salmonella and other diarrheal illnesses are common. This kind of “pre-exposure prophylaxis” could help prevent a great deal of suffering and lost productivity due to illness, the researchers say.

“Mucin mimics would particularly shine as preventatives, because that’s how the body evolved mucus — as part of this innate immune system to prevent infection,” Wheeler says.

The research was funded by the U.S. Army Research Office, the U.S. Army Institute for Collaborative Biotechnologies, the U.S. National Science Foundation, the U.S. National Institute of Health and Environmental Sciences, the U.S. National Institutes of Health, and the German Research Foundation.

Annotating regions of interest in medical images, a process known as segmentation, is often one of the first steps clinical researchers take when running a new study involving biomedical images.

For instance, to determine how the size of the brain’s hippocampus changes as patients age, the scientist first outlines each hippocampus in a series of brain scans. For many structures and image types, this is often a manual process that can be extremely time-consuming, especially if the regions being studied are challenging to delineate.

To streamline the process, MIT researchers developed an artificial intelligence-based system that enables a researcher to rapidly segment new biomedical imaging datasets by clicking, scribbling, and drawing boxes on the images. This new AI model uses these interactions to predict the segmentation.

As the user marks additional images, the number of interactions they need to perform decreases, eventually dropping to zero. The model can then segment each new image accurately without user input.

It can do this because the model’s architecture has been specially designed to use information from images it has already segmented to make new predictions.

Unlike other medical image segmentation models, this system allows the user to segment an entire dataset without repeating their work for each image.

In addition, the interactive tool does not require a presegmented image dataset for training, so users don’t need machine-learning expertise or extensive computational resources. They can use the system for a new segmentation task without retraining the model.

In the long run, this tool could accelerate studies of new treatment methods and reduce the cost of clinical trials and medical research. It could also be used by physicians to improve the efficiency of clinical applications, such as radiation treatment planning.

“Many scientists might only have time to segment a few images per day for their research because manual image segmentation is so time-consuming. Our hope is that this system will enable new science by allowing clinical researchers to conduct studies they were prohibited from doing before because of the lack of an efficient tool,” says Hallee Wong, an electrical engineering and computer science graduate student and lead author of a paper on this new tool.

She is joined on the paper by Jose Javier Gonzalez Ortiz PhD ’24; John Guttag, the Dugald C. Jackson Professor of Computer Science and Electrical Engineering; and senior author Adrian Dalca, an assistant professor at Harvard Medical School and MGH, and a research scientist in the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL). The research will be presented at the International Conference on Computer Vision.

Streamlining segmentation

There are primarily two methods researchers use to segment new sets of medical images. With interactive segmentation, they input an image into an AI system and use an interface to mark areas of interest. The model predicts the segmentation based on those interactions.

A tool previously developed by the MIT researchers, ScribblePrompt, allows users to do this, but they must repeat the process for each new image.

Another approach is to develop a task-specific AI model to automatically segment the images. This approach requires the user to manually segment hundreds of images to create a dataset, and then train a machine-learning model. That model predicts the segmentation for a new image. But the user must start the complex, machine-learning-based process from scratch for each new task, and there is no way to correct the model if it makes a mistake.

This new system, MultiverSeg, combines the best of each approach. It predicts a segmentation for a new image based on user interactions, like scribbles, but also keeps each segmented image in a context set that it refers to later.

When the user uploads a new image and marks areas of interest, the model draws on the examples in its context set to make a more accurate prediction, with less user input.

The researchers designed the model’s architecture to use a context set of any size, so the user doesn’t need to have a certain number of images. This gives MultiverSeg the flexibility to be used in a range of applications.

“At some point, for many tasks, you shouldn’t need to provide any interactions. If you have enough examples in the context set, the model can accurately predict the segmentation on its own,” Wong says.

The researchers carefully engineered and trained the model on a diverse collection of biomedical imaging data to ensure it had the ability to incrementally improve its predictions based on user input.

The user doesn’t need to retrain or customize the model for their data. To use MultiverSeg for a new task, one can upload a new medical image and start marking it.

When the researchers compared MultiverSeg to state-of-the-art tools for in-context and interactive image segmentation, it outperformed each baseline.

Fewer clicks, better results

Unlike these other tools, MultiverSeg requires less user input with each image. By the ninth new image, it needed only two clicks from the user to generate a segmentation more accurate than a model designed specifically for the task.

For some image types, like X-rays, the user might only need to segment one or two images manually before the model becomes accurate enough to make predictions on its own.

The tool’s interactivity also enables the user to make corrections to the model’s prediction, iterating until it reaches the desired level of accuracy. Compared to the researchers’ previous system, MultiverSeg reached 90 percent accuracy with roughly 2/3 the number of scribbles and 3/4 the number of clicks.

“With MultiverSeg, users can always provide more interactions to refine the AI predictions. This still dramatically accelerates the process because it is usually faster to correct something that exists than to start from scratch,” Wong says.

Moving forward, the researchers want to test this tool in real-world situations with clinical collaborators and improve it based on user feedback. They also want to enable MultiverSeg to segment 3D biomedical images.

This work is supported, in part, by Quanta Computer, Inc. and the National Institutes of Health, with hardware support from the Massachusetts Life Sciences Center.

People are increasingly turning to software to design complex material structures like airplane wings and medical implants. But as design models become more capable, our fabrication techniques haven’t kept up. Even 3D printers struggle to reliably produce the precise designs created by algorithms. The problem has led to a disconnect between the ways a material is expected to perform and how it actually works.

Now, MIT researchers have created a way for models to account for 3D printing’s limitations during the design process. In experiments, they showed their approach could be used to make materials that perform much more closely to the way they’re intended to.

“If you don’t account for these limitations, printers can either over- or under-deposit material by quite a lot, so your part becomes heavier or lighter than intended. It can also over- or underestimate the material performance significantly,” says Gilbert W. Winslow Associate Professor of Civil and Environmental Engineering Josephine Carstensen. “With our technique, you know what you’re getting in terms of performance because the numerical model and experimental results align very well.”

The approach is described in the journal Materials and Design, in an open-access paper co-authored by Carstensen and PhD student Hajin Kim-Tackowiak.

Matching theory with reality

Over the last decade, new design and fabrication technologies have transformed the way things are made, especially in industries like aerospace, automotive, and biomedical engineering, where materials must reach precise weight-to-strength ratios and other performance thresholds. In particular, 3D printing allows materials to be made with more complex internal structures.

“3D printing processes generally give us more flexibility because we don’t have to come up with forms or molds for things that would be made through more traditional means like injection molding,” Kim-Tackowiak explains.

As 3D printing has made production more precise, so have methods for designing complex material structures. One of the most advanced computational design techniques is known as topology optimization. Topology optimization has been used to generate new and often surprising material structures that can outperform conventional designs, in some cases approaching the theoretical limits of certain performance thresholds. It is currently being used to design materials with optimized stiffness and strength, maximized energy absorption, fluid permeability, and more.

But topology optimization often creates designs at extremely fine scales that 3D printers have struggled to reliably reproduce. The problem is the size of the print head that extrudes the material. If the design specifies a layer to be 0.5 millimeters thick, for instance, and the print head is only capable of extruding 1-millimeter-thick layers, the final design will be warped and imprecise.

Another problem has to do with the way 3D printers create parts, with a print head extruding a thin bead of material as it glides across the printing area, gradually building parts layer by layer. That can cause weak bonding between layers, making the part more prone to separation or failure.

The researchers sought to address the disconnect between expected and actual properties of materials that arise from those limitations.

“We thought, ‘We know these limitations in the beginning, and the field has gotten better at quantifying these limitations, so we might as well design from the get-go with that in mind,” Kim-Tackowiak says.

In previous work, Carstensen developed an algorithm that embedded information about the print nozzle size into design algorithms for beam structures. For this paper, the researchers built off that approach to incorporate the direction of the print head and the corresponding impact of weak bonding between layers. They also made it work with more complex, porous structures that can have extremely elastic properties.

The approach allows users to add variables to the design algorithms that account for the center of the bead being extruded from a print head and the exact location of the weaker bonding region between layers. The approach also automatically dictates the path the print head should take during production.

The researchers used their technique to create a series of repeating 2D designs with various sizes of hollow pores, or densities. They compared those creations to materials made using traditional topology optimization designs of the same densities.

In tests, the traditionally designed materials deviated from their intended mechanical performance more than materials designed using the researchers’ new technique at material densities under 70 percent. The researchers also found that conventional designs consistently over-deposited material during fabrication. Overall, the researchers’ approach led to parts with more reliable performance at most densities.

“One of the challenges of topology optimization has been that you need a lot of expertise to get good results, so that once you take the designs off the computer, the materials behave the way you thought they would,” Carstensen says. “We’re trying to make it easy to get these high-fidelity products.”

Scaling a new design approach

The researchers believe this is the first time a design technique has accounted for both the print head size and weak bonding between layers.

“When you design something, you should use as much context as possible,” Kim-Tackowiak says. “It was rewarding to see that putting more context into the design process makes your final materials more accurate. It means there are fewer surprises. Especially when we’re putting so much more computational resources into these designs, it’s nice to see we can correlate what comes out of the computer with what comes out of the production process.”

In future work, the researchers hope to improve their method for higher material densities and for different kinds of materials like cement and ceramics. Still, they said their approach offered an improvement over existing techniques, which often require experienced 3D printing specialists to help account for the limitations of the machines and materials.

“It was cool to see that just by putting in the size of your deposition and the bonding property values, you get designs that would have required the consultation of somebody who’s worked in the space for years,” Kim-Tackowiak says.

The researchers say the work paves the way to design with more materials.

“We’d like to see this enable the use of materials that people have disregarded because printing with them has led to issues,” Kim-Tackowiak says. “Now we can leverage those properties or work with those quirks as opposed to just not using all the material options we have at our disposal.”

Transistors, the building blocks of modern electronics, are typically made of silicon. Because it’s a semiconductor, this material can control the flow of electricity in a circuit. But silicon has fundamental physical limits that restrict how compact and energy-efficient a transistor can be.

MIT researchers have now replaced silicon with a magnetic semiconductor, creating a magnetic transistor that could enable smaller, faster, and more energy-efficient circuits. The material’s magnetism strongly influences its electronic behavior, leading to more efficient control of the flow of electricity. 

The team used a novel magnetic material and an optimization process that reduces the material’s defects, which boosts the transistor’s performance.

The material’s unique magnetic properties also allow for transistors with built-in memory, which would simplify circuit design and unlock new applications for high-performance electronics.

“People have known about magnets for thousands of years, but there are very limited ways to incorporate magnetism into electronics. We have shown a new way to efficiently utilize magnetism that opens up a lot of possibilities for future applications and research,” says Chung-Tao Chou, an MIT graduate student in the departments of Electrical Engineering and Computer Science (EECS) and Physics, and co-lead author of a paper on this advance.

Chou is joined on the paper by co-lead author Eugene Park, a graduate student in the Department of Materials Science and Engineering (DMSE); Julian Klein, a DMSE research scientist; Josep Ingla-Aynes, a postdoc in the MIT Plasma Science and Fusion Center; Jagadeesh S. Moodera, a senior research scientist in the Department of Physics; and senior authors Frances Ross, TDK Professor in DMSE; and Luqiao Liu, an associate professor in EECS, and a member of the Research Laboratory of Electronics; as well as others at the University of Chemistry and Technology in Prague. The paper appears today in Physical Review Letters.

Overcoming the limits

In an electronic device, silicon semiconductor transistors act like tiny light switches that turn a circuit on and off, or amplify weak signals in a communication system. They do this using a small input voltage.

But a fundamental physical limit of silicon semiconductors prevents a transistor from operating below a certain voltage, which hinders its energy efficiency.

To make more efficient electronics, researchers have spent decades working toward magnetic transistors that utilize electron spin to control the flow of electricity. Electron spin is a fundamental property that enables electrons to behave like tiny magnets.

So far, scientists have mostly been limited to using certain magnetic materials. These lack the favorable electronic properties of semiconductors, constraining device performance.

“In this work, we combine magnetism and semiconductor physics to realize useful spintronic devices,” Liu says.

The researchers replace the silicon in the surface layer of a transistor with chromium sulfur bromide, a two-dimensional material that acts as a magnetic semiconductor.

Due to the material’s structure, researchers can switch between two magnetic states very cleanly. This makes it ideal for use in a transistor that smoothly switches between “on” and “off.”

“One of the biggest challenges we faced was finding the right material. We tried many other materials that didn’t work,” Chou says.

They discovered that changing these magnetic states modifies the material’s electronic properties, enabling low-energy operation. And unlike many other 2D materials, chromium sulfur bromide remains stable in air.

To make a transistor, the researchers pattern electrodes onto a silicon substrate, then carefully align and transfer the 2D material on top. They use tape to pick up a tiny piece of material, only a few tens of nanometers thick, and place it onto the substrate.

“A lot of researchers will use solvents or glue to do the transfer, but transistors require a very clean surface. We eliminate all those risks by simplifying this step,” Chou says.

Leveraging magnetism

This lack of contamination enables their device to outperform existing magnetic transistors. Most others can only create a weak magnetic effect, changing the flow of current by a few percent or less. Their new transistor can switch or amplify the electric current by a factor of 10.

They use an external magnetic field to change the magnetic state of the material, switching the transistor using significantly less energy than would usually be required.

The material also allows them to control the magnetic states with electric current. This is important because engineers cannot apply magnetic fields to individual transistors in an electronic device. They need to control each one electrically.

The material’s magnetic properties could also enable transistors with built-in memory, simplifying the design of logic or memory circuits.

A typical memory device has a magnetic cell to store information and a transistor to read it out. Their method can combine both into one magnetic transistor.

“Now, not only are transistors turning on and off, they are also remembering information. And because we can switch the transistor with greater magnitude, the signal is much stronger so we can read out the information faster, and in a much more reliable way,” Liu says.

Building on this demonstration, the researchers plan to further study the use of electrical current to control the device. They are also working to make their method scalable so they can fabricate arrays of transistors.

This research was supported, in part, by the Semiconductor Research Corporation, the U.S. Defense Advanced Research Projects Agency (DARPA), the U.S. National Science Foundation (NSF), the U.S. Department of Energy, the U.S. Army Research Office, and the Czech Ministry of Education, Youth, and Sports. The work was partially carried out at the MIT.nano facilities.

At the center of nuclear reactors across the United States, a new type of chromium-coated fuel is being used to make the reactors more efficient and more resistant to accidents. The fuel is one of many innovations sprung from collaboration between researchers at MIT and the Idaho National Laboratory (INL) — a relationship that has altered the trajectory of the country’s nuclear industry.

Amid renewed excitement around nuclear energy in America, MIT’s research community is working to further develop next-generation fuels, accelerate the deployment of small modular reactors (SMRs), and enable the first nuclear reactor in space.

Researchers at MIT and INL have worked closely for decades, and the collaboration takes many forms, including joint research efforts, student and postdoc internships, and a standing agreement that lets INL employees spend extended periods on MIT’s campus researching and teaching classes. MIT is also a founding member of the Battelle Energy Alliance, which has managed the Idaho National Laboratory for the Department of Energy since 2005.

The collaboration gives MIT’s community a chance to work on the biggest problems facing America’s nuclear industry while bolstering INL’s research infrastructure.

“The Idaho National Laboratory is the lead lab for nuclear energy technology in the United States today — that’s why it’s essential that MIT works hand in hand with INL,” says Jacopo Buongiorno, the Battelle Energy Alliance Professor in Nuclear Science and Engineering at MIT. “Countless MIT students and postdocs have interned at INL over the years, and a memorandum of understanding that strengthened the collaboration between MIT and INL in 2019 has been extended twice.”

Ian Waitz, MIT’s vice president for research, adds, “The strong collaborative history between MIT and the Idaho National Laboratory enables us to jointly contribute practical technologies to enable the growth of clean, safe nuclear energy. It’s a clear example of how rigorous collaboration across sectors, and among the nation’s top research facilities, can advance U.S. economic prosperity, health, and well-being.”

Research with impact

Much of MIT’s joint research with INL involves tests and simulations of new nuclear materials, fuels, and instrumentation. One of the largest collaborations was part of a global push for more accident-tolerant fuels in the wake of the nuclear accident that followed the 2011 earthquake and tsunami in Fukushima, Japan.

In a series of studies involving INL and members of the nuclear energy industry, MIT researchers helped identify and evaluate alloy materials that could be deployed in the near term to not only bolster safety but also offer higher densities of fuel.

“These new alloys can withstand much more challenging conditions during abnormal occurrences without reacting chemically with steam, which could result in hydrogen explosions during accidents,” explains Buongiorno, who is also the director of science and technology at MIT’s Nuclear Reactor Laboratory and the director of MIT’s Center for Advanced Nuclear Energy Systems. “The fuels can take much more abuse without breaking apart in the reactor, resulting in a higher safety margin.”

The fuels tested at MIT were eventually adopted by power plants across the U.S., starting with the Byron Clean Energy Center in Ogle County, Illinois.

“We’re also developing new materials, fuels, and instrumentation,” Buongiorno says. “People don’t just come to MIT and say, ‘I have this idea, evaluate it for me.’ We collaborate with industry and national labs to develop the new ideas together, and then we put them to the test,  reproducing the environment in which these materials and fuels would operate in commercial power reactors. That capability is quite unique.”

Another major collaboration was led by Koroush Shirvan, MIT’s Atlantic Richfield Career Development Professor in Energy Studies. Shirvan’s team analyzed the costs associated with different reactor designs, eventually developing an open-source tool to help industry leaders evaluate the feasibility of different approaches.

“The reason we’re not building a single nuclear reactor in the U.S. right now is cost and financial risk,” Shirvan says. “The projects have gone over budget by a factor of two and their schedule has lengthened by a factor of 1.5, so we’ve been doing a lot of work assessing the risk drivers. There’s also a lot of different types of reactors proposed, so we’ve looked at their cost potential as well and how those costs change if you can mass manufacture them.”

Other INL-supported research of Shirvan’s involves exploring new manufacturing methods for nuclear fuels and testing materials for use in a nuclear reactor on the surface of the moon.

“You want materials that are lightweight for these nuclear reactors because you have to send them to space, but there isn’t much data around how those light materials perform in nuclear environments,” Shirvan says.

People and progress

Every summer, MIT students at every level travel to Idaho to conduct research in INL labs as interns.

“It’s an example of our students getting access to cutting-edge research facilities,” Shirvan says.

There are also several joint research appointments between the institutions. One such appointment is held by Sacit Cetiner, a distinguished scientist at INL who also currently runs the MIT and INL Joint Center for Reactor Instrumentation and Sensor Physics (CRISP) at MIT’s Nuclear Reactor Laboratory.

CRISP focuses its research on key technology areas in the field of instrumentation and controls, which have long stymied the bottom line of nuclear power generation.

“For the current light-water reactor fleet, operations and maintenance expenditures constitute a sizeable fraction of unit electricity generation cost,” says Cetiner. “In order to make advanced reactors economically competitive, it’s much more reasonable to address anticipated operational issues during the design phase. One such critical technology area is remote and autonomous operations. Working directly with INL, which manages the projects for the design and testing of several advanced reactors under a number of federal programs, gives our students, faculty, and researchers opportunities to make a real impact.”

The sharing of experts helps strengthen MIT and the nation’s nuclear workforce overall.

“MIT has a crucial role to play in advancing the country’s nuclear industry, whether that’s testing and developing new technologies or assessing the economic feasibility of new nuclear designs,” Buongiorno says.

MIT Department of Mathematics researchers David Roe ’06 and Andrew Sutherland ’90, PhD ’07 are among the inaugural recipients of the Renaissance Philanthropy and XTX Markets’ AI for Math grants. 

Four additional MIT alumni — Anshula Gandhi ’19, Viktor Kunčak SM ’01, PhD ’07; Gireeja Ranade ’07; and Damiano Testa PhD ’05 — were also honored for separate projects.

The first 29 winning projects will support mathematicians and researchers at universities and organizations working to develop artificial intelligence systems that help advance mathematical discovery and research across several key tasks.

Roe and Sutherland, along with Chris Birkbeck of the University of East Anglia, will use their grant to boost automated theorem proving by building connections between the L-Functions and Modular Forms Database (LMFDB) and the Lean4 mathematics library (mathlib).

“Automated theorem provers are quite technically involved, but their development is under-resourced,” says Sutherland. With AI technologies such as large language models (LLMs), the barrier to entry for these formal tools is dropping rapidly, making formal verification frameworks accessible to working mathematicians. 

Mathlib is a large, community-driven mathematical library for the Lean theorem prover, a formal system that verifies the correctness of every step in a proof. Mathlib currently contains on the order of 10⁵ mathematical results (such as lemmas, propositions, and theorems). The LMFDB, a massive, collaborative online resource that serves as a kind of “encyclopedia” of modern number theory, contains more than 10⁹ concrete statements. Sutherland and Roe are managing editors of the LMFDB.

Roe and Sutherland’s grant will be used for a project that aims to augment both systems, making the LMFDB’s results available within mathlib as assertions that have not yet been formally proved, and providing precise formal definitions of the numerical data stored within the LMFDB. This bridge will benefit both human mathematicians and AI agents, and provide a framework for connecting other mathematical databases to formal theorem-proving systems.

The main obstacles to automating mathematical discovery and proof are the limited amount of formalized math knowledge, the high cost of formalizing complex results, and the gap between what is computationally accessible and what is feasible to formalize.

To address these obstacles, the researchers will use the funding to build tools for accessing the LMFDB from mathlib, making a large database of unformalized mathematical knowledge accessible to a formal proof system. This approach enables proof assistants to identify specific targets for formalization without the need to formalize the entire LMFDB corpus in advance.

“Making a large database of unformalized number-theoretic facts available within mathlib will provide a powerful technique for mathematical discovery, because the set of facts an agent might wish to consider while searching for a theorem or proof is exponentially larger than the set of facts that eventually need to be formalized in actually proving the theorem,” says Roe.

The researchers note that proving new theorems at the frontier of mathematical knowledge often involves steps that rely on a nontrivial computation. For example, Andrew Wiles’ proof of Fermat’s Last Theorem uses what is known as the “3-5 trick” at a crucial point in the proof.

“This trick depends on the fact that the modular curve X_0(15) has only finitely many rational points, and none of those rational points correspond to a semi-stable elliptic curve,” according to Sutherland. “This fact was known well before Wiles’ work, and is easy to verify using computational tools available in modern computer algebra systems, but it is not something one can realistically prove using pencil and paper, nor is it necessarily easy to formalize.”

While formal theorem provers are being connected to computer algebra systems for more efficient verification, tapping into computational outputs in existing mathematical databases offers several other benefits.

Using stored results leverages the thousands of CPU-years of computation time already spent in creating the LMFDB, saving money that would be needed to redo these computations. Having precomputed information available also makes it feasible to search for examples or counterexamples without knowing ahead of time how broad the search can be. In addition, mathematical databases are curated repositories, not simply a random collection of facts. 

“The fact that number theorists emphasized the role of the conductor in databases of elliptic curves has already proved to be crucial to one notable mathematical discovery made using machine learning tools: murmurations,” says Sutherland.

“Our next steps are to build a team, engage with both the LMFDB and mathlib communities, start to formalize the definitions that underpin the elliptic curve, number field, and modular form sections of the LMFDB, and make it possible to run LMFDB searches from within mathlib,” says Roe. “If you are an MIT student interested in getting involved, feel free to reach out!” 

The artificial intelligence models that turn text into images are also useful for generating new materials. Over the last few years, generative materials models from companies like Google, Microsoft, and Meta have drawn on their training data to help researchers design tens of millions of new materials.

But when it comes to designing materials with exotic quantum properties like superconductivity or unique magnetic states, those models struggle. That’s too bad, because humans could use the help. For example, after a decade of research into a class of materials that could revolutionize quantum computing, called quantum spin liquids, only a dozen material candidates have been identified. The bottleneck means there are fewer materials to serve as the basis for technological breakthroughs.

Now, MIT researchers have developed a technique that lets popular generative materials models create promising quantum materials by following specific design rules. The rules, or constraints, steer models to create materials with unique structures that give rise to quantum properties.

“The models from these large companies generate materials optimized for stability,” says Mingda Li, MIT’s Class of 1947 Career Development Professor. “Our perspective is that’s not usually how materials science advances. We don’t need 10 million new materials to change the world. We just need one really good material.”

The approach is described today in a paper published by Nature Materials. The researchers applied their technique to generate millions of candidate materials consisting of geometric lattice structures associated with quantum properties. From that pool, they synthesized two actual materials with exotic magnetic traits.

“People in the quantum community really care about these geometric constraints, like the Kagome lattices that are two overlapping, upside-down triangles. We created materials with Kagome lattices because those materials can mimic the behavior of rare earth elements, so they are of high technical importance.” Li says.

Li is the senior author of the paper. His MIT co-authors include PhD students Ryotaro Okabe, Mouyang Cheng, Abhijatmedhi Chotrattanapituk, and Denisse Cordova Carrizales; postdoc Manasi Mandal; undergraduate researchers Kiran Mak and Bowen Yu; visiting scholar Nguyen Tuan Hung; Xiang Fu ’22, PhD ’24; and professor of electrical engineering and computer science Tommi Jaakkola, who is an affiliate of the Computer Science and Artificial Intelligence Laboratory (CSAIL) and Institute for Data, Systems, and Society. Additional co-authors include Yao Wang of Emory University, Weiwei Xie of Michigan State University, YQ Cheng of Oak Ridge National Laboratory, and Robert Cava of Princeton University.

Steering models toward impact

A material’s properties are determined by its structure, and quantum materials are no different. Certain atomic structures are more likely to give rise to exotic quantum properties than others. For instance, square lattices can serve as a platform for high-temperature superconductors, while other shapes known as Kagome and Lieb lattices can support the creation of materials that could be useful for quantum computing.

To help a popular class of generative models known as a diffusion models produce materials that conform to particular geometric patterns, the researchers created SCIGEN (short for Structural Constraint Integration in GENerative model). SCIGEN is a computer code that ensures diffusion models adhere to user-defined constraints at each iterative generation step. With SCIGEN, users can give any generative AI diffusion model geometric structural rules to follow as it generates materials.

AI diffusion models work by sampling from their training dataset to generate structures that reflect the distribution of structures found in the dataset. SCIGEN blocks generations that don’t align with the structural rules.

To test SCIGEN, the researchers applied it to a popular AI materials generation model known as DiffCSP. They had the SCIGEN-equipped model generate materials with unique geometric patterns known as Archimedean lattices, which are collections of 2D lattice tilings of different polygons. Archimedean lattices can lead to a range of quantum phenomena and have been the focus of much research.

“Archimedean lattices give rise to quantum spin liquids and so-called flat bands, which can mimic the properties of rare earths without rare earth elements, so they are extremely important,” says Cheng, a co-corresponding author of the work. “Other Archimedean lattice materials have large pores that could be used for carbon capture and other applications, so it’s a collection of special materials. In some cases, there are no known materials with that lattice, so I think it will be really interesting to find the first material that fits in that lattice.”

The model generated over 10 million material candidates with Archimedean lattices. One million of those materials survived a screening for stability. Using the supercomputers in Oak Ridge National Laboratory, the researchers then took a smaller sample of 26,000 materials and ran detailed simulations to understand how the materials’ underlying atoms behaved. The researchers found magnetism in 41 percent of those structures.

From that subset, the researchers synthesized two previously undiscovered compounds, TiPdBi and TiPbSb, at Xie and Cava’s labs. Subsequent experiments showed the AI model’s predictions largely aligned with the actual material’s properties.

“We wanted to discover new materials that could have a huge potential impact by incorporating these structures that have been known to give rise to quantum properties,” says Okabe, the paper’s first author. “We already know that these materials with specific geometric patterns are interesting, so it’s natural to start with them.”

Accelerating material breakthroughs

Quantum spin liquids could unlock quantum computing by enabling stable, error-resistant qubits that serve as the basis of quantum operations. But no quantum spin liquid materials have been confirmed. Xie and Cava believe SCIGEN could accelerate the search for these materials.

“There’s a big search for quantum computer materials and topological superconductors, and these are all related to the geometric patterns of materials,” Xie says. “But experimental progress has been very, very slow,” Cava adds. “Many of these quantum spin liquid materials are subject to constraints: They have to be in a triangular lattice or a Kagome lattice. If the materials satisfy those constraints, the quantum researchers get excited; it’s a necessary but not sufficient condition. So, by generating many, many materials like that, it immediately gives experimentalists hundreds or thousands more candidates to play with to accelerate quantum computer materials research.”

“This work presents a new tool, leveraging machine learning, that can predict which materials will have specific elements in a desired geometric pattern,” says Drexel University Professor Steve May, who was not involved in the research. “This should speed up the development of previously unexplored materials for applications in next-generation electronic, magnetic, or optical technologies.”

The researchers stress that experimentation is still critical to assess whether AI-generated materials can be synthesized and how their actual properties compare with model predictions. Future work on SCIGEN could incorporate additional design rules into generative models, including chemical and functional constraints.

“People who want to change the world care about material properties more than the stability and structure of materials,” Okabe says. “With our approach, the ratio of stable materials goes down, but it opens the door to generate a whole bunch of promising materials.”

The work was supported, in part, by the U.S. Department of Energy, the National Energy Research Scientific Computing Center, the National Science Foundation, and Oak Ridge National Laboratory.

When OpenAI introduced ChatGPT to the world in 2022, it brought generative artificial intelligence into the mainstream and started a snowball effect that led to its rapid integration into industry, scientific research, health care, and the everyday lives of people who use the technology.

What comes next for this powerful but imperfect tool?

With that question in mind, hundreds of researchers, business leaders, educators, and students gathered at MIT’s Kresge Auditorium for the inaugural MIT Generative AI Impact Consortium (MGAIC) Symposium on Sept. 17 to share insights and discuss the potential future of generative AI.

“This is a pivotal moment — generative AI is moving fast. It is our job to make sure that, as the technology keeps advancing, our collective wisdom keeps pace,” said MIT Provost Anantha Chandrakasan to kick off this first symposium of the MGAIC, a consortium of industry leaders and MIT researchers launched in February to harness the power of generative AI for the good of society.

Underscoring the critical need for this collaborative effort, MIT President Sally Kornbluth said that the world is counting on faculty, researchers, and business leaders like those in MGAIC to tackle the technological and ethical challenges of generative AI as the technology advances.

“Part of MIT’s responsibility is to keep these advances coming for the world. … How can we manage the magic [of generative AI] so that all of us can confidently rely on it for critical applications in the real world?” Kornbluth said.

To keynote speaker Yann LeCun, chief AI scientist at Meta, the most exciting and significant advances in generative AI will most likely not come from continued improvements or expansions of large language models like Llama, GPT, and Claude. Through training, these enormous generative models learn patterns in huge datasets to produce new outputs.

Instead, LuCun and others are working on the development of “world models” that learn the same way an infant does — by seeing and interacting with the world around them through sensory input.

“A 4-year-old has seen as much data through vision as the largest LLM. … The world model is going to become the key component of future AI systems,” he said.

A robot with this type of world model could learn to complete a new task on its own with no training. LeCun sees world models as the best approach for companies to make robots smart enough to be generally useful in the real world.

But even if future generative AI systems do get smarter and more human-like through the incorporation of world models, LeCun doesn’t worry about robots escaping from human control.

Scientists and engineers will need to design guardrails to keep future AI systems on track, but as a society, we have already been doing this for millennia by designing rules to align human behavior with the common good, he said.

“We are going to have to design these guardrails, but by construction, the system will not be able to escape those guardrails,” LeCun said.

Keynote speaker Tye Brady, chief technologist at Amazon Robotics, also discussed how generative AI could impact the future of robotics.

For instance, Amazon has already incorporated generative AI technology into many of its warehouses to optimize how robots travel and move material to streamline order processing.

He expects many future innovations will focus on the use of generative AI in collaborative robotics by building machines that allow humans to become more efficient.

“GenAI is probably the most impactful technology I have witnessed throughout my whole robotics career,” he said.

Other presenters and panelists discussed the impacts of generative AI in businesses, from largescale enterprises like Coca-Cola and Analog Devices to startups like health care AI company Abridge.

Several MIT faculty members also spoke about their latest research projects, including the use of AI to reduce noise in ecological image data, designing new AI systems that mitigate bias and hallucinations, and enabling LLMs to learn more about the visual world.

After a day spent exploring new generative AI technology and discussing its implications for the future, MGAIC faculty co-lead Vivek Farias, the Patrick J. McGovern Professor at MIT Sloan School of Management, said he hoped attendees left with “a sense of possibility, and urgency to make that possibility real.”

Cancer cells have one relentless goal: to grow and divide. While most stick together within the original tumor, some rogue cells break away to traverse to distant organs. There, they can lie dormant — undetectable and not dividing — for years, like landmines waiting to go off.

This migration of cancer cells, called metastasis, is especially common in breast cancer. For many patients, the disease can return months — or even decades — after initial treatment, this time in an entirely different organ.

Robert Weinberg, the Daniel K. Ludwig Professor for Cancer Research at MIT and a Whitehead Institute for Biomedical Research founding member, has spent decades unraveling the complex biology of metastasis and pursuing research that could improve survival rates among patients with metastatic breast cancer — or prevent metastasis altogether.

In his latest study, Weinberg, postdoc Jingwei Zhang, and colleagues ask a critical question: What causes these dormant cancer cells to erupt into a frenzy of growth and division? The group’s findings, published Sept. 1 in The Proceedings of the National Academy of Sciences (PNAS), point to a unique culprit.

This awakening of dormant cancer cells, they’ve discovered, isn’t a spontaneous process. Instead, the wake-up call comes from the inflamed tissue surrounding the cells. One trigger for this inflammation is bleomycin, a common chemotherapy drug that can scar and thicken lung tissue.

“The inflammation jolts the dormant cancer cells awake,” Weinberg says. “Once awakened, they start multiplying again, seeding new life-threatening tumors in the body.”

Decoding metastasis

There’s a lot that scientists still don’t know about metastasis, but this much is clear: Cancer cells must undergo a long and arduous journey to achieve it. The first step is to break away from their neighbors within the original tumor.

Normally, cells stick to one another using surface proteins that act as molecular “velcro,” but some cancer cells can acquire genetic changes that disrupt the production of these proteins and make them more mobile and invasive, allowing them to detach from the parent tumor. 

Once detached, they can penetrate blood vessels and lymphatic channels, which act as highways to distant organs.

While most cancer cells die at some point during this journey, a few persist. These cells exit the bloodstream and invade different tissues—lungs, liver, bone, and even the brain — to give birth to new, often more-aggressive tumors.

“Almost 90 percent of cancer-related deaths occur not from the original tumor, but when cancer cells spread to other parts of the body,” says Weinberg, who is a member of the Koch Institute for Integrative Cancer Research at MIT and the MIT Stem Cell Initiative. “This is why it’s so important to understand how these ‘sleeping’ cancer cells can wake up and start growing again.”

Setting up shop in new tissue comes with changes in surroundings — the “tumor microenvironment” — to which the cancer cells may not be well-suited. These cells face constant threats, including detection and attack by the immune system. 

To survive, they often enter a protective state of dormancy that puts a pause on growth and division. This dormant state also makes them resistant to conventional cancer treatments, which often target rapidly dividing cells.

To investigate what makes this dormancy reversible months or years down the line, researchers in the Weinberg Lab injected human breast cancer cells into mice. These cancer cells were modified to produce a fluorescent protein, allowing the scientists to track their behavior in the body.

The group then focused on cancer cells that had lodged themselves in the lung tissue. By examining them for specific proteins — Ki67, ITGB4, and p63 — that act as markers of cell activity and state, the researchers were able to confirm that these cells were in a non-dividing, dormant state.

Previous work from the Weinberg Lab had shown that inflammation in organ tissue can provoke dormant breast cancer cells to start growing again. In this study, the team tested bleomycin — a chemotherapy drug known to cause lung inflammation — that can be given to patients after surgery to lower the risk of cancer recurrence.

The researchers found that lung inflammation from bleomycin was sufficient to trigger the growth of large lung cancer colonies in treated mice — and to shift the character of these once-dormant cells to those that are more invasive and mobile.

Zeroing in on the tumor microenvironment, the team identified a type of immune cells, called M2 macrophages, as drivers of this process. These macrophages release molecules called epidermal growth factor receptor (EGFR) ligands, which bind to receptors on the surface of dormant cancer cells. This activates a cascade of signals that provoke dormant cancer cells to start multiplying rapidly. 

But EGFR signaling is only the initial spark that ignites the fire. “We found that once dormant cancer cells are awakened, they retain what we call an ‘awakening memory,’” Zhang says. “They no longer require ongoing inflammatory signals from the microenvironment to stay active [growing and multiplying] — they remember the awakened state.”

While signals related to inflammation are necessary to awaken dormant cancer cells, exactly how much signaling is needed remains unclear. “This aspect of cancer biology is particularly challenging, because multiple signals contribute to the state change in these dormant cells,” Zhang says.

The team has already identified one key player in the awakening process, but understanding the full set of signals and how each contributes is far more complex — a question they are continuing to investigate in their new work. 

Studying these pivotal changes in the lives of cancer cells — such as their transition from dormancy to active growth — will deepen our scientific understanding of metastasis and, as researchers in the Weinberg Lab hope, lead to more effective treatments for patients with metastatic cancers.

This work was supported in part by the MIT Stem Cell Initiative.

The last gasp of a primordial black hole may be the source of the highest-energy “ghost particle” detected to date, a new MIT study proposes.

In a paper appearing today in Physical Review Letters, MIT physicists put forth a strong theoretical case that a recently observed, highly energetic neutrino may have been the product of a primordial black hole exploding outside our solar system.

Neutrinos are sometimes referred to as ghost particles, for their invisible yet pervasive nature: They are the most abundant particle type in the universe, yet they leave barely a trace. Scientists recently identified signs of a neutrino with the highest energy ever recorded, but the source of such an unusually powerful particle has yet to be confirmed.

The MIT researchers propose that the mysterious neutrino may have come from the inevitable explosion of a primordial black hole. Primordial black holes (PBHs) are hypothetical black holes that are microscopic versions of the much more massive black holes that lie at the center of most galaxies. PBHs are theorized to have formed in the first moments following the Big Bang. Some scientists believe that primordial black holes could constitute most or all of the dark matter in the universe today.

Like their more massive counterparts, PBHs should leak energy and shrink over their lifetimes, in a process known as Hawking radiation, which was predicted by the physicist Stephen Hawking. The more a black hole radiates, the hotter it gets and the more high-energy particles it releases. This is a runaway process that should produce an incredibly violent explosion of the most energetic particles just before a black hole evaporates away.

The MIT physicists calculate that, if PBHs make up most of the dark matter in the universe, then a small subpopulation of them would be undergoing their final explosions today throughout the Milky Way galaxy. And, there should be a statistically significant possibility that such an explosion could have occurred relatively close to our solar system. The explosion would have released a burst of high-energy particles, including neutrinos, one of which could have had a good chance of hitting a detector on Earth.

If such a scenario had indeed occurred, the recent detection of the highest-energy neutrino would represent the first observation of Hawking radiation, which has long been assumed, but has never been directly observed from any black hole. What’s more, the event might indicate that primordial black holes exist and that they make up most of dark matter — a mysterious substance that comprises 85 percent of the total matter in the universe, the nature of which remains unknown.

“It turns out there’s this scenario where everything seems to line up, and not only can we show that most of the dark matter [in this scenario] is made of primordial black holes, but we can also produce these high-energy neutrinos from a fluke nearby PBH explosion,” says study lead author Alexandra Klipfel, a graduate student in MIT’s Department of Physics. “It’s something we can now try to look for and confirm with various experiments.”

The study’s other co-author is David Kaiser, professor of physics and the Germeshausen Professor of the History of Science at MIT.

High-energy tension

In February, scientists at the Cubic Kilometer Neutrino Telescope, or KM3NeT, reported the detection of the highest-energy neutrino recorded to date. KM3NeT is a large-scale underwater neutrino detector located at the bottom of the Mediterranean Sea, where the environment is meant to mute the effects of any particles other than neutrinos.

The scientists operating the detector picked up signatures of a passing neutrino with an energy of over 100 peta-electron-volts. One peta-electron volt is equivalent to the energy of 1 quadrillion electron volts.

“This is an incredibly high energy, far beyond anything humans are capable of accelerating particles up to,” Klipfel says. “There’s not much consensus on the origin of such high-energy particles.”

Similarly high-energy neutrinos, though not as high as what KM3NeT observed, have been detected by the IceCube Observatory — a neutrino detector embedded deep in the ice at the South Pole. IceCube has detected about half a dozen such neutrinos, whose unusually high energies have also eluded explanation. Whatever their source, the IceCube observations enable scientists to work out a plausible rate at which neutrinos of those energies typically hit Earth. If this rate were correct, however, it would be extremely unlikely to have seen the ultra-high-energy neutrino that KM3NeT recently detected. The two detectors’ discoveries, then, seemed to be what scientists call “in tension.”

Kaiser and Klipfel, who had been working on a separate project involving primordial black holes, wondered: Could a PBH have produced both the KM3NeT neutrino and the handful of IceCube neutrinos, under conditions in which PBHs comprise most of the dark matter in the galaxy? If they could show a chance existed, it would raise an even more exciting possibility — that both observatories observed not only high-energy neutrinos but also the remnants of Hawking radiation.

“Our best chance”

The first step the scientists took in their theoretical analysis was to calculate how many particles would be emitted by an exploding black hole. All black holes should slowly radiate over time. The larger a black hole, the colder it is, and the lower-energy particles it emits as it slowly evaporates. Thus, any particles that are emitted as Hawking radiation from heavy stellar-mass black holes would be near impossible to detect. By the same token, however, much smaller primordial black holes would be very hot and emit high-energy particles in a process that accelerates the closer the black hole gets to disappearing entirely.

“We don’t have any hope of detecting Hawking radiation from astrophysical black holes,” Klipfel says. “So if we ever want to see it, the smallest primordial black holes are our best chance.”

The researchers calculated the number and energies of particles that a black hole should emit, given its temperature and shrinking mass. In its final nanosecond, they estimate that once a black hole is smaller than an atom, it should emit a final burst of particles, including about 10²⁰ neutrinos, or about a sextillion of the particles, with energies of about 100 peta-electron-volts (around the energy that KM3NeT observed).

They used this result to calculate the number of PBH explosions that would have to occur in a galaxy in order to explain the reported IceCube results. They found that, in our region of the Milky Way galaxy, about 1,000 primordial black holes should be exploding per cubic parsec per year. (A parsec is a unit of distance equal to about 3 light years, which is more than 10 trillion kilometers.)

They then calculated the distance at which one such explosion in the Milky Way could have occurred, such that just a handful of the high-energy neutrinos could have reached Earth and produced the recent KM3NeT event. They find that a PBH would have to explode relatively close to our solar system — at a distance about 2,000 times further than the distance between the Earth and our sun.

The particles emitted from such a nearby explosion would radiate in all directions. However, the team found there is a small, 8 percent chance that an explosion can happen close enough to the solar system, once every 14 years, such that enough ultra-high-energy neutrinos hit the Earth.

“An 8 percent chance is not terribly high, but it’s well within the range for which we should take such chances seriously — all the more so because so far, no other explanation has been found that can account for both the unexplained very-high-energy neutrinos and the even more surprising ultra-high-energy neutrino event,” Kaiser says.

The team’s scenario seems to hold up, at least in theory. To confirm their idea will require many more detections of particles, including neutrinos at “insanely high energies.” Then, scientists can build up better statistics regarding such rare events.

“In that case, we could use all of our combined experience and instrumentation, to try to measure still-hypothetical Hawking radiation,” Kaiser says. “That would provide the first-of-its-kind evidence for one of the pillars of our understanding of black holes — and could account for these otherwise anomalous high-energy neutrino events as well. That’s a very exciting prospect!”

In tandem, other efforts to detect nearby PBHs could further bolster the hypothesis that these unusual objects make up most or all of the dark matter.

This work was supported, in part, by the National Science Foundation, MIT’s Center for Theoretical Physics – A Leinweber Institute, and the U.S. Department of Energy.

George Tynan followed a nonlinear path to fusion.

Following his undergraduate degree in aerospace engineering, Tynann's work in the industry spurred his interest in rocket propulsion technology. Because most methods for propulsion involve the manipulation of hot ionized matter, or plasmas, Tynan focused his attention on plasma physics.

It was then that he realized that plasmas could also drive nuclear fusion. “As a potential energy source, it could really be transformative, and the idea that I could work on something that could have that kind of impact on the future was really attractive to me,” he says.

That same drive, to realize the promise of fusion by researching both plasma physics and fusion engineering, drives Tynan today. It’s work he will be pursuing as the Norman C. Rasmussen Adjunct Professor in the Department of Nuclear Science and Engineering (NSE) and at MIT's Plasma Science and Fusion Center (PSFC).

An early interest in fluid flow

Tynan’s enthusiasm for science and engineering traces back to his childhood. His electrical engineer father found employment in the U.S. space program and moved the family to Cape Canaveral in Florida.

“This was in the ’60s, when we were launching Saturn V to the moon, and I got to watch all the launches from the beach,” Tynan remembers. That experience was formative and Tynan became fascinated with how fluids flow.

“I would stick my hand out the window and pretend it was an airplane wing and tilt it with oncoming wind flow and see how the force would change on my hand,” Tynan laughs. The interest eventually led to an undergraduate degree in aerospace engineering at California State Polytechnic University in Pomona.

The switch to a new career would happen after work in the private sector, when Tynan discovered an interest in the use of plasmas for propulsion systems. He moved to the University of California at Los Angeles for graduate school, and it was here that the realization that plasmas could also anchor fusion moved Tynan into this field.

This was in the ’80s, when climate change was not as much in the public consciousness as it is today. Even so, “I knew there’s not an infinite amount of oil and gas around, and that at some point we would have to have widespread adoption of nuclear-based sources,” Tynan remembers. He was also attracted by the sustained effort it would take to make fusion a reality.

Doctoral work

To create energy from fusion, it’s important to get an accurate measurement of the “energy confinement time,” which is a measure of how long it takes for the hot fuel to cool down when all heat sources are turned off. When Tynan started graduate school, this measure was still an empirical guess. He decided to focus his research on the physics of observable confinement time.

It was during this doctoral research that Tynan was able to study the fundamental differences in the behavior of turbulence in plasma as compared to conventional fluids. Typically, when an ordinary fluid is stirred with increasing vigor, the fluid’s motion eventually becomes chaotic or turbulent. However, plasmas can act in a surprising way: confined plasmas, when heated sufficiently strongly, would spontaneously quench the turbulent transport at the boundary of the plasma

An experiment in Germany had unexpectedly discovered this plasma behavior. While subsequent work on other experimental devices confirmed this surprising finding, all earlier experiments lacked the ability to measure the turbulence in detail.

Brian LaBombard, now a senior research scientist at PSFC, was a postdoc at UCLA at the time. Under LaBombard’s direction, Tynan developed a set of Langmuir probes, which are reasonably simple diagnostics for plasma turbulence studies, to further investigate this unusual phenomenon. It formed the basis for his doctoral dissertation. “I happened to be at the right place at the right time so I could study this turbulence quenching phenomenon in much more detail than anyone else could, up until that time,” Tynan says.

As a PhD student and then postdoc, Tynan studied the phenomenon in depth, shuttling between research facilities in Germany, Princeton University’s Plasma Physics Laboratory, and UCLA.

Fusion at UCSD

After completing his doctorate and postdoctoral work, Tynan worked at a startup for a few years when he learned that the University of California at San Diego was launching a new fusion research group at the engineering school. When they reached out, Tynan joined the faculty and built a research program focused on plasma turbulence and plasma-material interactions in fusion systems. Eventually, he became associate dean of engineering, and later, chair of the Department of Mechanical and Aerospace Engineering, serving in these roles for nearly a decade.

Tynan visited MIT on sabbatical in 2023, when his conversations with NSE faculty members Dennis Whyte, Zach Hartwig, and Michael Short excited him about the challenges the private sector faces in making fusion a reality. He saw opportunities to solve important problems at MIT that complemented his work at UC San Diego.

Tynan is excited to tackle what he calls, “the big physics and engineering challenges of fusion plasmas” at NSE: how to remove the heat and exhaust generated by burning plasma so it doesn’t damage the walls of the fusion device and the plasma does not choke on the helium ash. He also hopes to explore robust engineering solutions for practical fusion energy, with a particular focus on developing better materials for use in fusion devices that will make them longer-lasting, while  minimizing the production of radioactive waste.

“Ten or 15 years ago, I was somewhat pessimistic that I would ever see commercial exploitation of fusion in my lifetime,” Tynan says. But that outlook has changed, as he has seen collaborations between MIT and Commonwealth Fusion Systems (CFS) and other private-sector firms that seek to accelerate the timeline to the deployment of fusion in the real world.

In 2021, for example, MIT’s PSFC and CFS took a significant step toward commercial carbon-free power generation. They designed and built a high-temperature superconducting magnet, the strongest fusion magnet in the world.

The milestone was especially exciting because the promise of realizing the dream of fusion energy now felt closer. And being at MIT “seemed like a really quick way to get deeply connected with what’s going on in the efforts to develop fusion energy,” Tynan says.

In addition, “while on sabbatical at MIT, I saw how quickly research staff and students can capitalize on a suggestion of a new idea, and that intrigued me,” he adds.

Tynan brings his special blend of expertise to the table. In addition to extensive experience in plasma physics, he has spent a lot more time on hardcore engineering issues like materials, as well. “The key is to integrate the whole thing into a workable and viable system,” Tynan says.

To most, the Arctic can feel like an abstract place, difficult to imagine beyond images of ice and polar bears. But researcher David Whelihan of MIT Lincoln Laboratory's Advanced Undersea Systems and Technology Group is no stranger to the Arctic. Through Operation Ice Camp, a U.S. Navy–sponsored biennial mission to assess operational readiness in the Arctic region, he has traveled to this vast and remote wilderness twice over the past few years to test low-cost sensor nodes developed by the group to monitor loss in Arctic sea ice extent and thickness. The research team envisions establishing a network of such sensors across the Arctic that will persistently detect ice-fracturing events and correlate these events with environmental conditions to provide insights into why the sea ice is breaking up. Whelihan shared his perspectives on why the Arctic matters and what operating there is like.

Q: Why do we need to be able to operate in the Arctic?

A: Spanning approximately 5.5 million square miles, the Arctic is huge, and one of its salient features is that the ice covering much of the Arctic Ocean is decreasing in volume with every passing year. Melting ice opens up previously impassable areas, resulting in increasing interest from potential adversaries and allies alike for activities such as military operations, commercial shipping, and natural resource extraction. Through Alaska, the United States has approximately 1,060 miles of Arctic coastline that is becoming much more accessible because of reduced ice cover. So, U.S. operation in the Arctic is a matter of national security.  

Q: What are the technological limitations to Arctic operations?

A: The Arctic is an incredibly harsh environment. The cold kills battery life, so collecting sensor data at high rates over long periods of time is very difficult. The ice is dynamic and can easily swallow or crush sensors. In addition, most deployments involve "boots-on-the-ice," which is expensive and at times dangerous. One of the technological limitations is how to deploy sensors while keeping humans alive.

Q: How does the group's sensor node R&D work seek to support Arctic operations?

A: A lot of the work we put into our sensors pertains to deployability. Our ultimate goal is to free researchers from going onto the ice to deploy sensors. This goal will become increasingly necessary as the shrinking ice pack becomes more dynamic, unstable, and unpredictable. At the last Operation Ice Camp (OIC) in March 2024, we built and rapidly tested deployable and recoverable sensors, as well as novel concepts such as using UAVs (uncrewed aerial vehicles), or drones, as "data mules" that can fly out to and interrogate the sensors to see what they captured. We also built a prototype wearable system that cues automatic download of sensor data over Wi-Fi so that operators don't have to take off their gloves.

Q: The Arctic Circle is the northernmost region on Earth. How do you reach this remote place?

A: We usually fly on commercial airlines from Boston to Seattle to Anchorage to Prudhoe Bay on the North Slope of Alaska. From there, the Navy flies us on small prop planes, like Single and Twin Otters, about 200 miles north and lands us on an ice runway built by the Navy's Arctic Submarine Lab (ASL). The runway is part of a temporary camp that ASL establishes on floating sea ice for their operational readiness exercises conducted during OIC.

Q: Think back to the first time you stepped foot in the Arctic. Can you paint a picture of what you experienced?

A: My first experience was at Prudhoe Bay, coming out of the airport, which is a corrugated metal building with a single gate. Before you open the door to the outside, a sign warns you to be on the lookout for polar bears. Walking out into the sheer desolation and blinding whiteness of everything made me realize I was experiencing something very new.

When I flew out onto the ice and stepped out of the plane, I was amazed that the area could somehow be even more desolate. Bright white snowy ice goes in every direction, broken up by pressure ridges that form when ice sheets collide. The sun is low, and seems to move horizontally only. It is very hard to tell the time. The air temperature is really variable. On our first trip in 2022, it really wasn't (relatively) that cold — only around minus 5 or 10 degrees during the day. On our second trip in 2024, we were hit by minus 30 almost every day, and with winds of 20 to 25 miles per hour. The last night we were on the ice that year, it warmed up a bit to minus 10 to 20, but the winds kicked up and started blowing snow onto the heaters attached to our tents. Those heaters started failing one by one as the blowing snow covered them, blocking airflow. After our heater failed, I asked myself, while warm in my bed, whether I wanted to go outside to the command tent for help or try to make it until dawn in my thick sleeping bag. I picked the first option, but mostly because the heater control was beeping loudly right next to my bunk, so I couldn’t sleep anyway. Shout-out to the ASL staff who ran around fixing heaters all night!

Q: How do you survive in a place generally inhospitable to humans?

A: In partnership with the native population, ASL brings a lot of gear — from insulated, heated tents and communications equipment to large snowblowers to keep the runway clear. A few months before OIC, participants attend training on what conditions you will be exposed to and how to protect yourself through appropriate clothing, and how to use survival gear in case of an emergency.

Q: Do you have plans to return to the Arctic?  

A: We are hoping to go back this winter as part of OIC 2026! We plan to test a through-ice communication device. Communicating through 4 to 12 feet of ice is pretty tricky but could allow us to connect underwater drones and stationary sensors under the ice to the rest of the world. To support the through-ice communication system, we will repurpose our sensor-node boxes deployed during OIC 2024. If this setup works, those same boxes could be used as control centers for all sorts of undersea systems and relay information about the under-ice world back home via satellite.

Q: What lessons learned will you bring to your upcoming trip, and any potential future trips?

A: After the first trip, I had a visceral understanding of how hard operating there is. Prototyping of systems becomes a different game. Prototypes are often fragile, but fragility doesn't go over too well on the ice. So, there is a robustification step, which can take some time.

On this last trip, I realized that you have to really be careful with your energy expenditure and pace yourself. While the average adult may require about 2,000 calories a day, an Arctic explorer may burn several times more than that exerting themselves (we do a lot of walking around camp) and keeping warm. Usually, we live on the same freeze-dried food that you would take on camping trips. Each package only has so many calories, so you find yourself eating multiple of those and supplementing with lots of snacks such as Clif Bars or, my favorite, Babybel cheeses (which I bring myself). You also have to be really careful of dehydration. Your body's reaction to extreme cold is to reduce blood flow to your skin, which generally results in less liquid in your body. We have to drink constantly — water, cocoa, and coffee — to avoid dehydration.

We only have access to the ice every two years with the Navy, so we try to make the most of our time. In the several-day lead-up to our field expedition, my research partner Ben and I were really pushing ourselves to ready our sensor nodes for deployment and probably not eating and drinking as regularly as we should. When we ventured to our sensor deployment site about 5 kilometers outside of camp, I had to learn to slow down so I didn't sweat under my gear, as sweating in the extremely cold conditions can quickly lead to hypothermia. I also learned to pay more attention to exposed places on my face, as I got a bit of frostnip around my goggles.

Operating in the Arctic is a fine balance: you can't spend too much time out there, but you also can't rush.

A genome-editing technique known as prime editing holds potential for treating many diseases by transforming faulty genes into functional ones. However, the process carries a small chance of inserting errors that could be harmful.

MIT researchers have now found a way to dramatically lower the error rate of prime editing, using modified versions of the proteins involved in the process. This advance could make it easier to develop gene therapy treatments for a variety of diseases, the researchers say.

“This paper outlines a new approach to doing gene editing that doesn’t complicate the delivery system and doesn’t add additional steps, but results in a much more precise edit with fewer unwanted mutations,” says Phillip Sharp, an MIT Institute Professor Emeritus, a member of MIT’s Koch Institute for Integrative Cancer Research, and one of the senior authors of the new study.

With their new strategy, the MIT team was able to improve the error rate of prime editors from about one error in seven edits to one in 101 for the most-used editing mode, or from one error in 122 edits to one in 543 for a high-precision mode.

“For any drug, what you want is something that is effective, but with as few side effects as possible,” says Robert Langer, the David H. Koch Institute Professor at MIT, a member of the Koch Institute, and one of the senior authors of the new study. “For any disease where you might do genome editing, I would think this would ultimately be a safer, better way of doing it.”

Koch Institute research scientist Vikash Chauhan is the lead author of the paper, which appears today in Nature.

The potential for error

The earliest forms of gene therapy, first tested in the 1990s, involved delivering new genes carried by viruses. Subsequently, gene-editing techniques that use enzymes such as zinc finger nucleases to correct genes were developed. These nucleases are difficult to engineer, however, so adapting them to target different DNA sequences is a very laborious process.

Many years later, the CRISPR genome-editing system was discovered in bacteria, offering scientists a potentially much easier way to edit the genome. The CRISPR system consists of an enzyme called Cas9 that can cut double-stranded DNA at a particular spot, along with a guide RNA that tells Cas9 where to cut. Researchers have adapted this approach to cut out faulty gene sequences or to insert new ones, following an RNA template.

In 2019, researchers at the Broad Institute of MIT and Harvard reported the development of prime editing: a new system, based on CRISPR, that is more precise and has fewer off-target effects. A recent study reported that prime editors were successfully used to treat a patient with chronic granulomatous disease (CGD), a rare genetic disease that affects white blood cells.

“In principle, this technology could eventually be used to address many hundreds of genetic diseases by correcting small mutations directly in cells and tissues,” Chauhan says.

One of the advantages of prime editing is that it doesn’t require making a double-stranded cut in the target DNA. Instead, it uses a modified version of Cas9 that cuts just one of the complementary strands, opening up a flap where a new sequence can be inserted. A guide RNA delivered along with the prime editor serves as the template for the new sequence.

Once the new sequence has been copied, however, it must compete with the old DNA strand to be incorporated into the genome. If the old strand outcompetes the new one, the extra flap of new DNA hanging off may accidentally get incorporated somewhere else, giving rise to errors.

Many of these errors might be relatively harmless, but it’s possible that some could eventually lead to tumor development or other complications. With the most recent version of prime editors, this error rate ranges from one per seven edits to one per 121 edits for different editing modes.

“The technologies we have now are really a lot better than earlier gene therapy tools, but there’s always a chance for these unintended consequences,” Chauhan says.

Precise editing

To reduce those error rates, the MIT team decided to take advantage of a phenomenon they had observed in a 2023 study. In that paper, they found that while Cas9 usually cuts in the same DNA location every time, some mutated versions of the protein show a relaxation of those constraints. Instead of always cutting the same location, those Cas9 proteins would sometimes make their cut one or two bases further along the DNA sequence.

This relaxation, the researchers discovered, makes the old DNA strands less stable, so they get degraded, making it easier for the new strands to be incorporated without introducing any errors.

In the new study, the researchers were able to identify Cas9 mutations that dropped the error rate to 1/20th its original value. Then, by combining pairs of those mutations, they created a Cas9 editor that lowered the error rate even further, to 1/36th the original amount.

To make the editors even more accurate, the researchers incorporated their new Cas9 proteins into a prime editing system that has an RNA binding protein that stabilizes the ends of the RNA template more efficiently. This final editor, which the researchers call vPE, had an error rate just 1/60th of the original, ranging from one in 101 edits to one in 543 edits for different editing modes. These tests were performed in mouse and human cells.

The MIT team is now working on further improving the efficiency of prime editors, through further modifications of Cas9 and the RNA template. They are also working on ways to deliver the editors to specific tissues of the body, which is a longstanding challenge in gene therapy.

They also hope that other labs will begin using the new prime editing approach in their research studies. Prime editors are commonly used to explore many different questions, including how tissues develop, how populations of cancer cells evolve, and how cells respond to drug treatment.

“Genome editors are used extensively in research labs,” Chauhan says. “So the therapeutic aspect is exciting, but we are really excited to see how people start to integrate our editors into their research workflows.”

The research was funded by the Life Sciences Research Foundation, the National Institute of Biomedical Imaging and Bioengineering, the National Cancer Institute, and the Koch Institute Support (core) Grant from the National Cancer Institute.

Before batteries lose power, fail suddenly, or burst into flames, they tend to produce faint sounds over time that provide a signature of the degradation processes going on within their structure. But until now, nobody had figured out how to interpret exactly what those sounds meant, and how to distinguish between ordinary background noise and significant signs of possible trouble.

Now, a team of researchers at MIT’s Department of Chemical Engineering have done a detailed analysis of the sounds emanating from lithium ion batteries, and has been able to correlate particular sound patterns with specific degradation processes taking place inside the cells. The new findings could provide the basis for relatively simple, totally passive and nondestructive devices that could continuously monitor the health of battery systems, for example in electric vehicles or grid-scale storage facilities, to provide ways of predicting useful operating lifetimes and forecasting failures before they occur.

The findings were reported Sept. 5 in the journal Joule, in a paper by MIT graduate students Yash Samantaray and Alexander Cohen, former MIT research scientist Daniel Cogswell PhD ’10, and Chevron Professor of Chemical Engineering and professor of mathematics Martin Z. Bazant.

“In this study, through some careful scientific work, our team has managed to decode the acoustic emissions,” Bazant says. “We were able to classify them as coming from gas bubbles that are generated by side reactions, or by fractures from the expansion and contraction of the active material, and to find signatures of those signals even in noisy data.”

Samantaray explains that, “I think the core of this work is to look at a way to investigate internal battery mechanisms while they’re still charging and discharging, and to do this nondestructively.” He adds, “Out there in the world now, there are a few methods that exist, but most are very expensive and not really conducive to batteries in their normal format.”

To carry out their analysis, the team coupled electrochemical testing with recording of the acoustic emissions, under real-world charging and discharging conditions, using detailed signal processing to correlate the electrical and acoustic data. By doing so, he says, “we were able to come up with a very cost-effective and efficient method of actually understanding gas generation and fracture of materials.”

Gas generation and fracturing are two primary mechanisms of degradation and failure in batteries, so being able to detect and distinguish those processes, just by monitoring the sounds produced by the batteries, could be a significant tool for those managing battery systems.

Previous approaches have simply monitored the sounds and recorded times when the overall sound level exceeded some threshold. But in this work, by simultaneously monitoring the voltage and current as well as the sound characteristics, Bazant says, “We know that [sound] emissions happen at a certain potential [voltage], and that helps us identify what the process might be that is causing that emission.”

After these tests, they would then take the batteries apart and study them under an electron microscope to detect fracturing of the materials.

In addition, they took a wavelet transform — essentially, a way of encoding the frequency and duration of each signal that is captured, providing distinct signatures that can then be more easily extracted from background noise. “No one had done that before,” Bazant says, “so that was another breakthrough.”

Acoustic emissions are widely used in engineering, he points out, for example to monitor structures such as bridges for signs of incipient failure. “It’s a great way to monitor a system,” he says, “because those emissions are happening whether you’re listening to them or not,” so by listening, you can learn something about internal processes that would otherwise be invisible.

With batteries, he says, “we often have a hard time interpreting the voltage and current information as precisely as we’d like, to know what’s happening inside a cell. And so this offers another window into the cell’s state of health, including its remaining useful life, and safety, too.” In a related paper with Oak Ridge National Laboratory researchers, the team has shown that acoustic emissions can provide an early warning of thermal runaway, a situation that can lead to fires if not caught. The new study suggests that these sounds can be used to detect gas generation prior to combustion, “like seeing the first tiny bubbles in a pot of heated water, long before it boils,” says Bazant.

The next step will be to take this new knowledge of how certain sounds relate to specific conditions, and develop a practical, inexpensive monitoring system based on this understanding. “Now, we know what to look for, and how to correlate that with lifetime and health and safety,” Bazant says.

One possible application of this new understanding, Samantaray says, is “as a lab tool for groups that are trying to develop new materials or test new environments, so they can actually determine gas generation or active material fracturing without having to open up the battery.”

Bazant adds that the system could also be useful for quality control in battery manufacturing. “The most expensive and rate-limiting process in battery production is often the formation cycling,” he says. This is the process where batteries are cycled through charging and discharging to break them in, and part of that process involves chemical reactions that release some gas. The new system would allow detection of these gas formation signatures, he says, “and by sensing them, it may be easier to isolate well-formed cells from poorly formed cells very early, even before the useful life of the battery, when it’s being made,” he says.

The work was supported by the Toyota Research Institute, the Center for Battery Sustainability, the National Science Foundation, and the Department of Defense, and made use of the facilities of MIT.nano.

When researchers are building large language models (LLMs), they aim to maximize performance under a particular computational and financial budget. Since training a model can amount to millions of dollars, developers need to be judicious with cost-impacting decisions about, for instance, the model architecture, optimizers, and training datasets before committing to a model. To anticipate the quality and accuracy of a large model’s predictions, practitioners often turn to scaling laws: using smaller, cheaper models to try to approximate the performance of a much larger target model. The challenge, however, is that there are thousands of ways to create a scaling law.

New work from MIT and MIT-IBM Watson AI Lab researchers addresses this by amassing and releasing a collection of hundreds of models and metrics concerning training and performance to approximate more than a thousand scaling laws. From this, the team developed a meta-analysis and guide for how to select small models and estimate scaling laws for different LLM model families, so that the budget is optimally applied toward generating reliable performance predictions.

“The notion that you might want to try to build mathematical models of the training process is a couple of years old, but I think what was new here is that most of the work that people had been doing before is saying, ‘can we say something post-hoc about what happened when we trained all of these models, so that when we’re trying to figure out how to train a new large-scale model, we can make the best decisions about how to use our compute budget?’” says Jacob Andreas, associate professor in the Department of Electrical Engineering and Computer Science and principal investigator with the MIT-IBM Watson AI Lab.

The research was recently presented at the International Conference on Machine Learning by Andreas, along with MIT-IBM Watson AI Lab researchers Leshem Choshen and Yang Zhang of IBM Research.

Extrapolating performance

No matter how you slice it, developing LLMs is an expensive endeavor: from decision-making regarding the numbers of parameters and tokens, data selection and size, and training techniques to determining output accuracy and tuning to the target applications and tasks. Scaling laws offer a way to forecast model behavior by relating a large model’s loss to the performance of smaller, less-costly models from the same family, avoiding the need to fully train every candidate. Mainly, the differences between the smaller models are the number of parameters and token training size. According to Choshen, elucidating scaling laws not only enable better pre-training decisions, but also democratize the field by enabling researchers without vast resources to understand and build effective scaling laws.

The functional form of scaling laws is relatively simple, incorporating components from the small models that capture the number of parameters and their scaling effect, the number of training tokens and their scaling effect, and the baseline performance for the model family of interest. Together, they help researchers estimate a target large model’s performance loss; the smaller the loss, the better the target model’s outputs are likely to be.

These laws allow research teams to weigh trade-offs efficiently and to test how best to allocate limited resources. They’re particularly useful for evaluating scaling of a certain variable, like the number of tokens, and for A/B testing of different pre-training setups.

In general, scaling laws aren’t new; however, in the field of AI, they emerged as models grew and costs skyrocketed. “It’s like scaling laws just appeared at some point in the field,” says Choshen. “They started getting attention, but no one really tested how good they are and what you need to do to make a good scaling law.” Further, scaling laws were themselves also a black box, in a sense. “Whenever people have created scaling laws in the past, it has always just been one model, or one model family, and one dataset, and one developer,” says Andreas. “There hadn’t really been a lot of systematic meta-analysis, as everybody is individually training their own scaling laws. So, [we wanted to know,] are there high-level trends that you see across those things?”

Building better

To investigate this, Choshen, Andreas, and Zhang created a large dataset. They collected LLMs from 40 model families, including Pythia, OPT, OLMO, LLaMA, Bloom, T5-Pile, ModuleFormer mixture-of-experts, GPT, and other families. These included 485 unique, pre-trained models, and where available, data about their training checkpoints, computational cost (FLOPs), training epochs, and the seed, along with 1.9 million performance metrics of loss and downstream tasks. The models differed in their architectures, weights, and so on. Using these models, the researchers fit over 1,000 scaling laws and compared their accuracy across architectures, model sizes, and training regimes, as well as testing how the number of models, inclusion of intermediate training checkpoints, and partial training impacted the predictive power of scaling laws to target models. They used measurements of absolute relative error (ARE); this is the difference between the scaling law’s prediction and the observed loss of a large, trained model. With this, the team compared the scaling laws, and after analysis, distilled practical recommendations for AI practitioners about what makes effective scaling laws.

Their shared guidelines walk the developer through steps and options to consider and expectations. First, it’s critical to decide on a compute budget and target model accuracy. The team found that 4 percent ARE is about the best achievable accuracy one could expect due to random seed noise, but up to 20 percent ARE is still useful for decision-making. The researchers identified several factors that improve predictions, like including intermediate training checkpoints, rather than relying only on final losses; this made scaling laws more reliable. However, very early training data before 10 billion tokens are noisy, reduce accuracy, and should be discarded. They recommend prioritizing training more models across a spread of sizes to improve robustness of the scaling law’s prediction, not just larger models; selecting five models provides a solid starting point. 

Generally, including larger models improves prediction, but costs can be saved by partially training the target model to about 30 percent of its dataset and using that for extrapolation. If the budget is considerably constrained, developers should consider training one smaller model within the target model family and borrow scaling law parameters from a model family with similar architecture; however, this may not work for encoder–decoder models. Lastly, the MIT-IBM research group found that when scaling laws were compared across model families, there was strong correlation between two sets of hyperparameters, meaning that three of the five hyperparameters explained nearly all of the variation and could likely capture the model behavior. Together, these guidelines provide a systematic approach to making scaling law estimation more efficient, reliable, and accessible for AI researchers working under varying budget constraints.

Several surprises arose during this work: small models partially trained are still very predictive, and further, the intermediate training stages from a fully trained model can be used (as if they are individual models) for prediction of another target model. “Basically, you don’t pay anything in the training, because you already trained the full model, so the half-trained model, for instance, is just a byproduct of what you did,” says Choshen. Another feature Andreas pointed out was that, when aggregated, the variability across model families and different experiments jumped out and was noisier than expected. Unexpectedly, the researchers found that it’s possible to utilize the scaling laws on large models to predict performance down to smaller models. Other research in the field has hypothesized that smaller models were a “different beast” compared to large ones; however, Choshen disagrees. “If they’re totally different, they should have shown totally different behavior, and they don’t.”

While this work focused on model training time, the researchers plan to extend their analysis to model inference. Andreas says it’s not, “how does my model get better as I add more training data or more parameters, but instead as I let it think for longer, draw more samples. I think there are definitely lessons to be learned here about how to also build predictive models of how much thinking you need to do at run time.” He says the theory of inference time scaling laws might become even more critical because, “it’s not like I'm going to train one model and then be done. [Rather,] it’s every time a user comes to me, they’re going to have a new query, and I need to figure out how hard [my model needs] to think to come up with the best answer. So, being able to build those kinds of predictive models, like we’re doing in this paper, is even more important.”

This research was supported, in part, by the MIT-IBM Watson AI Lab and a Sloan Research Fellowship. 

The ground-shaking that an earthquake generates is only a fraction of the total energy that a quake releases. A quake can also generate a flash of heat, along with a domino-like fracturing of underground rocks. But exactly how much energy goes into each of these three processes is exceedingly difficult, if not impossible, to measure in the field.

Now MIT geologists have traced the energy that is released by “lab quakes” — miniature analogs of natural earthquakes that are carefully triggered in a controlled laboratory setting. For the first time, they have quantified the complete energy budget of such quakes, in terms of the fraction of energy that goes into heat, shaking, and fracturing.

They found that only about 10 percent of a lab quake’s energy causes physical shaking. An even smaller fraction — less than 1 percent — goes into breaking up rock and creating new surfaces. The overwhelming portion of a quake’s energy — on average 80 percent — goes into heating up the immediate region around a quake’s epicenter. In fact, the researchers observed that a lab quake can produce a temperature spike hot enough to melt surrounding material and turn it briefly into liquid melt.

The geologists also found that a quake’s energy budget depends on a region’s deformation history — the degree to which rocks have been shifted and disturbed by previous tectonic motions. The fractions of quake energy that produce heat, shaking, and rock fracturing can shift depending on what the region has experienced in the past.

“The deformation history — essentially what the rock remembers — really influences how destructive an earthquake could be,” says Daniel Ortega-Arroyo, a graduate student in MIT’s Department of Earth, Atmospheric and Planetary Sciences (EAPS). “That history affects a lot of the material properties in the rock, and it dictates to some degree how it is going to slip.”

The team’s lab quakes are a simplified analog of what occurs during a natural earthquake. Down the road, their results could help seismologists predict the likelihood of earthquakes in regions that are prone to seismic events. For instance, if scientists have an idea of how much shaking a quake generated in the past, they might be able to estimate the degree to which the quake’s energy also affected rocks deep underground by melting or breaking them apart. This in turn could reveal how much more or less vulnerable the region is to future quakes.

“We could never reproduce the complexity of the Earth, so we have to isolate the physics of what is happening, in these lab quakes,” says Matěj Peč, associate professor of geophysics at MIT. “We hope to understand these processes and try to extrapolate them to nature.”

Peč (pronounced “Peck”) and Ortega-Arroyo reported their results on Aug. 28 in the journal AGU Advances. Their MIT co-authors are Hoagy O’Ghaffari and Camilla Cattania, along with Zheng Gong and Roger Fu at Harvard University and Markus Ohl and Oliver Plümper at Utrecht University in the Netherlands.

Under the surface

Earthquakes are driven by energy that is stored up in rocks over millions of years. As tectonic plates slowly grind against each other, stress accumulates through the crust. When rocks are pushed past their material strength, they can suddenly slip along a narrow zone, creating a geologic fault. As rocks slip on either side of the fault, they produce seismic waves that ripple outward and upward.

We perceive an earthquake’s energy mainly in the form of ground shaking, which can be measured using seismometers and other ground-based instruments. But the other two major forms of a quake’s energy — heat and underground fracturing — are largely inaccessible with current technologies.

“Unlike the weather, where we can see daily patterns and measure a number of pertinent variables, it’s very hard to do that very deep in the Earth,” Ortega-Arroyo says. “We don’t know what’s happening to the rocks themselves, and the timescales over which earthquakes repeat within a fault zone are on the century-to-millenia timescales, making any sort of actionable forecast challenging.”

To get an idea of how an earthquake’s energy is partitioned, and how that energy budget might affect a region’s seismic risk, he and Peč went into the lab. Over the last seven years, Peč’s group at MIT has developed methods and instrumentation to simulate seismic events, at the microscale, in an effort to understand how earthquakes at the macroscale may play out.

“We are focusing on what’s happening on a really small scale, where we can control many aspects of failure and try to understand it before we can do any scaling to nature,” Ortega-Arroyo says.

Microshakes

For their new study, the team generated miniature lab quakes that simulate a seismic slipping of rocks along a fault zone. They worked with small samples of granite, which are representative of rocks in the seismogenic layer — the geologic region in the continental crust where earthquakes typically originate. They ground up the granite into a fine powder and mixed the crushed granite with a much finer powder of magnetic particles, which they used as a sort of internal temperature gauge. (A particle’s magnetic field strength will change in response to a fluctuation in temperature.)

The researchers placed samples of the powdered granite — each about 10 square millimeters and 1 millimeter thin — between two small pistons and wrapped the ensemble in a gold jacket. They then applied a strong magnetic field to orient the powder’s magnetic particles in the same initial direction and to the same field strength. They reasoned that any change in the particles’ orientation and field strength afterward should be a sign of how much heat that region experienced as a result of any seismic event.

Once samples were prepared, the team placed them one at a time into a custom-built apparatus that the researchers tuned to apply steadily increasing pressure, similar to the pressures that rocks experience in the Earth’s seismogenic layer, about 10 to 20 kilometers below the surface. They used custom-made piezoelectric sensors, developed by co-author O’Ghaffari, which they attached to either end of a sample to measure any shaking that occurred as they increased the stress on the sample.

They observed that at certain stresses, some samples slipped, producing a microscale seismic event similar to an earthquake. By analyzing the magnetic particles in the samples after the fact, they obtained an estimate of how much each sample was temporarily heated — a method developed in collaboration with Roger Fu’s lab at Harvard University. They also estimated the amount of shaking each sample experienced, using measurements from the piezoelectric sensor and numerical models. The researchers also examined each sample under the microscope, at different magnifications, to assess how the size of the granite grains changed — whether and how many grains broke into smaller pieces, for instance.

From all these measurements, the team was able to estimate each lab quake’s energy budget. On average, they found that about 80 percent of a quake’s energy goes into heat, while 10 percent generates shaking, and less than 1 percent goes into rock fracturing, or creating new, smaller particle surfaces. 

“In some instances we saw that, close to the fault, the sample went from room temperature to 1,200 degrees Celsius in a matter of microseconds, and then immediately cooled down once the motion stopped,” Ortega-Arroyo says. “And in one sample, we saw the fault move by about 100 microns, which implies slip velocities essentially about 10 meters per second. It moves very fast, though it doesn’t last very long.”

The researchers suspect that similar processes play out in actual, kilometer-scale quakes.

“Our experiments offer an integrated approach that provides one of the most complete views of the physics of earthquake-like ruptures in rocks to date,” Peč says. “This will provide clues on how to improve our current earthquake models and natural hazard mitigation.”

This research was supported, in part, by the National Science Foundation.

For pregnant women, ultrasounds are an informative (and sometimes necessary) procedure. They typically produce two-dimensional black-and-white scans of fetuses that can reveal key insights, including biological sex, approximate size, and abnormalities like heart issues or cleft lip. If your doctor wants a closer look, they may use magnetic resonance imaging (MRI), which uses magnetic fields to capture images that can be combined to create a 3D view of the fetus.

MRIs aren’t a catch-all, though; the 3D scans are difficult for doctors to interpret well enough to diagnose problems because our visual system is not accustomed to processing 3D volumetric scans (in other words, a wrap-around look that also shows us the inner structures of a subject). Enter machine learning, which could help model a fetus’s development more clearly and accurately from data — although no such algorithm has been able to model their somewhat random movements and various body shapes.

That is, until a new approach called “Fetal SMPL” from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), Boston Children’s Hospital (BCH), and Harvard Medical School presented clinicians with a more detailed picture of fetal health. It was adapted from “SMPL” (Skinned Multi-Person Linear model), a 3D model developed in computer graphics to capture adult body shapes and poses, as a way to represent fetal body shapes and poses accurately. Fetal SMPL was then trained on 20,000 MRI volumes to predict the location and size of a fetus and create sculpture-like 3D representations. Inside each model is a skeleton with 23 articulated joints called a “kinematic tree,” which the system uses to pose and move like the fetuses it saw during training.

The extensive, real-world scans that Fetal SMPL learned from helped it develop pinpoint accuracy. Imagine stepping into a stranger’s footprint while blindfolded, and not only does it fit perfectly, but you correctly guess what shoe they wore — similarly, the tool closely matched the position and size of fetuses in MRI frames it hadn’t seen before. Fetal SMPL was only misaligned by an average of about 3.1 millimeters, a gap smaller than a single grain of rice.

The approach could enable doctors to precisely measure things like the size of a baby’s head or abdomen and compare these metrics with healthy fetuses at the same age. Fetal SMPL has demonstrated its clinical potential in early tests, where it achieved accurate alignment results on a small group of real-world scans.

“It can be challenging to estimate the shape and pose of a fetus because they’re crammed into the tight confines of the uterus,” says lead author, MIT PhD student, and CSAIL researcher Yingcheng Liu SM ’21. “Our approach overcomes this challenge using a system of interconnected bones under the surface of the 3D model, which represent the fetal body and its motions realistically. Then, it relies on a coordinate descent algorithm to make a prediction, essentially alternating between guessing pose and shape from tricky data until it finds a reliable estimate.”

In utero

Fetal SMPL was tested on shape and pose accuracy against the closest baseline the researchers could find: a system that models infant growth called “SMIL.” Since babies out of the womb are larger than fetuses, the team shrank those models by 75 percent to level the playing field.

The system outperformed this baseline on a dataset of fetal MRIs between the gestational ages of 24 and 37 weeks taken at Boston Children’s Hospital. Fetal SMPL was able to recreate real scans more precisely, as its models closely lined up with real MRIs.

The method was efficient at lining up their models to images, only needing three iterations to arrive at a reasonable alignment. In an experiment that counted how many incorrect guesses Fetal SMPL had made before arriving at a final estimate, its accuracy plateaued from the fourth step onward.

The researchers have just begun testing their system in the real world, where it produced similarly accurate models in initial clinical tests. While these results are promising, the team notes that they’ll need to apply their results to larger populations, different gestational ages, and a variety of disease cases to better understand the system’s capabilities.

Only skin deep

Liu also notes that their system only helps analyze what doctors can see on the surface of a fetus, since only bone-like structures lie beneath the skin of the models. To better monitor babies’ internal health, such as liver, lung, and muscle development, the team intends to make their tool volumetric, modeling the fetus’s inner anatomy from scans. Such upgrades would make the models more human-like, but the current version of Fetal SMPL already presents a precise (and unique) upgrade to 3D fetal health analysis.

“This study introduces a method specifically designed for fetal MRI that effectively captures fetal movements, enhancing the assessment of fetal development and health,” says Kiho Im, Harvard Medical School associate professor of pediatrics and staff scientist in the Division of Newborn Medicine at BCH’s Fetal-Neonatal Neuroimaging and Developmental Science Center. Im, who was not involved with the paper, adds that this approach “will not only improve the diagnostic utility of fetal MRI, but also provide insights into the early functional development of the fetal brain in relation to body movements.”

“This work reaches a pioneering milestone by extending parametric surface human body models for the earliest shapes of human life: fetuses,” says Sergi Pujades, an associate professor at University Grenoble Alpes, who wasn’t involved in the research. “It allows us to detangle the shape and motion of a human, which has already proven to be key in understanding how adult body shape relates to metabolic conditions and how infant motion relates to neurodevelopmental disorders. In addition, the fact that the fetal model stems from, and is compatible with, the adult (SMPL) and infant (SMIL) body models, will allow us to study human shape and pose evolution over long periods of time. This is an unprecedented opportunity to further quantify how human shape growth and motion are affected by different conditions.”

Liu wrote the paper with three CSAIL members: Peiqi Wang SM ’22, PhD ’25; MIT PhD student Sebastian Diaz; and senior author Polina Golland, the Sunlin and Priscilla Chou Professor of Electrical Engineering and Computer Science, a principal investigator in MIT CSAIL, and the leader of the Medical Vision Group. BCH assistant professor of pediatrics Esra Abaci Turk, Inria researcher Benjamin Billot, and Harvard Medical School professor of pediatrics and professor of radiology Patricia Ellen Grant are also authors on the paper. This work was supported, in part, by the National Institutes of Health and the MIT CSAIL-Wistron Program.

The researchers will present their work at the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) in September.

Researchers at the Antimicrobial Resistance (AMR) interdisciplinary research group of the Singapore-MIT Alliance for Research and Technology (SMART), MIT’s research enterprise in Singapore, have developed a powerful tool capable of scanning thousands of biological samples to detect transfer ribonucleic acid (tRNA) modifications — tiny chemical changes to RNA molecules that help control how cells grow, adapt to stress, and respond to diseases such as cancer and antibiotic‑resistant infections. This tool opens up new possibilities for science, health care, and industry — from accelerating disease research and enabling more precise diagnostics to guiding the development of more effective medical treatments for diseases such as cancer and antibiotic-resistant infections.

For this study, the SMART AMR team worked in collaboration with researchers at MIT, Nanyang Technological University in Singapore, the University of Florida, the University at Albany in New York, and Lodz University of Technology in Poland.

Addressing current limitations in RNA modification profiling

Cancer and infectious diseases are complicated health conditions in which cells are forced to function abnormally by mutations in their genetic material or by instructions from an invading microorganism. The SMART-led research team is among the world’s leaders in understanding how the epitranscriptome — the over 170 different chemical modifications of all forms of RNA — controls growth of normal cells and how cells respond to stressful changes in the environment, such as loss of nutrients or exposure to toxic chemicals. The researchers are also studying how this system is corrupted in cancer or exploited by viruses, bacteria, and parasites in infectious diseases.

Current molecular methods used to study the expansive epitranscriptome and all of the thousands of different types of modified RNA are often slow, labor-intensive, costly, and involve hazardous chemicals, which limits research capacity and speed.

To solve this problem, the SMART team developed a new tool that enables fast, automated profiling of tRNA modifications — molecular changes that regulate how cells survive, adapt to stress, and respond to disease. This capability allows scientists to map cell regulatory networks, discover novel enzymes, and link molecular patterns to disease mechanisms, paving the way for better drug discovery and development, and more accurate disease diagnostics. 

Unlocking the complexity of RNA modifications

SMART’s open-access research, recently published in Nucleic Acids Research and titled “tRNA modification profiling reveals epitranscriptome regulatory networks in Pseudomonas aeruginosa,” shows that the tool has already enabled the discovery of previously unknown RNA-modifying enzymes and the mapping of complex gene regulatory networks. These networks are crucial for cellular adaptation to stress and disease, providing important insights into how RNA modifications control bacterial survival mechanisms. 

Using robotic liquid handlers, researchers extracted tRNA from more than 5,700 genetically modified strains of Pseudomonas aeruginosa, a bacterium that causes infections such as pneumonia, urinary tract infections, bloodstream infections, and wound infections. Samples were enzymatically digested and analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS), a technique that separates molecules based on their physical properties and identifies them with high precision and sensitivity. 

As part of the study, the process generated over 200,000 data points in a high-resolution approach that revealed new tRNA-modifying enzymes and simplified gene networks controlling how cells respond and adapt to stress. For example, the data revealed that the methylthiotransferase MiaB, one of the enzymes responsible for tRNA modification ms2i6A, was found to be sensitive to the availability of iron and sulfur and to metabolic changes when oxygen is low. Discoveries like this highlight how cells respond to environmental stresses, and could lead to future development of therapies or diagnostics.

SMART’s automated system was specially designed to profile tRNA modifications across thousands of samples rapidly and safely. Unlike traditional methods, this tool integrates robotics to automate sample preparation and analysis, eliminating the need for hazardous chemical handling and reducing costs. This advancement increases safety, throughput, and affordability, enabling routine large-scale use in research and clinical labs.

A faster and automated way to study RNA

As the first system capable of quantitative, system‑wide profiling of tRNA modifications at this scale, the tool provides a unique and comprehensive view of the epitranscriptome — the complete set of RNA chemical modifications within cells. This capability allows researchers to validate hypotheses about RNA modifications, uncover novel biology, and identify promising molecular targets for developing new therapies.

“This pioneering tool marks a transformative advance in decoding the complex language of RNA modifications that regulate cellular responses,” says Professor Peter Dedon, co-lead principal investigator at SMART AMR, professor of biological engineering at MIT, and corresponding author of the paper. “Leveraging AMR’s expertise in mass spectrometry and RNA epitranscriptomics, our research uncovers new methods to detect complex gene networks critical for understanding and treating cancer, as well as antibiotic-resistant infections. By enabling rapid, large-scale analysis, the tool accelerates both fundamental scientific discovery and the development of targeted diagnostics and therapies that will address urgent global health challenges.”

Accelerating research, industry, and health-care applications

This versatile tool has broad applications across scientific research, industry, and health care. It enables large-scale studies of gene regulation, RNA biology, and cellular responses to environmental and therapeutic challenges. The pharmaceutical and biotech industry can harness it for drug discovery and biomarker screening, efficiently evaluating how potential drugs affect RNA modifications and cellular behavior. This aids the development of targeted therapies and personalized medical treatments.

“This is the first tool that can rapidly and quantitatively profile RNA modifications across thousands of samples,” says Jingjing Sun, research scientist at SMART AMR and first author of the paper. “It has not only allowed us to discover new RNA-modifying enzymes and gene networks, but also opens the door to identifying biomarkers and therapeutic targets for diseases such as cancer and antibiotic-resistant infections. For the first time, large-scale epitranscriptomic analysis is practical and accessible.”

Looking ahead: advancing clinical and pharmaceutical applications

Moving forward, SMART AMR plans to expand the tool’s capabilities to analyze RNA modifications in human cells and tissues, moving beyond microbial models to deepen understanding of disease mechanisms in humans. Future efforts will focus on integrating the platform into clinical research to accelerate the discovery of biomarkers and therapeutic targets. The translation of the technology into an epitranscriptome-wide analysis tool that can be used in pharmaceutical and health-care settings will drive the development of more effective and personalized treatments.

The research conducted at SMART is supported by the National Research Foundation Singapore under its Campus for Research Excellence and Technological Enterprise program.

At MIT, a few scribbles on a whiteboard can turn into a potentially transformational cancer treatment.

This scenario came to fruition this week when the U.S. Food and Drug Administration approved a system for treating an aggressive form of bladder cancer. More than a decade ago, the system started as an idea in the lab of MIT Professor Michael Cima at the Koch Institute for Integrative Cancer Research, enabled by funding from the National Institutes of Health and MIT’s Deshpande Center.

The work that started with a few researchers at MIT turned into a startup, TARIS Biomedical LLC, that was co-founded by Cima and David H. Koch Institute Professor Robert Langer, and acquired by Johnson & Johnson in 2019. In developing the core concept of a device for local drug delivery to the bladder — which represents a new paradigm in bladder cancer treatment — the MIT team approached drug delivery like an engineering problem.

“We spoke to urologists and sketched out the problems with past treatments to get to a set of design parameters,” says Cima, a David H. Koch Professor of Engineering and professor of materials science and engineering. “Part of our criteria was it had to fit into urologists’ existing procedures. We wanted urologists to know what to do with the system without even reading the instructions for use. That’s pretty much how it came out.”

To date, the system has been used in patients thousands of times. In one study involving people with high-risk, non-muscle-invasive bladder cancer whose disease had proven resistant to standard care, doctors could find no evidence of cancer in 82.4 percent of patients treated with the system. More than 50 percent of those patients were still cancer-free nine months after treatment.

The results are extremely gratifying for the team of researchers that worked on it at MIT, including Langer and Heejin Lee SM ’04, PhD ’09, who developed the system as part of his PhD thesis. And Cima says far more people deserve credit than just the ones who scribbled on his whiteboard all those years ago.

“Drug products like this take an enormous amount of effort,” says Cima. “There are probably more than 1,000 people that have been involved in developing and commercializing the system: the MIT inventors, the urologists they consulted, the scientists at TARIS, the scientists at Johnson & Johnson — and that’s not including all the patients who participated in clinical trials. I also want to emphasize the importance of the MIT ecosystem, and the importance of giving people the resources to pursue arguably crazy ideas. We need to continue to support those kinds of activities.”

In the mid 2000s, Langer connected Cima with a urologist at Boston Children’s Hospital who was seeking a new treatment for a painful bladder disease known as interstitial cystitis. The standard treatment required frequent drug infusions into a patient’s bladder through a catheter, which provided only temporary relief.

A group of researchers including Cima; Lee; Hong Linh Ho Duc SM ’05, PhD ’09; Grace Kim PhD ’08; and Karen Daniel PhD ’09 began speaking with urologists and people who had run failed clinical trials involving bladder treatments to understand what went wrong. All that information went on Cima’s whiteboard over the course of several weeks. Fortunately, Cima also scribbled “Do not erase!”

“We learned a lot in the process of writing everything down,” Cima says. “We learned what not to build and what to avoid.”

With the problem well-defined, Cima received a grant from MIT’s Deshpande Center for Technological Innovation, which allowed Lee to work on designing a better solution as part of his PhD thesis.

One of the key advances the group made was using a special alloy that gave the device “shape memory” so that it could be straightened out and inserted into the bladder through a catheter. Then it would fold up, preventing it from being expelled during urination.

The new design was able to slowly release drugs over a two-week period — far longer than any other approach — and could then be removed using a thin, flexible tube commonly used in urology, called a cystoscope. The progress was enough for Cima and Langer, who are both serial entrepreneurs, to found TARIS Biomedical and license the technology from MIT. Lee and three other MIT graduates joined the company.

“It was a real pleasure working with Mike Cima, our students, and colleagues on this novel drug delivery system, which is already changing patients’ lives,” Langer says, “It’s a great example of how research at the Koch Institute starts with basic science and engineering and ends up with new treatments for cancer patients.”

The FDA’s approval of the system for the treatment of certain patients with high-risk, non-muscle-invasive bladder cancer now means that patients with this disease may have a better treatment option. Moving forward, Cima hopes the system continues to be explored to treat other diseases.

Whether you’re an artist, advertising specialist, or just looking to spruce up your home, turning everyday objects into dynamic displays is a great way to make them more visually engaging. For example, you could turn a kids’ book into a handheld cartoon of sorts, making the reading experience more immersive and memorable for a child.

But now, thanks to MIT researchers, it’s also possible to make dynamic displays without using electronics, using barrier-grid animations (or scanimations), which use printed materials instead. This visual trick involves sliding a patterned sheet across an image to create the illusion of a moving image. The secret of barrier-grid animations lies in its name: An overlay called a barrier (or grid) often resembling a picket fence moves across, rotates around, or tilts toward an image to reveal frames in an animated sequence. That underlying picture is a combination of each still, sliced and interwoven to present a different snapshot depending on the overlay’s position.

While tools exist to help artists create barrier-grid animations, they’re typically used to create barrier patterns that have straight lines. Building off of previous work in creating images that appear to move, researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed a tool that allows users to explore more unconventional designs. From zigzags to circular patterns, the team’s “FabObscura” software turns unique concepts into printable scanimations, helping users add dynamic animations to things like pictures, toys, and decor.

MIT Department of Electrical Engineering and Computer Science (EECS) PhD student and CSAIL researcher Ticha Sethapakdi SM ’19, a lead author on a paper presenting FabObscura, says that the system is a one-size-fits-all tool for customizing barrier-grid animations. This versatility extends to unconventional, elaborate overlay designs, like pointed, angled lines to animate a picture you might put on your desk, or the swirling, hypnotic appearance of a radial pattern you could spin over an image placed on a coin or a Frisbee.

“Our system can turn a seemingly static, abstract image into an attention-catching animation,” says Sethapakdi. “The tool lowers the barrier to entry to creating these barrier-grid animations, while helping users express a variety of designs that would’ve been very time-consuming to explore by hand.”

Behind these novel scanimations is a key finding: Barrier patterns can be expressed as any continuous mathematical function — not just straight lines. Users can type these equations into a text box within the FabObscura program, and then see how it graphs out the shape and movement of a barrier pattern. If you wanted a traditional horizontal pattern, you’d enter in a constant function, where the output is the same no matter the input, much like drawing a straight line across a graph. For a wavy design, you’d use a sine function, which is smooth and resembles a mountain range when plotted out. The system’s interface includes helpful examples of these equations to guide users toward their preferred pattern.

A simple interface for elaborate ideas

FabObscura works for all known types of barrier-grid animations, supporting a variety of user interactions. The system enables the creation of a display with an appearance that changes depending on your viewpoint. FabObscura also allows you to create displays that you can animate by sliding or rotating a barrier over an image.

To produce these designs, users can upload a folder of frames of an animation (perhaps a few stills of a horse running), or choose from a few preset sequences (like an eye blinking) and specify the angle your barrier will move. After previewing your design, you can fabricate the barrier and picture onto separate transparent sheets (or print the image on paper) using a standard 2D printer, such as an inkjet. Your image can then be placed and secured on flat, handheld items such as picture frames, phones, and books.

You can enter separate equations if you want two sequences on one surface, which the researchers call “nested animations.” Depending on how you move the barrier, you’ll see a different story being told. For example, CSAIL researchers created a car that rotates when you move its sheet vertically, but transforms into a spinning motorcycle when you slide the grid horizontally.

These customizations lead to unique household items, too. The researchers designed an interactive coaster that you can switch from displaying a “coffee” icon to symbols of a martini and a glass of water by pressing your fingers down on the edges of its surface. The team also spruced up a jar of sunflower seeds, producing a flower animation on the lid that blooms when twisted off.

Artists, including graphic designers and printmakers, could also use this tool to make dynamic pieces without needing to connect any wires. The tool saves them crucial time to explore creative, low-power designs, such as a clock with a mouse that runs along as it ticks. FabObscura could produce animated food packaging, or even reconfigurable signage for places like construction sites or stores that notify people when a particular area is closed or a machine isn’t working.

Keep it crisp

FabObscura’s barrier-grid creations do come with certain trade-offs. While nested animations are novel and more dynamic than a single-layer scanimation, their visual quality isn’t as strong. The researchers wrote design guidelines to address these challenges, recommending users upload fewer frames for nested animations to keep the interlaced image simple and stick to high-contrast images for a crisper presentation.

In the future, the researchers intend to expand what users can upload to FabObscura, like being able to drop in a video file that the program can then select the best frames from. This would lead to even more expressive barrier-grid animations.

FabObscura might also step into a new dimension: 3D. While the system is currently optimized for flat, handheld surfaces, CSAIL researchers are considering implementing their work into larger, more complex objects, possibly using 3D printers to fabricate even more elaborate illusions.

Sethapakdi wrote the paper with several CSAIL affiliates: Zhejiang University PhD student and visiting researcher Mingming Li; MIT EECS PhD student Maxine Perroni-Scharf; MIT postdoc Jiaji Li; MIT associate professors Arvind Satyanarayan and Justin Solomon; and senior author and MIT Associate Professor Stefanie Mueller, leader of the Human-Computer Interaction (HCI) Engineering Group at CSAIL. Their work will be presented at the ACM Symposium on User Interface Software and Technology (UIST) this month.

The U.S. Department of Energy’s National Nuclear Security Administration (DOE/NNSA) recently announced that it has selected MIT to establish a new research center dedicated to advancing the predictive simulation of extreme environments, such as those encountered in hypersonic flight and atmospheric re-entry. The center will be part of the fourth phase of NNSA's Predictive Science Academic Alliance Program (PSAAP-IV), which supports frontier research advancing the predictive capabilities of high-performance computing for open science and engineering applications relevant to national security mission spaces.

The Center for the Exascale Simulation of Coupled High-Enthalpy Fluid–Solid Interactions (CHEFSI) — a joint effort of the MIT Center for Computational Science and Engineering, the MIT Schwarzman College of Computing, and the MIT Institute for Soldier Nanotechnologies (ISN) — plans to harness cutting-edge exascale supercomputers and next-generation algorithms to simulate with unprecedented detail how extremely hot, fast-moving gaseous and solid materials interact. The understanding of these extreme environments — characterized by temperatures of more than 1,500 degrees Celsius and speeds as high as Mach 25 — and their effect on vehicles is central to national security, space exploration, and the development of advanced thermal protection systems.

“CHEFSI will capitalize on MIT’s deep strengths in predictive modeling, high-performance computing, and STEM education to help ensure the United States remains at the forefront of scientific and technological innovation,” says Ian A. Waitz, MIT’s vice president for research. “The center’s particular relevance to national security and advanced technologies exemplifies MIT’s commitment to advancing research with broad societal benefit.”

CHEFSI is one of five new Predictive Simulation Centers announced by the NNSA as part of a program expected to provide up to $17.5 million to each center over five years.

CHEFSI’s research aims to couple detailed simulations of high-enthalpy gas flows with models of the chemical, thermal, and mechanical behavior of solid materials, capturing phenomena such as oxidation, nitridation, ablation, and fracture. Advanced computational models — validated by carefully designed experiments — can address the limitations of flight testing by providing critical insights into material performance and failure.

“By integrating high-fidelity physics models with artificial intelligence-based surrogate models, experimental validation, and state-of-the-art exascale computational tools, CHEFSI will help us understand and predict how thermal protection systems perform under some of the harshest conditions encountered in engineering systems,” says Raúl Radovitzky, the Jerome C. Hunsaker Professor of Aeronautics and Astronautics, associate director of the ISN, and director of CHEFSI. “This knowledge will help in the design of resilient systems for applications ranging from reusable spacecraft to hypersonic vehicles.”

Radovitzky will be joined on the center’s leadership team by Youssef Marzouk, the Breene M. Kerr (1951) Professor of Aeronautics and Astronautics, co-director of the MIT Center for Computational Science and Engineering (CCSE), and recently named the associate dean of the MIT Schwarzman College of Computing; and Nicolas Hadjiconstantinou, the Quentin Berg (1937) Professor of Mechanical Engineering and co-director of CCSE, who will serve as associate directors. The center co-principal investigators include MIT faculty members across the departments of Aeronautics and Astronautics, Electrical Engineering and Computer Science, Materials Science and Engineering, Mathematics, and Mechanical Engineering. Franklin Hadley will lead center operations, with administration and finance under the purview of Joshua Freedman. Hadley and Freedman are both members of the ISN headquarters team. 

CHEFSI expects to collaborate extensively with the DoE/NNSA national laboratories — Lawrence Livermore National Laboratory, Los Alamos National Laboratory, and Sandia National Laboratories — and, in doing so, offer graduate students and postdocs immersive research experiences and internships at these facilities.

The following article is adapted from a press release issued by the Laser Interferometer Gravitational-wave Observatory (LIGO) Laboratory. LIGO is funded by the National Science Foundation and operated by Caltech and MIT, which conceived and built the project.

On Sept. 14, 2015, a signal arrived on Earth, carrying information about a pair of remote black holes that had spiraled together and merged. The signal had traveled about 1.3 billion years to reach us at the speed of light — but it was not made of light. It was a different kind of signal: a quivering of space-time called gravitational waves first predicted by Albert Einstein 100 years prior. On that day 10 years ago, the twin detectors of the U.S. National Science Foundation Laser Interferometer Gravitational-wave Observatory (NSF LIGO) made the first-ever direct detection of gravitational waves, whispers in the cosmos that had gone unheard until that moment.

The historic discovery meant that researchers could now sense the universe through three different means. Light waves, such as X-rays, optical, radio, and other wavelengths of light, as well as high-energy particles called cosmic rays and neutrinos, had been captured before, but this was the first time anyone had witnessed a cosmic event through the gravitational warping of space-time. For this achievement, first dreamed up more than 40 years prior, three of the team’s founders won the 2017 Nobel Prize in Physics: MIT’s Rainer Weiss, professor emeritus of physics (who recently passed away at age 92); Caltech’s Barry Barish, the Ronald and Maxine Linde Professor of Physics, Emeritus; and Caltech’s Kip Thorne, the Richard P. Feynman Professor of Theoretical Physics, Emeritus.

Today, LIGO, which consists of detectors in both Hanford, Washington, and Livingston, Louisiana, routinely observes roughly one black hole merger every three days. LIGO now operates in coordination with two international partners, the Virgo gravitational-wave detector in Italy and KAGRA in Japan. Together, the gravitational-wave-hunting network, known as the LVK (LIGO, Virgo, KAGRA), has captured a total of about 300 black hole mergers, some of which are confirmed while others await further analysis. During the network’s current science run, the fourth since the first run in 2015, the LVK has discovered more than 200 candidate black hole mergers, more than double the number caught in the first three runs.

The dramatic rise in the number of LVK discoveries over the past decade is owed to several improvements to their detectors — some of which involve cutting-edge quantum precision engineering. The LVK detectors remain by far the most precise rulers for making measurements ever created by humans. The space-time distortions induced by gravitational waves are incredibly miniscule. For instance, LIGO detects changes in space-time smaller than 1/10,000 the width of a proton. That’s 1/700 trillionth the width of a human hair.

“Rai Weiss proposed the concept of LIGO in 1972, and I thought, ‘This doesn’t have much chance at all of working,’” recalls Thorne, an expert on the theory of black holes. “It took me three years of thinking about it on and off and discussing ideas with Rai and Vladimir Braginsky [a Russian physicist], to be convinced this had a significant possibility of success. The technical difficulty of reducing the unwanted noise that interferes with the desired signal was enormous. We had to invent a whole new technology. NSF was just superb at shepherding this project through technical reviews and hurdles.”

Nergis Mavalvala, the Curtis and Kathleen Marble Professor of Astrophysics at MIT and dean of the MIT School of Science, says that the challenges the team overcame to make the first discovery are still very much at play. “From the exquisite precision of the LIGO detectors to the astrophysical theories of gravitational-wave sources, to the complex data analyses, all these hurdles had to be overcome, and we continue to improve in all of these areas,” Mavalvala says. “As the detectors get better, we hunger for farther, fainter sources. LIGO continues to be a technological marvel.”

The clearest signal yet

LIGO’s improved sensitivity is exemplified in a recent discovery of a black hole merger referred to as GW250114. (The numbers denote the date the gravitational-wave signal arrived at Earth: January 14, 2025.) The event was not that different from LIGO’s first-ever detection (called GW150914) — both involve colliding black holes about 1.3 billion light-years away with masses between 30 to 40 times that of our sun. But thanks to 10 years of technological advances reducing instrumental noise, the GW250114 signal is dramatically clearer.

“We can hear it loud and clear, and that lets us test the fundamental laws of physics,” says LIGO team member Katerina Chatziioannou, Caltech assistant professor of physics and William H. Hurt Scholar, and one of the authors of a new study on GW250114 published in the Physical Review Letters.

By analyzing the frequencies of gravitational waves emitted by the merger, the LVK team provided the best observational evidence captured to date for what is known as the black hole area theorem, an idea put forth by Stephen Hawking in 1971 that says the total surface areas of black holes cannot decrease. When black holes merge, their masses combine, increasing the surface area. But they also lose energy in the form of gravitational waves. Additionally, the merger can cause the combined black hole to increase its spin, which leads to it having a smaller area. The black hole area theorem states that despite these competing factors, the total surface area must grow in size.

Later, Hawking and physicist Jacob Bekenstein concluded that a black hole’s area is proportional to its entropy, or degree of disorder. The findings paved the way for later groundbreaking work in the field of quantum gravity, which attempts to unite two pillars of modern physics: general relativity and quantum physics.

In essence, the LIGO detection allowed the team to “hear” two black holes growing as they merged into one, verifying Hawking’s theorem. (Virgo and KAGRA were offline during this particular observation.) The initial black holes had a total surface area of 240,000 square kilometers (roughly the size of Oregon), while the final area was about 400,000 square kilometers (roughly the size of California) — a clear increase. This is the second test of the black hole area theorem; an initial test was performed in 2021 using data from the first GW150914 signal, but because that data were not as clean, the results had a confidence level of 95 percent compared to 99.999 percent for the new data.

Thorne recalls Hawking phoning him to ask whether LIGO might be able to test his theorem immediately after he learned of the 2015 gravitational-wave detection. Hawking died in 2018 and sadly did not live to see his theory observationally verified. “If Hawking were alive, he would have reveled in seeing the area of the merged black holes increase,” Thorne says.

The trickiest part of this type of analysis had to do with determining the final surface area of the merged black hole. The surface areas of pre-merger black holes can be more readily gleaned as the pair spiral together, roiling space-time and producing gravitational waves. But after the black holes coalesce, the signal is not as clear-cut. During this so-called ringdown phase, the final black hole vibrates like a struck bell.

In the new study, the researchers precisely measured the details of the ringdown phase, which allowed them to calculate the mass and spin of the black hole and, subsequently, determine its surface area. More specifically, they were able, for the first time, to confidently pick out two distinct gravitational-wave modes in the ringdown phase. The modes are like characteristic sounds a bell would make when struck; they have somewhat similar frequencies but die out at different rates, which makes them hard to identify. The improved data for GW250114 meant that the team could extract the modes, demonstrating that the black hole’s ringdown occurred exactly as predicted by math models based on the Teukolsky formalism — devised in 1972 by Saul Teukolsky, now a professor at Caltech and Cornell University.

Another study from the LVK, submitted to Physical Review Letters today, places limits on a predicted third, higher-pitched tone in the GW250114 signal, and performs some of the most stringent tests yet of general relativity’s accuracy in describing merging black holes.

“A decade of improvements allowed us to make this exquisite measurement,” Chatziioannou says. “It took both of our detectors, in Washington and Louisiana, to do this. I don’t know what will happen in 10 more years, but in the first 10 years, we have made tremendous improvements to LIGO’s sensitivity. This not only means we are accelerating the rate at which we discover new black holes, but we are also capturing detailed data that expand the scope of what we know about the fundamental properties of black holes.”

Jenne Driggers, detection lead senior scientist at LIGO Hanford, adds, “It takes a global village to achieve our scientific goals. From our exquisite instruments, to calibrating the data very precisely, vetting and providing assurances about the fidelity of the data quality, searching the data for astrophysical signals, and packaging all that into something that telescopes can read and act upon quickly, there are a lot of specialized tasks that come together to make LIGO the great success that it is.”

Pushing the limits

LIGO and Virgo have also unveiled neutron stars over the past decade. Like black holes, neutron stars form from the explosive deaths of massive stars, but they weigh less and glow with light. Of note, in August 2017, LIGO and Virgo witnessed an epic collision between a pair of neutron stars — a kilonova — that sent gold and other heavy elements flying into space and drew the gaze of dozens of telescopes around the world, which captured light ranging from high-energy gamma rays to low-energy radio waves. The “multi-messenger” astronomy event marked the first time that both light and gravitational waves had been captured in a single cosmic event. Today, the LVK continues to alert the astronomical community to potential neutron star collisions, who then use telescopes to search the skies for signs of kilonovae.

“The LVK has made big strides in recent years to make sure we’re getting high-quality data and alerts out to the public in under a minute, so that astronomers can look for multi-messenger signatures from our gravitational-wave candidates,” Driggers says.

“The global LVK network is essential to gravitational-wave astronomy,” says Gianluca Gemme, Virgo spokesperson and director of research at the National Institute of Nuclear Physics in Italy. “With three or more detectors operating in unison, we can pinpoint cosmic events with greater accuracy, extract richer astrophysical information, and enable rapid alerts for multi-messenger follow-up. Virgo is proud to contribute to this worldwide scientific endeavor.”

Other LVK scientific discoveries include the first detection of collisions between one neutron star and one black hole; asymmetrical mergers, in which one black hole is significantly more massive than its partner black hole; the discovery of the lightest black holes known, challenging the idea that there is a “mass gap” between neutron stars and black holes; and the most massive black hole merger seen yet with a merged mass of 225 solar masses. For reference, the previous record holder for the most massive merger had a combined mass of 140 solar masses.

Even in the decades before LIGO began taking data, scientists were building foundations that made the field of gravitational-wave science possible. Breakthroughs in computer simulations of black hole mergers, for example, allow the team to extract and analyze the feeble gravitational-wave signals generated across the universe.

LIGO’s technological achievements, beginning as far back as the 1980s, include several far-reaching innovations, such as a new way to stabilize lasers using the so-called Pound–Drever–Hall technique. Invented in 1983 and named for contributing physicists Robert Vivian Pound, the late Ronald Drever of Caltech (a founder of LIGO), and John Lewis Hall, this technique is widely used today in other fields, such as the development of atomic clocks and quantum computers. Other innovations include cutting-edge mirror coatings that almost perfectly reflect laser light; “quantum squeezing” tools that enable LIGO to surpass sensitivity limits imposed by quantum physics; and new artificial intelligence methods that could further hush certain types of unwanted noise.

“What we are ultimately doing inside LIGO is protecting quantum information and making sure it doesn’t get destroyed by external factors,” Mavalvala says. “The techniques we are developing are pillars of quantum engineering and have applications across a broad range of devices, such as quantum computers and quantum sensors.”

In the coming years, the scientists and engineers of LVK hope to further fine-tune their machines, expanding their reach deeper and deeper into space. They also plan to use the knowledge they have gained to build another gravitational-wave detector, LIGO India. Having a third LIGO observatory would greatly improve the precision with which the LVK network can localize gravitational-wave sources.

Looking farther into the future, the team is working on a concept for an even larger detector, called Cosmic Explorer, which would have arms 40 kilometers long. (The twin LIGO observatories have 4-kilometer arms.) A European project, called Einstein Telescope, also has plans to build one or two huge underground interferometers with arms more than 10 kilometers long. Observatories on this scale would allow scientists to hear the earliest black hole mergers in the universe.

“Just 10 short years ago, LIGO opened our eyes for the first time to gravitational waves and changed the way humanity sees the cosmos,” says Aamir Ali, a program director in the NSF Division of Physics, which has supported LIGO since its inception. “There’s a whole universe to explore through this completely new lens and these latest discoveries show LIGO is just getting started.”

The LIGO-Virgo-KAGRA Collaboration

LIGO is funded by the U.S. National Science Foundation and operated by Caltech and MIT, which together conceived and built the project. Financial support for the Advanced LIGO project was led by NSF with Germany (Max Planck Society), the United Kingdom (Science and Technology Facilities Council), and Australia (Australian Research Council) making significant commitments and contributions to the project. More than 1,600 scientists from around the world participate in the effort through the LIGO Scientific Collaboration, which includes the GEO Collaboration. Additional partners are listed at my.ligo.org/census.php.

The Virgo Collaboration is currently composed of approximately 1,000 members from 175 institutions in 20 different (mainly European) countries. The European Gravitational Observatory (EGO) hosts the Virgo detector near Pisa, Italy, and is funded by the French National Center for Scientific Research, the National Institute of Nuclear Physics in Italy, the National Institute of Subatomic Physics in the Netherlands, The Research Foundation – Flanders, and the Belgian Fund for Scientific Research. A list of the Virgo Collaboration groups can be found on the project website.

KAGRA is the laser interferometer with 3-kilometer arm length in Kamioka, Gifu, Japan. The host institute is the Institute for Cosmic Ray Research of the University of Tokyo, and the project is co-hosted by the National Astronomical Observatory of Japan and the High Energy Accelerator Research Organization. The KAGRA collaboration is composed of more than 400 members from 128 institutes in 17 countries/regions. KAGRA’s information for general audiences is at the website gwcenter.icrr.u-tokyo.ac.jp/en/. Resources for researchers are accessible at gwwiki.icrr.u-tokyo.ac.jp/JGWwiki/KAGRA. 

A new study from MIT neuroscientists reveals how rare variants of a gene called ABCA7 may contribute to the development of Alzheimer’s in some of the people who carry it.

Dysfunctional versions of the ABCA7 gene, which are found in a very small proportion of the population, contribute strongly to Alzheimer’s risk. In the new study, the researchers discovered that these mutations can disrupt the metabolism of lipids that play an important role in cell membranes.

This disruption makes neurons hyperexcitable and leads them into a stressed state that can damage DNA and other cellular components. These effects, the researchers found, could be reversed by treating neurons with choline, an important building block precursor needed to make cell membranes.

“We found pretty strikingly that when we treated these cells with choline, a lot of the transcriptional defects were reversed. We also found that the hyperexcitability phenotype and elevated amyloid beta peptides that we observed in neurons that lost ABCA7 was reduced after treatment,” says Djuna von Maydell, an MIT graduate student and the lead author of the study.

Li-Huei Tsai, director of MIT’s Picower Institute for Learning and Memory and the Picower Professor in the MIT Department of Brain and Cognitive Sciences, is the senior author of the paper, which appears today in Nature.

Membrane dysfunction

Genomic studies of Alzheimer’s patients have found that people who carry variants of ABCA7 that generate reduced levels of functional ABCA7 protein have about double the odds of developing Alzheimer’s as people who don’t have those variants.

ABCA7 encodes a protein that transports lipids across cell membranes. Lipid metabolism is also the primary target of a more common Alzheimer’s risk factor known as APOE4. In previous work, Tsai’s lab has shown that APOE4, which is found in about half of all Alzheimer’s patients, disrupts brain cells’ ability to metabolize lipids and respond to stress.

To explore how ABCA7 variants might contribute to Alzheimer’s risk, the researchers obtained tissue samples from the Religious Orders Study/Memory and Aging Project (ROSMAP), a longitudinal study that has tracked memory, motor, and other age-related changes in older people since 1994. Of about 1,200 samples in the dataset that had genetic information available, the researchers obtained 12 from people who carried a rare variant of ABCA7.

The researchers performed single-cell RNA sequencing of neurons from these ABCA7 carriers, allowing them to determine which other genes are affected when ABCA7 is missing. They found that the most significantly affected genes fell into three clusters related to lipid metabolism, DNA damage, and oxidative phosphorylation (the metabolic process that cells use to capture energy as ATP).

To investigate how those alterations could affect neuron function, the researchers introduced ABCA7 variants into neurons derived from induced pluripotent stem cells.

These cells showed many of the same gene expression changes as the cells from the patient samples, especially among genes linked to oxidative phosphorylation. Further experiments showed that the “safety valve” that normally lets mitochondria limit excess build-up of electrical charge was less active. This can lead to oxidative stress, a state that occurs when too many cell-damaging free radicals build up in tissues.

Using these engineered cells, the researchers also analyzed the effects of ABCA7 variants on lipid metabolism. Cells with the variants altered metabolism of a molecule called phosphatidylcholine, which could lead to membrane stiffness and may explain why the mitochondrial membranes of the cells were unable to function normally.

A boost in choline

Those findings raised the possibility that intervening in phosphatidylcholine metabolism might reverse some of the cellular effects of ABCA7 loss. To test that idea, the researchers treated neurons with ABCA7 mutations with a molecule called CDP-choline, a precursor of phosphatidylcholine.

As these cells began producing new phosphatidylcholine (both saturated and unsaturated forms), their mitochondrial membrane potentials also returned to normal, and their oxidative stress levels went down.

The researchers then used induced pluripotent stem cells to generate 3D tissue organoids made of neurons with the ABCA7 variant. These organoids developed higher levels of amyloid beta proteins, which form the plaques seen in the brains of Alzheimer’s patients. However, those levels returned to normal when the organoids were treated with CDP-choline. The treatment also reduced neurons’ hyperexcitability.

In a 2021 paper, Tsai’s lab found that CDP-choline treatment could also reverse many of the effects of another Alzheimer’s-linked gene variant, APOE4, in mice. She is now working with researchers at the University of Texas and MD Anderson Cancer Center on a clinical trial exploring how choline supplements affect people who carry the APOE4 gene.

Choline is naturally found in foods such as eggs, meat, fish, and some beans and nuts. Boosting choline intake with supplements may offer a way for many people to reduce their risk of Alzheimer’s disease, Tsai says.

“From APOE4 to ABCA7 loss of function, my lab demonstrates that disruption of lipid homeostasis leads to the development of Alzheimer’s-related pathology, and that restoring lipid homeostasis, such as through choline supplementation, can ameliorate these pathological phenotypes,” she says.

In addition to the rare variants of ABCA7 that the researchers studied in this paper, there is also a more common variant that is found at a frequency of about 18 percent in the population. This variant was thought to be harmless, but the MIT team showed that cells with this variant exhibited many of the same gene alterations in lipid metabolism that they found in cells with the rare ABCA7 variants.

“There’s more work to be done in this direction, but this suggests that ABCA7 dysfunction might play an important role in a much larger part of the population than just people who carry the rare variants,” von Maydell says.

The research was funded, in part, by the Cure Alzheimer’s Fund, the Freedom Together Foundation, the Carol and Gene Ludwig Family Foundation, James D. Cook, and the National Institutes of Health.

When cells are healthy, we don’t expect them to suddenly change cell types. A skin cell on your hand won’t naturally morph into a brain cell, and vice versa. That’s thanks to epigenetic memory, which enables the expression of various genes to “lock in” throughout a cell’s lifetime. Failure of this memory can lead to diseases, such as cancer.

Traditionally, scientists have thought that epigenetic memory locks genes either “on” or “off” — either fully activated or fully repressed, like a permanent Lite-Brite pattern. But MIT engineers have found that the picture has many more shades.

In a new study appearing today in Cell Genomics, the team reports that a cell’s memory is set not by on/off switching but through a more graded, dimmer-like dial of gene expression.

The researchers carried out experiments in which they set the expression of a single gene at different levels in different cells. While conventional wisdom would assume the gene should eventually switch on or off, the researchers found that the gene’s original expression persisted: Cells whose gene expression was set along a spectrum between on and off remained in this in-between state.

The results suggest that epigenetic memory — the process by which cells retain gene expression and “remember” their identity — is not binary but instead analog, which allows for a spectrum of gene expression and associated cell identities.

“Our finding opens the possibility that cells commit to their final identity by locking genes at specific levels of gene expression instead of just on and off,” says study author Domitilla Del Vecchio, professor of mechanical and biological engineering at MIT. “The consequence is that there may be many more cell types in our body than we know and recognize today, that may have important functions and could underlie healthy or diseased states.”

The study’s MIT lead authors are Sebastian Palacios and Simone Bruno, with additional co-authors.

Beyond binary

Every cell shares the same genome, which can be thought of as the starting ingredient for life. As a cell takes shape, it differentiates into one type or another, through the expression of genes in its genome. Some genes are activated, while others are repressed. The combination steers a cell toward one identity versus another.

A process of DNA methylation, by which certain molecules attach to the genes’ DNA, helps lock their expression in place. DNA methylation assists a cell to “remember” its unique pattern of gene expression, which ultimately establishes the cell’s identity.

Del Vecchio’s group at MIT applies mathematics and genetic engineering to understand cellular molecular processes and to engineer cells with new capabilities. In previous work, her group was experimenting with DNA methylation and ways to lock the expression of certain genes in ovarian cells.

“The textbook understanding was that DNA methylation had a role to lock genes in either an on or off state,” Del Vecchio says. “We thought this was the dogma. But then we started seeing results that were not consistent with that.”

While many of the cells in their experiment exhibited an all-or-nothing expression of genes, a significant number of cells appeared to freeze genes in an in-between state — neither entirely on or off.

“We found there was a spectrum of cells that expressed any level between on and off,” Palacios says. “And we thought, how is this possible?”

Shades of blue

In their new study, the team aimed to see whether the in-between gene expression they observed was a fluke or a more established property of cells that until now has gone unnoticed.

“It could be that scientists disregarded cells that don’t have a clear commitment, because they assumed this was a transient state,” Del Vecchio says. “But actually these in-between cell types may be permanent states that could have important functions.”

To test their idea, the researchers ran experiments with hamster ovarian cells — a line of cells commonly used in the laboratory. In each cell, an engineered gene was initially set to a different level of expression. The gene was turned fully on in some cells, completely off in others, and set somewhere in between on and off for the remaining cells.

The team paired the engineered gene with a fluorescent marker that lights up with a brightness corresponding to the gene’s level of expression. The researchers introduced, for a short time, an enzyme that triggers the gene’s DNA methylation, a natural gene-locking mechanism. They then monitored the cells over five months to see whether the modification would lock the genes in place at their in-between expression levels, or whether the genes would migrate toward fully on or off states before locking in.

“Our fluorescent marker is blue, and we see cells glow across the entire spectrum, from really shiny blue, to dimmer and dimmer, to no blue at all,” Del Vecchio says. “Every intensity level is maintained over time, which means gene expression is graded, or analog, and not binary. We were very surprised, because we thought after such a long time, the gene would veer off, to be either fully on or off, but it did not.”

The findings open new avenues into engineering more complex artificial tissues and organs by tuning the expression of certain genes in a cell’s genome, like a dial on a radio, rather than a switch. The results also complicate the picture of how a cell’s epigenetic memory works to establish its identity. It opens up the possibility that cell modifications such as those exhibited in therapy-resistant tumors could be treated in a more precise fashion.

“Del Vecchio and colleagues have beautifully shown how analog memory arises through chemical modifications to the DNA itself,” says Michael Elowitz, professor of biology and biological engineering at the California institute of Technology, who was not involved in the study. “As a result, we can now imagine repurposing this natural analog memory mechanism, invented by evolution, in the field of synthetic biology, where it could help allow us to program permanent and precise multicellular behaviors.”

“One of the things that enables the complexity in humans is epigenetic memory,” Palacios says. “And we find that it is not what we thought. For me, that’s actually mind-blowing. And I think we’re going to find that this analog memory is relevant for many different processes across biology.”

This research was supported, in part, by the National Science Foundation, MODULUS, and a Vannevar Bush Faculty Fellowship through the U.S. Office of Naval Research.

Using tiny particles shaped like bottlebrushes, MIT chemists have found a way to deliver a large range of chemotherapy drugs directly to tumor cells.

To guide them to the right location, each particle contains an antibody that targets a specific tumor protein. This antibody is tethered to bottlebrush-shaped polymer chains carrying dozens or hundreds of drug molecules — a much larger payload than can be delivered by any existing antibody-drug conjugates.

In mouse models of breast and ovarian cancer, the researchers found that treatment with these conjugated particles could eliminate most tumors. In the future, the particles could be modified to target other types of cancer, by swapping in different antibodies.

“We are excited about the potential to open up a new landscape of payloads and payload combinations with this technology, that could ultimately provide more effective therapies for cancer patients,” says Jeremiah Johnson, the A. Thomas Geurtin Professor of Chemistry at MIT, a member of the Koch Institute for Integrative Cancer Research, and the senior author of the new study.

MIT postdoc Bin Liu is the lead author of the paper, which appears today in Nature Biotechnology.

A bigger drug payload

Antibody-drug conjugates (ADCs) are a promising type of cancer treatment that consist of a cancer-targeting antibody attached to a chemotherapy drug. At least 15 ADCs have been approved by the FDA to treat several different types of cancer.

This approach allows specific targeting of a cancer drug to a tumor, which helps to prevent some of the side effects that occur when chemotherapy drugs are given intravenously. However, one drawback to currently approved ADCs is that only a handful of drug molecules can be attached to each antibody. That means they can only be used with very potent drugs — usually DNA-damaging agents or drugs that interfere with cell division.

To try to use a broader range of drugs, which are often less potent, Johnson and his colleagues decided to adapt bottlebrush particles that they had previously invented. These particles consist of a polymer backbone that are attached to tens to hundreds of “prodrug” molecules — inactive drug molecules that are activated upon release within the body. This structure allows the particles to deliver a wide range of drug molecules, and the particles can be designed to carry multiple drugs in specific ratios.

Using a technique called click chemistry, the researchers showed that they could attach one, two, or three of their bottlebrush polymers to a single tumor-targeting antibody, creating an antibody-bottlebrush conjugate (ABC). This means that just one antibody can carry hundreds of prodrug molecules. The currently approved ADCs can carry a maximum of about eight drug molecules.

The huge number of payloads in the ABC particles allows the researchers to incorporate less potent cancer drugs such as doxorubicin or paclitaxel, which enhances the customizability of the particles and the variety of drug combinations that can be used.

“We can use antibody-bottlebrush conjugates to increase the drug loading, and in that case, we can use less potent drugs,” Liu says. “In the future, we can very easily copolymerize with multiple drugs together to achieve combination therapy.”

The prodrug molecules are attached to the polymer backbone by cleavable linkers. After the particles reach a tumor site, some of these linkers are broken right away, allowing the drugs to kill nearby cancer cells even if they don’t express the target antibody. Other particles are absorbed into cells with the target antibody before releasing their toxic payload.

Effective treatment

For this study, the researchers created ABC particles carrying a few different types of drugs: microtubule inhibitors called MMAE and paclitaxel, and two DNA-damaging agents, doxorubicin and SN-38. They also designed ABC particles carrying an experimental type of drug known as PROTAC (proteolysis-targeting chimera), which can selectively degrade disease-causing proteins inside cells.

Each bottlebrush was tethered to an antibody targeting either HER2, a protein often overexpressed in breast cancer, or MUC1, which is commonly found in ovarian, lung, and other types of cancer.

The researchers tested each of the ABCs in mouse models of breast or ovarian cancer and found that in most cases, the ABC particles were able to eradicate the tumors. This treatment was significantly more effective than giving the same bottlebrush prodrugs by injection, without being conjugated to a targeting antibody.

“We used a very low dose, almost 100 times lower compared to the traditional small-molecule drug, and the ABC still can achieve much better efficacy compared to the small-molecule drug given on its own,” Liu says.

These ABCs also performed better than two FDA-approved ADCs, T-DXd and TDM-1, which both use HER2 to target cells. T-DXd carries deruxtecan, which interferes with DNA replication, and TDM-1 carries emtansine, a microtubule inhibitor.

In future work, the MIT team plans to try delivering combinations of drugs that work by different mechanisms, which could enhance their overall effectiveness. Among these could be immunotherapy drugs such as STING activators.

The researchers are also working on swapping in different antibodies, such as antibodies targeting EGFR, which is widely expressed in many tumors. More than 100 antibodies have been approved to treat cancer and other diseases, and in theory any of those could be conjugated to cancer drugs to create a targeted therapy.

The research was funded in part by the National Institutes of Health, the Ludwig Center at MIT, and the Koch Institute Frontier Research Program. 

Most people recognize Alzheimer’s disease from its devastating symptoms such as memory loss, while new drugs target pathological aspects of disease manifestations, such as plaques of amyloid proteins. Now, a sweeping new open-access study in the Sept. 4 edition of Cell by MIT researchers shows the importance of understanding the disease as a battle over how well brain cells control the expression of their genes. The study paints a high-resolution picture of a desperate struggle to maintain healthy gene expression and gene regulation, where the consequences of failure or success are nothing less than the loss or preservation of cell function and cognition.

The study presents a first-of-its-kind, multimodal atlas of combined gene expression and gene regulation spanning 3.5 million cells from six brain regions, obtained by profiling 384 post-mortem brain samples across 111 donors. The researchers profiled both the “transcriptome,” showing which genes are expressed into RNA, and the “epigenome,” the set of chromosomal modifications that establish which DNA regions are accessible and thus utilized between different cell types.

The resulting atlas revealed many insights showing that the progression of Alzheimer’s is characterized by two major epigenomic trends. The first is that vulnerable cells in key brain regions suffer a breakdown of the rigorous nuclear “compartments” they normally maintain to ensure some parts of the genome are open for expression but others remain locked away. The second major finding is that susceptible cells experience a loss of “epigenomic information,” meaning they lose their grip on the unique pattern of gene regulation and expression that gives them their specific identity and enables their healthy function.

Accompanying the evidence of compromised compartmentalization and the erosion of epigenomic information are many specific findings pinpointing molecular circuitry that breaks down by cell type, by region, and gene network. They found, for instance, that when epigenomic conditions deteriorate, that opens the door to expression of many genes associated with disease, whereas if cells manage to keep their epigenomic house in order, they can keep disease-associated genes in check. Moreover, the researchers clearly saw that when the epigenomic breakdowns were occurring people lost cognitive ability, but where epigenomic stability remained, so did cognition.

“To understand the circuitry, the logic responsible for gene expression changes in Alzheimer’s disease [AD], we needed to understand the regulation and upstream control of all the changes that are happening, and that’s where the epigenome comes in,” says senior author Manolis Kellis, a professor in the Computer Science and Artificial Intelligence Lab and head of MIT’s Computational Biology Group. “This is the first large-scale, single-cell, multi-region gene-regulatory atlas of AD, systematically dissecting the dynamics of epigenomic and transcriptomic programs across disease progression and resilience.”

By providing that detailed examination of the epigenomic mechanisms of Alzheimer’s progression, the study provides a blueprint for devising new Alzheimer’s treatments that can target factors underlying the broad erosion of epigenomic control or the specific manifestations that affect key cell types such as neurons and supporting glial cells.

“The key to developing new and more effective treatments for Alzheimer’s disease depends on deepening our understanding of the mechanisms that contribute to the breakdowns of cellular and network function in the brain,” says Picower Professor and co-corresponding author Li-Huei Tsai, director of The Picower Institute for Learning and Memory and a founding member of MIT’s Aging Brain Initiative, along with Kellis. “This new data advances our understanding of how epigenomic factors drive disease.”

Kellis Lab members Zunpeng Liu and Shanshan Zhang are the study’s co-lead authors.

Compromised compartments and eroded information

Among the post-mortem brain samples in the study, 57 came from donors to the Religious Orders Study or the Rush Memory and Aging Project (collectively known as “ROSMAP”) who did not have AD pathology or symptoms, while 33 came from donors with early-stage pathology and 21 came from donors at a late stage. The samples therefore provided rich information about the symptoms and pathology each donor was experiencing before death.

In the new study, Liu and Zhang combined analyses of single-cell RNA sequencing of the samples, which measures which genes are being expressed in each cell, and ATACseq, which measures whether chromosomal regions are accessible for gene expression. Considered together, these transcriptomic and epigenomic measures enabled the researchers to understand the molecular details of how gene expression is regulated across seven broad classes of brain cells (e.g., neurons or other glial cell types) and 67 subtypes of cell types (e.g., 17 kinds of excitatory neurons or six kinds of inhibitory ones).

The researchers annotated more than 1 million gene-regulatory control regions that different cells employ to establish their specific identities and functionality using epigenomic marking. Then, by comparing the cells from Alzheimer’s brains to the ones without, and accounting for stage of pathology and cognitive symptoms, they could produce rigorous associations between the erosion of these epigenomic markings, and ultimately loss of function.

For instance, they saw that among people who advanced to late-stage AD, normally repressive compartments opened up for more expression and compartments that were normally more open during health became more repressed. Worryingly, when the normally repressive compartments of brain cells opened up, they became more afflicted with disease.

“For Alzheimer’s patients, repressive compartments opened up, and gene expression levels increased, which was associated with decreased cognitive function,” explains Liu.

But when cells managed to keep their compartments in order such that they expressed the genes they were supposed to, people remained cognitively intact.

Meanwhile, based on the cells’ expression of their regulatory elements, the researchers created an epigenomic information score for each cell. Generally, information declined as pathology progressed, but that was particularly notable among cells in the two brain regions affected earliest in Alzheimer’s: the entorhinal cortex and the hippocampus. The analyses also highlighted specific cell types that were especially vulnerable including microglia that play immune and other roles, oligodendrocytes that produce myelin insulation for neurons, and particular kinds of excitatory neurons.

Risk genes and “chromatin guardians”

Detailed analyses in the paper highlighted how epigenomic regulation tracked with disease-related problems, Liu notes. The e4 variant of the APOE gene, for instance, is widely understood to be the single biggest genetic risk factor for Alzheimer’s. In APOE4 brains, microglia initially responded to the emerging disease pathology with an increase in their epigenomic information, suggesting that they were stepping up to their unique responsibility to fight off disease. But as the disease progressed, the cells exhibited a sharp drop off in information, a sign of deterioration and degeneration. This turnabout was strongest in people who had two copies of APOE4, rather than just one. The findings, Kellis said, suggest that APOE4 might destabilize the genome of microglia, causing them to burn out.

Another example is the fate of neurons expressing the gene RELN and its protein Reelin. Prior studies, including by Kellis and Tsai, have shown that RELN- expressing neurons in the entorhinal cortex and hippocampus are especially vulnerable in Alzheimer’s, but promote resilience if they survive. The new study sheds new light on their fate by demonstrating that they exhibit early and severe epigenomic information loss as disease advances, but that in people who remained cognitively resilient the neurons maintained epigenomic information.

In yet another example, the researchers tracked what they colloquially call “chromatin guardians” because their expression sustains and regulates cells’ epigenomic programs. For instance, cells with greater epigenomic erosion and advanced AD progression displayed increased chromatin accessibility in areas that were supposed to be locked down by Polycomb repression genes or other gene expression silencers. While resilient cells expressed genes promoting neural connectivity, epigenomically eroded cells expressed genes linked to inflammation and oxidative stress.

“The message is clear: Alzheimer’s is not only about plaques and tangles, but about the erosion of nuclear order itself,” Kellis says. “Cognitive decline emerges when chromatin guardians lose ground to the forces of erosion, switching from resilience to vulnerability at the most fundamental level of genome regulation.

“And when our brain cells lose their epigenomic memory marks and epigenomic information at the lowest level deep inside our neurons and microglia, it seems that Alheimer’s patients also lose their memory and cognition at the highest level.”

Other authors of the paper are Benjamin T. James, Kyriaki Galani, Riley J. Mangan, Stuart Benjamin Fass, Chuqian Liang, Manoj M. Wagle, Carles A. Boix, Yosuke Tanigawa, Sukwon Yun, Yena Sung, Xushen Xiong, Na Sun, Lei Hou, Martin Wohlwend, Mufan Qiu, Xikun Han, Lei Xiong, Efthalia Preka, Lei Huang, William F. Li, Li-Lun Ho, Amy Grayson, Julio Mantero, Alexey Kozlenkov, Hansruedi Mathys, Tianlong Chen, Stella Dracheva, and David A. Bennett.

Funding for the research came from the National Institutes of Health, the National Science Foundation, the Cure Alzheimer’s Fund, the Freedom Together Foundation, the Robert A. and Renee E. Belfer Family Foundation, Eduardo Eurnekian, and Joseph P. DiSabato.

At any given moment, trillions of particles called neutrinos are streaming through our bodies and every material in our surroundings, without noticeable effect. Smaller than electrons and lighter than photons, these ghostly entities are the most abundant particles with mass in the universe.

The exact mass of a neutrino is a big unknown. The particle is so small, and interacts so rarely with matter, that it is incredibly difficult to measure. Scientists attempt to do so by harnessing nuclear reactors and massive particle accelerators to generate unstable atoms, which then decay into various byproducts including neutrinos. In this way, physicists can manufacture beams of neutrinos that they can probe for properties including the particle’s mass.

Now MIT physicists propose a much more compact and efficient way to generate neutrinos that could be realized in a tabletop experiment.

In a paper appearing in Physical Review Letters, the physicists introduce the concept for a “neutrino laser” — a burst of neutrinos that could be produced by laser-cooling a gas of radioactive atoms down to temperatures colder than interstellar space. At such frigid temps, the team predicts the atoms should behave as one quantum entity, and radioactively decay in sync.

The decay of radioactive atoms naturally releases neutrinos, and the physicists say that in a coherent, quantum state this decay should accelerate, along with the production of neutrinos. This quantum effect should produce an amplified beam of neutrinos, broadly similar to how photons are amplified to produce conventional laser light.

“In our concept for a neutrino laser, the neutrinos would be emitted at a much faster rate than they normally would, sort of like a laser emits photons very fast,” says study co-author Ben Jones PhD ’15, an associate professor of physics at the University of Texas at Arlington.

As an example, the team calculated that such a neutrino laser could be realized by trapping 1 million atoms of rubidium-83. Normally, the radioactive atoms have a half-life of about 82 days, meaning that half the atoms decay, shedding an equivalent number of neutrinos, every 82 days. The physicists show that, by cooling rubidium-83 to a coherent, quantum state, the atoms should undergo radioactive decay in mere minutes.

“This is a novel way to accelerate radioactive decay and the production of neutrinos, which to my knowledge, has never been done,” says co-author Joseph Formaggio, professor of physics at MIT.

The team hopes to build a small tabletop demonstration to test their idea. If it works, they envision a neutrino laser could be used as a new form of communication, by which the particles could be sent directly through the Earth to underground stations and habitats. The neutrino laser could also be an efficient source of radioisotopes, which, along with neutrinos, are byproducts of radioactive decay. Such radioisotopes could be used to enhance medical imaging and cancer diagnostics.

Coherent condensate

For every atom in the universe, there are about a billion neutrinos. A large fraction of these invisible particles may have formed in the first moments following the Big Bang, and they persist in what physicists call the “cosmic neutrino background.” Neutrinos are also produced whenever atomic nuclei fuse together or break apart, such as in the fusion reactions in the sun’s core, and in the normal decay of radioactive materials.

Several years ago, Formaggio and Jones separately considered a novel possibility: What if a natural process of neutrino production could be enhanced through quantum coherence? Initial explorations revealed fundamental roadblocks in realizing this. Years later, while discussing the properties of ultracold tritium (an unstable isotope of hydrogen that undergoes radioactive decay) they asked: Could the production of neutrinos be enhanced if radioactive atoms such as tritium could be made so cold that they could be brought into a quantum state known as a Bose-Einstein condensate?

A Bose-Einstein condensate, or BEC, is a state of matter that forms when a gas of certain particles is cooled down to near absolute zero. At this point, the particles are brought down to their lowest energy level and stop moving as individuals. In this deep freeze, the particles can start to “feel” each others’ quantum effects, and can act as one coherent entity — a unique phase that can result in exotic physics.

BECs have been realized in a number of atomic species. (One of the first instances was with sodium atoms, by MIT’s Wolfgang Ketterle, who shared the 2001 Nobel Prize in Physics for the result.) However, no one has made a BEC from radioactive atoms. To do so would be exceptionally challenging, as most radioisotopes have short half-lives and would decay entirely before they could be sufficiently cooled to form a BEC.

Nevertheless, Formaggio wondered, if radioactive atoms could be made into a BEC, would this enhance the production of neutrinos in some way? In trying to work out the quantum mechanical calculations, he found initially that no such effect was likely.

“It turned out to be a red herring — we can’t accelerate the process of radioactive decay, and neutrino production, just by making a Bose-Einstein condensate,” Formaggio says.

In sync with optics

Several years later, Jones revisited the idea, with an added ingredient: superradiance — a phenomenon of quantum optics that occurs when a collection of light-emitting atoms is stimulated to behave in sync. In this coherent phase, it’s predicted that the atoms should emit a burst of photons that is “superradiant,” or more radiant than when the atoms are normally out of sync.

Jones proposed to Formaggio that perhaps a similar superradiant effect is possible in a radioactive Bose-Einstein condensate, which could then result in a similar burst of neutrinos. The physicists went to the drawing board to work out the equations of quantum mechanics governing how light-emitting atoms morph from a coherent starting state into a superradiant state. They used the same equations to work out what radioactive atoms in a coherent BEC state would do.

“The outcome is: You get a lot more photons more quickly, and when you apply the same rules to something that gives you neutrinos, it will give you a whole bunch more neutrinos more quickly,” Formaggio explains. “That’s when the pieces clicked together, that superradiance in a radioactive condensate could enable this accelerated, laser-like neutrino emission.”

To test their concept in theory, the team calculated how neutrinos would be produced from a cloud of 1 million super-cooled rubidium-83 atoms. They found that, in the coherent BEC state, the atoms radioactively decayed at an accelerating rate, releasing a laser-like beam of neutrinos within minutes.

Now that the physicists have shown in theory that a neutrino laser is possible, they plan to test the idea with a small tabletop setup.

“It should be enough to take this radioactive material, vaporize it, trap it with lasers, cool it down, and then turn it into a Bose-Einstein condensate,” Jones says. “Then it should start doing this superradiance spontaneously.”

The pair acknowledge that such an experiment will require a number of precautions and careful manipulation.

“If it turns out that we can show it in the lab, then people can think about: Can we use this as a neutrino detector? Or a new form of communication?” Formaggio says. “That’s when the fun really starts.”

In the search for habitable exoplanets, atmospheric conditions play a key role in determining if a planet can sustain liquid water. Suitable candidates often sit in the “Goldilocks zone,” a distance that is neither too close nor too far from their host star to allow liquid water. With the launch of the James Webb Space Telescope (JWST), astronomers are collecting improved observations of exoplanet atmospheres that will help determine which exoplanets are good candidates for further study.

In an open-access paper published today in The Astrophysical Journal Letters, astronomers used JWST to take a closer look at the atmosphere of the exoplanet TRAPPIST-1e, located in the TRAPPIST-1 system. While they haven’t found definitive proof of what it is made of — or if it even has an atmosphere — they were able to rule out several possibilities.

“The idea is: If we assume that the planet is not airless, can we constrain different atmospheric scenarios? Do those scenarios still allow for liquid water at the surface?” says Ana Glidden, a postdoc in the MIT Department of Earth, Atmospheric and Planetary Sciences (EAPS) and the MIT Kavli Institute for Astrophysics and Space Research, and the first author on the paper. The answers they found were yes.

The new data rule out a hydrogen-dominated atmosphere, and place tighter constraints on other atmospheric conditions that are commonly created through secondary-generation, such as volcanic eruptions and outgassing from the planet’s interior. The data were consistent enough to still allow for the possibility of a surface ocean.

“TRAPPIST-1e remains one of our most compelling habitable-zone planets, and these new results take us a step closer to knowing what kind of world it is,” says Sara Seager, Class of 1941 Professor of Planetary Science at MIT and co-author on the study. “The evidence pointing away from Venus- and Mars-like atmospheres sharpens our focus on the scenarios still in play.”

The study’s co-authors also include collaborators from the University of Arizona, Johns Hopkins University, University of Michigan, the Space Telescope Science Institute, and members of the JWST-TST DREAMS Team.

Improved observations

Exoplanet atmospheres are studied using a technique called transmission spectroscopy. When a planet passes in front of its host star, the starlight is filtered through the planet’s atmosphere. Astronomers can determine which molecules are present in the atmosphere by seeing how the light changes at different wavelengths.

“Each molecule has a spectral fingerprint. You can compare your observations with those fingerprints to suss out which molecules may be present,” says Glidden.

JWST has a larger wavelength coverage and higher spectral resolution than its predecessor, the Hubble Space Telescope, which makes it possible to observe molecules like carbon dioxide and methane that are more commonly found in our own solar system. However, the improved observations have also highlighted the problem of stellar contamination, where changes in the host star’s temperature due to things like sunspots and solar flares make it difficult to interpret data.

“Stellar activity strongly interferes with the planetary interpretation of the data because we can only observe a potential atmosphere through starlight,” says Glidden. “It is challenging to separate out which signals come from the star versus from the planet itself.”

Ruling out atmospheric conditions

The researchers used a novel approach to mitigate for stellar activity and, as a result, “any signal you can see varying visit-to-visit is most likely from the star, while anything that’s consistent between the visits is most likely the planet,” says Glidden.

The researchers were then able to compare the results to several different possible atmospheric scenarios. They found that carbon dioxide-rich atmospheres, like those of Mars and Venus, are unlikely, while a warm, nitrogen-rich atmosphere similar to Saturn’s moon Titan remains possible. The evidence, however, is too weak to determine if any atmosphere was present, let alone detecting a specific type of gas. Additional, ongoing observations that are already in the works will help to narrow down the possibilities.

“With our initial observations, we have showcased the gains made with JWST. Our follow-up program will help us to further refine our understanding of one of our best habitable-zone planets,” says Glidden.

The first comprehensive map of mouse brain activity has been unveiled by a large international collaboration of neuroscientists. 

Researchers from the International Brain Laboratory (IBL), including MIT neuroscientist Ila Fiete, published their open-access findings today in two papers in Nature, revealing insights into how decision-making unfolds across the entire brain in mice at single-cell resolution. This brain-wide activity map challenges the traditional hierarchical view of information processing in the brain and shows that decision-making is distributed across many regions in a highly coordinated way.

“This is the first time anyone has produced a full, brain-wide map of the activity of single neurons during decision-making,” explains co-founder of IBL Alexandre Pouget. “The scale is unprecedented as we recorded from over half-a-million neurons across mice in 12 labs, covering 279 brain areas, which together represent 95 percent of the mouse brain volume. The decision-making activity, and particularly reward, lit up the brain like a Christmas tree,” adds Pouget, who is also a group leader at the University of Geneva in Switzerland.

Modeling decision-making

The brain map was made possible by a major international collaboration of neuroscientists from multiple universities, including MIT. Researchers across 12 labs used state-of-the-art silicon electrodes, called neuropixels probes, for simultaneous neural recordings to measure brain activity while mice were carrying out a decision-making task.

“Participating in the International Brain Laboratory has added new ways for our group to contribute to science,” says Fiete, who is also a professor of brain and cognitive sciences, an associate investigator at the McGovern Institute for Brain Research, and director of the K. Lisa Yang ICoN Center at MIT. “Our lab has helped standardize methods to analyze and generate robust conclusions from data. As computational neuroscientists interested in building models of how the brain works, access to brain-wide recordings is incredible: the traditional approach of recording from one or a few brain areas limited our ability to build and test theories, resulting in fragmented models. Now, we have the delightful but formidable task to make sense of how all parts of the brain coordinate to perform a behavior. Surprisingly, having a full view of the brain leads to simplifications in the models of decision-making,” says Fiete.

The labs collected data from mice performing a decision-making task with sensory, motor, and cognitive components. In the task, a mouse sits in front of a screen and a light appears on the left or right side. If the mouse then responds by moving a small wheel in the correct direction, it receives a reward.

In some trials, the light is so faint that the animal must guess which way to turn the wheel, for which it can use prior knowledge: the light tends to appear more frequently on one side for a number of trials, before the high-frequency side switches. Well-trained mice learn to use this information to help them make correct guesses. These challenging trials therefore allowed the researchers to study how prior expectations influence perception and decision-making.

Brain-wide results

The first paper, “A brain-wide map of neural activity during complex behaviour,” showed that decision-making signals are surprisingly distributed across the brain, not localized to specific regions. This adds brain-wide evidence to a growing number of studies that challenge the traditional hierarchical model of brain function, and emphasizes that there is constant communication across brain areas during decision-making, movement onset, and even reward. This means that neuroscientists will need to take a more holistic, brain-wide approach when studying complex behaviors in the future.

“The unprecedented breadth of our recordings pulls back the curtain on how the entire brain performs the whole arc of sensory processing, cognitive decision-making, and movement generation,” says Fiete. “Structuring a collaboration that collects a large standardized dataset which single labs could not assemble is a revolutionary new direction for systems neuroscience, initiating the field into the hyper-collaborative mode that has contributed to leaps forward in particle physics and human genetics. Beyond our own conclusions, the dataset and associated technologies, which were released much earlier as part of the IBL mission, have already become a massively used resource for the entire neuroscience community.”

The second paper, “Brain-wide representations of prior information,” showed that prior expectations — our beliefs about what is likely to happen based on our recent experience — are encoded throughout the brain. Surprisingly, these expectations are not only found in cognitive areas, but also brain areas that process sensory information and control actions. For example, expectations are even encoded in early sensory areas such as the thalamus, the brain’s first relay for visual input from the eye. This supports the view that the brain acts as a prediction machine, but with expectations encoded across multiple brain structures playing a central role in guiding behavior responses. These findings could have implications for understanding conditions such as schizophrenia and autism, which are thought to be caused by differences in the way expectations are updated in the brain.

“Much remains to be unpacked: If it is possible to find a signal in a brain area, does it mean that this area is generating the signal, or simply reflecting a signal generated somewhere else? How strongly is our perception of the world shaped by our expectations? Now we can generate some quantitative answers and begin the next phase experiments to learn about the origins of the expectation signals by intervening to modulate their activity,” says Fiete.

Looking ahead, the team at IBL plan to expand beyond their initial focus on decision-making to explore a broader range of neuroscience questions. With renewed funding in hand, IBL aims to expand its research scope and continue to support large-scale, standardized experiments.

New model of collaborative neuroscience

Officially launched in 2017, IBL introduced a new model of collaboration in neuroscience that uses a standardized set of tools and data processing pipelines shared across multiple labs, enabling the collection of massive datasets while ensuring data alignment and reproducibility. This approach to democratize and accelerate science draws inspiration from large-scale collaborations in physics and biology, such as CERN and the Human Genome Project.

All data from these studies, along with detailed specifications of the tools and protocols used for data collection, are openly accessible to the global scientific community for further analysis and research. Summaries of these resources can be viewed and downloaded on the IBL website under the sections: Data, Tools, Protocols.

This research was supported by grants from Wellcome, the Simons Foundation, the National Institutes of Health, the National Science Foundation, the Gatsby Charitable Foundation, and by the Max Planck Society and the Humboldt Foundation.

3D printing has come a long way since its invention in 1983 by Chuck Hull, who pioneered stereolithography, a technique that solidifies liquid resin into solid objects using ultraviolet lasers. Over the decades, 3D printers have evolved from experimental curiosities into tools capable of producing everything from custom prosthetics to complex food designs, architectural models, and even functioning human organs. 

But as the technology matures, its environmental footprint has become increasingly difficult to set aside. The vast majority of consumer and industrial 3D printing still relies on petroleum-based plastic filament. And while “greener” alternatives made from biodegradable or recycled materials exist, they come with a serious trade-off: they’re often not as strong. These eco-friendly filaments tend to become brittle under stress, making them ill-suited for structural applications or load-bearing parts — exactly where strength matters most.

This trade-off between sustainability and mechanical performance prompted researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Hasso Plattner Institute to ask: Is it possible to build objects that are mostly eco-friendly, but still strong where it counts?

Their answer is SustainaPrint, a new software and hardware toolkit designed to help users strategically combine strong and weak filaments to get the best of both worlds. Instead of printing an entire object with high-performance plastic, the system analyzes a model through finite element analysis simulations, predicts where the object is most likely to experience stress, and then reinforces just those zones with stronger material. The rest of the part can be printed using greener, weaker filament, reducing plastic use while preserving structural integrity.

“Our hope is that SustainaPrint can be used in industrial and distributed manufacturing settings one day, where local material stocks may vary in quality and composition,” says MIT PhD student and CSAIL researcher Maxine Perroni-Scharf, who is a lead author on a paper presenting the project. “In these contexts, the testing toolkit could help ensure the reliability of available filaments, while the software’s reinforcement strategy could reduce overall material consumption without sacrificing function.” 

For their experiments, the team used Polymaker’s PolyTerra PLA as the eco-friendly filament, and standard or Tough PLA from Ultimaker for reinforcement. They used a 20 percent reinforcement threshold to show that even a small amount of strong plastic goes a long way. Using this ratio, SustainaPrint was able to recover up to 70 percent of the strength of an object printed entirely with high-performance plastic.

They printed dozens of objects, from simple mechanical shapes like rings and beams to more functional household items such as headphone stands, wall hooks, and plant pots. Each object was printed three ways: once using only eco-friendly filament, once using only strong PLA, and once with the hybrid SustainaPrint configuration. The printed parts were then mechanically tested by pulling, bending, or otherwise breaking them to measure how much force each configuration could withstand. 

In many cases, the hybrid prints held up nearly as well as the full-strength versions. For example, in one test involving a dome-like shape, the hybrid version outperformed the version printed entirely in Tough PLA. The team believes this may be due to the reinforced version’s ability to distribute stress more evenly, avoiding the brittle failure sometimes caused by excessive stiffness.

“This indicates that in certain geometries and loading conditions, mixing materials strategically may actually outperform a single homogenous material,” says Perroni-Scharf. “It’s a reminder that real-world mechanical behavior is full of complexity, especially in 3D printing, where interlayer adhesion and tool path decisions can affect performance in unexpected ways.”

A lean, green, eco-friendly printing machine

SustainaPrint starts off by letting a user upload their 3D model into a custom interface. By selecting fixed regions and areas where forces will be applied, the software then uses an approach called “Finite Element Analysis” to simulate how the object will deform under stress. It then creates a map showing pressure distribution inside the structure, highlighting areas under compression or tension, and applies heuristics to segment the object into two categories: those that need reinforcement, and those that don’t.

Recognizing the need for accessible and low-cost testing, the team also developed a DIY testing toolkit to help users assess strength before printing. The kit has a 3D-printable device with modules for measuring both tensile and flexural strength. Users can pair the device with common items like pull-up bars or digital scales to get rough, but reliable performance metrics. The team benchmarked their results against manufacturer data and found that their measurements consistently fell within one standard deviation, even for filaments that had undergone multiple recycling cycles.

Although the current system is designed for dual-extrusion printers, the researchers believe that with some manual filament swapping and calibration, it could be adapted for single-extruder setups, too. In current form, the system simplifies the modeling process by allowing just one force and one fixed boundary per simulation. While this covers a wide range of common use cases, the team sees future work expanding the software to support more complex and dynamic loading conditions. The team also sees potential in using AI to infer the object’s intended use based on its geometry, which could allow for fully automated stress modeling without manual input of forces or boundaries.

3D for free

The researchers plan to release SustainaPrint open-source, making both the software and testing toolkit available for public use and modification. Another initiative they aspire to bring to life in the future: education. “In a classroom, SustainaPrint isn’t just a tool, it’s a way to teach students about material science, structural engineering, and sustainable design, all in one project,” says Perroni-Scharf. “It turns these abstract concepts into something tangible.”

As 3D printing becomes more embedded in how we manufacture and prototype everything from consumer goods to emergency equipment, sustainability concerns will only grow. With tools like SustainaPrint, those concerns no longer need to come at the expense of performance. Instead, they can become part of the design process: built into the very geometry of the things we make.

Co-author Patrick Baudisch, who is a professor at the Hasso Plattner Institute, adds that “the project addresses a key question: What is the point of collecting material for the purpose of recycling, when there is no plan to actually ever use that material? Maxine presents the missing link between the theoretical/abstract idea of 3D printing material recycling and what it actually takes to make this idea relevant.”

Perroni-Scharf and Baudisch wrote the paper with CSAIL research assistant Jennifer Xiao; MIT Department of Electrical Engineering and Computer Science master’s student Cole Paulin ’24; master’s student Ray Wang SM ’25 and PhD student Ticha Sethapakdi SM ’19 (both CSAIL members); Hasso Plattner Institute PhD student Muhammad Abdullah; and Associate Professor Stefanie Mueller, lead of the Human-Computer Interaction Engineering Group at CSAIL.

The researchers’ work was supported by a Designing for Sustainability Grant from the Designing for Sustainability MIT-HPI Research Program. Their work will be presented at the ACM Symposium on User Interface Software and Technology in September.

Many attempts have been made to harness the power of new artificial intelligence and large language models (LLMs) to try to predict the outcomes of new chemical reactions. These have had limited success, in part because until now they have not been grounded in an understanding of fundamental physical principles, such as the laws of conservation of mass. Now, a team of researchers at MIT has come up with a way of incorporating these physical constraints on a reaction prediction model, and thus greatly improving the accuracy and reliability of its outputs.

The new work was reported Aug. 20 in the journal Nature, in a paper by recent postdoc Joonyoung Joung (now an assistant professor at Kookmin University, South Korea); former software engineer Mun Hong Fong (now at Duke University); chemical engineering graduate student Nicholas Casetti; postdoc Jordan Liles; physics undergraduate student Ne Dassanayake; and senior author Connor Coley, who is the Class of 1957 Career Development Professor in the MIT departments of Chemical Engineering and Electrical Engineering and Computer Science.

“The prediction of reaction outcomes is a very important task,” Joung explains. For example, if you want to make a new drug, “you need to know how to make it. So, this requires us to know what product is likely” to result from a given set of chemical inputs to a reaction. But most previous efforts to carry out such predictions look only at a set of inputs and a set of outputs, without looking at the intermediate steps or considering the constraints of ensuring that no mass is gained or lost in the process, which is not possible in actual reactions.

Joung points out that while large language models such as ChatGPT have been very successful in many areas of research, these models do not provide a way to limit their outputs to physically realistic possibilities, such as by requiring them to adhere to conservation of mass. These models use computational “tokens,” which in this case represent individual atoms, but “if you don’t conserve the tokens, the LLM model starts to make new atoms, or deletes atoms in the reaction.” Instead of being grounded in real scientific understanding, “this is kind of like alchemy,” he says. While many attempts at reaction prediction only look at the final products, “we want to track all the chemicals, and how the chemicals are transformed” throughout the reaction process from start to end, he says.

In order to address the problem, the team made use of a method developed back in the 1970s by chemist Ivar Ugi, which uses a bond-electron matrix to represent the electrons in a reaction. They used this system as the basis for their new program, called FlowER (Flow matching for Electron Redistribution), which allows them to explicitly keep track of all the electrons in the reaction to ensure that none are spuriously added or deleted in the process.

The system uses a matrix to represent the electrons in a reaction, and uses nonzero values to represent bonds or lone electron pairs and zeros to represent a lack thereof. “That helps us to conserve both atoms and electrons at the same time,” says Fong. This representation, he says, was one of the key elements to including mass conservation in their prediction system.

The system they developed is still at an early stage, Coley says. “The system as it stands is a demonstration — a proof of concept that this generative approach of flow matching is very well suited to the task of chemical reaction prediction.” While the team is excited about this promising approach, he says, “we’re aware that it does have specific limitations as far as the breadth of different chemistries that it’s seen.” Although the model was trained using data on more than a million chemical reactions, obtained from a U.S. Patent Office database, those data do not include certain metals and some kinds of catalytic reactions, he says.

“We’re incredibly excited about the fact that we can get such reliable predictions of chemical mechanisms” from the existing system, he says. “It conserves mass, it conserves electrons, but we certainly acknowledge that there’s a lot more expansion and robustness to work on in the coming years as well.”

But even in its present form, which is being made freely available through the online platform GitHub, “we think it will make accurate predictions and be helpful as a tool for assessing reactivity and mapping out reaction pathways,” Coley says. “If we’re looking toward the future of really advancing the state of the art of mechanistic understanding and helping to invent new reactions, we’re not quite there. But we hope this will be a steppingstone toward that.”

“It’s all open source,” says Fong. “The models, the data, all of them are up there,” including a previous dataset developed by Joung that exhaustively lists the mechanistic steps of known reactions. “I think we are one of the pioneering groups making this dataset, and making it available open-source, and making this usable for everyone,” he says.

The FlowER model matches or outperforms existing approaches in finding standard mechanistic pathways, the team says, and makes it possible to generalize to previously unseen reaction types. They say the model could potentially be relevant for predicting reactions for medicinal chemistry, materials discovery, combustion, atmospheric chemistry, and electrochemical systems.

In their comparisons with existing reaction prediction systems, Coley says, “using the architecture choices that we’ve made, we get this massive increase in validity and conservation, and we get a matching or a little bit better accuracy in terms of performance.”

He adds that “what’s unique about our approach is that while we are using these textbook understandings of mechanisms to generate this dataset, we’re anchoring the reactants and products of the overall reaction in experimentally validated data from the patent literature.” They are inferring the underlying mechanisms, he says, rather than just making them up. “We’re imputing them from experimental data, and that’s not something that has been done and shared at this kind of scale before.”

The next step, he says, is “we are quite interested in expanding the model’s understanding of metals and catalytic cycles. We’ve just scratched the surface in this first paper,” and most of the reactions included so far don’t include metals or catalysts, “so that’s a direction we’re quite interested in.”

In the long term, he says, “a lot of the excitement is in using this kind of system to help discover new complex reactions and help elucidate new mechanisms. I think that the long-term potential impact is big, but this is of course just a first step.”

The work was supported by the Machine Learning for Pharmaceutical Discovery and Synthesis consortium and the National Science Foundation.

Synthetic data are artificially generated by algorithms to mimic the statistical properties of actual data, without containing any information from real-world sources. While concrete numbers are hard to pin down, some estimates suggest that more than 60 percent of data used for AI applications in 2024 was synthetic, and this figure is expected to grow across industries.

Because synthetic data don’t contain real-world information, they hold the promise of safeguarding privacy while reducing the cost and increasing the speed at which new AI models are developed. But using synthetic data requires careful evaluation, planning, and checks and balances to prevent loss of performance when AI models are deployed.       

To unpack some pros and cons of using synthetic data, MIT News spoke with Kalyan Veeramachaneni, a principal research scientist in the Laboratory for Information and Decision Systems and co-founder of DataCebo whose open-core platform, the Synthetic Data Vault, helps users generate and test synthetic data.

Q: How are synthetic data created?

A: Synthetic data are algorithmically generated but do not come from a real situation. Their value lies in their statistical similarity to real data. If we’re talking about language, for instance, synthetic data look very much as if a human had written those sentences. While researchers have created synthetic data for a long time, what has changed in the past few years is our ability to build generative models out of data and use them to create realistic synthetic data. We can take a little bit of real data and build a generative model from that, which we can use to create as much synthetic data as we want. Plus, the model creates synthetic data in a way that captures all the underlying rules and infinite patterns that exist in the real data.

There are essentially four different data modalities: language, video or images, audio, and tabular data. All four of them have slightly different ways of building the generative models to create synthetic data. An LLM, for instance, is nothing but a generative model from which you are sampling synthetic data when you ask it a question.      

A lot of language and image data are publicly available on the internet. But tabular data, which is the data collected when we interact with physical and social systems, is often locked up behind enterprise firewalls. Much of it is sensitive or private, such as customer transactions stored by a bank. For this type of data, platforms like the Synthetic Data Vault provide software that can be used to build generative models. Those models then create synthetic data that preserve customer privacy and can be shared more widely.      

One powerful thing about this generative modeling approach for synthesizing data is that enterprises can now build a customized, local model for their own data. Generative AI automates what used to be a manual process.

Q: What are some benefits of using synthetic data, and which use-cases and applications are they particularly well-suited for?

A: One fundamental application which has grown tremendously over the past decade is using synthetic data to test software applications. There is data-driven logic behind many software applications, so you need data to test that software and its functionality. In the past, people have resorted to manually generating data, but now we can use generative models to create as much data as we need.

Users can also create specific data for application testing. Say I work for an e-commerce company. I can generate synthetic data that mimics real customers who live in Ohio and made transactions pertaining to one particular product in February or March.

Because synthetic data aren’t drawn from real situations, they are also privacy-preserving. One of the biggest problems in software testing has been getting access to sensitive real data for testing software in non-production environments, due to privacy concerns. Another immediate benefit is in performance testing. You can create a billion transactions from a generative model and test how fast your system can process them.

Another application where synthetic data hold a lot of promise is in training machine-learning models. Sometimes, we want an AI model to help us predict an event that is less frequent. A bank may want to use an AI model to predict fraudulent transactions, but there may be too few real examples to train a model that can identify fraud accurately. Synthetic data provide data augmentation — additional data examples that are similar to the real data. These can significantly improve the accuracy of AI models.

Also, sometimes users don’t have time or the financial resources to collect all the data. For instance, collecting data about customer intent would require conducting many surveys. If you end up with limited data and then try to train a model, it won’t perform well. You can augment by adding synthetic data to train those models better.

Q. What are some of the risks or potential pitfalls of using synthetic data, and are there steps users can take to prevent or mitigate those problems?

A. One of the biggest questions people often have in their mind is, if the data are synthetically created, why should I trust them? Determining whether you can trust the data often comes down to evaluating the overall system where you are using them.

There are a lot of aspects of synthetic data we have been able to evaluate for a long time. For instance, there are existing methods to measure how close synthetic data are to real data, and we can measure their quality and whether they preserve privacy. But there are other important considerations if you are using those synthetic data to train a machine-learning model for a new use case. How would you know the data are going to lead to models that still make valid conclusions?

New efficacy metrics are emerging, and the emphasis is now on efficacy for a particular task. You must really dig into your workflow to ensure the synthetic data you add to the system still allow you to draw valid conclusions. That is something that must be done carefully on an application-by-application basis.

Bias can also be an issue. Since it is created from a small amount of real data, the same bias that exists in the real data can carry over into the synthetic data. Just like with real data, you would need to purposefully make sure the bias is removed through different sampling techniques, which can create balanced datasets. It takes some careful planning, but you can calibrate the data generation to prevent the proliferation of bias.

To help with the evaluation process, our group created the Synthetic Data Metrics Library. We worried that people would use synthetic data in their environment and it would give different conclusions in the real world. We created a metrics and evaluation library to ensure checks and balances. The machine learning community has faced a lot of challenges in ensuring models can generalize to new situations. The use of synthetic data adds a whole new dimension to that problem.

I expect that the old systems of working with data, whether to build software applications, answer analytical questions, or train models, will dramatically change as we get more sophisticated at building these generative models. A lot of things we have never been able to do before will now be possible.

If your hand lotion is a bit runnier than usual coming out of the bottle, it might have something to do with the goop’s “mechanical memory.”

Soft gels and lotions are made by mixing ingredients until they form a stable and uniform substance. But even after a gel has set, it can hold onto “memories,” or residual stress, from the mixing process. Over time, the material can give in to these embedded stresses and slide back into its former, premixed state. Mechanical memory is, in part, why hand lotion separates and gets runny over time. 

Now, an MIT engineer has devised a simple way to measure the degree of residual stress in soft materials after they have been mixed, and found that common products like hair gel and shaving cream have longer mechanical memories, holding onto residual stresses for longer periods of time than manufacturers might have assumed.

In a study appearing today in Physical Review Letters, Crystal Owens, a postdoc in MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), presents a new protocol for measuring residual stress in soft, gel-like materials, using a standard benchtop rheometer.

Applying this protocol to everyday soft materials, Owens found that if a gel is made by mixing it in one direction, once it settles into a stable and uniform state, it effectively holds onto the memory of the direction in which it is mixed. Even after several days, the gel will hold some internal stress that, if released, will cause the gel to shift in the direction opposite to how it was initially mixed, reverting back to its earlier state.

“This is one reason different batches of cosmetics or food behave differently even if they underwent ‘identical’ manufacturing,” Owens says. “Understanding and measuring these hidden stresses during processing could help manufacturers design better products that last longer and perform more predictably.”

A soft glass

Hand lotion, hair gel, and shaving cream all fall under the category of “soft glassy materials” — materials that exhibit properties of both solids and liquids.

“Anything you can pour into your hand and it forms a soft mound is going to be considered a soft glass,” Owens explains. “In materials science, it’s considered a soft version of something that has the same amorphous structure as glass.”

In other words, a soft glassy material is a strange amalgam of a solid and a liquid. It can be poured out like a liquid, and it can hold its shape like a solid. Once they are made, these materials exist in a delicate balance between solid and liquid. And Owens wondered: For how long?

“What happens to these materials after very long times? Do they finally relax or do they never relax?” Owens says. “From a physics perspective, that’s a very interesting concept: What is the essential state of these materials?”

Twist and hold

In the manufacturing of soft glassy materials such as hair gel and shampoo, ingredients are first mixed into a uniform product. Quality control engineers then let a sample sit for about a minute — a period of time that they assume is enough to allow any residual stresses from the mixing process dissipate. In that time, the material should settle into a steady, stable state, ready for use.

But Owens suspected that the materials may hold some degree of stress from the production process long after they’ve appeared to settle.

“Residual stress is a low level of stress that’s trapped inside a material after it’s come to a steady state,” Owens says. “This sort of stress has not been measured in these sorts of materials.”

To test her hypothesis, she carried out experiments with two common soft glassy materials: hair gel and shaving cream. She made measurements of each material in a rheometer — an instrument consisting of two rotating plates that can twist and press a material together at precisely controlled pressures and forces that relate directly to the material’s internal stresses and strains.

In her experiments, she placed each material in the rheometer and spun the instrument’s top plate around to mix the material. Then she let the material settle, and then settle some more — much longer than one minute. During this time, she observed the amount of force it took the rheometer to hold the material in place. She reasoned that the greater the rheometer’s force, the more it must be counteracting any stress within the material that would otherwise cause it to shift out of its current state.

Over multiple experiments using this new protocol, Owens found that different types of soft glassy materials held a significant amount of residual stress, long after most researchers would assume the stress had dissipated. What’s more, she found that the degree of stress that a material retained was a reflection of the direction in which it was initially mixed, and when it was mixed.

“The material can effectively ‘remember’ which direction it was mixed, and how long ago,” Owens says. “And it turns out they hold this memory of their past, a lot longer than we used to think.”

In addition to the protocol she has developed to measure residual stress, Owens has developed a model to estimate how a material will change over time, given the degree of residual stress that it holds. Using this model, she says scientists might design materials with “short-term memory,” or very little residual stress, such that they remain stable over longer periods.

One material where she sees room for such improvement is asphalt — a substance that is first mixed, then poured in molten form over a surface where it then cools and settles over time. She suspects that residual stresses from the mixing of asphalt may contribute to cracks forming in pavement over time. Reducing these stresses at the start of the process could lead to longer-lasting, more resilient roads.

“People are inventing new types of asphalt all the time to be more eco-friendly, and all of these will have different levels of residual stress that will need some control,” she says. “There’s plenty of room to explore.”

This research was supported, in part, by MIT’s Postdoctoral Fellowship for Engineering Excellence and an MIT Mathworks Fellowship.

One in every eight people — 970 million globally — live with mental illness, according to the World Health Organization, with depression and anxiety being the most common mental health conditions worldwide. Existing therapies for complex psychiatric disorders like depression, anxiety, and schizophrenia have limitations, and federal funding to address these shortcomings is growing increasingly uncertain.

Patricia and James Poitras ’63 have committed $8 million to the Poitras Center for Psychiatric Disorders Research to launch pioneering research initiatives aimed at uncovering the brain basis of major mental illness and accelerating the development of novel treatments.

“Federal funding rarely supports the kind of bold, early-stage research that has the potential to transform our understanding of psychiatric illness. Pat and I want to help fill that gap — giving researchers the freedom to follow their most promising leads, even when the path forward isn’t guaranteed,” says James Poitras, who is chair of the McGovern Institute for Brain Research board.

Their latest gift builds upon their legacy of philanthropic support for psychiatric disorders research at MIT, which now exceeds $46 million.

“With deep gratitude for Jim and Pat’s visionary support, we are eager to launch a bold set of studies aimed at unraveling the neural and cognitive underpinnings of major mental illnesses,” says Professor Robert Desimone, director of the McGovern Institute, home to the Poitras Center. “Together, these projects represent a powerful step toward transforming how we understand and treat mental illness.”

A legacy of support

Soon after joining the McGovern Institute Leadership Board in 2006, the Poitrases made a $20 million commitment to establish the Poitras Center for Psychiatric Disorders Research at MIT. The center’s goal, to improve human health by addressing the root causes of complex psychiatric disorders, is deeply personal to them both.

“We had decided many years ago that our philanthropic efforts would be directed towards psychiatric research. We could not have imagined then that this perfect synergy between research at MIT’s McGovern Institute and our own philanthropic goals would develop,” recalls Patricia. 

The center supports research at the McGovern Institute and collaborative projects with institutions such as the Broad Institute of MIT and Harvard, McLean Hospital, Mass General Brigham, and other clinical research centers. Since its establishment in 2007, the center has enabled advances in psychiatric research including the development of a machine learning “risk calculator” for bipolar disorder, the use of brain imaging to predict treatment outcomes for anxiety, and studies demonstrating that mindfulness can improve mental health in adolescents.

For the past decade, the Poitrases have also fueled breakthroughs in the lab of McGovern investigator and MIT Professor Feng Zhang, backing the invention of powerful CRISPR systems and other molecular tools that are transforming biology and medicine. Their support has enabled the Zhang team to engineer new delivery vehicles for gene therapy, including vehicles capable of carrying genetic payloads that were once out of reach. The lab has also advanced innovative RNA-guided gene engineering tools such as NovaIscB, published in Nature Biotechnology in May 2025. These revolutionary genome editing and delivery technologies hold promise for the next generation of therapies needed for serious psychiatric illness.

In addition to fueling research in the center, the Poitras family has gifted two endowed professorships — the James and Patricia Poitras Professor of Neuroscience at MIT, currently held by Feng Zhang, and the James W. (1963) and Patricia T. Poitras Professor of Brain and Cognitive Sciences at MIT, held by Guoping Feng — and an annual postdoctoral fellowship at the McGovern Institute.

New initiatives at the Poitras Center

The Poitras family’s latest commitment to the Poitras Center will launch an ambitious set of new projects that bring together neuroscientists, clinicians, and computational experts to probe underpinnings of complex psychiatric disorders including schizophrenia, anxiety, and depression. These efforts reflect the center’s core mission: to speed scientific discovery and therapeutic innovation in the field of psychiatric brain disorders research.

McGovern cognitive neuroscientists Evelina Fedorenko PhD ’07, an associate professor, and Nancy Kanwisher ’80, PhD ’86, the Walter A. Rosenblith Professor of Cognitive Neuroscience — in collaboration with psychiatrist Ann Shinn of McLean Hospital — will explore how altered inner speech and reasoning contribute to the symptoms of schizophrenia. They will collect functional MRI data from individuals diagnosed with schizophrenia and matched controls as they perform reasoning tasks. The goal is to identify the brain activity patterns that underlie impaired reasoning in schizophrenia, a core cognitive disruption in the disorder.

A complementary line of investigation will focus on the role of inner speech — the “voice in our head” that shapes thought and self-awareness. The team will conduct a large-scale online behavioral study of neurotypical individuals to analyze how inner speech characteristics correlate with schizophrenia-spectrum traits. This will be followed by neuroimaging work comparing brain architecture among individuals with strong or weak inner voices and people with schizophrenia, with the aim of discovering neural markers linked to self-talk and disrupted cognition.

A different project led by McGovern neuroscientist and MIT Associate Professor Mark Harnett and 2024–2026 Poitras Center Postdoctoral Fellow Cynthia Rais focuses on how ketamine — an increasingly used antidepressant — alters brain circuits to produce rapid and sustained improvements in mood. Despite its clinical success, ketamine’s mechanisms of action remain poorly understood. The Harnett lab is using sophisticated tools to track how ketamine affects synaptic communication and large-scale brain network dynamics, particularly in models of treatment-resistant depression. By mapping these changes at both the cellular and systems levels, the team hopes to reveal how ketamine lifts mood so quickly — and inform the development of safer, longer-lasting antidepressants.

Guoping Feng is leveraging a new animal model of depression to uncover the brain circuits that drive major depressive disorder. The new animal model provides a powerful system for studying the intricacies of mood regulation. Feng’s team is using state-of-the-art molecular tools to identify the specific genes and cell types involved in this circuit, with the goal of developing targeted treatments that can fine-tune these emotional pathways.

“This is one of the most promising models we have for understanding depression at a mechanistic level,” says Feng, who is also associate director of the McGovern Institute. “It gives us a clear target for future therapies.”

Another novel approach to treating mood disorders comes from the lab of James DiCarlo, the Peter de Florez Professor of Neuroscience at MIT, who is exploring the brain’s visual-emotional interface as a therapeutic tool for anxiety. The amygdala, a key emotional center in the brain, is heavily influenced by visual input. DiCarlo’s lab is using advanced computational models to design visual scenes that may subtly shift emotional processing in the brain — essentially using sight to regulate mood. Unlike traditional therapies, this strategy could offer a noninvasive, drug-free option for individuals suffering from anxiety.

Together, these projects exemplify the kind of interdisciplinary, high-impact research that the Poitras Center was established to support.

“Mental illness affects not just individuals, but entire families who often struggle in silence and uncertainty,” adds Patricia Poitras. “Our hope is that Poitras Center scientists will continue to make important advancements and spark novel treatments for complex mental health disorders and, most of all, give families living with these conditions a renewed sense of hope for the future.”

A new and powerful particle detector just passed a critical test in its goal to decipher the ingredients of the early universe.

The sPHENIX detector is the newest experiment at Brookhaven National Laboratory’s Relativistic Heavy Ion Collider (RHIC) and is designed to precisely measure products of high-speed particle collisions. From the aftermath, scientists hope to reconstruct the properties of quark-gluon plasma (QGP) — a white-hot soup of subatomic particles known as quarks and gluons that is thought to have sprung into existence in the few microseconds following the Big Bang. Just as quickly, the mysterious plasma disappeared, cooling and combining to form the protons and neutrons that make up today’s ordinary matter.

Now, the sPHENIX detector has made a key measurement that proves it has the precision to help piece together the primordial properties of quark-gluon plasma.

In a paper in the Journal of High Energy Physics, scientists including physicists at MIT report that sPHENIX precisely measured the number and energy of particles that streamed out from gold ions that collided at close to the speed of light.

Straight ahead

This test is considered in physics to be a “standard candle,” meaning that the measurement is a well-established constant that can be used to gauge a detector’s precision.

In particular, sPHENIX successfully measured the number of charged particles that are produced when two gold ions collide, and determined how this number changes when the ions collide head-on, versus just glancing by. The detector’s measurements revealed that head-on collisions produced 10 times more charged particles, which were also 10 times more energetic, compared to less straight-on collisions.

“This indicates the detector works as it should,” says Gunther Roland, professor of physics at MIT, who is a member and former spokesperson for the sPHENIX Collaboration. “It’s as if you sent a new telescope up in space after you’ve spent 10 years building it, and it snaps the first picture. It’s not necessarily a picture of something completely new, but it proves that it’s now ready to start doing new science.”

“With this strong foundation, sPHENIX is well-positioned to advance the study of the quark-gluon plasma with greater precision and improved resolution,” adds Hao-Ren Jheng, a graduate student in physics at MIT and a lead co-author of the new paper. “Probing the evolution, structure, and properties of the QGP will help us reconstruct the conditions of the early universe.”

The paper’s co-authors are all members of the sPHENIX Collaboration, which comprises over 300 scientists from multiple institutions around the world, including Roland, Jheng, and physicists at MIT’s Bates Research and Engineering Center.

“Gone in an instant”

Particle colliders such as Brookhaven’s RHIC are designed to accelerate particles at “relativistic” speeds, meaning close to the speed of light. When these particles are flung around in opposite, circulating beams and brought back together, any smash-ups that occur can release an enormous amount of energy. In the right conditions, this energy can very briefly exist in the form of quark-gluon plasma — the same stuff that sprung out of the Big Bang.

Just as in the early universe, quark-gluon plasma doesn’t hang around for very long in particle colliders. If and when QGP is produced, it exists for just 10 to the minus 22, or about a sextillionth, of a second. In this moment, quark-gluon plasma is incredibly hot, up to several trillion degrees Celsius, and behaves as a “perfect fluid,” moving as one entity rather than as a collection of random particles. Almost immediately, this exotic behavior disappears, and the plasma cools and transitions into more ordinary particles such as protons and neutrons, which stream out from the main collision.

“You never see the QGP itself — you just see its ashes, so to speak, in the form of the particles that come from its decay,” Roland says. “With sPHENIX, we want to measure these particles to reconstruct the properties of the QGP, which is essentially gone in an instant.”

“One in a billion”

The sPHENIX detector is the next generation of Brookhaven’s original Pioneering High Energy Nuclear Interaction eXperiment, or PHENIX, which measured collisions of heavy ions generated by RHIC. In 2021, sPHENIX was installed in place of its predecessor, as a faster and more powerful version, designed to detect quark-gluon plasma’s more subtle and ephemeral signatures.

The detector itself is about the size of a two-story house and weighs around 1,000 tons. It sits at the intersection of RHIC’s two main collider beams, where relativistic particles, accelerated from opposite directions, meet and collide, producing particles that fly out into the detector. The sPHENIX detector is able to catch and measure 15,000 particle collisions per second, thanks to its novel, layered components, including the MVTX, or micro-vertex — a subdetector that was designed, built, and installed by scientists at MIT’s Bates Research and Engineering Center.

Together, the detector’s systems enable sPHENIX to act as a giant 3D camera that can track the number, energy, and paths of individual particles during an explosion of particles generated by a single collision.

“SPHENIX takes advantage of developments in detector technology since RHIC switched on 25 years ago, to collect data at the fastest possible rate,” says MIT postdoc Cameron Dean, who was a main contributor to the new study’s analysis. “This allows us to probe incredibly rare processes for the first time.”

In the fall of 2024, scientists ran the detector through the “standard candle” test to gauge its speed and precision. Over three weeks, they gathered data from sPHENIX as the main collider accelerated and smashed together beams of gold ions traveling at the speed of light. Their analysis of the data showed that sPHENIX accurately measured the number of charged particles produced in individual gold ion collisions, as well as the particles’ energies. What’s more, the detector was sensitive to a collision’s “head-on-ness,” and could observe that head-on collisions produced more particles with greater energy, compared to less direct collisions.

“This measurement provides clear evidence that the detector is functioning as intended,” Jheng says.

“The fun for sPHENIX is just beginning,” Dean adds. “We are currently back colliding particles and expect to do so for several more months. With all our data, we can look for the one-in-a-billion rare process that could give us insights on things like the density of QGP, the diffusion of particles through ultra-dense matter, and how much energy it takes to bind different particles together.”

This work was supported, in part, by the U.S. Department of Energy Office of Science, and the National Science Foundation.

Every year, global health experts are faced with a high-stakes decision: Which influenza strains should go into the next seasonal vaccine? The choice must be made months in advance, long before flu season even begins, and it can often feel like a race against the clock. If the selected strains match those that circulate, the vaccine will likely be highly effective. But if the prediction is off, protection can drop significantly, leading to (potentially preventable) illness and strain on health care systems.

This challenge became even more familiar to scientists in the years during the Covid-19 pandemic. Think back to the time (and time and time again), when new variants emerged just as vaccines were being rolled out. Influenza behaves like a similar, rowdy cousin, mutating constantly and unpredictably. That makes it hard to stay ahead, and therefore harder to design vaccines that remain protective.

To reduce this uncertainty, scientists at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the MIT Abdul Latif Jameel Clinic for Machine Learning in Health set out to make vaccine selection more accurate and less reliant on guesswork. They created an AI system called VaxSeer, designed to predict dominant flu strains and identify the most protective vaccine candidates, months ahead of time. The tool uses deep learning models trained on decades of viral sequences and lab test results to simulate how the flu virus might evolve and how the vaccines will respond.

Traditional evolution models often analyze the effect of single amino acid mutations independently. “VaxSeer adopts a large protein language model to learn the relationship between dominance and the combinatorial effects of mutations,” explains Wenxian Shi, a PhD student in MIT’s Department of Electrical Engineering and Computer Science, researcher at CSAIL, and lead author of a new paper on the work. “Unlike existing protein language models that assume a static distribution of viral variants, we model dynamic dominance shifts, making it better suited for rapidly evolving viruses like influenza.”

An open-access report on the study was published today in Nature Medicine.

The future of flu

VaxSeer has two core prediction engines: one that estimates how likely each viral strain is to spread (dominance), and another that estimates how effectively a vaccine will neutralize that strain (antigenicity). Together, they produce a predicted coverage score: a forward-looking measure of how well a given vaccine is likely to perform against future viruses.

The scale of the score could be from an infinite negative to 0. The closer the score to 0, the better the antigenic match of vaccine strains to the circulating viruses. (You can imagine it as the negative of some kind of “distance.”)

In a 10-year retrospective study, the researchers evaluated VaxSeer’s recommendations against those made by the World Health Organization (WHO) for two major flu subtypes: A/H3N2 and A/H1N1. For A/H3N2, VaxSeer’s choices outperformed the WHO’s in nine out of 10 seasons, based on retrospective empirical coverage scores (a surrogate metric of the vaccine effectiveness, calculated from the observed dominance from past seasons and experimental HI test results). The team used this to evaluate vaccine selections, as the effectiveness is only available for vaccines actually given to the population. 

For A/H1N1, it outperformed or matched the WHO in six out of 10 seasons. In one notable case, for the 2016 flu season, VaxSeer identified a strain that wasn’t chosen by the WHO until the following year. The model’s predictions also showed strong correlation with real-world vaccine effectiveness estimates, as reported by the CDC, Canada’s Sentinel Practitioner Surveillance Network, and Europe’s I-MOVE program. VaxSeer’s predicted coverage scores aligned closely with public health data on flu-related illnesses and medical visits prevented by vaccination.

So how exactly does VaxSeer make sense of all these data? Intuitively, the model first estimates how rapidly a viral strain spreads over time using a protein language model, and then determines its dominance by accounting for competition among different strains.

Once the model has calculated its insights, they’re plugged into a mathematical framework based on something called ordinary differential equations to simulate viral spread over time. For antigenicity, the system estimates how well a given vaccine strain will perform in a common lab test called the hemagglutination inhibition assay. This measures how effectively antibodies can inhibit the virus from binding to human red blood cells, which is a widely used proxy for antigenic match/antigenicity. 

Outpacing evolution

“By modeling how viruses evolve and how vaccines interact with them, AI tools like VaxSeer could help health officials make better, faster decisions — and stay one step ahead in the race between infection and immunity,” says Shi. 

VaxSeer currently focuses only on the flu virus’s HA (hemagglutinin) protein,the major antigen of influenza. Future versions could incorporate other proteins like NA (neuraminidase), and factors like immune history, manufacturing constraints, or dosage levels. Applying the system to other viruses would also require large, high-quality datasets that track both viral evolution and immune responses — data that aren’t always publicly available. The team, however is currently working on the methods that can predict viral evolution in low-data regimes building on relations between viral families

“Given the speed of viral evolution, current therapeutic development often lags behind. VaxSeer is our attempt to catch up,” says Regina Barzilay, the School of Engineering Distinguished Professor for AI and Health at MIT, AI lead of Jameel Clinic, and CSAIL principal investigator. 

“This paper is impressive, but what excites me perhaps even more is the team’s ongoing work on predicting viral evolution in low-data settings,” says Assistant Professor Jon Stokes of the Department of Biochemistry and Biomedical Sciences at McMaster University in Hamilton, Ontario. “The implications go far beyond influenza. Imagine being able to anticipate how antibiotic-resistant bacteria or drug-resistant cancers might evolve, both of which can adapt rapidly. This kind of predictive modeling opens up a powerful new way of thinking about how diseases change, giving us the opportunity to stay one step ahead and design clinical interventions before escape becomes a major problem.”

Shi and Barzilay wrote the paper with MIT CSAIL postdoc Jeremy Wohlwend ’16, MEng ’17, PhD ’25 and recent CSAIL affiliate Menghua Wu ’19, MEng ’20, PhD ’25. Their work was supported, in part, by the U.S. Defense Threat Reduction Agency and MIT Jameel Clinic.

Today’s electric vehicle boom is tomorrow’s mountain of electronic waste. And while myriad efforts are underway to improve battery recycling, many EV batteries still end up in landfills.

A research team from MIT wants to help change that with a new kind of self-assembling battery material that quickly breaks apart when submerged in a simple organic liquid. In a new paper published in Nature Chemistry, the researchers showed the material can work as the electrolyte in a functioning, solid-state battery cell and then revert back to its original molecular components in minutes.

The approach offers an alternative to shredding the battery into a mixed, hard-to-recycle mass. Instead, because the electrolyte serves as the battery’s connecting layer, when the new material returns to its original molecular form, the entire battery disassembles to accelerate the recycling process.

“So far in the battery industry, we’ve focused on high-performing materials and designs, and only later tried to figure out how to recycle batteries made with complex structures and hard-to-recycle materials,” says the paper’s first author Yukio Cho PhD ’23. “Our approach is to start with easily recyclable materials and figure out how to make them battery-compatible. Designing batteries for recyclability from the beginning is a new approach.”

Joining Cho on the paper are PhD candidate Cole Fincher, Ty Christoff-Tempesta PhD ’22, Kyocera Professor of Ceramics Yet-Ming Chiang, Visiting Associate Professor Julia Ortony, Xiaobing Zuo, and Guillaume Lamour.

Better batteries

There’s a scene in one of the “Harry Potter” films where Professor Dumbledore cleans a dilapidated home with the flick of the wrist and a spell. Cho says that image stuck with him as a kid. (What better way to clean your room?) When he saw a talk by Ortony on engineering molecules so that they could assemble into complex structures and then revert back to their original form, he wondered if it could be used to make battery recycling work like magic.

That would be a paradigm shift for the battery industry. Today, batteries require harsh chemicals, high heat, and complex processing to recycle. There are three main parts of a battery: the positively charged cathode, the negatively charged electrode, and the electrolyte that shuttles lithium ions between them. The electrolytes in most lithium-ion batteries are highly flammable and degrade over time into toxic byproducts that require specialized handling.

To simplify the recycling process, the researchers decided to make a more sustainable electrolyte. For that, they turned to a class of molecules that self-assemble in water, named aramid amphiphiles (AAs), whose chemical structures and stability mimic that of Kevlar. The researchers further designed the AAs to contain polyethylene glycol (PEG), which can conduct lithium ions, on one end of each molecule. When the molecules are exposed to water, they spontaneously form nanoribbons with ion-conducting PEG surfaces and bases that imitate the robustness of Kevlar through tight hydrogen bonding. The result is a mechanically stable nanoribbon structure that conducts ions across its surface.

“The material is composed of two parts,” Cho explains. “The first part is this flexible chain that gives us a nest, or host, for lithium ions to jump around. The second part is this strong organic material component that is used in the Kevlar, which is a bulletproof material. Those make the whole structure stable.”

When added to water, the nanoribbons self-assemble to form millions of nanoribbons that can be hot-pressed into a solid-state material.

“Within five minutes of being added to water, the solution becomes gel-like, indicating there are so many nanofibers formed in the liquid that they start to entangle each other,” Cho says. “What’s exciting is we can make this material at scale because of the self-assembly behavior.”

The team tested the material’s strength and toughness, finding it could endure the stresses associated with making and running the battery. They also constructed a solid-state battery cell that used lithium iron phosphate for the cathode and lithium titanium oxide as the anode, both common materials in today’s batteries. The nanoribbons moved lithium ions successfully between the electrodes, but a side-effect known as polarization limited the movement of lithium ions into the battery’s electrodes during fast bouts of charging and discharging, hampering its performance compared to today’s gold-standard commercial batteries.

“The lithium ions moved along the nanofiber all right, but getting the lithium ion from the nanofibers to the metal oxide seems to be the most sluggish point of the process,” Cho says.

When they immersed the battery cell into organic solvents, the material immediately dissolved, with each part of the battery falling away for easier recycling. Cho compared the materials’ reaction to cotton candy being submerged in water.

“The electrolyte holds the two battery electrodes together and provides the lithium-ion pathways,” Cho says. “So, when you want to recycle the battery, the entire electrolyte layer can fall off naturally and you can recycle the electrodes separately.”

Validating a new approach

Cho says the material is a proof of concept that demonstrates the recycle-first approach.

“We don’t want to say we solved all the problems with this material,” Cho says. “Our battery performance was not fantastic because we used only this material as the entire electrolyte for the paper, but what we’re picturing is using this material as one layer in the battery electrolyte. It doesn’t have to be the entire electrolyte to kick off the recycling process.”

Cho also sees a lot of room for optimizing the material’s performance with further experiments.

Now, the researchers are exploring ways to integrate these kinds of materials into existing battery designs as well as implementing the ideas into new battery chemistries.

“It’s very challenging to convince existing vendors to do something very differently,” Cho says. “But with new battery materials that may come out in five or 10 years, it could be easier to integrate this into new designs in the beginning.”

Cho also believes the approach could help reshore lithium supplies by reusing materials from batteries that are already in the U.S.

“People are starting to realize how important this is,” Cho says. “If we can start to recycle lithium-ion batteries from battery waste at scale, it’ll have the same effect as opening lithium mines in the U.S. Also, each battery requires a certain amount of lithium, so extrapolating out the growth of electric vehicles, we need to reuse this material to avoid massive lithium price spikes.”

The work was supported, in part, by the National Science Foundation and the U.S. Department of Energy. This work was performed, in part, using the MIT.nano Characterization facilities.

In World War II, Britain was fighting for its survival against German aerial bombardment. Yet Britain was importing dyes from Germany at the same time. This sounds curious, to put it mildly. How can two countries at war with each other also be trading goods?

Examples of this abound, actually. Britain also traded with its enemies for almost all of World War I. India and Pakistan conducted trade with each other during the First Kashmir War, from 1947 to 1949, and during the India-Pakistan War of 1965. Croatia and then-Yugoslavia traded with each other while fighting in 1992.

“States do in fact trade with their enemies during wars,” says MIT political scientist Mariya Grinberg. “There is a lot of variation in which products get traded, and in which wars, and there are differences in how long trade lasts into a war. But it does happen.”

Indeed, as Grinberg has found, state leaders tend to calculate whether trade can give them an advantage by boosting their own economies while not supplying their enemies with anything too useful in the near term.

“At its heart, wartime trade is all about the tradeoff between military benefits and economic costs,” Grinberg says. “Severing trade denies the enemy access to your products that could increase their military capabilities, but it also incurs a cost to you because you’re losing trade and neutral states could take over your long-term market share.” Therefore, many countries try trading with their wartime foes.

Grinberg explores this topic in a groundbreaking new book, the first one on the subject, “Trade in War: Economic Cooperation Across Enemy Lines,” published this month by Cornell University Press. It is also the first book by Grinberg, an assistant professor of political science at MIT.

Calculating time and utility

“Trade in War” has its roots in research Grinberg started as a doctoral student at the University of Chicago, where she noticed that wartime trade was a phenomenon not yet incorporated into theories of state behavior.

Grinberg wanted to learn about it comprehensively, so, as she quips, “I did what academics usually do: I went to the work of historians and said, ‘Historians, what have you got for me?’”

Modern wartime trading began during the Crimean War, which pitted Russia against France, Britain, the Ottoman Empire, and other allies. Before the war’s start in 1854, France had paid for many Russian goods that could not be shipped because ice in the Baltic Sea was late to thaw. To rescue its produce, France then persuaded Britain and Russia to adopt “neutral rights,” codified in the 1856 Declaration of Paris, which formalized the idea that goods in wartime could be shipped via neutral parties (sometimes acting as intermediaries for warring countries).

“This mental image that everyone has, that we don’t trade with our enemies during war, is actually an artifact of the world without any neutral rights,” Grinberg says. “Once we develop neutral rights, all bets are off, and now we have wartime trade.”

Overall, Grinberg’s systematic analysis of wartime trade shows that it needs to be understood on the level of particular goods. During wartime, states calculate how much it would hurt their own economies to stop trade of certain items; how useful specific products would be to enemies during war, and in what time frame; and how long a war is going to last.

“There are two conditions under which we can see wartime trade,” Grinberg says. “Trade is permitted when it does not help the enemy win the war, and it’s permitted when ending it would damage the state’s long-term economic security, beyond the current war.”

Therefore a state might export diamonds, knowing an adversary would need to resell such products over time to finance any military activities. Conversely, states will not trade products that can quickly convert into military use.

“The tradeoff is not the same for all products,” Grinberg says. “All products can be converted into something of military utility, but they vary in how long that takes. If I’m expecting to fight a short war, things that take a long time for my opponent to convert into military capabilities won’t help them win the current war, so they’re safer to trade.” Moreover, she adds, “States tend to prioritize maintaining their long-term economic stability, as long as the stakes don’t hit too close to home.”

This calculus helps explain some seemingly inexplicable wartime trade decisions. In 1917, three years into World War I, Germany started trading dyes to Britain. As it happens, dyes have military uses, for example as coatings for equipment. And World War I, infamously, was lasting far beyond initial expectations. But as of 1917, German planners thought the introduction of unrestricted submarine warfare would bring the war to a halt in their favor within a few months, so they approved the dye exports. That calculation was wrong, but it fits the framework Grinberg has developed.

States: Usually wrong about the length of wars

“Trade in War” has received praise from other scholars in the field. Michael Mastanduno of Dartmouth College has said the book “is a masterful contribution to our understanding of how states manage trade-offs across economics and security in foreign policy.”

For her part, Grinberg notes that her work holds multiple implications for international relations — one being that trade relationships do not prevent hostilities from unfolding, as some have theorized.

“We can’t expect even strong trade relations to deter a conflict,” Grinberg says. “On the other hand, when we learn our assumptions about the world are not necessarily correct, we can try to find different levers to deter war.”

Grinberg has also observed that states are not good, by any measure, at projecting how long they will be at war.

“States very infrequently get forecasts about the length of war right,” Grinberg says. That fact has formed the basis of a second, ongoing Grinberg book project.

“Now I’m studying why states go to war unprepared, why they think their wars are going to end quickly,” Grinberg says. “If people just read history, they will learn almost all of human history works against this assumption.”

At the same time, Grinberg thinks there is much more that scholars could learn specifically about trade and economic relations among warring countries — and hopes her book will spur additional work on the subject.

“I’m almost certain that I’ve only just begun to scratch the surface with this book,” she says. 

Our cells produce a variety of proteins, each with a specific role that, in many cases, means that they need to be in a particular part of the cell where that role is needed. One of the ways that cells ensure certain proteins end up in the right location at the right time is through localized translation, a process that ensures that proteins are made — or translated — close to where they will be needed. MIT professor of biology and Whitehead Institute for Biomedical Research member Jonathan Weissman and colleagues have studied localized translation in order to understand how it affects cell functions and allows cells to quickly respond to changing conditions.

Now, Weissman, who is also a Howard Hughes Medical Institute Investigator, and postdoc in his lab Jingchuan Luo have expanded our knowledge of localized translation at mitochondria, structures that generate energy for the cell. In an open-access paper published today in Cell, they share a new tool, LOCL-TL, for studying localized translation in close detail, and describe the discoveries it enabled about two classes of proteins that are locally translated at mitochondria.

The importance of localized translation at mitochondria relates to their unusual origin. Mitochondria were once bacteria that lived within our ancestors’ cells. Over time, the bacteria lost their autonomy and became part of the larger cells, which included migrating most of their genes into the larger cell’s genome in the nucleus. Cells evolved processes to ensure that proteins needed by mitochondria that are encoded in genes in the larger cell’s genome get transported to the mitochondria. Mitochondria retain a few genes in their own genome, so production of proteins from the mitochondrial genome and that of the larger cell’s genome must be coordinated to avoid mismatched production of mitochondrial parts. Localized translation may help cells to manage the interplay between mitochondrial and nuclear protein production — among other purposes.

How to detect local protein production

For a protein to be made, genetic code stored in DNA is read into RNA, and then the RNA is read or translated by a ribosome, a cellular machine that builds a protein according to the RNA code. Weissman’s lab previously developed a method to study localized translation by tagging ribosomes near a structure of interest, and then capturing the tagged ribosomes in action and observing the proteins they are making. This approach, called proximity-specific ribosome profiling, allows researchers to see what proteins are being made where in the cell. The challenge that Luo faced was how to tweak this method to capture only ribosomes at work near mitochondria.

Ribosomes work quickly, so a ribosome that gets tagged while making a protein at the mitochondria can move on to making other proteins elsewhere in the cell in a matter of minutes. The only way researchers can guarantee that the ribosomes they capture are still working on proteins made near the mitochondria is if the experiment happens very quickly.

Weissman and colleagues had previously solved this time sensitivity problem in yeast cells with a ribosome-tagging tool called BirA that is activated by the presence of the molecule biotin. BirA is fused to the cellular structure of interest, and tags ribosomes it can touch — but only once activated. Researchers keep the cell depleted of biotin until they are ready to capture the ribosomes, to limit the time when tagging occurs. However, this approach does not work with mitochondria in mammalian cells because they need biotin to function normally, so it cannot be depleted.

Luo and Weissman adapted the existing tool to respond to blue light instead of biotin. The new tool, LOV-BirA, is fused to the mitochondrion’s outer membrane. Cells are kept in the dark until the researchers are ready. Then they expose the cells to blue light, activating LOV-BirA to tag ribosomes. They give it a few minutes and then quickly extract the ribosomes. This approach proved very accurate at capturing only ribosomes working at mitochondria.

The researchers then used a method originally developed by the Weissman lab to extract the sections of RNA inside of the ribosomes. This allows them to see exactly how far along in the process of making a protein the ribosome is when captured, which can reveal whether the entire protein is made at the mitochondria, or whether it is partly produced elsewhere and only gets completed at the mitochondria.

“One advantage of our tool is the granularity it provides,” Luo says. “Being able to see what section of the protein is locally translated helps us understand more about how localized translation is regulated, which can then allow us to understand its dysregulation in disease and to control localized translation in future studies.”

Two protein groups are made at mitochondria

Using these approaches, the researchers found that about 20 percent of the genes needed in mitochondria that are located in the main cellular genome are locally translated at mitochondria. These proteins can be divided into two distinct groups with different evolutionary histories and mechanisms for localized translation.

One group consists of relatively long proteins, each containing more than 400 amino acids or protein building blocks. These proteins tend to be of bacterial origin — present in the ancestor of mitochondria — and they are locally translated in both mammalian and yeast cells, suggesting that their localized translation has been maintained through a long evolutionary history.

Like many mitochondrial proteins encoded in the nucleus, these proteins contain a mitochondrial targeting sequence (MTS), a ZIP code that tells the cell where to bring them. The researchers discovered that most proteins containing an MTS also contain a nearby inhibitory sequence that prevents transportation until they are done being made. This group of locally translated proteins lacks the inhibitory sequence, so they are brought to the mitochondria during their production.

Production of these longer proteins begins anywhere in the cell, and then after approximately the first 250 amino acids are made, they get transported to the mitochondria. While the rest of the protein gets made, it is simultaneously fed into a channel that brings it inside the mitochondrion. This ties up the channel for a long time, limiting import of other proteins, so cells can only afford to do this simultaneous production and import for select proteins. The researchers hypothesize that these bacterial-origin proteins are given priority as an ancient mechanism to ensure that they are accurately produced and placed within mitochondria.

The second locally translated group consists of short proteins, each less than 200 amino acids long. These proteins are more recently evolved, and correspondingly, the researchers found that the mechanism for their localized translation is not shared by yeast. Their mitochondrial recruitment happens at the RNA level. Two sequences within regulatory sections of each RNA molecule that do not encode the final protein instead code for the cell’s machinery to recruit the RNAs to the mitochondria.

The researchers searched for molecules that might be involved in this recruitment, and identified the RNA binding protein AKAP1, which exists at mitochondria. When they eliminated AKAP1, the short proteins were translated indiscriminately around the cell. This provided an opportunity to learn more about the effects of localized translation, by seeing what happens in its absence. When the short proteins were not locally translated, this led to the loss of various mitochondrial proteins, including those involved in oxidative phosphorylation, our cells’ main energy generation pathway.

In future research, Weissman and Luo will delve deeper into how localized translation affects mitochondrial function and dysfunction in disease. The researchers also intend to use LOCL-TL to study localized translation in other cellular processes, including in relation to embryonic development, neural plasticity, and disease.

“This approach should be broadly applicable to different cellular structures and cell types, providing many opportunities to understand how localized translation contributes to biological processes,” Weissman says. “We’re particularly interested in what we can learn about the roles it may play in diseases including neurodegeneration, cardiovascular diseases, and cancers.”

MIT researchers have developed a technique that enables real-time, 3D monitoring of corrosion, cracking, and other material failure processes inside a nuclear reactor environment.

This could allow engineers and scientists to design safer nuclear reactors that also deliver higher performance for applications like electricity generation and naval vessel propulsion.

During their experiments, the researchers utilized extremely powerful X-rays to mimic the behavior of neutrons interacting with a material inside a nuclear reactor.

They found that adding a buffer layer of silicon dioxide between the material and its substrate, and keeping the material under the X-ray beam for a longer period of time, improves the stability of the sample. This allows for real-time monitoring of material failure processes.

By reconstructing 3D image data on the structure of a material as it fails, researchers could design more resilient materials that can better withstand the stress caused by irradiation inside a nuclear reactor.

“If we can improve materials for a nuclear reactor, it means we can extend the life of that reactor. It also means the materials will take longer to fail, so we can get more use out of a nuclear reactor than we do now. The technique we’ve demonstrated here allows to push the boundary in understanding how materials fail in real-time,” says Ericmoore Jossou, who has shared appointments in the Department of Nuclear Science and Engineering (NSE), where he is the John Clark Hardwick Professor, and the Department of Electrical Engineering and Computer Science (EECS), and the MIT Schwarzman College of Computing.

Jossou, senior author of a study on this technique, is joined on the paper by lead author David Simonne, an NSE postdoc; Riley Hultquist, a graduate student in NSE; Jiangtao Zhao, of the European Synchrotron; and Andrea Resta, of Synchrotron SOLEIL. The research was published Tuesday by the journal Scripta Materiala.

“Only with this technique can we measure strain with a nanoscale resolution during corrosion processes. Our goal is to bring such novel ideas to the nuclear science community while using synchrotrons both as an X-ray probe and radiation source,” adds Simonne.

Real-time imaging

Studying real-time failure of materials used in advanced nuclear reactors has long been a goal of Jossou’s research group.

Usually, researchers can only learn about such material failures after the fact, by removing the material from its environment and imaging it with a high-resolution instrument.

“We are interested in watching the process as it happens. If we can do that, we can follow the material from beginning to end and see when and how it fails. That helps us understand a material much better,” he says.

They simulate the process by firing an extremely focused X-ray beam at a sample to mimic the environment inside a nuclear reactor. The researchers must use a special type of high-intensity X-ray, which is only found in a handful of experimental facilities worldwide.

For these experiments they studied nickel, a material incorporated into alloys that are commonly used in advanced nuclear reactors. But before they could start the X-ray equipment, they had to prepare a sample.

To do this, the researchers used a process called solid state dewetting, which involves putting a thin film of the material onto a substrate and heating it to an extremely high temperature in a furnace until it transforms into single crystals.

“We thought making the samples was going to be a walk in the park, but it wasn’t,” Jossou says.

As the nickel heated up, it interacted with the silicon substrate and formed a new chemical compound, essentially derailing the entire experiment. After much trial-and-error, the researchers found that adding a thin layer of silicon dioxide between the nickel and substrate prevented this reaction.

But when crystals formed on top of the buffer layer, they were highly strained. This means the individual atoms had moved slightly to new positions, causing distortions in the crystal structure.

Phase retrieval algorithms can typically recover the 3D size and shape of a crystal in real-time, but if there is too much strain in the material, the algorithms will fail.

However, the team was surprised to find that keeping the X-ray beam trained on the sample for a longer period of time caused the strain to slowly relax, due to the silicon buffer layer. After a few extra minutes of X-rays, the sample was stable enough that they could utilize phase retrieval algorithms to accurately recover the 3D shape and size of the crystal.

“No one had been able to do that before. Now that we can make this crystal, we can image electrochemical processes like corrosion in real time, watching the crystal fail in 3D under conditions that are very similar to inside a nuclear reactor. This has far-reaching impacts,” he says.

They experimented with a different substrate, such as niobium doped strontium titanate, and found that only a silicon dioxide buffered silicon wafer created this unique effect.

An unexpected result

As they fine-tuned the experiment, the researchers discovered something else.

They could also use the X-ray beam to precisely control the amount of strain in the material, which could have implications for the development of microelectronics.

In the microelectronics community, engineers often introduce strain to deform a material’s crystal structure in a way that boosts its electrical or optical properties.

“With our technique, engineers can use X-rays to tune the strain in microelectronics while they are manufacturing them. While this was not our goal with these experiments, it is like getting two results for the price of one,” he adds.

In the future, the researchers want to apply this technique to more complex materials like steel and other metal alloys used in nuclear reactors and aerospace applications. They also want to see how changing the thickness of the silicon dioxide buffer layer impacts their ability to control the strain in a crystal sample.

“This discovery is significant for two reasons. First, it provides fundamental insight into how nanoscale materials respond to radiation — a question of growing importance for energy technologies, microelectronics, and quantum materials. Second, it highlights the critical role of the substrate in strain relaxation, showing that the supporting surface can determine whether particles retain or release strain when exposed to focused X-ray beams,” says Edwin Fohtung, an associate professor at the Rensselaer Polytechnic Institute, who was not involved with this work.

This work was funded, in part, by the MIT Faculty Startup Fund and the U.S. Department of Energy. The sample preparation was carried out, in part, at the MIT.nano facilities.

Crime and public safety are among the most pressing concerns across communities in the United States. Violence fractures lives and carries staggering costs; the economic burden of gun violence alone tops $100 billion each year. More than 5 million people live under supervision through incarceration, probation, or parole, while countless more experience the collateral consequences of arrests and criminal charges. Achieving lasting public safety requires confronting both crime itself and the collateral consequences of the U.S. criminal justice system.

To help meet these dual challenges, J-PAL North America — a regional office of MIT’s Abdul Latif Jameel Poverty Action Lab (J-PAL) — with generous grant support from Arnold Ventures, launched the Initiative for Effective US Crime Policy (IECP). This initiative will generate rigorous evidence on strategies to make communities safer, reduce discrimination, and improve outcomes at every stage of the criminal justice process.

“There are a lot of open questions. We desperately need to be trying new solutions, but we need to try them in a way that enables us to learn whether they work,” notes Jennifer Doleac, executive vice president of criminal justice at Arnold Ventures. “There is a path forward for us to step up and make a concerted effort to make sure that we are being very strategic in how we spend our time and where we are directing our resources.”

Building on more than a decade of pioneering randomized evaluations, J‑PAL North America’s IECP will fund rigorous new studies in the criminal justice space, offer hands-on technical assistance, and connect researchers with practitioners. By reviewing both established and emerging evidence, the initiative will also help decision-makers focus resources on interventions that demonstrably improve public safety.

“Through this initiative, we aim to expand the use of rigorous existing evidence and help scale interventions that are proven to improve outcomes, from prevention to reintegration,” says Sara Heller, associate professor of economics at the University of Michigan and co-chair of IECP. “At the same time, IECP seeks to fill critical gaps in the evidence base by supporting new research on what works to improve the criminal justice system in the United States.”

A platform for collaboration

In June at the MIT Museum, IECP convened over 70 researchers, policymakers, and practitioners to identify research priorities and catalyze collaboration. Speakers explored the structural drivers of violence, effective pathways for translating evidence into policy, and strategies for establishing successful partnerships between researchers and practitioners.

Speakers also reflected on the value and limits of existing evidence and discussed areas in which randomized evaluations can help address the most pressing questions. Randomized evaluations have contributed powerful insights in areas such as summer youth employment programs, reminders to increase court appearances, hot-spot policing, and the use of body-worn cameras. Yet many important questions remain unanswered.

“We know randomized evaluations can answer hard policy questions, but only if we ask the right questions, with the right lens, at the right scale,” says Amanda Agan, associate professor at Cornell University and co-chair of IECP. “This convening was a call to push further: to design studies that are not only rigorous, but also relevant to the lived experiences of communities and the structural forces that shape public safety."

How to take part

Are you a practitioner with a promising idea in the criminal justice space, a policymaker planning a new program, a researcher developing a real-world intervention, or a funder investing in rigorous empirical evidence? IECP supports research partnerships to advance scalable, evidence-based solutions in the criminal legal system by funding impact evaluations, connecting researchers and practitioners, and supporting the design of randomized evaluations and the dissemination of evidence.   

To learn more about this initiative, please contact iecp@povertyactionlab.org.

Environmental scientists are increasingly using enormous artificial intelligence models to make predictions about changes in weather and climate, but a new study by MIT researchers shows that bigger models are not always better.

The team demonstrates that, in certain climate scenarios, much simpler, physics-based models can generate more accurate predictions than state-of-the-art deep-learning models.

Their analysis also reveals that a benchmarking technique commonly used to evaluate machine-learning techniques for climate predictions can be distorted by natural variations in the data, like fluctuations in weather patterns. This could lead someone to believe a deep-learning model makes more accurate predictions when that is not the case.

The researchers developed a more robust way of evaluating these techniques, which shows that, while simple models are more accurate when estimating regional surface temperatures, deep-learning approaches can be the best choice for estimating local rainfall.

They used these results to enhance a simulation tool known as a climate emulator, which can rapidly simulate the effect of human activities onto a future climate.

The researchers see their work as a “cautionary tale” about the risk of deploying large AI models for climate science. While deep-learning models have shown incredible success in domains such as natural language, climate science contains a proven set of physical laws and approximations, and the challenge becomes how to incorporate those into AI models.

“We are trying to develop models that are going to be useful and relevant for the kinds of things that decision-makers need going forward when making climate policy choices. While it might be attractive to use the latest, big-picture machine-learning model on a climate problem, what this study shows is that stepping back and really thinking about the problem fundamentals is important and useful,” says study senior author Noelle Selin, a professor in the MIT Institute for Data, Systems, and Society (IDSS) and the Department of Earth, Atmospheric and Planetary Sciences (EAPS), and director of the Center for Sustainability Science and Strategy.

Selin’s co-authors are lead author Björn Lütjens, a former EAPS postdoc who is now a research scientist at IBM Research; senior author Raffaele Ferrari, the Cecil and Ida Green Professor of Oceanography in EAPS and co-director of the Lorenz Center; and Duncan Watson-Parris, assistant professor at the University of California at San Diego. Selin and Ferrari are also co-principal investigators of the Bringing Computation to the Climate Challenge project, out of which this research emerged. The paper appears today in the Journal of Advances in Modeling Earth Systems.

Comparing emulators

Because the Earth’s climate is so complex, running a state-of-the-art climate model to predict how pollution levels will impact environmental factors like temperature can take weeks on the world’s most powerful supercomputers.

Scientists often create climate emulators, simpler approximations of a state-of-the art climate model, which are faster and more accessible. A policymaker could use a climate emulator to see how alternative assumptions on greenhouse gas emissions would affect future temperatures, helping them develop regulations.

But an emulator isn’t very useful if it makes inaccurate predictions about the local impacts of climate change. While deep learning has become increasingly popular for emulation, few studies have explored whether these models perform better than tried-and-true approaches.

The MIT researchers performed such a study. They compared a traditional technique called linear pattern scaling (LPS) with a deep-learning model using a common benchmark dataset for evaluating climate emulators.

Their results showed that LPS outperformed deep-learning models on predicting nearly all parameters they tested, including temperature and precipitation.

“Large AI methods are very appealing to scientists, but they rarely solve a completely new problem, so implementing an existing solution first is necessary to find out whether the complex machine-learning approach actually improves upon it,” says Lütjens.

Some initial results seemed to fly in the face of the researchers’ domain knowledge. The powerful deep-learning model should have been more accurate when making predictions about precipitation, since those data don’t follow a linear pattern.

They found that the high amount of natural variability in climate model runs can cause the deep learning model to perform poorly on unpredictable long-term oscillations, like El Niño/La Niña. This skews the benchmarking scores in favor of LPS, which averages out those oscillations.

Constructing a new evaluation

From there, the researchers constructed a new evaluation with more data that address natural climate variability. With this new evaluation, the deep-learning model performed slightly better than LPS for local precipitation, but LPS was still more accurate for temperature predictions.

“It is important to use the modeling tool that is right for the problem, but in order to do that you also have to set up the problem the right way in the first place,” Selin says.

Based on these results, the researchers incorporated LPS into a climate emulation platform to predict local temperature changes in different emission scenarios.

“We are not advocating that LPS should always be the goal. It still has limitations. For instance, LPS doesn’t predict variability or extreme weather events,” Ferrari adds.

Rather, they hope their results emphasize the need to develop better benchmarking techniques, which could provide a fuller picture of which climate emulation technique is best suited for a particular situation.

“With an improved climate emulation benchmark, we could use more complex machine-learning methods to explore problems that are currently very hard to address, like the impacts of aerosols or estimations of extreme precipitation,” Lütjens says.

Ultimately, more accurate benchmarking techniques will help ensure policymakers are making decisions based on the best available information.

The researchers hope others build on their analysis, perhaps by studying additional improvements to climate emulation methods and benchmarks. Such research could explore impact-oriented metrics like drought indicators and wildfire risks, or new variables like regional wind speeds.

This research is funded, in part, by Schmidt Sciences, LLC, and is part of the MIT Climate Grand Challenges team for “Bringing Computation to the Climate Challenge.”

Back in the 17th century, German astronomer Johannes Kepler figured out the laws of motion that made it possible to accurately predict where our solar system’s planets would appear in the sky as they orbit the sun. But it wasn’t until decades later, when Isaac Newton formulated the universal laws of gravitation, that the underlying principles were understood. Although they were inspired by Kepler’s laws, they went much further, and made it possible to apply the same formulas to everything from the trajectory of a cannon ball to the way the moon’s pull controls the tides on Earth — or how to launch a satellite from Earth to the surface of the moon or planets.

Today’s sophisticated artificial intelligence systems have gotten very good at making the kind of specific predictions that resemble Kepler’s orbit predictions. But do they know why these predictions work, with the kind of deep understanding that comes from basic principles like Newton’s laws? As the world grows ever-more dependent on these kinds of AI systems, researchers are struggling to try to measure just how they do what they do, and how deep their understanding of the real world actually is.

Now, researchers in MIT’s Laboratory for Information and Decision Systems (LIDS) and at Harvard University have devised a new approach to assessing how deeply these predictive systems understand their subject matter, and whether they can apply knowledge from one domain to a slightly different one. And by and large the answer at this point, in the examples they studied, is — not so much.

The findings were presented at the International Conference on Machine Learning, in Vancouver, British Columbia, last month by Harvard postdoc Keyon Vafa, MIT graduate student in electrical engineering and computer science and LIDS affiliate Peter G. Chang, MIT assistant professor and LIDS principal investigator Ashesh Rambachan, and MIT professor, LIDS principal investigator, and senior author Sendhil Mullainathan.

“Humans all the time have been able to make this transition from good predictions to world models,” says Vafa, the study’s lead author. So the question their team was addressing was, “have foundation models — has AI — been able to make that leap from predictions to world models? And we’re not asking are they capable, or can they, or will they. It’s just, have they done it so far?” he says.

“We know how to test whether an algorithm predicts well. But what we need is a way to test for whether it has understood well,” says Mullainathan, the Peter de Florez Professor with dual appointments in the MIT departments of Economics and Electrical Engineering and Computer Science and the senior author on the study. “Even defining what understanding means was a challenge.” 

In the Kepler versus Newton analogy, Vafa says, “they both had models that worked really well on one task, and that worked essentially the same way on that task. What Newton offered was ideas that were able to generalize to new tasks.” That capability, when applied to the predictions made by various AI systems, would entail having it develop a world model so it can “transcend the task that you’re working on and be able to generalize to new kinds of problems and paradigms.”

Another analogy that helps to illustrate the point is the difference between centuries of accumulated knowledge of how to selectively breed crops and animals, versus Gregor Mendel’s insight into the underlying laws of genetic inheritance.

“There is a lot of excitement in the field about using foundation models to not just perform tasks, but to learn something about the world,” for example in the natural sciences, he says. “It would need to adapt, have a world model to adapt to any possible task.”

Are AI systems anywhere near the ability to reach such generalizations? To test the question, the team looked at different examples of predictive AI systems, at different levels of complexity. On the very simplest of examples, the systems succeeded in creating a realistic model of the simulated system, but as the examples got more complex that ability faded fast.

The team developed a new metric, a way of measuring quantitatively how well a system approximates real-world conditions. They call the measurement inductive bias — that is, a tendency or bias toward responses that reflect reality, based on inferences developed from looking at vast amounts of data on specific cases.

The simplest level of examples they looked at was known as a lattice model. In a one-dimensional lattice, something can move only along a line. Vafa compares it to a frog jumping between lily pads in a row. As the frog jumps or sits, it calls out what it’s doing — right, left, or stay. If it reaches the last lily pad in the row, it can only stay or go back. If someone, or an AI system, can just hear the calls, without knowing anything about the number of lily pads, can it figure out the configuration? The answer is yes: Predictive models do well at reconstructing the “world” in such a simple case. But even with lattices, as you increase the number of dimensions, the systems no longer can make that leap.

“For example, in a two-state or three-state lattice, we showed that the model does have a pretty good inductive bias toward the actual state,” says Chang. “But as we increase the number of states, then it starts to have a divergence from real-world models.”

A more complex problem is a system that can play the board game Othello, which involves players alternately placing black or white disks on a grid. The AI models can accurately predict what moves are allowable at a given point, but it turns out they do badly at inferring what the overall arrangement of pieces on the board is, including ones that are currently blocked from play.

The team then looked at five different categories of predictive models actually in use, and again, the more complex the systems involved, the more poorly the predictive modes performed at matching the true underlying world model.

With this new metric of inductive bias, “our hope is to provide a kind of test bed where you can evaluate different models, different training approaches, on problems where we know what the true world model is,” Vafa says. If it performs well on these cases where we already know the underlying reality, then we can have greater faith that its predictions may be useful even in cases “where we don’t really know what the truth is,” he says.

People are already trying to use these kinds of predictive AI systems to aid in scientific discovery, including such things as properties of chemical compounds that have never actually been created, or of potential pharmaceutical compounds, or for predicting the folding behavior and properties of unknown protein molecules. “For the more realistic problems,” Vafa says, “even for something like basic mechanics, we found that there seems to be a long way to go.”

Chang says, “There’s been a lot of hype around foundation models, where people are trying to build domain-specific foundation models — biology-based foundation models, physics-based foundation models, robotics foundation models, foundation models for other types of domains where people have been collecting a ton of data” and training these models to make predictions, “and then hoping that it acquires some knowledge of the domain itself, to be used for other downstream tasks.”

This work shows there’s a long way to go, but it also helps to show a path forward. “Our paper suggests that we can apply our metrics to evaluate how much the representation is learning, so that we can come up with better ways of training foundation models, or at least evaluate the models that we’re training currently,” Chang says. “As an engineering field, once we have a metric for something, people are really, really good at optimizing that metric.”

For both research and medical purposes, researchers have spent decades pushing the limits of microscopy to produce ever deeper and sharper images of brain activity, not only in the cortex but also in regions underneath, such as the hippocampus. In a new study, a team of MIT scientists and engineers demonstrates a new microscope system capable of peering exceptionally deep into brain tissues to detect the molecular activity of individual cells by using sound.

“The major advance here is to enable us to image deeper at single-cell resolution,” says neuroscientist Mriganka Sur, a corresponding author along with mechanical engineering professor Peter So and principal research scientist Brian Anthony. Sur is the Paul and Lilah Newton Professor in The Picower Institute for Learning and Memory and the Department of Brain and Cognitive Sciences at MIT.

In the journal Light: Science and Applications, the team demonstrates that they could detect NAD(P)H, a molecule tightly associated with cell metabolism in general and electrical activity in neurons in particular, all the way through samples such as a 1.1-millimeter “cerebral organoid,” a 3D-mini brain-like tissue generated from human stem cells, and a 0.7-milimeter-thick slice of mouse brain tissue.

In fact, says co-lead author and mechanical engineering postdoc W. David Lee, who conceived the microscope’s innovative design, the system could have peered far deeper, but the test samples weren’t big enough to demonstrate that.

“That’s when we hit the glass on the other side,” he says. “I think we’re pretty confident about going deeper.”

Still, a depth of 1.1 milimeters is more than five times deeper than other microscope technologies can resolve NAD(P)H within dense brain tissue. The new system achieved the depth and sharpness by combining several advanced technologies to precisely and efficiently excite the molecule and then to detect the resulting energy, all without having to add any external labels, either via added chemicals or genetically engineered fluorescence.

Rather than focusing the required NAD(P)H excitation energy on a neuron with near ultraviolet light at its normal peak absorption, the scope accomplishes the excitation by focusing an intense, extremely short burst of light (a quadrillionth of a second long) at three times the normal absorption wavelength. Such “three-photon” excitation penetrates deep into tissue with less scattering by brain tissue because of the longer wavelength of the light (“like fog lamps,” Sur says). Meanwhile, although the excitation produces a weak fluorescent signal of light from NAD(P)H, most of the absorbed energy produces a localized (about 10 microns) thermal expansion within the cell, which produces sound waves that travel relatively easily through tissue compared to the fluorescence emission. A sensitive ultrasound microphone in the microscope detects those waves and, with enough sound data, software turns them into high-resolution images (much like a sonogram does). Imaging created in this way is “three-photon photoacoustic imaging.”

“We merged all these techniques — three-photon, label-free, photoacoustic detection,” says co-lead author Tatsuya Osaki, a research scientist in the Picower Institute in Sur’s lab. “We integrated all these cutting-edge techniques into one process to establish this ‘Multiphoton-In and Acoustic-Out’ platform.”

Lee and Osaki combined with research scientist Xiang Zhang and postdoc Rebecca Zubajlo to lead the study, in which the team demonstrated reliable detection of the sound signal through the samples. So far, the team has produced visual images from the sound at various depths as they refine their signal processing.

In the study, the team also shows simultaneous “third-harmonic generation” imaging, which comes from the three-photon stimulation and finely renders cellular structures, alongside their photoacoustic imaging, which detects NAD(P)H. They also note that their photoacoustic method could detect other molecules such as the genetically encoded calcium indicator GCaMP, that neuroscientists use to signal neural electrical activity.

With the concept of label-free, multiphoton, photoacoustic microscopy (LF-MP-PAM) established in the paper, the team is now looking ahead to neuroscience and clinical applications.

For instance, through the company Precision Healing, Inc., which he founded and sold, Lee has already established that NAD(P)H imaging can inform wound care. In the brain, levels of the molecule are known to vary in conditions such as Alzheimer’s disease, Rett syndrome, and seizures, making it a potentially valuable biomarker. Because the new system is label-free (i.e., no added chemicals or altered genes), it could be used in humans, for instance, during brain surgeries.

The next step for the team is to demonstrate it in a living animal, rather than just in in vitro and ex-vivo tissues. The technical challenge there is that the microphone can no longer be on the opposite side of the sample from the light source (as it was in the current study). It has to be on top, just like the light source.

Lee says he expects that full imaging at depths of 2 milimeters in live brains is entirely feasible, given the results in the new study.

“In principle, it should work,” he says.

Mercedes Balcells and Elazer Edelman are also authors of the paper. Funding for the research came from sources including the National Institutes of Health, the Simon Center for the Social Brain, the lab of Peter So, The Picower Institute for Learning and Memory, and the Freedom Together Foundation.

A fast radio burst is an immense flash of radio emission that lasts for just a few milliseconds, during which it can momentarily outshine every other radio source in its galaxy. These flares can be so bright that their light can be seen from halfway across the universe, several billion light years away.

The sources of these brief and dazzling signals are unknown. But scientists now have a chance to study a fast radio burst (FRB) in unprecedented detail. An international team of scientists including physicists at MIT have detected a near and ultrabright fast radio burst some 130 million light-years from Earth in the constellation Ursa Major. It is one of the closest FRBs detected to date. It is also the brightest — so bright that the signal has garnered the informal moniker, RBFLOAT, for “radio brightest flash of all time.”

The burst’s brightness, paired with its proximity, is giving scientists the closest look yet at FRBs and the environments from which they emerge.

“Cosmically speaking, this fast radio burst is just in our neighborhood,” says Kiyoshi Masui, associate professor of physics and affiliate of MIT’s Kavli Institute for Astrophysics and Space Research. “This means we get this chance to study a pretty normal FRB in exquisite detail.”

Masui and his colleagues report their findings today in the Astrophysical Journal Letters.

Diverse bursts

The clarity of the new detection is thanks to a significant upgrade to The Canadian Hydrogen Intensity Mapping Experiment (CHIME), a large array of halfpipe-shaped antennae based in British Columbia. CHIME was originally designed to detect and map the distribution of hydrogen across the universe. The telescope is also sensitive to ultrafast and bright radio emissions. Since it started observations in 2018, CHIME has detected about 4,000 fast radio bursts, from all parts of the sky. But the telescope had not been able to precisely pinpoint the location of each fast radio burst, until now.

CHIME recently got a significant boost in precision, in the form of CHIME Outriggers — three miniature versions of CHIME, each sited in different parts of North America. Together, the telescopes work as one continent-sized system that can focus in on any bright flash that CHIME detects, to pin down its location in the sky with extreme precision.

“Imagine we are in New York and there’s a firefly in Florida that is bright for a thousandth of a second, which is usually how quick FRBs are,” says MIT Kavli graduate student Shion Andrew. “Localizing an FRB to a specific part of its host galaxy is analogous to figuring out not just what tree the firefly came from, but which branch it’s sitting on.”

The new fast radio burst is the first detection made using the combination of CHIME and the completed CHIME Outriggers. Together, the telescope array identified the FRB and determined not only the specific galaxy, but also the region of the galaxy from where the burst originated. It appears that the burst arose from the edge of the galaxy, just outside of a star-forming region. The precise localization of the FRB is allowing scientists to study the environment around the signal for clues to what brews up such bursts.

“As we’re getting these much more precise looks at FRBs, we’re better able to see the diversity of environments they’re coming from,” says MIT physics postdoc Adam Lanman.

Lanman, Andrew, and Masui are members of the CHIME Collaboration — which includes scientists from multiple institutions around the world — and are authors of the new paper detailing the discovery of the new FRB detection.

An older edge

Each of CHIME’s Outrigger stations continuously monitors the same swath of sky as the parent CHIME array. Both CHIME and the Outriggers “listen” for radio flashes, at incredibly short, millisecond timescales. Even over several minutes, such precision monitoring can amount to a huge amount of data. If CHIME detects no FRB signal, the Outriggers automatically delete the last 40 seconds of data to make room for the next span of measurements.

On March 16, 2025, CHIME detected an ultrabright flash of radio emissions, which automatically triggered the CHIME Outriggers to record the data. Initially, the flash was so bright that astronomers were unsure whether it was an FRB or simply a terrestrial event caused, for instance, by a burst of cellular communications.

That notion was put to rest as the CHIME Outrigger telescopes focused in on the flash and pinned down its location to NGC4141 — a spiral galaxy in the constellation Ursa Major about 130 million light years away, which happens to be surprisingly close to our own Milky Way. The detection is one of the closest and brightest fast radio bursts detected to date.

Follow-up observations in the same region revealed that the burst came from the very edge of an active region of star formation. While it’s still a mystery as to what source could produce FRBs, scientists’ leading hypothesis points to magnetars — young neutron stars with extremely powerful magnetic fields that can spin out high-energy flares across the electromagnetic spectrum, including in the radio band. Physicists suspect that magnetars are found in the center of star formation regions, where the youngest, most active stars are forged. The location of the new FRB, just outside a star-forming region in its galaxy, may suggest that the source of the burst is a slightly older magnetar.

“These are mostly hints,” Masui says. “But the precise localization of this burst is letting us dive into the details of how old an FRB source could be. If it were right in the middle, it would only be thousands of years old — very young for a star. This one, being on the edge, may have had a little more time to bake.”

No repeats

In addition to pinpointing where the new FRB was in the sky, the scientists also looked back through CHIME data to see whether any similar flares occurred in the same region in the past. Since the first FRB was discovered in 2007, astronomers have detected over 4,000 radio flares. Most of these bursts are one-offs. But a few percent have been observed to repeat, flashing every so often. And an even smaller fraction of these repeaters flash in a pattern, like a rhythmic heartbeat, before flaring out. A central question surrounding fast radio bursts is whether repeaters and nonrepeaters come from different origins.

The scientists looked through CHIME’s six years of data and came up empty: This new FRB appears to be a one-off, at least in the last six years. The findings are particularly exciting, given the burst’s proximity. Because it is so close and so bright, scientists can probe the environment in and around the burst for clues to what might produce a nonrepeating FRB.

“Right now we’re in the middle of this story of whether repeating and nonrepeating FRBs are different. These observations are putting together bits and pieces of the puzzle,” Masui says.

“There’s evidence to suggest that not all FRB progenitors are the same,” Andrew adds. “We’re on track to localize hundreds of FRBs every year. The hope is that a larger sample of FRBs localized to their host environments can help reveal the full diversity of these populations.”

The construction of the CHIME Outriggers was funded by the Gordon and Betty Moore Foundation and the U.S. National Science Foundation. The construction of CHIME was funded by the Canada Foundation for Innovation and provinces of Quebec, Ontario, and British Columbia.

Rising global temperatures affect human activity in many ways. Now, a new study illuminates an important dimension of the problem: Very hot days are associated with more negative moods, as shown by a large-scale look at social media postings.

Overall, the study examines 1.2 billion social media posts from 157 countries over the span of a year. The research finds that when the temperature rises above 95 degrees Fahrenheit, or 35 degrees Celsius, expressed sentiments become about 25 percent more negative in lower-income countries and about 8 percent more negative in better-off countries. Extreme heat affects people emotionally, not just physically.

“Our study reveals that rising temperatures don’t just threaten physical health or economic productivity — they also affect how people feel, every day, all over the world,” says Siqi Zheng, a professor in MIT’s Department of Urban Studies and Planning (DUSP) and Center for Real Estate (CRE), and co-author of a new paper detailing the results. “This work opens up a new frontier in understanding how climate stress is shaping human well-being at a planetary scale.”

The paper, “Unequal Impacts of Rising Temperatures on Global Human Sentiment,” is published today in the journal One Earth. The authors are Jianghao Wang, of the Chinese Academy of Sciences; Nicolas Guetta-Jeanrenaud SM ’22, a graduate of MIT’s Technology and Policy Program (TPP) and Institute for Data, Systems, and Society; Juan Palacios, a visiting assistant professor at MIT’s Sustainable Urbanization Lab (SUL) and an assistant professor Maastricht University; Yichun Fan, of SUL and Duke University; Devika Kakkar, of Harvard University; Nick Obradovich, of SUL and the Laureate Institute for Brain Research in Tulsa; and Zheng, who is the STL Champion Professor of Urban and Real Estate Sustainability at CRE and DUSP. Zheng is also the faculty director of CRE and founded the Sustainable Urbanization Lab in 2019.

Social media as a window

To conduct the study, the researchers evaluated 1.2 billion posts from the social media platforms Twitter and Weibo, all of which appeared in 2019. They used a natural language processing technique called Bidirectional Encoder Representations from Transformers (BERT), to analyze 65 languages across the 157 countries in the study.

Each social media post was given a sentiment rating from 0.0 (for very negative posts) to 1.0 (for very positive posts). The posts were then aggregated geographically to 2,988 locations and evaluated in correlation with area weather. From this method, the researchers could then deduce the connection between extreme temperatures and expressed sentiment.

“Social media data provides us with an unprecedented window into human emotions across cultures and continents,” Wang says. “This approach allows us to measure emotional impacts of climate change at a scale that traditional surveys simply cannot achieve, giving us real-time insights into how temperature affects human sentiment worldwide.”

To assess the effects of temperatures on sentiment in higher-income and middle-to-lower-income settings, the scholars also used a World Bank cutoff level of gross national income per-capita annual income of $13,845, finding that in places with incomes below that, the effects of heat on mood were triple those found in economically more robust settings.

“Thanks to the global coverage of our data, we find that people in low- and middle-income countries experience sentiment declines from extreme heat that are three times greater than those in high-income countries,” Fan says. “This underscores the importance of incorporating adaptation into future climate impact projections.”

In the long run

Using long-term global climate models, and expecting some adaptation to heat, the researchers also produced a long-range estimate of the effects of extreme temperatures on sentiment by the year 2100. Extending the current findings to that time frame, they project a 2.3 percent worsening of people’s emotional well-being based on high temperatures alone by then — although that is a far-range projection.

“It's clear now, with our present study adding to findings from prior studies, that weather alters sentiment on a global scale,” Obradovich says. “And as weather and climates change, helping individuals become more resilient to shocks to their emotional states will be an important component of overall societal adaptation.”

The researchers note that there are many nuances to the subject, and room for continued research in this area. For one thing, social media users are not likely to be a perfectly representative portion of the population, with young children and the elderly almost certainly using social media less than other people. However, as the researchers observe in the paper, the very young and elderly are probably particularly vulnerable to heat shocks, making the response to hot weather possible even larger than their study can capture.

The research is part of the Global Sentiment project led by the MIT Sustainable Urbanization Lab, and the study’s dataset is publicly available. Zheng and other co-authors have previously investigated these dynamics using social media, although never before at this scale.

“We hope this resource helps researchers, policymakers, and communities better prepare for a warming world,” Zheng says.

The research was supported, in part, by Zheng’s chaired professorship research fund, and grants Wang received from the National Natural Science Foundation of China and the Chinese Academy of Sciences. 

A multiyear program at MIT Lincoln Laboratory to characterize how biological and chemical vapors and aerosols disperse through the New York City subway system is coming to a close. The program, part of the U.S. Department of Homeland Security (DHS) Science and Technology Directorate's Urban Area Security Initiative, builds on other efforts at Lincoln Laboratory to detect chemical and biological threats, validate air dispersion models, and improve emergency protocols in urban areas in case of an airborne attack. The results of this program will inform the New York Metropolitan Transportation Authority (MTA) on how best to install an efficient, cost-effective system for airborne threat detection and mitigation throughout the subway. On a broader scale, the study will help the national security community understand pragmatic chemical and biological defense options for mass transit, critical facilities, and special events.

Trina Vian from the laboratory's Counter–Weapons of Mass Destruction (WMD) Systems Group led this project, which she says had as much to do with air flow and sensors as it did with MTA protocols and NYC commuters. "There are real dangers associated with panic during an alarm. People can get hurt during mass evacuation, or lose trust in a system and the authorities that administer that system, if there are false alarms," she says. "A novel aspect of our project was to investigate effective low-regret response options, meaning those with little operational consequence to responding to a false alarm."

Currently, depending on the severity of the alarm, the MTA's response can include stopping service and evacuating passengers and employees.

A complex environment for testing

For the program, which started in 2019, Vian and her team collected data on how chemical and biological sensors performed in the subway, what factors affected sensor accuracy, and how different mitigation protocols fared in stopping an airborne threat from spreading and removing the threat from a contaminated location. For their tests, they released batches of a safe, custom-developed aerosol simulant within Grand Central Station that they could track with DNA barcodes. Each batch had a different barcode, which allowed the team to differentiate among them and quantitatively assess different combinations of mitigation strategies.

To control and isolate air flow, the team tested static air curtains as well as air filtration systems. They also tested a spray knockdown system developed by Sandia National Laboratories designed to reduce and isolate particulate hazards in large volume areas. The system sprays a fine water mist into the tunnels that attaches to threat particulates and uses gravity to rain out the threat material. The spray contains droplets of a particular size and concentration, and with an applied electrostatic field. The original idea for the system was adapted from the coal mining industry, which used liquid sprayers to reduce the amount of inhalable soot.

The tests were done in a busy environment, and the team was required to complete trainings on MTA protocols such as track safety and how to interact with the public.

"We had long and sometimes very dirty days," says Jason Han of the Counter–WMD Systems Group, who collected measurements in the tunnels and analyzed the data. "We all wore bright orange contractor safety vests, which made people think we were official employees of the MTA. We would often get approached by people asking for directions!"

At times, issues such as power outages or database errors could disrupt data capture.

"We learned fairly early on that we had to capture daily data backups and keep a daily evolving master list of unique sensor identifiers and locations," says fellow team member Cassie Smith. "We developed workflows and wrote scripts to help automate the process, which ensured successful sensor data capture and attribution."

The team also worked closely with the MTA to make sure their tests and data capture ran smoothly. "The MTA was great at helping us maintain the test bed, doing as much as they could in our physical absence," Vian says.

Calling on industry

Another crucial aspect of the program was to connect with the greater chemical and biological industrial community to solicit their sensors for testing. These partnerships reduced the cost for DHS to bring new sensing technologies into the project, and, in return, participants gained a testing and data collection opportunity within the challenging NYC subway environment.

The team ultimately fielded 16 different sensors, each with varying degrees of maturity, that operated through a range of methods, such as ultraviolet laser–induced fluorescence, polymerase chain reaction, and long-wave infrared spectrometry.

"The partners appreciated the unique data they got and the opportunity to work with the MTA and experience an environment and customer base that they may not have anticipated before," Vian says.

The team finished testing in 2024 and has delivered the final report to the DHS. The MTA will use the report to help expand their PROTECT chemical detection system (originally developed by Argonne National Laboratory) from Grand Central Station into adjacent stations. They expect to complete this work in 2026.

"The value of this program cannot be overstated. This partnership with DHS and MIT Lincoln Laboratory has led to the identification of the best-suited systems for the MTA’s unique operating environment," says Michael Gemelli, director of chemical, biological, radiological, and nuclear/WMD detection and mitigation at the New York MTA.

"Other transit authorities can leverage these results to start building effective chemical and biological defense systems for their own specific spaces and threat priorities," adds Benjamin Ervin, leader of Lincoln Laboratory's Counter–WMD Systems Group. "Specific test and evaluation within the operational environment of interest, however, is always recommended to ensure defense system objectives are met."

Building these types of decision-making reports for airborne chemical and biological sensing has been a part of Lincoln Laboratory's mission since the mid-1990s. The laboratory also helped to define priorities in the field when DHS was forming in the early 2000s.

Beyond this study, the Lincoln Laboratory is leading several other projects focused on forecasting the impact of novel chemical and biological threats within multiple domains — military, space, agriculture, health, etc. — and on prototyping rapid, autonomous, high-confidence biological identification capabilities for the homeland to provide actionable evidence of hazardous environments.

From toddlers’ timeouts to criminals’ prison sentences, punishment reinforces social norms, making it known that an offender has done something unacceptable. At least, that is usually the intent — but the strategy can backfire. When a punishment is perceived as too harsh, observers can be left with the impression that an authority figure is motivated by something other than justice.

It can be hard to predict what people will take away from a particular punishment, because everyone makes their own inferences not just about the acceptability of the act that led to the punishment, but also the legitimacy of the authority who imposed it. A new computational model developed by scientists at MIT’s McGovern Institute for Brain Research makes sense of these complicated cognitive processes, recreating the ways people learn from punishment and revealing how their reasoning is shaped by their prior beliefs.

Their work, reported Aug. 4 in the journal PNAS, explains how a single punishment can send different messages to different people, and even strengthen the opposing viewpoints of groups who hold different opinions about authorities or social norms.

“The key intuition in this model is the fact that you have to be evaluating simultaneously both the norm to be learned and the authority who’s punishing,” says McGovern investigator and John W. Jarve Professor of Brain and Cognitive Sciences Rebecca Saxe, who led the research. “One really important consequence of that is even where nobody disagrees about the facts — everybody knows what action happened, who punished it, and what they did to punish it — different observers of the same situation could come to different conclusions.”

For example, she says, a child who is sent to timeout after biting a sibling might interpret the event differently than the parent. One might see the punishment as proportional and important, teaching the child not to bite. But if the biting, to the toddler, seemed a reasonable tactic in the midst of a squabble, the punishment might be seen as unfair, and the lesson will be lost.

People draw on their own knowledge and opinions when they evaluate these situations — but to study how the brain interprets punishment, Saxe and graduate student Setayesh Radkani wanted to take those personal ideas out of the equation. They needed a clear understanding of the beliefs that people held when they observed a punishment, so they could learn how different kinds of information altered their perceptions. So Radkani set up scenarios in imaginary villages where authorities punished individuals for actions that had no obvious analog in the real world.

Participants observed these scenarios in a series of experiments, with different information offered in each one. In some cases, for example, participants were told that the person being punished was either an ally or a competitor of the authority, whereas in other cases, the authority’s possible bias was left ambiguous.

“That gives us a really controlled setup to vary prior beliefs,” Radkani explains. “We could ask what people learn from observing punitive decisions with different severities, in response to acts that vary in their level of wrongness, by authorities that vary in their level of different motives.”

For each scenario, participants were asked to evaluate four factors: how much the authority figure cared about justice; the selfishness of the authority; the authority’s bias for or against the individual being punished; and the wrongness of the punished act. The research team asked these questions when participants were first introduced to the hypothetical society, then tracked how their responses changed after they observed the punishment. Across the scenarios, participants’ initial beliefs about the authority and the wrongness of the act shaped the extent to which those beliefs shifted after they observed the punishment.

Radkani was able to replicate these nuanced interpretations using a cognitive model framed around an idea that Saxe’s team has long used to think about how people interpret the actions of others. That is, to make inferences about others’ intentions and beliefs, we assume that people choose actions that they expect will help them achieve their goals.

To apply that concept to the punishment scenarios, Radkani developed a model that evaluates the meaning of a punishment (an action aimed at achieving a goal of the authority) by considering the harm associated with that punishment; its costs or benefits to the authority; and its proportionality to the violation. By assessing these factors, along with prior beliefs about the authority and the punished act, the model was able to predict people’s responses to the hypothetical punishment scenarios, supporting the idea that people use a similar mental model. “You need to have them consider those things, or you can’t make sense of how people understand punishment when they observe it,” Saxe says.

Even though the team designed their experiments to preclude preconceived ideas about the people and actions in their imaginary villages, not everyone drew the same conclusions from the punishments they observed. Saxe’s group found that participants’ general attitudes toward authority influenced their interpretation of events. Those with more authoritarian attitudes — assessed through a standard survey — tended to judge punished acts as more wrong and authorities as more motivated by justice than other observers.

“If we differ from other people, there’s a knee-jerk tendency to say, ‘either they have different evidence from us, or they’re crazy,’” Saxe says. Instead, she says, “It’s part of the way humans think about each other’s actions.”

“When a group of people who start out with different prior beliefs get shared evidence, they will not end up necessarily with shared beliefs. That’s true even if everybody is behaving rationally,” says Saxe.

This way of thinking also means that the same action can simultaneously strengthen opposing viewpoints. The Saxe lab’s modeling and experiments showed that when those viewpoints shape individuals’ interpretations of future punishments, the groups’ opinions will continue to diverge. For instance, a punishment that seems too harsh to a group who suspects an authority is biased can make that group even more skeptical of the authority’s future actions. Meanwhile, people who see the same punishment as fair and the authority as just will be more likely to conclude that the authority figure’s future actions are also just. 

“You will get a vicious cycle of polarization, staying and actually spreading to new things,” says Radkani.

The researchers say their findings point toward strategies for communicating social norms through punishment. “It is exactly sensible in our model to do everything you can to make your action look like it’s coming out of a place of care for the long-term outcome of this individual, and that it’s proportional to the norm violation they did,” Saxe says. “That is your best shot at getting a punishment interpreted pedagogically, rather than as evidence that you’re a bully.”

Nevertheless, she says that won’t always be enough. “If the beliefs are strong the other way, it’s very hard to punish and still sustain a belief that you were motivated by justice.”

Joining Saxe and Radkani on the paper is Joshua Tenenbaum, MIT professor of brain and cognitive sciences. The study was funded, in part, by the Patrick J McGovern Foundation.

The U.S. Food and Drug Administration’s recent approval of the first CRISPR-Cas9–based gene therapy has marked a major milestone in biomedicine, validating genome editing as a promising treatment strategy for disorders like sickle cell disease, muscular dystrophy, and certain cancers.

CRISPR-Cas9, often likened to “molecular scissors,” allows scientists to cut DNA at targeted sites to snip, repair, or replace genes. But despite its power, Cas9 poses a critical safety risk: The active enzyme can linger in cells and cause unintended DNA breaks — so-called off-target effects — which may trigger harmful mutations in healthy genes.

Now, researchers in the labs of Ronald T. Raines, MIT professor of chemistry, and Amit Choudhary, professor of medicine at Harvard Medical School, have engineered a precise way to turn Cas9 off after its job is done — significantly reducing off-target effects and improving the clinical safety of gene editing. Their findings are detailed in a new paper published in the Proceedings of the National Academy of Sciences (PNAS).

“To ‘turn off’ Cas9 after it achieves its intended genome-editing outcome, we developed the first cell-permeable anti-CRISPR protein system,” says Raines, the Roger and Georges Firmenich Professor of Natural Products Chemistry. “Our technology reduces the off-target activity of Cas9 and increases its genome-editing specificity and clinical utility.”

The new tool — called LF_N-Acr/PA — uses a protein-based delivery system to ferry anti-CRISPR proteins into human cells rapidly and efficiently. While natural Type II anti-CRISPR proteins (Acrs) are known to inhibit Cas9, their use in therapy has been limited because they’re often too bulky or charged to enter cells, and conventional delivery methods are too slow or ineffective.

LF_N-Acr/PA overcomes these hurdles using a component derived from anthrax toxin to introduce Acrs into cells within minutes. Even at picomolar concentrations, the system shuts down Cas9 activity with remarkable speed and precision — boosting genome-editing specificity up to 40 percent.

Bradley L. Pentelute, MIT professor of chemistry, is an expert on the anthrax delivery system, and is also an author of the paper.

The implications of this advance are wide-ranging. With patent applications filed, LF_N-Acr/PA represents a faster, safer, and more controllable means of harnessing CRISPR-Cas9, opening the door to more-refined gene therapies with fewer unintended consequences.

The research was supported by the National Institutes of Health and a Gilliam Fellowship from the Howard Hughes Medical Institute awarded to lead author Axel O. Vera, a graduate student in the Department of Chemistry.

Materials research thrives across MIT, spanning disciplines and departments. Recent breakthroughs include strategies for securing sustainable supplies of nickel — critical to clean-energy technologies (Department of Materials Science and Engineering); the discovery of unexpected magnetism in atomically thin quantum materials (Department of Physics); and the development of adhesive coatings that reduce scarring around medical implants (departments of Mechanical Engineering and Civil and Environmental Engineering).

At the center of these efforts is the Materials Research Laboratory (MRL), a hub that connects and supports the Institute’s materials research community. “MRL serves as a home for the entire materials research community at MIT,” says C. Cem Tasan, who became director in April 2025. “Our goal is to make it easier for our faculty to conduct their extraordinary research,” adds Tasan, the POSCO Associate Professor of Metallurgy in the Department of Materials Science and Engineering.

A storied history

Established in 2017, the MRL brings together more than 30 researchers and builds on a 48-year legacy of innovation. It was formed through the merger of the MIT Materials Processing Center (MPC) and the Center for Materials Science and Engineering (CMSE), two institutions that helped lay the foundation for MIT’s global leadership in materials science.

Over the years, research supported by MPC and CMSE has led to transformative technologies and successful spinout companies. Notable examples include amsc, based on advances in superconductivity; OmniGuide, which developed cutting-edge optical fiber technologies; and QD Vision, a pioneer in quantum dot technology acquired by Samsung in 2016. Another landmark achievement was the development of the first germanium laser to operate at room temperature — a breakthrough now used in optical communications.

Enabling research through partnership and support

MRL is launching targeted initiatives to connect MIT researchers with industry partners around specific technical challenges. Each initiative will be led by a junior faculty member working closely with MRL to identify a problem that aligns with their research expertise and is relevant to industry needs.

Through multi-year collaborations with participating companies, faculty can explore early-stage solutions in partnership with postdocs or graduate students. These initiatives are designed to be agile and interdisciplinary, with the potential to grow into major, long-term research programs.

Behind-the-scenes support, front-line impact

MRL provides critical infrastructure that enables faculty to focus on discovery, not logistics. “MRL works silently in the background, where every problem a principal investigator has related to the administration of materials research is solved with efficiency, good organization, and minimum effort,” says Tasan.

This quiet but powerful support spans multiple areas:

• The finance team manages grants and helps secure new funding opportunities.
• The human resources team supports the hiring of postdocs.
• The communications team amplifies the lab’s impact through compelling stories shared with the public and funding agencies.
• The events team plans and coordinates conferences, seminars, and symposia that foster collaboration within the MIT community and with external partners.

Together, these functions ensure that research at MRL runs smoothly and effectively — from initial idea to lasting innovation.

Leadership with a vision

Tasan, who also leads a research group focused on metallurgy, says he took on the directorship because “I thrive on new challenges.” He also saw the role as an opportunity to contribute more broadly to MIT. 

“I believe MRL can play an even greater role in advancing materials research across the Institute, and I’m excited to help make that happen,” he says.

Optical frequency combs are specially designed lasers that act like rulers to accurately and rapidly measure specific frequencies of light. They can be used to detect and identify chemicals and pollutants with extremely high precision.

Frequency combs would be ideal for remote sensors or portable spectrometers because they can enable accurate, real-time monitoring of multiple chemicals without complex moving parts or external equipment.

But developing frequency combs with high enough bandwidth for these applications has been a challenge. Often, researchers must add bulky components that limit scalability and performance.

Now, a team of MIT researchers has demonstrated a compact, fully integrated device that uses a carefully crafted mirror to generate a stable frequency comb with very broad bandwidth. The mirror they developed, along with an on-chip measurement platform, offers the scalability and flexibility needed for mass-producible remote sensors and portable spectrometers. This development could enable more accurate environmental monitors that can identify multiple harmful chemicals from trace gases in the atmosphere.

“The broader the bandwidth a spectrometer has, the more powerful it is, but dispersion is in the way. Here we took the hardest problem that limits bandwidth and made it the centerpiece of our study, addressing every step to ensure robust frequency comb operation,” says Qing Hu, Distinguished Professor in Electrical Engineering and Computer Science at MIT, principal investigator in the Research Laboratory of Electronics, and senior author on an open-access paper describing the work.

He is joined on the paper by lead author Tianyi Zeng PhD ’23; as well as Yamac Dikmelik of General Dynamics Mission Systems; Feng Xie and Kevin Lascola of Thorlabs Quantum Electronics; and David Burghoff SM ’09, PhD ’14, an assistant professor at the University of Texas at Austin. The research appears today in Light: Science and Applications.

Broadband combs

An optical frequency comb produces a spectrum of equally spaced laser lines, which resemble the teeth of a comb.

Scientists can generate frequency combs using several types of lasers for different wavelengths. By using a laser that produces long wave infrared radiation, such as a quantum cascade laser, they can use frequency combs for high-resolution sensing and spectroscopy.

In dual-comb spectroscopy (DCS), the beam of one frequency comb travels straight through the system and strikes a detector at the other end. The beam of the second frequency comb passes through a chemical sample before striking the same detector. Using the results from both combs, scientists can faithfully replicate the chemical features of the sample at much lower frequencies, where signals can be easily analyzed.

The frequency combs must have high bandwidth, or they will only be able to detect a small frequency range of chemical compounds, which could lead to false alarms or inaccurate results.

Dispersion is the most important factor that limits a frequency comb’s bandwidth. If there is dispersion, the laser lines are not evenly spaced, which is incompatible with the formation of frequency combs.

“With long wave infrared radiation, the dispersion will be very high. There is no way to get around it, so we have to find a way to compensate for it or counteract it by engineering our system,” Hu says.

Many existing approaches aren’t flexible enough to be used in different scenarios or don’t enable high enough bandwidth.

Hu’s group previously solved this problem in a different type of frequency comb, one that used terahertz waves, by developing a double-chirped mirror (DCM).

A DCM is a special type of optical mirror that has multiple layers with thicknesses that change gradually from one end to the other. They found that this DCM, which has a corrugated structure, could effectively compensate for dispersion when used with a terahertz laser.

“We tried to borrow this trick and apply it to an infrared comb, but we ran into lots of challenges,” Hu says.

Because infrared waves are 10 times shorter than terahertz waves, fabricating the new mirror required an extreme level of precision. At the same time, they needed to coat the entire DCM in a thick layer of gold to remove the heat under laser operation. Plus, their dispersion measurement system, designed for terahertz waves, wouldn’t work with infrared waves, which have frequencies that are about 10 times higher than terahertz.

“After more than two years of trying to implement this scheme, we reached a dead end,” Hu says.

A new solution

Ready to throw in the towel, the team realized something they had missed. They had designed the mirror with corrugation to compensate for the lossy terahertz laser, but infrared radiation sources aren’t as lossy.

This meant they could use a standard DCM design to compensate for dispersion, which is compatible with infrared radiation. However, they still needed to create curved mirror layers to capture the beam of the laser, which made fabrication much more difficult than usual.

“The adjacent layers of mirror differ only by tens of nanometers. That level of precision precludes standard photolithography techniques. On top of that, we still had to etch very deeply into the notoriously stubborn material stacks. Achieving those critical dimensions and etch depths was key to unlocking broadband comb performance,” Zeng says. In addition to precisely fabricating the DCM, they integrated the mirror directly onto the laser, making the device extremely compact. The team also developed a high-resolution, on-chip dispersion measurement platform that doesn’t require bulky external equipment.

“Our approach is flexible. As long as we can use our platform to measure the dispersion, we can design and fabricate a DCM that compensates for it,” Hu adds.

Taken together, the DCM and on-chip measurement platform enabled the team to generate stable infrared laser frequency combs that had far greater bandwidth than can usually be achieved without a DCM.

In the future, the researchers want to extend their approach to other laser platforms that could generate combs with even greater bandwidth and higher power for more demanding applications.

“These researchers developed an ingenious nanophotonic dispersion compensation scheme based on an integrated air–dielectric double-chirped mirror. This approach provides unprecedented control over dispersion, enabling broadband comb formation at room temperature in the long-wave infrared. Their work opens the door to practical, chip-scale frequency combs for applications ranging from chemical sensing to free-space communications,” says Jacob B. Khurgin, a professor at the Johns Hopkins University Whiting School of Engineering, who was not involved with this paper.

This work is funded, in part, by the U.S. Defense Advanced Research Projects Agency (DARPA) and the Gordon and Betty Moore Foundation. This work was carried out, in part, using facilities at MIT.nano.

Using machine learning, MIT chemical engineers have created a computational model that can predict how well any given molecule will dissolve in an organic solvent — a key step in the synthesis of nearly any pharmaceutical. This type of prediction could make it much easier to develop new ways to produce drugs and other useful molecules.

The new model, which predicts how much of a solute will dissolve in a particular solvent, should help chemists to choose the right solvent for any given reaction in their synthesis, the researchers say. Common organic solvents include ethanol and acetone, and there are hundreds of others that can also be used in chemical reactions.

“Predicting solubility really is a rate-limiting step in synthetic planning and manufacturing of chemicals, especially drugs, so there’s been a longstanding interest in being able to make better predictions of solubility,” says Lucas Attia, an MIT graduate student and one of the lead authors of the new study.

The researchers have made their model freely available, and many companies and labs have already started using it. The model could be particularly useful for identifying solvents that are less hazardous than some of the most commonly used industrial solvents, the researchers say.

“There are some solvents which are known to dissolve most things. They’re really useful, but they’re damaging to the environment, and they’re damaging to people, so many companies require that you have to minimize the amount of those solvents that you use,” says Jackson Burns, an MIT graduate student who is also a lead author of the paper. “Our model is extremely useful in being able to identify the next-best solvent, which is hopefully much less damaging to the environment.”

William Green, the Hoyt Hottel Professor of Chemical Engineering and director of the MIT Energy Initiative, is the senior author of the study, which appears today in Nature Communications. Patrick Doyle, the Robert T. Haslam Professor of Chemical Engineering, is also an author of the paper.

Solving solubility

The new model grew out of a project that Attia and Burns worked on together in an MIT course on applying machine learning to chemical engineering problems. Traditionally, chemists have predicted solubility with a tool known as the Abraham Solvation Model, which can be used to estimate a molecule’s overall solubility by adding up the contributions of chemical structures within the molecule. While these predictions are useful, their accuracy is limited.

In the past few years, researchers have begun using machine learning to try to make more accurate solubility predictions. Before Burns and Attia began working on their new model, the state-of-the-art model for predicting solubility was a model developed in Green’s lab in 2022.

That model, known as SolProp, works by predicting a set of related properties and combining them, using thermodynamics, to ultimately predict the solubility. However, the model has difficulty predicting solubility for solutes that it hasn’t seen before.

“For drug and chemical discovery pipelines where you’re developing a new molecule, you want to be able to predict ahead of time what its solubility looks like,” Attia says.

Part of the reason that existing solubility models haven’t worked well is because there wasn’t a comprehensive dataset to train them on. However, in 2023 a new dataset called BigSolDB was released, which compiled data from nearly 800 published papers, including information on solubility for about 800 molecules dissolved about more than 100 organic solvents that are commonly used in synthetic chemistry.

Attia and Burns decided to try training two different types of models on this data. Both of these models represent the chemical structures of molecules using numerical representations known as embeddings, which incorporate information such as the number of atoms in a molecule and which atoms are bound to which other atoms. Models can then use these representations to predict a variety of chemical properties.

One of the models used in this study, known as FastProp and developed by Burns and others in Green’s lab, incorporates “static embeddings.” This means that the model already knows the embedding for each molecule before it starts doing any kind of analysis.

The other model, ChemProp, learns an embedding for each molecule during the training, at the same time that it learns to associate the features of the embedding with a trait such as solubility. This model, developed across multiple MIT labs, has already been used for tasks such as antibiotic discovery, lipid nanoparticle design, and predicting chemical reaction rates.

The researchers trained both types of models on over 40,000 data points from BigSolDB, including information on the effects of temperature, which plays a significant role in solubility. Then, they tested the models on about 1,000 solutes that had been withheld from the training data. They found that the models’ predictions were two to three times more accurate than those of SolProp, the previous best model, and the new models were especially accurate at predicting variations in solubility due to temperature.

“Being able to accurately reproduce those small variations in solubility due to temperature, even when the overarching experimental noise is very large, was a really positive sign that the network had correctly learned an underlying solubility prediction function,” Burns says.

Accurate predictions

The researchers had expected that the model based on ChemProp, which is able to learn new representations as it goes along, would be able to make more accurate predictions. However, to their surprise, they found that the two models performed essentially the same. That suggests that the main limitation on their performance is the quality of the data, and that the models are performing as well as theoretically possible based on the data that they’re using, the researchers say.

“ChemProp should always outperform any static embedding when you have sufficient data,” Burns says. “We were blown away to see that the static and learned embeddings were statistically indistinguishable in performance across all the different subsets, which indicates to us that that the data limitations that are present in this space dominated the model performance.”

The models could become more accurate, the researchers say, if better training and testing data were available — ideally, data obtained by one person or a group of people all trained to perform the experiments the same way.

“One of the big limitations of using these kinds of compiled datasets is that different labs use different methods and experimental conditions when they perform solubility tests. That contributes to this variability between different datasets,” Attia says.

Because the model based on FastProp makes its predictions faster and has code that is easier for other users to adapt, the researchers decided to make that one, known as FastSolv, available to the public. Multiple pharmaceutical companies have already begun using it.

“There are applications throughout the drug discovery pipeline,” Burns says. “We’re also excited to see, outside of formulation and drug discovery, where people may use this model.”

The research was funded, in part, by the U.S. Department of Energy.

Within the past few years, models that can predict the structure or function of proteins have been widely used for a variety of biological applications, such as identifying drug targets and designing new therapeutic antibodies.

These models, which are based on large language models (LLMs), can make very accurate predictions of a protein’s suitability for a given application. However, there’s no way to determine how these models make their predictions or which protein features play the most important role in those decisions.

In a new study, MIT researchers have used a novel technique to open up that “black box” and allow them to determine what features a protein language model takes into account when making predictions. Understanding what is happening inside that black box could help researchers to choose better models for a particular task, helping to streamline the process of identifying new drugs or vaccine targets.

“Our work has broad implications for enhanced explainability in downstream tasks that rely on these representations,” says Bonnie Berger, the Simons Professor of Mathematics, head of the Computation and Biology group in MIT’s Computer Science and Artificial Intelligence Laboratory, and the senior author of the study. “Additionally, identifying features that protein language models track has the potential to reveal novel biological insights from these representations.”

Onkar Gujral, an MIT graduate student, is the lead author of the open-access study, which appears this week in the Proceedings of the National Academy of Sciences. Mihir Bafna, an MIT graduate student in electrical engineering and computer science, and Eric Alm, an MIT professor of biological engineering, are also authors of the paper.

Opening the black box

In 2018, Berger and former MIT graduate student Tristan Bepler PhD ’20 introduced the first protein language model. Their model, like subsequent protein models that accelerated the development of AlphaFold, such as ESM2 and OmegaFold, was based on LLMs. These models, which include ChatGPT, can analyze huge amounts of text and figure out which words are most likely to appear together.

Protein language models use a similar approach, but instead of analyzing words, they analyze amino acid sequences. Researchers have used these models to predict the structure and function of proteins, and for applications such as identifying proteins that might bind to particular drugs.

In a 2021 study, Berger and colleagues used a protein language model to predict which sections of viral surface proteins are less likely to mutate in a way that enables viral escape. This allowed them to identify possible targets for vaccines against influenza, HIV, and SARS-CoV-2.

However, in all of these studies, it has been impossible to know how the models were making their predictions.

“We would get out some prediction at the end, but we had absolutely no idea what was happening in the individual components of this black box,” Berger says.

In the new study, the researchers wanted to dig into how protein language models make their predictions. Just like LLMs, protein language models encode information as representations that consist of a pattern of activation of different “nodes” within a neural network. These nodes are analogous to the networks of neurons that store memories and other information within the brain.

The inner workings of LLMs are not easy to interpret, but within the past couple of years, researchers have begun using a type of algorithm known as a sparse autoencoder to help shed some light on how those models make their predictions. The new study from Berger’s lab is the first to use this algorithm on protein language models.

Sparse autoencoders work by adjusting how a protein is represented within a neural network. Typically, a given protein will be represented by a pattern of activation of a constrained number of neurons, for example, 480. A sparse autoencoder will expand that representation into a much larger number of nodes, say 20,000.

When information about a protein is encoded by only 480 neurons, each node lights up for multiple features, making it very difficult to know what features each node is encoding. However, when the neural network is expanded to 20,000 nodes, this extra space along with a sparsity constraint gives the information room to “spread out.” Now, a feature of the protein that was previously encoded by multiple nodes can occupy a single node.

“In a sparse representation, the neurons lighting up are doing so in a more meaningful manner,” Gujral says. “Before the sparse representations are created, the networks pack information so tightly together that it's hard to interpret the neurons.”

Interpretable models

Once the researchers obtained sparse representations of many proteins, they used an AI assistant called Claude (related to the popular Anthropic chatbot of the same name), to analyze the representations. In this case, they asked Claude to compare the sparse representations with the known features of each protein, such as molecular function, protein family, or location within a cell.

By analyzing thousands of representations, Claude can determine which nodes correspond to specific protein features, then describe them in plain English. For example, the algorithm might say, “This neuron appears to be detecting proteins involved in transmembrane transport of ions or amino acids, particularly those located in the plasma membrane.”

This process makes the nodes far more “interpretable,” meaning the researchers can tell what each node is encoding. They found that the features most likely to be encoded by these nodes were protein family and certain functions, including several different metabolic and biosynthetic processes.

“When you train a sparse autoencoder, you aren’t training it to be interpretable, but it turns out that by incentivizing the representation to be really sparse, that ends up resulting in interpretability,” Gujral says.

Understanding what features a particular protein model is encoding could help researchers choose the right model for a particular task, or tweak the type of input they give the model, to generate the best results. Additionally, analyzing the features that a model encodes could one day help biologists to learn more about the proteins that they are studying.

“At some point when the models get a lot more powerful, you could learn more biology than you already know, from opening up the models,” Gujral says.

The research was funded by the National Institutes of Health. 

MIT researchers have developed a reconfigurable antenna that dynamically adjusts its frequency range by changing its physical shape, making it more versatile for communications and sensing than static antennas.

A user can stretch, bend, or compress the antenna to make reversible changes to its radiation properties, enabling a device to operate in a wider frequency range without the need for complex, moving parts. With an adjustable frequency range, a reconfigurable antenna could adapt to changing environmental conditions and reduce the need for multiple antennas.

The word “antenna” may draw to mind metal rods like the “bunny ears” on top of old television sets, but the MIT team instead worked with metamaterials — engineered materials whose mechanical properties, such as stiffness and strength, depend on the geometric arrangement of the material’s components.

The result is a simplified design for a reconfigurable antenna that could be used for applications like energy transfer in wearable devices, motion tracking and sensing for augmented reality, or wireless communication across a wide range of network protocols.

In addition, the researchers developed an editing tool so users can generate customized metamaterial antennas, which can be fabricated using a laser cutter.

“Usually, when we think of antennas, we think of static antennas — they are fabricated to have specific properties and that is it. However, by using auxetic metamaterials, which can deform into three different geometric states, we can seamlessly change the properties of the antenna by changing its geometry, without fabricating a new structure. In addition, we can use changes in the antenna’s radio frequency properties, due to changes in the metamaterial geometry, as a new method of sensing for interaction design,” says lead author Marwa AlAlawi, a mechanical engineering graduate student at MIT.

Her co-authors include Regina Zheng and Katherine Yan, both MIT undergraduate students; Ticha Sethapakdi, an MIT graduate student in electrical engineering and computer science; Soo Yeon Ahn of the Gwangju Institute of Science and Technology in Korea; and co-senior authors Junyi Zhu, assistant professor at the University of Michigan; and Stefanie Mueller, the TIBCO Career Development Associate Professor in MIT’s departments of Electrical Engineering and Computer Science and Mechanical Engineering and leader of the Human-Computer Interaction Group at the Computer Science and Artificial Intelligence Lab. The research will be presented at the ACM Symposium on User Interface Software and Technology.

Making sense of antennas

While traditional antennas radiate and receive radio signals, in this work, the researchers looked at how the devices can act as sensors. The team’s goal was to develop a mechanical element that can also be used as an antenna for sensing.

To do this, they leveraged the antenna’s “resonance frequency,” which is the frequency at which the antenna is most efficient.

An antenna’s resonance frequency will shift due to changes in its shape. (Think about extending the left “bunny ear” to reduce TV static.) Researchers can capture these shifts for sensing. For instance, a reconfigurable antenna could be used in this way to detect the expansion of a person’s chest, to monitor their respiration.

To design a versatile reconfigurable antenna, the researchers used metamaterials. These engineered materials, which can be programmed to adopt different shapes, are composed of a periodic arrangement of unit cells that can be rotated, compressed, stretched, or bent.

By deforming the metamaterial structure, one can shift the antenna’s resonance frequency.

“In order to trigger changes in resonance frequency, we either need to change the antenna’s effective length or introduce slits and holes into it. Metamaterials allow us to get those different states from only one structure,” AlAlawi says.

The device, dubbed the meta-antenna, is composed of a dielectric layer of material sandwiched between two conductive layers.

To fabricate a meta-antenna, the researchers cut the dielectric laser out of a rubber sheet with a laser cutter. Then they added a patch on top of the dielectric layer using conductive spray paint, creating a resonating “patch antenna.”

But they found that even the most flexible conductive material couldn’t withstand the amount of deformation the antenna would experience.

“We did a lot of trial and error to determine that, if we coat the structure with flexible acrylic paint, it protects the hinges so they don’t break prematurely,” AlAlawi explains.

A means for makers

With the fabrication problem solved, the researchers built a tool that enables users to design and produce metamaterial antennas for specific applications.

The user can define the size of the antenna patch, choose a thickness for the dielectric layer, and set the length to width ratio of the metamaterial unit cells. Then the system automatically simulates the antenna’s resonance frequency range.

“The beauty of metamaterials is that, because it is an interconnected system of linkages, the geometric structure allows us to reduce the complexity of a mechanical system,” AlAlawi says.

Using the design tool, the researchers incorporated meta-antennas into several smart devices, including a curtain that dynamically adjusts household lighting and headphones that seamlessly transition between noise-cancelling and transparent modes.

For the smart headphone, for instance, when the meta-antenna expands and bends, it shifts the resonance frequency by 2.6 percent, which switches the headphone mode. The team’s experiments also showed that meta-antenna structures are durable enough to withstand more than 10,000 compressions.

Because the antenna patch can be patterned onto any surface, it could be used with more complex structures. For instance, the antenna could be incorporated into smart textiles that perform noninvasive biomedical sensing or temperature monitoring.

In the future, the researchers want to design three-dimensional meta-antennas for a wider range of applications. They also want to add more functions to the design tool, improve the durability and flexibility of the metamaterial structure, experiment with different symmetric metamaterial patterns, and streamline some manual fabrication steps.

This research was funded, in part, by the Bahrain Crown Prince International Scholarship and the Gwangju Institute of Science and Technology.

Using artificial intelligence, MIT researchers have come up with a new way to design nanoparticles that can more efficiently deliver RNA vaccines and other types of RNA therapies.

After training a machine-learning model to analyze thousands of existing delivery particles, the researchers used it to predict new materials that would work even better. The model also enabled the researchers to identify particles that would work well in different types of cells, and to discover ways to incorporate new types of materials into the particles.

“What we did was apply machine-learning tools to help accelerate the identification of optimal ingredient mixtures in lipid nanoparticles to help target a different cell type or help incorporate different materials, much faster than previously was possible,” says Giovanni Traverso, an associate professor of mechanical engineering at MIT, a gastroenterologist at Brigham and Women’s Hospital, and the senior author of the study.

This approach could dramatically speed the process of developing new RNA vaccines, as well as therapies that could be used to treat obesity, diabetes, and other metabolic disorders, the researchers say.

Alvin Chan, a former MIT postdoc who is now an assistant professor at Nanyang Technological University, and Ameya Kirtane, a former MIT postdoc who is now an assistant professor at the University of Minnesota, are the lead authors of the new open-access study, which appears today in Nature Nanotechnology.

Particle predictions

RNA vaccines, such as the vaccines for SARS-CoV-2, are usually packaged in lipid nanoparticles (LNPs) for delivery. These particles protect mRNA from being broken down in the body and help it to enter cells once injected.

Creating particles that handle these jobs more efficiently could help researchers to develop even more effective vaccines. Better delivery vehicles could also make it easier to develop mRNA therapies that encode genes for proteins that could help to treat a variety of diseases.

In 2024, Traverso’s lab launched a multiyear research program, funded by the U.S. Advanced Research Projects Agency for Health (ARPA-H), to develop new ingestible devices that could achieve oral delivery of RNA treatments and vaccines.

“Part of what we’re trying to do is develop ways of producing more protein, for example, for therapeutic applications. Maximizing the efficiency is important to be able to boost how much we can have the cells produce,” Traverso says.

A typical LNP consists of four components — a cholesterol, a helper lipid, an ionizable lipid, and a lipid that is attached to polyethylene glycol (PEG). Different variants of each of these components can be swapped in to create a huge number of possible combinations. Changing up these formulations and testing each one individually is very time-consuming, so Traverso, Chan, and their colleagues decided to turn to artificial intelligence to help speed up the process.

“Most AI models in drug discovery focus on optimizing a single compound at a time, but that approach doesn’t work for lipid nanoparticles, which are made of multiple interacting components,” Chan says. “To tackle this, we developed a new model called COMET, inspired by the same transformer architecture that powers large language models like ChatGPT. Just as those models understand how words combine to form meaning, COMET learns how different chemical components come together in a nanoparticle to influence its properties — like how well it can deliver RNA into cells.”

To generate training data for their machine-learning model, the researchers created a library of about 3,000 different LNP formulations. The team tested each of these 3,000 particles in the lab to see how efficiently they could deliver their payload to cells, then fed all of this data into a machine-learning model.

After the model was trained, the researchers asked it to predict new formulations that would work better than existing LNPs. They tested those predictions by using the new formulations to deliver mRNA encoding a fluorescent protein to mouse skin cells grown in a lab dish. They found that the LNPs predicted by the model did indeed work better than the particles in the training data, and in some cases better than LNP formulations that are used commercially.

Accelerated development

Once the researchers showed that the model could accurately predict particles that would efficiently deliver mRNA, they began asking additional questions. First, they wondered if they could train the model on nanoparticles that incorporate a fifth component: a type of polymer known as branched poly beta amino esters (PBAEs).

Research by Traverso and his colleagues has shown that these polymers can effectively deliver nucleic acids on their own, so they wanted to explore whether adding them to LNPs could improve LNP performance. The MIT team created a set of about 300 LNPs that also include these polymers, which they used to train the model. The resulting model could then predict additional formulations with PBAEs that would work better.

Next, the researchers set out to train the model to make predictions about LNPs that would work best in different types of cells, including a type of cell called Caco-2, which is derived from colorectal cancer cells. Again, the model was able to predict LNPs that would efficiently deliver mRNA to these cells.

Lastly, the researchers used the model to predict which LNPs could best withstand lyophilization — a freeze-drying process often used to extend the shelf-life of medicines.

“This is a tool that allows us to adapt it to a whole different set of questions and help accelerate development. We did a large training set that went into the model, but then you can do much more focused experiments and get outputs that are helpful on very different kinds of questions,” Traverso says.

He and his colleagues are now working on incorporating some of these particles into potential treatments for diabetes and obesity, which are two of the primary targets of the ARPA-H funded project. Therapeutics that could be delivered using this approach include GLP-1 mimics with similar effects to Ozempic.

This research was funded by the GO Nano Marble Center at the Koch Institute, the Karl van Tassel Career Development Professorship, the MIT Department of Mechanical Engineering, Brigham and Women’s Hospital, and ARPA-H.

Graphite is a key structural component in some of the world’s oldest nuclear reactors and many of the next-generation designs being built today. But it also condenses and swells in response to radiation — and the mechanism behind those changes has proven difficult to study.

Now, MIT researchers and collaborators have uncovered a link between properties of graphite and how the material behaves in response to radiation. The findings could lead to more accurate, less destructive ways of predicting the lifespan of graphite materials used in reactors around the world.

“We did some basic science to understand what leads to swelling and, eventually, failure in graphite structures,” says MIT Research Scientist Boris Khaykovich, senior author of the new study. “More research will be needed to put this into practice, but the paper proposes an attractive idea for industry: that you might not need to break hundreds of irradiated samples to understand their failure point.”

Specifically, the study shows a connection between the size of the pores within graphite and the way the material swells and shrinks in volume, leading to degradation.

“The lifetime of nuclear graphite is limited by irradiation-induced swelling,” says co-author and MIT Research Scientist Lance Snead. “Porosity is a controlling factor in this swelling, and while graphite has been extensively studied for nuclear applications since the Manhattan Project, we still do not have a clear understanding of the porosity in both mechanical properties and swelling. This work addresses that.”

The open-access paper appears this week in Interdisciplinary Materials. It is co-authored by Khaykovich, Snead, MIT Research Scientist Sean Fayfar, former MIT research fellow Durgesh Rai, Stony Brook University Assistant Professor David Sprouster, Oak Ridge National Laboratory Staff Scientist Anne Campbell, and Argonne National Laboratory Physicist Jan Ilavsky.

A long-studied, complex material

Ever since 1942, when physicists and engineers built the world’s first nuclear reactor on a converted squash court at the University of Chicago, graphite has played a central role in the generation of nuclear energy. That first reactor, dubbed the Chicago Pile, was constructed from about 40,000 graphite blocks, many of which contained nuggets of uranium.

Today graphite is a vital component of many operating nuclear reactors and is expected to play a central role in next-generation reactor designs like molten-salt and high-temperature gas reactors. That’s because graphite is a good neutron moderator, slowing down the neutrons released by nuclear fission so they are more likely to create fissions themselves and sustain a chain reaction.

“The simplicity of graphite makes it valuable,” Khaykovich explains. “It’s made of carbon, and it’s relatively well-known how to make it cleanly. Graphite is a very mature technology. It’s simple, stable, and we know it works.”

But graphite also has its complexities.

“We call graphite a composite even though it’s made up of only carbon atoms,” Khaykovich says. “It includes ‘filler particles’ that are more crystalline, then there is a matrix called a ‘binder’ that is less crystalline, then there are pores that span in length from nanometers to many microns.”

Each graphite grade has its own composite structure, but they all contain fractals, or shapes that look the same at different scales.

Those complexities have made it hard to predict how graphite will respond to radiation in microscopic detail, although it’s been known for decades that when graphite is irradiated, it first densifies, reducing its volume by up to 10 percent, before swelling and cracking. The volume fluctuation is caused by changes to graphite’s porosity and lattice stress.

“Graphite deteriorates under radiation, as any material does,” Khaykovich says. “So, on the one hand we have a material that’s extremely well-known, and on the other hand, we have a material that is immensely complicated, with a behavior that’s impossible to predict through computer simulations.”

For the study, the researchers received irradiated graphite samples from Oak Ridge National Laboratory. Co-authors Campbell and Snead were involved in irradiating the samples some 20 years ago. The samples are a grade of graphite known as G347A.

The research team used an analysis technique known as X-ray scattering, which uses the scattered intensity of an X-ray beam to analyze the properties of material. Specifically, they looked at the distribution of sizes and surface areas of the sample’s pores, or what are known as the material’s fractal dimensions.

“When you look at the scattering intensity, you see a large range of porosity,” Fayfar says. “Graphite has porosity over such large scales, and you have this fractal self-similarity: The pores in very small sizes look similar to pores spanning microns, so we used fractal models to relate different morphologies across length scales.”

Fractal models had been used on graphite samples before, but not on irradiated samples to see how the material’s pore structures changed. The researchers found that when graphite is first exposed to radiation, its pores get filled as the material degrades.

“But what was quite surprising to us is the [size distribution of the pores] turned back around,” Fayfar says. “We had this recovery process that matched our overall volume plots, which was quite odd. It seems like after graphite is irradiated for so long, it starts recovering. It’s sort of an annealing process where you create some new pores, then the pores smooth out and get slightly bigger. That was a big surprise.”

The researchers found that the size distribution of the pores closely follows the volume change caused by radiation damage.

“Finding a strong correlation between the [size distribution of pores] and the graphite’s volume changes is a new finding, and it helps connect to the failure of the material under irradiation,” Khaykovich says. “It’s important for people to know how graphite parts will fail when they are under stress and how failure probability changes under irradiation.”

From research to reactors

The researchers plan to study other graphite grades and explore further how pore sizes in irradiated graphite correlate with the probability of failure. They speculate that a statistical technique known as the Weibull Distribution could be used to predict graphite’s time until failure. The Weibull Distribution is already used to describe the probability of failure in ceramics and other porous materials like metal alloys.

Khaykovich also speculated that the findings could contribute to our understanding of why materials densify and swell under irradiation.

“There’s no quantitative model of densification that takes into account what’s happening at these tiny scales in graphite,” Khaykovich says. “Graphite irradiation densification reminds me of sand or sugar, where when you crush big pieces into smaller grains, they densify. For nuclear graphite, the crushing force is the energy that neutrons bring in, causing large pores to get filled with smaller, crushed pieces. But more energy and agitation create still more pores, and so graphite swells again. It’s not a perfect analogy, but I believe analogies bring progress for understanding these materials.”

The researchers describe the paper as an important step toward informing graphite production and use in nuclear reactors of the future.

“Graphite has been studied for a very long time, and we’ve developed a lot of strong intuitions about how it will respond in different environments, but when you’re building a nuclear reactor, details matter,” Khaykovich says. “People want numbers. They need to know how much thermal conductivity will change, how much cracking and volume change will happen. If components are changing volume, at some point you need to take that into account.”

This work was supported, in part, by the U.S. Department of Energy.

With help from artificial intelligence, MIT researchers have designed novel antibiotics that can combat two hard-to-treat infections: drug-resistant Neisseria gonorrhoeae and multi-drug-resistant Staphylococcus aureus (MRSA).

Using generative AI algorithms, the research team designed more than 36 million possible compounds and computationally screened them for antimicrobial properties. The top candidates they discovered are structurally distinct from any existing antibiotics, and they appear to work by novel mechanisms that disrupt bacterial cell membranes.

This approach allowed the researchers to generate and evaluate theoretical compounds that have never been seen before — a strategy that they now hope to apply to identify and design compounds with activity against other species of bacteria.

“We’re excited about the new possibilities that this project opens up for antibiotics development. Our work shows the power of AI from a drug design standpoint, and enables us to exploit much larger chemical spaces that were previously inaccessible,” says James Collins, the Termeer Professor of Medical Engineering and Science in MIT’s Institute for Medical Engineering and Science (IMES) and Department of Biological Engineering, and a member of the Broad Institute.

Collins is the senior author of the study, which appears today in Cell. The paper’s lead authors are MIT postdoc Aarti Krishnan, former postdoc Melis Anahtar ’08, and Jacqueline Valeri PhD ’23.

Exploring chemical space

Over the past 45 years, a few dozen new antibiotics have been approved by the FDA, but most of these are variants of existing antibiotics. At the same time, bacterial resistance to many of these drugs has been growing. Globally, it is estimated that drug-resistant bacterial infections cause nearly 5 million deaths per year.

In hopes of finding new antibiotics to fight this growing problem, Collins and others at MIT’s Antibiotics-AI Project have harnessed the power of AI to screen huge libraries of existing chemical compounds. This work has yielded several promising drug candidates, including halicin and abaucin.

To build on that progress, Collins and his colleagues decided to expand their search into molecules that can’t be found in any chemical libraries. By using AI to generate hypothetically possible molecules that don’t exist or haven’t been discovered, they realized that it should be possible to explore a much greater diversity of potential drug compounds.

In their new study, the researchers employed two different approaches: First, they directed generative AI algorithms to design molecules based on a specific chemical fragment that showed antimicrobial activity, and second, they let the algorithms freely generate molecules, without having to include a specific fragment.

For the fragment-based approach, the researchers sought to identify molecules that could kill N. gonorrhoeae, a Gram-negative bacterium that causes gonorrhea. They began by assembling a library of about 45 million known chemical fragments, consisting of all possible combinations of 11 atoms of carbon, nitrogen, oxygen, fluorine, chlorine, and sulfur, along with fragments from Enamine’s REadily AccessibLe (REAL) space.

Then, they screened the library using machine-learning models that Collins’ lab has previously trained to predict antibacterial activity against N. gonorrhoeae. This resulted in nearly 4 million fragments. They narrowed down that pool by removing any fragments predicted to be cytotoxic to human cells, displayed chemical liabilities, and were known to be similar to existing antibiotics. This left them with about 1 million candidates.

“We wanted to get rid of anything that would look like an existing antibiotic, to help address the antimicrobial resistance crisis in a fundamentally different way. By venturing into underexplored areas of chemical space, our goal was to uncover novel mechanisms of action,” Krishnan says.

Through several rounds of additional experiments and computational analysis, the researchers identified a fragment they called F1 that appeared to have promising activity against N. gonorrhoeae. They used this fragment as the basis for generating additional compounds, using two different generative AI algorithms.

One of those algorithms, known as chemically reasonable mutations (CReM), works by starting with a particular molecule containing F1 and then generating new molecules by adding, replacing, or deleting atoms and chemical groups. The second algorithm, F-VAE (fragment-based variational autoencoder), takes a chemical fragment and builds it into a complete molecule. It does so by learning patterns of how fragments are commonly modified, based on its pretraining on more than 1 million molecules from the ChEMBL database.

Those two algorithms generated about 7 million candidates containing F1, which the researchers then computationally screened for activity against N. gonorrhoeae. This screen yielded about 1,000 compounds, and the researchers selected 80 of those to see if they could be produced by chemical synthesis vendors. Only two of these could be synthesized, and one of them, named NG1, was very effective at killing N. gonorrhoeae in a lab dish and in a mouse model of drug-resistant gonorrhea infection.

Additional experiments revealed that NG1 interacts with a protein called LptA, a novel drug target involved in the synthesis of the bacterial outer membrane. It appears that the drug works by interfering with membrane synthesis, which is fatal to cells.

Unconstrained design

In a second round of studies, the researchers explored the potential of using generative AI to freely design molecules, using Gram-positive bacteria, S. aureus as their target.

Again, the researchers used CReM and VAE to generate molecules, but this time with no constraints other than the general rules of how atoms can join to form chemically plausible molecules. Together, the models generated more than 29 million compounds. The researchers then applied the same filters that they did to the N. gonorrhoeae candidates, but focusing on S. aureus, eventually narrowing the pool down to about 90 compounds.

They were able to synthesize and test 22 of these molecules, and six of them showed strong antibacterial activity against multi-drug-resistant S. aureus grown in a lab dish. They also found that the top candidate, named DN1, was able to clear a methicillin-resistant S. aureus (MRSA) skin infection in a mouse model. These molecules also appear to interfere with bacterial cell membranes, but with broader effects not limited to interaction with one specific protein.

Phare Bio, a nonprofit that is also part of the Antibiotics-AI Project, is now working on further modifying NG1 and DN1 to make them suitable for additional testing.

“In a collaboration with Phare Bio, we are exploring analogs, as well as working on advancing the best candidates preclinically, through medicinal chemistry work,” Collins says. “We are also excited about applying the platforms that Aarti and the team have developed toward other bacterial pathogens of interest, notably Mycobacterium tuberculosis and Pseudomonas aeruginosa.”

The research was funded, in part, by the U.S. Defense Threat Reduction Agency, the National Institutes of Health, the Audacious Project, Flu Lab, the Sea Grape Foundation, Rosamund Zander and Hansjorg Wyss for the Wyss Foundation, and an anonymous donor.

Is this movie review a rave or a pan? Is this news story about business or technology? Is this online chatbot conversation veering off into giving financial advice? Is this online medical information site giving out misinformation?

These kinds of automated conversations, whether they involve seeking a movie or restaurant review or getting information about your bank account or health records, are becoming increasingly prevalent. More than ever, such evaluations are being made by highly sophisticated algorithms, known as text classifiers, rather than by human beings. But how can we tell how accurate these classifications really are?

Now, a team at MIT’s Laboratory for Information and Decision Systems (LIDS) has come up with an innovative approach to not only measure how well these classifiers are doing their job, but then go one step further and show how to make them more accurate.

The new evaluation and remediation software was led and developed by Lei Xu alongside the research conducted by Sarah Alnegheimish, Kalyan Veeramachaneni, a principal research scientist at LIDS and senior author, with two others. The software package is being made freely available for download by anyone who wants to use it.

A standard method for testing these classification systems is to create what are known as synthetic examples — sentences that closely resemble ones that have already been classified. For example, researchers might take a sentence that has already been tagged by a classifier program as being a rave review, and see if changing a word or a few words while retaining the same meaning could fool the classifier into deeming it a pan. Or a sentence that was determined to be misinformation might get misclassified as accurate. This ability to fool the classifiers makes these adversarial examples.

People have tried various ways to find the vulnerabilities in these classifiers, Veeramachaneni says. But existing methods of finding these vulnerabilities have a hard time with this task and miss many examples that they should catch, he says.

Increasingly, companies are trying to use such evaluation tools in real time, monitoring the output of chatbots used for various purposes to try to make sure they are not putting out improper responses. For example, a bank might use a chatbot to respond to routine customer queries such as checking account balances or applying for a credit card, but it wants to ensure that its responses could never be interpreted as financial advice, which could expose the company to liability. “Before showing the chatbot’s response to the end user, they want to use the text classifier to detect whether it’s giving financial advice or not,” Veeramachaneni says. But then it’s important to test that classifier to see how reliable its evaluations are.

“These chatbots, or summarization engines or whatnot are being set up across the board,” he says, to deal with external customers and within an organization as well, for example providing information about HR issues. It’s important to put these text classifiers into the loop to detect things that they are not supposed to say, and filter those out before the output gets transmitted to the user.

That’s where the use of adversarial examples comes in — those sentences that have already been classified but then produce a different response when they are slightly modified while retaining the same meaning. How can people confirm that the meaning is the same? By using another large language model (LLM) that interprets and compares meanings. So, if the LLM says the two sentences mean the same thing, but the classifier labels them differently, “that is a sentence that is adversarial — it can fool the classifier,” Veeramachaneni says. And when the researchers examined these adversarial sentences, “we found that most of the time, this was just a one-word change,” although the people using LLMs to generate these alternate sentences often didn’t realize that.

Further investigation, using LLMs to analyze many thousands of examples, showed that certain specific words had an outsized influence in changing the classifications, and therefore the testing of a classifier’s accuracy could focus on this small subset of words that seem to make the most difference. They found that one-tenth of 1 percent of all the 30,000 words in the system’s vocabulary could account for almost half of all these reversals of classification, in some specific applications.

Lei Xu PhD ’23, a recent graduate from LIDS who performed much of the analysis as part of his thesis work, “used a lot of interesting estimation techniques to figure out what are the most powerful words that can change the overall classification, that can fool the classifier,” Veeramachaneni says. The goal is to make it possible to do much more narrowly targeted searches, rather than combing through all possible word substitutions, thus making the computational task of generating adversarial examples much more manageable. “He’s using large language models, interestingly enough, as a way to understand the power of a single word.”

Then, also using LLMs, he searches for other words that are closely related to these powerful words, and so on, allowing for an overall ranking of words according to their influence on the outcomes. Once these adversarial sentences have been found, they can be used in turn to retrain the classifier to take them into account, increasing the robustness of the classifier against those mistakes.

Making classifiers more accurate may not sound like a big deal if it’s just a matter of classifying news articles into categories, or deciding whether reviews of anything from movies to restaurants are positive or negative. But increasingly, classifiers are being used in settings where the outcomes really do matter, whether preventing the inadvertent release of sensitive medical, financial, or security information, or helping to guide important research, such as into properties of chemical compounds or the folding of proteins for biomedical applications, or in identifying and blocking hate speech or known misinformation.

As a result of this research, the team introduced a new metric, which they call p, which provides a measure of how robust a given classifier is against single-word attacks. And because of the importance of such misclassifications, the research team has made its products available as open access for anyone to use. The package consists of two components: SP-Attack, which generates adversarial sentences to test classifiers in any particular application, and SP-Defense, which aims to improve the robustness of the classifier by generating and using adversarial sentences to retrain the model.

In some tests, where competing methods of testing classifier outputs allowed a 66 percent success rate by adversarial attacks, this team’s system cut that attack success rate almost in half, to 33.7 percent. In other applications, the improvement was as little as a 2 percent difference, but even that can be quite important, Veeramachaneni says, since these systems are being used for so many billions of interactions that even a small percentage can affect millions of transactions.

The team’s results were published on July 7 in the journal Expert Systems in a paper by Xu, Veeramachaneni, and Alnegheimish of LIDS, along with Laure Berti-Equille at IRD in Marseille, France, and Alfredo Cuesta-Infante at the Universidad Rey Juan Carlos, in Spain. 

Around the world, about 2 billion people suffer from iron deficiency, which can lead to anemia, impaired brain development in children, and increased infant mortality.

To combat that problem, MIT researchers have come up with a new way to fortify foods and beverages with iron, using small crystalline particles. These particles, known as metal-organic frameworks, could be sprinkled on food, added to staple foods such as bread, or incorporated into drinks like coffee and tea.

“We’re creating a solution that can be seamlessly added to staple foods across different regions,” says Ana Jaklenec, a principal investigator at MIT’s Koch Institute for Integrative Cancer Research. “What’s considered a staple in Senegal isn’t the same as in India or the U.S., so our goal was to develop something that doesn’t react with the food itself. That way, we don’t have to reformulate for every context — it can be incorporated into a wide range of foods and beverages without compromise.”

The particles designed in this study can also carry iodine, another critical nutrient. The particles could also be adapted to carry important minerals such as zinc, calcium, or magnesium.

“We are very excited about this new approach and what we believe is a novel application of metal-organic frameworks to potentially advance nutrition, particularly in the developing world,” says Robert Langer, the David H. Koch Institute Professor at MIT and a member of the Koch Institute.

Jaklenec and Langer are the senior authors of the study, which appears today in the journal Matter. MIT postdoc Xin Yang and Linzixuan (Rhoda) Zhang PhD ’24 are the lead authors of the paper.

Iron stabilization

Food fortification can be a successful way to combat nutrient deficiencies, but this approach is often challenging because many nutrients are fragile and break down during storage or cooking. When iron is added to foods, it can react with other molecules in the food, giving the food a metallic taste.

In previous work, Jaklenec’s lab has shown that encapsulating nutrients in polymers can protect them from breaking down or reacting with other molecules. In a small clinical trial, the researchers found that women who ate bread fortified with encapsulated iron were able to absorb the iron from the food.

However, one drawback to this approach is that the polymer adds a lot of bulk to the material, limiting the amount of iron or other nutrients that end up in the food.

“Encapsulating iron in polymers significantly improves its stability and reactivity, making it easier to add to food,” Jaklenec says. “But to be effective, it requires a substantial amount of polymer. That limits how much iron you can deliver in a typical serving, making it difficult to meet daily nutritional targets through fortified foods alone.”

To overcome that challenge, Yang came up with a new idea: Instead of encapsulating iron in a polymer, they could use iron itself as a building block for a crystalline particle known as a metal-organic framework, or MOF (pronounced “moff”).

MOFs consist of metal atoms joined by organic molecules called ligands to create a rigid, cage-like structure. Depending on the combination of metals and ligands chosen, they can be used for a wide variety of applications.

“We thought maybe we could synthesize a metal-organic framework with food-grade ligands and food-grade micronutrients,” Yang says. “Metal-organic frameworks have very high porosity, so they can load a lot of cargo. That’s why we thought we could leverage this platform to make a new metal-organic framework that could be used in the food industry.”

In this case, the researchers designed a MOF consisting of iron bound to a ligand called fumaric acid, which is often used as a food additive to enhance flavor or help preserve food.

This structure prevents iron from reacting with polyphenols — compounds commonly found in foods such as whole grains and nuts, as well as coffee and tea. When iron does react with those compounds, it forms a metal polyphenol complex that cannot be absorbed by the body.

The MOFs’ structure also allows them to remain stable until they reach an acidic environment, such as the stomach, where they break down and release their iron payload.

Double-fortified salts

The researchers also decided to include iodine in their MOF particle, which they call NuMOF. Iodized salt has been very successful at preventing iodine deficiency, and many efforts are now underway to create “double-fortified salts” that would also contain iron.

Delivering these nutrients together has proven difficult because iron and iodine can react with each other, making each one less likely to be absorbed by the body. In this study, the MIT team showed that once they formed their iron-containing MOF particles, they could load them with iodine, in a way that the iron and iodine do not react with each other.

In tests of the particles’ stability, the researchers found that the NuMOFs could withstand long-term storage, high heat and humidity, and boiling water.

Throughout these tests, the particles maintained their structure. When the researchers then fed the particles to mice, they found that both iron and iodine became available in the bloodstream within several hours of the NuMOF consumption.

The researchers are now working on launching a company that is developing coffee and other beverages fortified with iron and iodine. They also hope to continue working toward a double-fortified salt that could be consumed on its own or incorporated into staple food products.

The research was partially supported by J-WAFS Fellowships for Water and Food Solutions. 

Other authors of the paper include Fangzheng Chen, Wenhao Gao, Zhiling Zheng, Tian Wang, Erika Yan Wang, Behnaz Eshaghi, and Sydney MacDonald.

This research was conducted, in part, using MIT.nano’s facilities.

Water is essential for life on Earth. So, the liquid must be a requirement for life on other worlds. For decades, scientists’ definition of habitability on other planets has rested on this assumption.

But what makes some planets habitable might have very little to do with water. In fact, an entirely different type of liquid could conceivably support life in worlds where water can barely exist. That’s a possibility that MIT scientists raise in a study appearing this week in the Proceedings of the National Academy of Sciences.

From lab experiments, the researchers found that a type of fluid known as an ionic liquid can readily form from chemical ingredients that are also expected to be found on the surface of some rocky planets and moons. Ionic liquids are salts that exist in liquid form below about 100 degrees Celsius. The team’s experiments showed that a mixture of sulfuric acid and certain nitrogen-containing organic compounds produced such a liquid. On rocky planets, sulfuric acid may be a byproduct of volcanic activity, while nitrogen-containing compounds have been detected on several asteroids and planets in our solar system, suggesting the compounds may be present in other planetary systems.

The scientists propose that, even on planets that are too warm or that have atmospheres are too low-pressure to support liquid water, there could still be pockets of ionic liquid. And where there is liquid, there may be potential for life, though likely not anything that resembles Earth’s water-based beings.

Ionic liquids have extremely low vapor pressure and do not evaporate; they can form and persist at higher temperatures and lower pressures than what liquid water can tolerate. The researchers note that ionic liquid can be a hospitable environment for some biomolecules, such as certain proteins that can remain stable in the fluid.

“We consider water to be required for life because that is what’s needed for Earth life. But if we look at a more general definition, we see that what we need is a liquid in which metabolism for life can take place,” says Rachana Agrawal, who led the study as a postdoc in MIT’s Department of Earth, Atmospheric and Planetary Sciences. “Now if we include ionic liquid as a possibility, this can dramatically increase the habitability zone for all rocky worlds.”

The study’s MIT co-authors are Sara Seager, the Class of 1941 Professor of Planetary Sciences in the Department of Earth, Atmospheric and Planetary Sciences and a professor in the departments of Physics and of Aeronautics and Astronautics, along with Iaroslav Iakubivskyi, Weston Buchanan, Ana Glidden, and Jingcheng Huang. Co-authors also include Maxwell Seager of Worcester Polytechnic Institute, William Bains of Cardiff University, and Janusz Petkowski of Wroclaw University of Science and Technology, in Poland.

A liquid leap

The team’s work with ionic liquid grew out of an effort to search for signs of life on Venus, where clouds of sulfuric acid envelope the planet in a noxious haze. Despite its toxicity, Venus’ clouds may contain signs of life — a notion that scientists plan to test with upcoming missions to the planet’s atmosphere.

Agrawal and Seager, who is leading the Morning Star Missions to Venus, were investigating ways to collect and evaporate sulfuric acid. If a mission collects samples from Venus’ clouds, sulfuric acid would have to be evaporated away in order to reveal any residual organic compounds that could then be analyzed for signs of life.

The researchers were using their custom, low-pressure system designed to evaporate away excess sulfuric acid, to test evaporation of a solution of the acid and an organic compound, glycine. They found that in every case, while most of the liquid sulfuric acid evaporated, a stubborn layer of liquid always remained. They soon realized that sulfuric acid was chemically reacting with glycine, resulting in an exchange of hydrogen atoms from the acid to the organic compound. The result was a fluid mixture of salts, or ions, known as an ionic liquid, that persists as a liquid across a wide range of temperatures and pressures.

This accidental finding kickstarted an idea: Could ionic liquid form on planets that are too warm and host atmospheres too thin for water to exist?

“From there, we took the leap of imagination of what this could mean,” Agrawal says. “Sulfuric acid is found on Earth from volcanoes, and organic compounds have been found on asteroids and other planetary bodies. So, this led us to wonder if ionic liquids could potentially form and exist naturally on exoplanets.”

Rocky oases

On Earth, ionic liquids are mainly synthesized for industrial purposes. They do not occur naturally, except for in one specific case, in which the liquid is generated from the mixing of venoms produced by two rival species of ants.

The team set out to investigate what conditions ionic liquid could be naturally produced in, and over what range of temperatures and pressures. In the lab, they mixed sulfuric acid with various nitrogen-containing organic compounds. In previous work, Seager’s team had found that the compounds, some of which can be considered ingredients associated with life, are surprisingly stable in sulfuric acid.

“In high school, you learn that an acid wants to donate a proton,” Seager says. “And oddly enough, we knew from our past work with sulfuric acid (the main component of Venus’ clouds) and nitrogen-containing compounds, that a nitrogen wants to receive a hydrogen. It’s like one person’s trash is another person’s treasure.”

The reaction could produce a bit of ionic liquid if the sulfuric acid and nitrogen-containing organics were in a one-to-one ratio — a ratio that was not a focus of the prior work. For their new study, Seager and Agrawal mixed sulfuric acid with over 30 different nitrogen-containing organic compounds, across a range of temperatures and pressures, then observed whether ionic liquid formed when they evaporated away the sulfuric acid in various vials. They also mixed the ingredients onto basalt rocks, which are known to exist on the surface of many rocky planets.

The team found that the reactions produced ionic liquid at temperatures up to 180 degrees Celsius and at extremely low pressures — much lower than that of the Earth’s atmosphere. Their results suggest that ionic liquid could naturally form on other planets where liquid water cannot exist, under the right conditions.

“We were just astonished that the ionic liquid forms under so many different conditions,” Seager says. “If you put the sulfuric acid and the organic on a rock, the excess sulfuric acid seeps into the rock pores, but you’re still left with a drop of ionic liquid on the rock. Whatever we tried, ionic liquid still formed.”

“We’re envisioning a planet warmer than Earth, that doesn’t have water, and at some point in its past or currently, it has to have had sulfuric acid, formed from volcanic outgassing,” Seager says. “This sulfuric acid has to flow over a little pocket of organics. And organic deposits are extremely common in the solar system.”

Then, she says, the resulting pockets of liquid could stay on the planet’s surface, potentially for years or millenia, where they could theoretically serve as small oases for simple forms of ionic-liquid-based life. Going forward, Seager’s team plans to investigate further, to see what biomolecules, and ingredients for life, might survive, and thrive, in ionic liquid.

“We just opened up a Pandora’s box of new research,” Seager says. “It’s been a real journey.”

This research was supported, in part, by the Sloan Foundation and the Volkswagen Foundation.

The cost of solar panels has dropped by more than 99 percent since the 1970s, enabling widespread adoption of photovoltaic systems that convert sunlight into electricity.

A new MIT study drills down on specific innovations that enabled such dramatic cost reductions, revealing that technical advances across a web of diverse research efforts and industries played a pivotal role.

The findings could help renewable energy companies make more effective R&D investment decisions and aid policymakers in identifying areas to prioritize to spur growth in manufacturing and deployment.

The researchers’ modeling approach shows that key innovations often originated outside the solar sector, including advances in semiconductor fabrication, metallurgy, glass manufacturing, oil and gas drilling, construction processes, and even legal domains.

“Our results show just how intricate the process of cost improvement is, and how much scientific and engineering advances, often at a very basic level, are at the heart of these cost reductions. A lot of knowledge was drawn from different domains and industries, and this network of knowledge is what makes these technologies improve,” says study senior author Jessika Trancik, a professor in MIT’s Institute for Data, Systems, and Society.

Trancik is joined on the paper by co-lead authors Goksin Kavlak, a former IDSS graduate student and postdoc who is now a senior energy associate at the Brattle Group; Magdalena Klemun, a former IDSS graduate student and postdoc who is now an assistant professor at Johns Hopkins University; former MIT postdoc Ajinkya Kamat; as well as Brittany Smith and Robert Margolis of the National Renewable Energy Laboratory. The research appears today in PLOS ONE.

Identifying innovations

This work builds on mathematical models that the researchers previously developed that tease out the effects of engineering technologies on the cost of photovoltaic (PV) modules and systems.

In this study, the researchers aimed to dig even deeper into the scientific advances that drove those cost declines.

They combined their quantitative cost model with a detailed, qualitative analysis of innovations that affected the costs of PV system materials, manufacturing steps, and deployment processes.

“Our quantitative cost model guided the qualitative analysis, allowing us to look closely at innovations in areas that are hard to measure due to a lack of quantitative data,” Kavlak says.

Building on earlier work identifying key cost drivers — such as the number of solar cells per module, wiring efficiency, and silicon wafer area — the researchers conducted a structured scan of the literature for innovations likely to affect these drivers. Next, they grouped these innovations to identify patterns, revealing clusters that reduced costs by improving materials or prefabricating components to streamline manufacturing and installation. Finally, the team tracked industry origins and timing for each innovation, and consulted domain experts to zero in on the most significant innovations.

All told, they identified 81 unique innovations that affected PV system costs since 1970, from improvements in antireflective coated glass to the implementation of fully online permitting interfaces.

“With innovations, you can always go to a deeper level, down to things like raw materials processing techniques, so it was challenging to know when to stop. Having that quantitative model to ground our qualitative analysis really helped,” Trancik says.

They chose to separate PV module costs from so-called balance-of-system (BOS) costs, which cover things like mounting systems, inverters, and wiring.

PV modules, which are wired together to form solar panels, are mass-produced and can be exported, while many BOS components are designed, built, and sold at the local level.

“By examining innovations both at the BOS level and within the modules, we identify the different types of innovations that have emerged in these two parts of PV technology,” Kavlak says.

BOS costs depend more on soft technologies, nonphysical elements such as permitting procedures, which have contributed significantly less to PV’s past cost improvement compared to hardware innovations.

“Often, it comes down to delays. Time is money, and if you have delays on construction sites and unpredictable processes, that affects these balance-of-system costs,” Trancik says.

Innovations such as automated permitting software, which flags code-compliant systems for fast-track approval, show promise. Though not yet quantified in this study, the team’s framework could support future analysis of their economic impact and similar innovations that streamline deployment processes.

Interconnected industries

The researchers found that innovations from the semiconductor, electronics, metallurgy, and petroleum industries played a major role in reducing both PV and BOS costs, but BOS costs were also impacted by innovations in software engineering and electric utilities.

Noninnovation factors, like efficiency gains from bulk purchasing and the accumulation of knowledge in the solar power industry, also reduced some cost variables.

In addition, while most PV panel innovations originated in research organizations or industry, many BOS innovations were developed by city governments, U.S. states, or professional associations.

“I knew there was a lot going on with this technology, but the diversity of all these fields and how closely linked they are, and the fact that we can clearly see that network through this analysis, was interesting,” Trancik says.

“PV was very well-positioned to absorb innovations from other industries — thanks to the right timing, physical compatibility, and supportive policies to adapt innovations for PV applications,” Klemun adds.

The analysis also reveals the role greater computing power could play in reducing BOS costs through advances like automated engineering review systems and remote site assessment software.

“In terms of knowledge spillovers, what we've seen so far in PV may really just be the beginning,” Klemun says, pointing to the expanding role of robotics and AI-driven digital tools in driving future cost reductions and quality improvements.

In addition to their qualitative analysis, the researchers demonstrated how this methodology could be used to estimate the quantitative impact of a particular innovation if one has the numerical data to plug into the cost equation.

For instance, using information about material prices and manufacturing procedures, they estimate that wire sawing, a technique which was introduced in the 1980s, led to an overall PV system cost decrease of $5 per watt by reducing silicon losses and increasing throughput during fabrication.

“Through this retrospective analysis, you learn something valuable for future strategy because you can see what worked and what didn’t work, and the models can also be applied prospectively. It is also useful to know what adjacent sectors may help support improvement in a particular technology,” Trancik says.

Moving forward, the researchers plan to apply this methodology to a wide range of technologies, including other renewable energy systems. They also want to further study soft technology to identify innovations or processes that could accelerate cost reductions.

“Although the process of technological innovation may seem like a black box, we’ve shown that you can study it just like any other phenomena,” Trancik says.

This research is funded, in part, by the U.S. Department of Energy Solar Energy Technologies Office.

Any motorist who has ever waited through multiple cycles for a traffic light to turn green knows how annoying signalized intersections can be. But sitting at intersections isn’t just a drag on drivers’ patience — unproductive vehicle idling could contribute as much as 15 percent of the carbon dioxide emissions from U.S. land transportation.

A large-scale modeling study led by MIT researchers reveals that eco-driving measures, which can involve dynamically adjusting vehicle speeds to reduce stopping and excessive acceleration, could significantly reduce those CO₂ emissions.

Using a powerful artificial intelligence method called deep reinforcement learning, the researchers conducted an in-depth impact assessment of the factors affecting vehicle emissions in three major U.S. cities.

Their analysis indicates that fully adopting eco-driving measures could cut annual city-wide intersection carbon emissions by 11 to 22 percent, without slowing traffic throughput or affecting vehicle and traffic safety.

Even if only 10 percent of vehicles on the road employ eco-driving, it would result in 25 to 50 percent of the total reduction in CO2 emissions, the researchers found.

In addition, dynamically optimizing speed limits at about 20 percent of intersections provides 70 percent of the total emission benefits. This indicates that eco-driving measures could be implemented gradually while still having measurable, positive impacts on mitigating climate change and improving public health.

“Vehicle-based control strategies like eco-driving can move the needle on climate change reduction. We’ve shown here that modern machine-learning tools, like deep reinforcement learning, can accelerate the kinds of analysis that support sociotechnical decision making. This is just the tip of the iceberg,” says senior author Cathy Wu, the Class of 1954 Career Development Associate Professor in Civil and Environmental Engineering (CEE) and the Institute for Data, Systems, and Society (IDSS) at MIT, and a member of the Laboratory for Information and Decision Systems (LIDS).

She is joined on the paper by lead author Vindula Jayawardana, an MIT graduate student; as well as MIT graduate students Ao Qu, Cameron Hickert, and Edgar Sanchez; MIT undergraduate Catherine Tang; Baptiste Freydt, a graduate student at ETH Zurich; and Mark Taylor and Blaine Leonard of the Utah Department of Transportation. The research appears in Transportation Research Part C: Emerging Technologies.

A multi-part modeling study

Traffic control measures typically call to mind fixed infrastructure, like stop signs and traffic signals. But as vehicles become more technologically advanced, it presents an opportunity for eco-driving, which is a catch-all term for vehicle-based traffic control measures like the use of dynamic speeds to reduce energy consumption.

In the near term, eco-driving could involve speed guidance in the form of vehicle dashboards or smartphone apps. In the longer term, eco-driving could involve intelligent speed commands that directly control the acceleration of semi-autonomous and fully autonomous vehicles through vehicle-to-infrastructure communication systems.

“Most prior work has focused on how to implement eco-driving. We shifted the frame to consider the question of should we implement eco-driving. If we were to deploy this technology at scale, would it make a difference?” Wu says.

To answer that question, the researchers embarked on a multifaceted modeling study that would take the better part of four years to complete.

They began by identifying 33 factors that influence vehicle emissions, including temperature, road grade, intersection topology, age of the vehicle, traffic demand, vehicle types, driver behavior, traffic signal timing, road geometry, etc.

“One of the biggest challenges was making sure we were diligent and didn’t leave out any major factors,” Wu says.

Then they used data from OpenStreetMap, U.S. geological surveys, and other sources to create digital replicas of more than 6,000 signalized intersections in three cities — Atlanta, San Francisco, and Los Angeles — and simulated more than a million traffic scenarios.

The researchers used deep reinforcement learning to optimize each scenario for eco-driving to achieve the maximum emissions benefits.

Reinforcement learning optimizes the vehicles’ driving behavior through trial-and-error interactions with a high-fidelity traffic simulator, rewarding vehicle behaviors that are more energy-efficient while penalizing those that are not.

The researchers cast the problem as a decentralized cooperative multi-agent control problem, where the vehicles cooperate to achieve overall energy efficiency, even among non-participating vehicles, and they act in a decentralized manner, avoiding the need for costly communication between vehicles.

However, training vehicle behaviors that generalize across diverse intersection traffic scenarios was a major challenge. The researchers observed that some scenarios are more similar to one another than others, such as scenarios with the same number of lanes or the same number of traffic signal phases.

As such, the researchers trained separate reinforcement learning models for different clusters of traffic scenarios, yielding better emission benefits overall.

But even with the help of AI, analyzing citywide traffic at the network level would be so computationally intensive it could take another decade to unravel, Wu says.

Instead, they broke the problem down and solved each eco-driving scenario at the individual intersection level.

“We carefully constrained the impact of eco-driving control at each intersection on neighboring intersections. In this way, we dramatically simplified the problem, which enabled us to perform this analysis at scale, without introducing unknown network effects,” she says.

Significant emissions benefits

When they analyzed the results, the researchers found that full adoption of eco-driving could result in intersection emissions reductions of between 11 and 22 percent.

These benefits differ depending on the layout of a city’s streets. A denser city like San Francisco has less room to implement eco-driving between intersections, offering a possible explanation for reduced emission savings, while Atlanta could see greater benefits given its higher speed limits.

Even if only 10 percent of vehicles employ eco-driving, a city could still realize 25 to 50 percent of the total emissions benefit because of car-following dynamics: Non-eco-driving vehicles would follow controlled eco-driving vehicles as they optimize speed to pass smoothly through intersections, reducing their carbon emissions as well.

In some cases, eco-driving could also increase vehicle throughput by minimizing emissions. However, Wu cautions that increasing throughput could result in more drivers taking to the roads, reducing emissions benefits.

And while their analysis of widely used safety metrics known as surrogate safety measures, such as time to collision, suggest that eco-driving is as safe as human driving, it could cause unexpected behavior in human drivers. More research is needed to fully understand potential safety impacts, Wu says.

Their results also show that eco-driving could provide even greater benefits when combined with alternative transportation decarbonization solutions. For instance, 20 percent eco-driving adoption in San Francisco would cut emission levels by 7 percent, but when combined with the projected adoption of hybrid and electric vehicles, it would cut emissions by 17 percent.

“This is a first attempt to systematically quantify network-wide environmental benefits of eco-driving. This is a great research effort that will serve as a key reference for others to build on in the assessment of eco-driving systems,” says Hesham Rakha, the Samuel L. Pritchard Professor of Engineering at Virginia Tech, who was not involved with this research.

And while the researchers focus on carbon emissions, the benefits are highly correlated with improvements in fuel consumption, energy use, and air quality.

“This is almost a free intervention. We already have smartphones in our cars, and we are rapidly adopting cars with more advanced automation features. For something to scale quickly in practice, it must be relatively simple to implement and shovel-ready. Eco-driving fits that bill,” Wu says.

This work is funded, in part, by Amazon and the Utah Department of Transportation.

A new strategy for strengthening polymer materials could lead to more durable plastics and cut down on plastic waste, according to researchers at MIT and Duke University.

Using machine learning, the researchers identified crosslinker molecules that can be added to polymer materials, allowing them to withstand more force before tearing. These crosslinkers belong to a class of molecules known as mechanophores, which change their shape or other properties in response to mechanical force.

“These molecules can be useful for making polymers that would be stronger in response to force. You apply some stress to them, and rather than cracking or breaking, you instead see something that has higher resilience,” says Heather Kulik, the Lammot du Pont Professor of Chemical Engineering at MIT, who is also a professor of chemistry and the senior author of the study.

The crosslinkers that the researchers identified in this study are iron-containing compounds known as ferrocenes, which until now had not been broadly explored for their potential as mechanophores. Experimentally evaluating a single mechanophore can take weeks, but the researchers showed that they could use a machine-learning model to dramatically speed up this process.

MIT postdoc Ilia Kevlishvili is the lead author of the open-access paper, which appeared Friday in ACS Central Science. Other authors include Jafer Vakil, a Duke graduate student; David Kastner and Xiao Huang, both MIT graduate students; and Stephen Craig, a professor of chemistry at Duke.

The weakest link

Mechanophores are molecules that respond to force in unique ways, typically by changing their color, structure, or other properties. In the new study, the MIT and Duke team wanted to investigate whether they could be used to help make polymers more resilient to damage.

The new work builds on a 2023 study from Craig and Jeremiah Johnson, the A. Thomas Guertin Professor of Chemistry at MIT, and their colleagues. In that work, the researchers found that, surprisingly, incorporating weak crosslinkers into a polymer network can make the overall material stronger. When materials with these weak crosslinkers are stretched to the breaking point, any cracks propagating through the material try to avoid the stronger bonds and go through the weaker bonds instead. This means the crack has to break more bonds than it would if all of the bonds were the same strength.

To find new ways to exploit that phenomenon, Craig and Kulik joined forces to try to identify mechanophores that could be used as weak crosslinkers.

“We had this new mechanistic insight and opportunity, but it came with a big challenge: Of all possible compositions of matter, how do we zero in on the ones with the greatest potential?” Craig says. “Full credit to Heather and Ilia for both identifying this challenge and devising an approach to meet it.”

Discovering and characterizing mechanophores is a difficult task that requires either time-consuming experiments or computationally intense simulations of molecular interactions. Most of the known mechanophores are organic compounds, such as cyclobutane, which was used as a crosslinker in the 2023 study.

In the new study, the researchers wanted to focus on molecules known as ferrocenes, which are believed to hold potential as mechanophores. Ferrocenes are organometallic compounds that have an iron atom sandwiched between two carbon-containing rings. Those rings can have different chemical groups added to them, which alter their chemical and mechanical properties.

Many ferrocenes are used as pharmaceuticals or catalysts, and a handful are known to be good mechanophores, but most have not been evaluated for that use. Experimental tests on a single potential mechanophore can take several weeks, and computational simulations, while faster, still take a couple of days. Evaluating thousands of candidates using these strategies is a daunting task.

Realizing that a machine-learning approach could dramatically speed up the characterization of these molecules, the MIT and Duke team decided to use a neural network to identify ferrocenes that could be promising mechanophores.

They began with information from a database known as the Cambridge Structural Database, which contains the structures of 5,000 different ferrocenes that have already been synthesized.

“We knew that we didn’t have to worry about the question of synthesizability, at least from the perspective of the mechanophore itself. This allowed us to pick a really large space to explore with a lot of chemical diversity, that also would be synthetically realizable,” Kevlishvili says.

First, the researchers performed computational simulations for about 400 of these compounds, allowing them to calculate how much force is necessary to pull atoms apart within each molecule. For this application, they were looking for molecules that would break apart quickly, as these weak links could make polymer materials more resistant to tearing.

Then they used this data, along with information on the structure of each compound, to train a machine-learning model. This model was able to predict the force needed to activate the mechanophore, which in turn influences resistance to tearing, for the remaining 4,500 compounds in the database, plus an additional 7,000 compounds that are similar to those in the database but have some atoms rearranged.

The researchers discovered two main features that seemed likely to increase tear resistance. One was interactions between the chemical groups that are attached to the ferrocene rings. Additionally, the presence of large, bulky molecules attached to both rings of the ferrocene made the molecule more likely to break apart in response to applied forces.

While the first of these features was not surprising, the second trait was not something a chemist would have predicted beforehand, and could not have been detected without AI, the researchers say. “This was something truly surprising,” Kulik says.

Tougher plastics

Once the researchers identified about 100 promising candidates, Craig’s lab at Duke synthesized a polymer material incorporating one of them, known as m-TMS-Fc. Within the material, m-TMS-Fc acts as a crosslinker, connecting the polymer strands that make up polyacrylate, a type of plastic.

By applying force to each polymer until it tore, the researchers found that the weak m-TMS-Fc linker produced a strong, tear-resistant polymer. This polymer turned out to be about four times tougher than polymers made with standard ferrocene as the crosslinker.

“That really has big implications because if we think of all the plastics that we use and all the plastic waste accumulation, if you make materials tougher, that means their lifetime will be longer. They will be usable for a longer period of time, which could reduce plastic production in the long term,” Kevlishvili says.

The researchers now hope to use their machine-learning approach to identify mechanophores with other desirable properties, such as the ability to change color or become catalytically active in response to force. Such materials could be used as stress sensors or switchable catalysts, and they could also be useful for biomedical applications such as drug delivery.

In those studies, the researchers plan to focus on ferrocenes and other metal-containing mechanophores that have already been synthesized but whose properties are not fully understood.

“Transition metal mechanophores are relatively underexplored, and they’re probably a little bit more challenging to make,” Kulik says. “This computational workflow can be broadly used to enlarge the space of mechanophores that people have studied.”

The research was funded by the National Science Foundation Center for the Chemistry of Molecularly Optimized Networks (MONET).

M.C. Escher’s artwork is a gateway into a world of depth-defying optical illusions, featuring “impossible objects” that break the laws of physics with convoluted geometries. What you perceive his illustrations to be depends on your point of view — for example, a person seemingly walking upstairs may be heading down the steps if you tilt your head sideways. 

Computer graphics scientists and designers can recreate these illusions in 3D, but only by bending or cutting a real shape and positioning it at a particular angle. This workaround has downsides, though: Changing the smoothness or lighting of the structure will expose that it isn’t actually an optical illusion, which also means you can’t accurately solve geometry problems on it.

Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed a unique approach to represent “impossible” objects in a more versatile way. Their “Meschers” tool converts images and 3D models into 2.5-dimensional structures, creating Escher-like depictions of things like windows, buildings, and even donuts. The approach helps users relight, smooth out, and study unique geometries while preserving their optical illusion.

This tool could assist geometry researchers with calculating the distance between two points on a curved impossible surface (“geodesics”) and simulating how heat dissipates over it (“heat diffusion”). It could also help artists and computer graphics scientists create physics-breaking designs in multiple dimensions.

Lead author and MIT PhD student Ana Dodik aims to design computer graphics tools that aren’t limited to replicating reality, enabling artists to express their intent independently of whether a shape can be realized in the physical world. “Using Meschers, we’ve unlocked a new class of shapes for artists to work with on the computer,” she says. “They could also help perception scientists understand the point at which an object truly becomes impossible.”

Dodik and her colleagues will present their paper at the SIGGRAPH conference in August.

Making impossible objects possible

Impossible objects can’t be fully replicated in 3D. Their constituent parts often look plausible, but these parts don’t glue together properly when assembled in 3D. But what can be computationally imitated, as the CSAIL researchers found out, is the process of how we perceive these shapes.

Take the Penrose Triangle, for instance. The object as a whole is physically impossible because the depths don’t “add up,” but we can recognize real-world 3D shapes (like its three L-shaped corners) within it. These smaller regions can be realized in 3D — a property called “local consistency” — but when we try to assemble them together, they don’t form a globally consistent shape.

The Meschers approach models’ locally consistent regions without forcing them to be globally consistent, piecing together an Escher-esque structure. Behind the scenes, Meschers represents impossible objects as if we know their x and y coordinates in the image, as well as differences in z coordinates (depth) between neighboring pixels; the tool uses these differences in depth to reason about impossible objects indirectly.

The many uses of Meschers

In addition to rendering impossible objects, Meschers can subdivide their structures into smaller shapes for more precise geometry calculations and smoothing operations. This process enabled the researchers to reduce visual imperfections of impossible shapes, such as a red heart outline they thinned out.

The researchers also tested their tool on an “impossibagel,” where a bagel is shaded in a physically impossible way. Meschers helped Dodik and her colleagues simulate heat diffusion and calculate geodesic distances between different points of the model.

“Imagine you’re an ant traversing this bagel, and you want to know how long it’ll take you to get across, for example,” says Dodik. “In the same way, our tool could help mathematicians analyze the underlying geometry of impossible shapes up close, much like how we study real-world ones.”

Much like a magician, the tool can create optical illusions out of otherwise practical objects, making it easier for computer graphics artists to create impossible objects. It can also use “inverse rendering” tools to convert drawings and images of impossible objects into high-dimensional designs. 

“Meschers demonstrates how computer graphics tools don’t have to be constrained by the rules of physical reality,” says senior author Justin Solomon, associate professor of electrical engineering and computer science and leader of the CSAIL Geometric Data Processing Group. “Incredibly, artists using Meschers can reason about shapes that we will never find in the real world.”

Meschers can also aid computer graphics artists with tweaking the shading of their creations, while still preserving an optical illusion. This versatility would allow creatives to change the lighting of their art to depict a wider variety of scenes (like a sunrise or sunset) — as Meschers demonstrated by relighting a model of a dog on a skateboard.

Despite its versatility, Meschers is just the start for Dodik and her colleagues. The team is considering designing an interface to make the tool easier to use while building more elaborate scenes. They’re also working with perception scientists to see how the computer graphics tool can be used more broadly.

Dodik and Solomon wrote the paper with CSAIL affiliates Isabella Yu ’24, SM ’25; PhD student Kartik Chandra SM ’23; MIT professors Jonathan Ragan-Kelley and Joshua Tenenbaum; and MIT Assistant Professor Vincent Sitzmann. 

Their work was supported, in part, by the MIT Presidential Fellowship, the Mathworks Fellowship, the Hertz Foundation, the U.S. National Science Foundation, the Schmidt Sciences AI2050 fellowship, MIT Quest for Intelligence, the U.S. Army Research Office, U.S. Air Force Office of Scientific Research, SystemsThatLearn@CSAIL initiative, Google, the MIT–IBM Watson AI Laboratory, from the Toyota–CSAIL Joint Research Center, Adobe Systems, the Singapore Defence Science and Technology Agency, and the U.S. Intelligence Advanced Research Projects Activity.

In the push to shrink and enhance technologies that control light, MIT researchers have unveiled a new platform that pushes the limits of modern optics through nanophotonics, the manipulation of light on the nanoscale, or billionths of a meter.

The result is a class of ultracompact optical devices that are not only smaller and more efficient than existing technologies, but also dynamically tunable, or switchable, from one optical mode to another. Until now, this has been an elusive combination in nanophotonics.

The work is reported in the July 8 issue of Nature Photonics.

“This work marks a significant step toward a future in which nanophotonic devices are not only compact and efficient, but also reprogrammable and adaptive, capable of dynamically responding to external inputs. The  marriage of emerging quantum materials and established nanophotonics architectures will surely bring advances to both fields,” says Riccardo Comin, MIT’s Class of 1947 Career Development Associate Professor of Physics and leader of the work. Comin is also affiliated with MIT’s Materials Research Laboratory and Research Laboratory of Electronics (RLE).

Comin’s colleagues on the work are Ahmet Kemal Demir, an MIT graduate student in physics; Luca Nessi, a former MIT postdoc who is now a postdoc at Politecnico di Milano; Sachin Vaidya, a postdoc in RLE; Connor A. Occhialini PhD ’24, who is now a postdoc at Columbia University; and Marin Soljačić, the Cecil and Ida Green Professor of Physics at MIT.

Demir and Nessi are co-first authors of the Nature Photonics paper.

Toward new nanophotonic materials

Nanophotonics has traditionally relied on materials like silicon, silicon nitride, or titanium dioxide. These are the building blocks of devices that guide and confine light using structures such as waveguides, resonators, and photonic crystals. The latter are periodic arrangements of materials that control how light propagates, much like how a semiconductor crystal affects electron motion.

While highly effective, these materials are constrained by two major limitations. The first involves their refractive indices. These are a measure of how strongly a material interacts with light; the higher the refractive index, the more the material “grabs” or interacts with the light, bending it more sharply and slowing it down more. The refractive indices of silicon and other traditional nanophotonic materials are often modest, which limits how tightly light can be confined and how small optical devices can be made.

A second major limitation of traditional nanophotonic materials: once a structure is fabricated, its optical behavior is essentially fixed. There is usually no way to significantly reconfigure how it responds to light without physically altering it. “Tunability is essential for many next-gen photonics applications, enabling adaptive imaging, precision sensing, reconfigurable light sources, and trainable optical neural networks,” says Vaidya.

Introducing chromium sulfide bromide

These are the longstanding challenges that chromium sulfide bromide (CrSBr) is poised to solve. CrSBr is a layered quantum material with a rare combination of magnetic order and strong optical response. Central to its unique optical properties are excitons: quasiparticles formed when a material absorbs light and an electron is excited, leaving behind a positively charged “hole.” The electron and hole remain bound together by electrostatic attraction, forming a sort of neutral particle that can strongly interact with light.

In CrSBr, excitons dominate the optical response and are highly sensitive to magnetic fields, which means they can be manipulated using external controls.

Because of these excitons, CrSBr exhibits an exceptionally large refractive index that allows researchers to sculpt the material to fabricate optical structures like photonic crystals that are up to an order of magnitude thinner than those made from traditional materials. “We can make optical structures as thin as 6 nanometers, or just seven layers of atoms stacked on top of each other,” says Demir.

And crucially, by applying a modest magnetic field, the MIT researchers were able to continuously and reversibly switch the optical mode. In other words, they demonstrated the ability to dynamically change how light flows through the nanostructure, all without any moving parts or changes in temperature. “This degree of control is enabled by a giant, magnetically induced shift in the refractive index, far beyond what is typically achievable in established photonic materials,” says Demir.

In fact, the interaction between light and excitons in CrSBr is so strong that it leads to the formation of polaritons, hybrid light-matter particles that inherit properties from both components. These polaritons enable new forms of photonic behavior, such as enhanced nonlinearities and new regimes of quantum light transport. And unlike conventional systems that require external optical cavities to reach this regime, CrSBr supports polaritons intrinsically.

While this demonstration uses standalone CrSBr flakes, the material can also be integrated into existing photonic platforms, such as integrated photonic circuits. This makes CrSBr immediately relevant to real-world applications, where it can serve as a tunable layer or component in otherwise passive devices.

The MIT results were achieved at very cold temperatures of up to 132 kelvins (-222 degrees Fahrenheit). Although this is below room temperature, there are compelling use cases, such as quantum simulation, nonlinear optics, and reconfigurable polaritonic platforms, where the unparalleled tunability of CrSBr could justify operation in cryogenic environments.

In other words, says Demir, “CrSBr is so unique with respect to other common materials that even going down to cryogenic temperatures will be worth the trouble, hopefully.”

That said, the team is also exploring related materials with higher magnetic ordering temperatures to enable similar functionality at more accessible conditions.

This work was supported by the U.S. Department of Energy, the U.S. Army Research Office, and a MathWorks Science Fellowship. The work was performed in part at MIT.nano.

A fundamental problem for governments is getting citizens to comply with their laws and policies. They can’t monitor everyone and catch all the rule-breakers. “It’s a logistical impossibility,” says Lily L. Tsai, MIT’s Ford Professor of Political Science and the director and founder of the MIT Governance Lab.

Instead, governments need citizens to choose to follow the rules of their own accord. “As a government, you have to rely on them to voluntarily comply with the laws, policies, and regulations that are put into place,” Tsai says.

One particularly important thing governments need citizens to do is pay their taxes. In a paper in the October issue of the journal World Development, Tsai and her co-authors, including Minh Trinh ’22, a graduate of the Department of Political Science, look at different factors that might affect compliance with property tax laws in China. They found that study participants in an in-person tax-paying experiment were more likely to pay their taxes if government officials were monitoring and punishing corruption.

“When people think that government authorities are motivated by the public good, have moral character, and have integrity, then the requests that those authorities make of citizens are more likely to seem legitimate, and so they’re more likely to pay their taxes,” Tsai says.

In China, only two cities, Chongqing and Shanghai, collect property taxes. Officials have been concerned that citizens might resist property taxes because homeownership is the main source of urban household wealth in China. Private homeownership accounts for 64 percent of household wealth in China, compared to only 29 percent in the United States.

Tsai and her co-authors wanted to test how governments might make people more willing to pay their property taxes. Researchers have theorized that citizens are more likely to comply with tax laws when they feel like they’re getting something in return from the government. The government can be responsive to citizens’ demands for public services, for example. Or the government can punish officials who are corrupt or perform poorly.

In the first part of the study, a survey of Chinese citizens, respondents expressed preferences for different hypothetical property tax policies. The results suggested that participants wanted the government to be responsive to their needs and to hold officials accountable. People preferred a policy that allowed for citizen input on the use of tax revenue over one that did not, and a policy that allowed for the sanctioning of corrupt officials garnered more support than a policy that did not.

Survey participants also preferred a lighter penalty for not paying their taxes over a harsher penalty, and they supported a tax exemption for first apartments. Interestingly to the researchers, policies that allowed for government responsiveness and accountability received roughly the same support as these policies with economic benefits. “This is evidence to show that we should really pay attention to non-economic factors, because they can have similar magnitudes of impact on tax-paying behavior,” Tsai says.

For the second stage of the study, researchers recruited people for a lab experiment in Shanghai (one of the two cities that collects property taxes). Participants played a game on an iPad in which they chose repeatedly whether or not to pay property taxes. At the end of the game, they received an amount of real money that varied depending on how they and other participants played the game.

Participants were then randomly split into different groups. In one group, participants were given an opportunity to voice their preference for how their property tax revenue was used. Some were told the government incorporated their feedback, while others were told their preferences were not considered — in other words, participants learned whether or not the government was responsive to their needs. In another group, participants learned that a corrupt official had stolen money from property tax revenue. Some were told that the official had been caught and punished, while others were told the official got away with stealing.

The researchers measured whether game players’ willingness to pay property taxes changed after receiving this new information. They found that while the willingness of players who learned the government was responsive to their needs did not change significantly, players who learned the government punished corrupt officials paid their property taxes more frequently.

“It was kind of amazing to see that people care a lot about whether or not higher-level authorities are making sure that tax dollars are not being wasted through corruption,” Tsai says. She argues in her 2021 book, “When People Want Punishment: Retributive Justice and the Puzzle of Authoritarian Popularity,” that when authorities are willing to punish their own officials, it may signal to people that leaders have moral integrity and share the values of ordinary people, making them appear more legitimate.

While the researchers expected to see government responsiveness affect tax payment as well, Tsai says it’s not totally surprising that for people living in places without direct channels for citizen input, the opportunity to participate in the decision-making process in a lab setting might not resonate as strongly.

The findings don’t mean that government responsiveness isn’t important. But they suggest that even when there aren’t opportunities for citizens to make their voices heard, there are other ways for governments to appear legitimate and get people to comply with rules voluntarily.

As the strength of democratic institutions declines globally, scholars wonder whether perceptions of governments’ legitimacy will decline at the same time. “These findings suggest that maybe that’s not necessarily the case,” Tsai says.

Imagine a ball bouncing down a flight of stairs. Now think about a cascade of water flowing down those same stairs. The ball and the water behave very differently, and it turns out that your brain has different regions for processing visual information about each type of physical matter.

In a new study, MIT neuroscientists have identified parts of the brain’s visual cortex that respond preferentially when you look at “things” — that is, rigid or deformable objects like a bouncing ball. Other brain regions are more activated when looking at “stuff” — liquids or granular substances such as sand.

This distinction, which has never been seen in the brain before, may help the brain plan how to interact with different kinds of physical materials, the researchers say.

“When you’re looking at some fluid or gooey stuff, you engage with it in different way than you do with a rigid object. With a rigid object, you might pick it up or grasp it, whereas with fluid or gooey stuff, you probably are going to have to use a tool to deal with it,” says Nancy Kanwisher, the Walter A. Rosenblith Professor of Cognitive Neuroscience; a member of the McGovern Institute for Brain Research and MIT’s Center for Brains, Minds, and Machines; and the senior author of the study.

MIT postdoc Vivian Paulun, who is joining the faculty of the University of Wisconsin at Madison this fall, is the lead author of the paper, which appears today in the journal Current Biology. RT Pramod, an MIT postdoc, and Josh Tenenbaum, an MIT professor of brain and cognitive sciences, are also authors of the study.

Stuff vs. things

Decades of brain imaging studies, including early work by Kanwisher, have revealed regions in the brain’s ventral visual pathway that are involved in recognizing the shapes of 3D objects, including an area called the lateral occipital complex (LOC). A region in the brain’s dorsal visual pathway, known as the frontoparietal physics network (FPN), analyzes the physical properties of materials, such as mass or stability.

Although scientists have learned a great deal about how these pathways respond to different features of objects, the vast majority of these studies have been done with solid objects, or “things.”

“Nobody has asked how we perceive what we call ‘stuff’ — that is, liquids or sand, honey, water, all sorts of gooey things. And so we decided to study that,” Paulun says.

These gooey materials behave very differently from solids. They flow rather than bounce, and interacting with them usually requires containers and tools such as spoons. The researchers wondered if these physical features might require the brain to devote specialized regions to interpreting them.

To explore how the brain processes these materials, Paulun used a software program designed for visual effects artists to create more than 100 video clips showing different types of things or stuff interacting with the physical environment. In these videos, the materials could be seen sloshing or tumbling inside a transparent box, being dropped onto another object, or bouncing or flowing down a set of stairs.

The researchers used functional magnetic resonance imaging (fMRI) to scan the visual cortex of people as they watched the videos. They found that both the LOC and the FPN respond to “things” and “stuff,” but that each pathway has distinctive subregions that respond more strongly to one or the other.

“Both the ventral and the dorsal visual pathway seem to have this subdivision, with one part responding more strongly to ‘things,’ and the other responding more strongly to ‘stuff,’” Paulun says. “We haven’t seen this before because nobody has asked that before.”

Roland Fleming, a professor of experimental psychology at Justus Liebig University of Geissen, described the findings as a “major breakthrough in the scientific understanding of how our brains represent the physical properties of our surrounding world.”

“We’ve known the distinction exists for a long time psychologically, but this is the first time that it’s been really mapped onto separate cortical structures in the brain. Now we can investigate the different computations that the distinct brain regions use to process and represent objects and materials,” says Fleming, who was not involved in the study.

Physical interactions

The findings suggest that the brain may have different ways of representing these two categories of material, similar to the artificial physics engines that are used to create video game graphics. These engines usually represent a 3D object as a mesh, while fluids are represented as sets of particles that can be rearranged.

“The interesting hypothesis that we can draw from this is that maybe the brain, similar to artificial game engines, has separate computations for representing and simulating ‘stuff’ and ‘things.’ And that would be something to test in the future,” Paulun says.

The researchers also hypothesize that these regions may have developed to help the brain understand important distinctions that allow it to plan how to interact with the physical world. To further explore this possibility, the researchers plan to study whether the areas involved in processing rigid objects are also active when a brain circuit involved in planning to grasp objects is active.

They also hope to look at whether any of the areas within the FPN correlate with the processing of more specific features of materials, such as the viscosity of liquids or the bounciness of objects. And in the LOC, they plan to study how the brain represents changes in the shape of fluids and deformable substances.

The research was funded by the German Research Foundation, the U.S. National Institutes of Health, and a U.S. National Science Foundation grant to the Center for Brains, Minds, and Machines.

Salt creeping, a phenomenon that occurs in both natural and industrial processes, describes the collection and migration of salt crystals from evaporating solutions onto surfaces. Once they start collecting, the crystals climb, spreading away from the solution. This creeping behavior, according to researchers, can cause damage or be harnessed for good, depending on the context. New research published June 30 in the journal Langmuir is the first to show salt creeping at a single-crystal scale and beneath a liquid’s meniscus.

“The work not only explains how salt creeping begins, but why it begins and when it does,” says Joseph Phelim Mooney, a postdoc in the MIT Device Research Laboratory and one of the authors of the new study. “We hope this level of insight helps others, whether they’re tackling water scarcity, preserving ancient murals, or designing longer-lasting infrastructure.”

The work is the first to directly visualize how salt crystals grow and interact with surfaces underneath a liquid meniscus, something that’s been theorized for decades but never actually imaged or confirmed at this level, and it offers fundamental insights that could impact a wide range of fields — from mineral extraction and desalination to anti-fouling coatings, membrane design for separation science, and even art conservation, where salt damage is a major threat to heritage materials.

In civil engineering applications, for example, the research can help explain why and when salt crystals start growing across surfaces like concrete, stone, or building materials. “These crystals can exert pressure and cause cracking or flaking, reducing the long-term durability of structures,” says Mooney. “By pinpointing the moment when salt begins to creep, engineers can better design protective coatings or drainage systems to prevent this form of degradation.”

For a field like art conservation, where salt can be devastating to murals, frescoes, and ancient artifacts, often forming beneath the surface before visible damage appears, the work can help identify the exact conditions that cause salt to start moving and spreading, allowing conservators to act earlier and more precisely to protect heritage objects.

The work began during Mooney’s Marie Curie Fellowship at MIT. “I was focused on improving desalination systems and quickly ran into [salt buildup as] a major roadblock,” he says. “[Salt] was everywhere, coating surfaces, clogging flow paths, and undermining the efficiency of our designs. I realized we didn’t fully understand how or why salt starts creeping across surfaces in the first place.”

That experience led Mooney to team up with colleagues to dig into the fundamentals of salt crystallization at the air–liquid–solid interface. “We wanted to zoom in, to really see the moment salt begins to move, so we turned to in situ X-ray microscopy,” he says. “What we found gave us a whole new way to think about surface fouling, material degradation, and controlled crystallization.”

The new research may, in fact, allow better control of a crystallization processes required to remove salt from water in zero-liquid discharge systems. It can also be used to explain how and when scaling happens on equipment surfaces, and may support emerging climate technologies that depend on smart control of evaporation and crystallization.

The work also supports mineral and salt extraction applications, where salt creeping can be both a bottleneck and an opportunity. In these applications, Mooney says, “by understanding the precise physics of salt formation at surfaces, operators can optimize crystal growth, improving recovery rates and reducing material losses.”

Mooney’s co-authors on the paper include fellow MIT Device Lab researchers Omer Refet Caylan, Bachir El Fil (now an associate professor at Georgia Tech), and Lenan Zhang (now an associate professor at Cornell University); Jeff Punch and Vanessa Egan of the University of Limerick; and Jintong Gao of Cornell.

The research was conducted using in situ X-ray microscopy. Mooney says the team’s big realization moment occurred when they were able to observe a single salt crystal pinning itself to the surface, which kicked off a cascading chain reaction of growth.

“People had speculated about this, but we captured it on X-ray for the first time. It felt like watching the microscopic moment where everything tips, the ignition points of a self-propagating process,” says Mooney. “Even more surprising was what followed: The salt crystal didn’t just grow passively to fill the available space. It pierced through the liquid-air interface and reshaped the meniscus itself, setting up the perfect conditions for the next crystal. That subtle, recursive mechanism had never been visually documented before — and seeing it play out in real time completely changed how we thought about salt crystallization.”

The paper, “In Situ X-ray Microscopy Unraveling the Onset of Salt Creeping at a Single-Crystal Level,” is available now in the journal Langmuir. Research was conducted in MIT.nano. 

If you rotate an image of a molecular structure, a human can tell the rotated image is still the same molecule, but a machine-learning model might think it is a new data point. In computer science parlance, the molecule is “symmetric,” meaning the fundamental structure of that molecule remains the same if it undergoes certain transformations, like rotation.

If a drug discovery model doesn’t understand symmetry, it could make inaccurate predictions about molecular properties. But despite some empirical successes, it’s been unclear whether there is a computationally efficient method to train a good model that is guaranteed to respect symmetry.

A new study by MIT researchers answers this question, and shows the first method for machine learning with symmetry that is provably efficient in terms of both the amount of computation and data needed.

These results clarify a foundational question, and they could aid researchers in the development of more powerful machine-learning models that are designed to handle symmetry. Such models would be useful in a variety of applications, from discovering new materials to identifying astronomical anomalies to unraveling complex climate patterns.

“These symmetries are important because they are some sort of information that nature is telling us about the data, and we should take it into account in our machine-learning models. We’ve now shown that it is possible to do machine-learning with symmetric data in an efficient way,” says Behrooz Tahmasebi, an MIT graduate student and co-lead author of this study.

He is joined on the paper by co-lead author and MIT graduate student Ashkan Soleymani; Stefanie Jegelka, an associate professor of electrical engineering and computer science (EECS) and a member of the Institute for Data, Systems, and Society (IDSS) and the Computer Science and Artificial Intelligence Laboratory (CSAIL); and senior author Patrick Jaillet, the Dugald C. Jackson Professor of Electrical Engineering and Computer Science and a principal investigator in the Laboratory for Information and Decision Systems (LIDS). The research was recently presented at the International Conference on Machine Learning.

Studying symmetry

Symmetric data appear in many domains, especially the natural sciences and physics. A model that recognizes symmetries is able to identify an object, like a car, no matter where that object is placed in an image, for example.

Unless a machine-learning model is designed to handle symmetry, it could be less accurate and prone to failure when faced with new symmetric data in real-world situations. On the flip side, models that take advantage of symmetry could be faster and require fewer data for training.

But training a model to process symmetric data is no easy task.

One common approach is called data augmentation, where researchers transform each symmetric data point into multiple data points to help the model generalize better to new data. For instance, one could rotate a molecular structure many times to produce new training data, but if researchers want the model to be guaranteed to respect symmetry, this can be computationally prohibitive.

An alternative approach is to encode symmetry into the model’s architecture. A well-known example of this is a graph neural network (GNN), which inherently handles symmetric data because of how it is designed.

“Graph neural networks are fast and efficient, and they take care of symmetry quite well, but nobody really knows what these models are learning or why they work. Understanding GNNs is a main motivation of our work, so we started with a theoretical evaluation of what happens when data are symmetric,” Tahmasebi says.

They explored the statistical-computational tradeoff in machine learning with symmetric data. This tradeoff means methods that require fewer data can be more computationally expensive, so researchers need to find the right balance.

Building on this theoretical evaluation, the researchers designed an efficient algorithm for machine learning with symmetric data.

Mathematical combinations

To do this, they borrowed ideas from algebra to shrink and simplify the problem. Then, they reformulated the problem using ideas from geometry that effectively capture symmetry.

Finally, they combined the algebra and the geometry into an optimization problem that can be solved efficiently, resulting in their new algorithm.

“Most of the theory and applications were focusing on either algebra or geometry. Here we just combined them,” Tahmasebi says.

The algorithm requires fewer data samples for training than classical approaches, which would improve a model’s accuracy and ability to adapt to new applications.

By proving that scientists can develop efficient algorithms for machine learning with symmetry, and demonstrating how it can be done, these results could lead to the development of new neural network architectures that could be more accurate and less resource-intensive than current models.

Scientists could also use this analysis as a starting point to examine the inner workings of GNNs, and how their operations differ from the algorithm the MIT researchers developed.

“Once we know that better, we can design more interpretable, more robust, and more efficient neural network architectures,” adds Soleymani.

This research is funded, in part, by the National Research Foundation of Singapore, DSO National Laboratories of Singapore, the U.S. Office of Naval Research, the U.S. National Science Foundation, and an Alexander von Humboldt Professorship.

Researchers from MIT and elsewhere have designed a novel transmitter chip that significantly improves the energy efficiency of wireless communications, which could boost the range and battery life of a connected device.

Their approach employs a unique modulation scheme to encode digital data into a wireless signal, which reduces the amount of error in the transmission and leads to more reliable communications.

The compact, flexible system could be incorporated into existing internet-of-things devices to provide immediate gains, while also meeting the more stringent efficiency requirements of future 6G technologies.

The versatility of the chip could make it well-suited for a range of applications that require careful management of energy for communications, such as industrial sensors that continuously monitor factory conditions and smart appliances that provide real-time notifications.

“By thinking outside the box, we created a more efficient, intelligent circuit for next-generation devices that is also even better than the state-of-the-art for legacy architectures. This is just one example of how adopting a modular approach to allow for adaptability can drive innovation at every level,” says Muriel Médard, the School of Science NEC Professor of Software Science and Engineering, a professor in the MIT Department of Electrical Engineering and Computer Science (EECS), and co-author of a paper on the new transmitter.

Médard’s co-authors include Timur Zirtiloglu, the lead author and a graduate student at Boston University; Arman Tan, a graduate student at BU; Basak Ozaydin, an MIT graduate student in EECS; Ken Duffy, a professor at Northeastern University; and Rabia Tugce Yazicigil, associate professor of electrical and computer engineering at BU. The research was recently presented at the IEEE Radio Frequency Circuits Symposium.

Optimizing transmissions

In wireless devices, a transmitter converts digital data into an electromagnetic signal that is sent over the airwaves to a receiver. The transmitter does this by mapping digital bits to symbols that represent the amplitude and phase of the electromagnetic signal, which is a process called modulation.

Traditional systems transmit signals that are evenly spaced by creating a uniform pattern of symbols, which helps avoid interference. But this uniform structure lacks adaptability and can be inefficient, since wireless channel conditions are dynamic and often change rapidly.

As an alternative, optimal modulation schemes follow a non-uniform pattern that can adapt to changing channel conditions, maximizing the amount of data transmitted while minimizing energy usage.

But while optimal modulation can be more energy efficient, it is also more susceptible to errors, especially in crowded wireless environments. When the signals aren’t uniform in length, it can be harder for the receiver to distinguish between symbols and noise that squeezed into the transmission.

To overcome this problem, the MIT transmitter adds a small amount of padding, in the form of extra bits between symbols, so that every transmission is the same length.

This helps the receiver identify the beginning and end of each transmission, preventing misinterpretation of the message. However, the device enjoys the energy efficiency gains of using a non-uniform, optimal modulation scheme.

This approach works because of a technique the researchers previously developed known as GRAND, which is a universal decoding algorithm that crack any code by guessing the noise that affected the transmission.

Here, they employ a GRAND-inspired algorithm to adjust the length of the received transmission by guessing the extra bits that have been added. In this way, the receiver can effectively reconstruct the original message.

“Now, thanks to GRAND, we can have a transmitter that is capable of doing these more efficient transmissions with non-uniform constellations of data, and we can see the gains,” Médard says.

A flexible circuit

The new chip, which has a compact architecture that allows the researchers to integrate additional efficiency-boosting methods, enabled transmissions with only about one-quarter the amount of signal error of methods that use optimal modulation.

Surprisingly, the device also achieved significantly lower error rates than transmitters that use traditional modulation.

“The traditional approach has become so ingrained that it was challenging to not get lured back to the status quo, especially since we were changing things that we often take for granted and concepts we’ve been teaching for decades,” Médard says.

This innovative architecture could be used to improve the energy efficiency and reliability of current wireless communication devices, while also offering the flexibility to be incorporated into future devices that employ optimal modulation.

Next, the researchers want to adapt their approach to leverage additional techniques that could boost efficiency and reduce the error rates in wireless transmissions.

“This optimal modulation transmitter radio frequency integrated circuit is a game-changing innovation over the traditional RF signal modulation. It’s set to play a major role for the next generation of wireless connectivity such as 6G and Wi-Fi,” says Rocco Tam, NXP Fellow for Wireless Connectivity SoC Research and Development at NXP Semiconductors, who was not involved with this research.

This work is supported, in part, by the U.S. Defense Advanced Research Projects Agency (DARPA), the National Science Foundation (NSF), and the Texas Analog Center for Excellence. 

A lot of attention has been paid to how climate change can drive biodiversity loss. Now, MIT researchers have shown the reverse is also true: Reductions in biodiversity can jeopardize one of Earth’s most powerful levers for mitigating climate change.

In a paper published in PNAS, the researchers showed that following deforestation, naturally-regrowing tropical forests, with healthy populations of seed-dispersing animals, can absorb up to four times more carbon than similar forests with fewer seed-dispersing animals.

Because tropical forests are currently Earth’s largest land-based carbon sink, the findings improve our understanding of a potent tool to fight climate change.

“The results underscore the importance of animals in maintaining healthy, carbon-rich tropical forests,” says Evan Fricke, a research scientist in the MIT Department of Civil and Environmental Engineering and the lead author of the new study. “When seed-dispersing animals decline, we risk weakening the climate-mitigating power of tropical forests.”

Fricke’s co-authors on the paper include César Terrer, the Tianfu Career Development Associate Professor at MIT; Charles Harvey, an MIT professor of civil and environmental engineering; and Susan Cook-Patton of The Nature Conservancy.

The study combines a wide array of data on animal biodiversity, movement, and seed dispersal across thousands of animal species, along with carbon accumulation data from thousands of tropical forest sites.

The researchers say the results are the clearest evidence yet that seed-dispersing animals play an important role in forests’ ability to absorb carbon, and that the findings underscore the need to address biodiversity loss and climate change as connected parts of a delicate ecosystem rather as separate problems in isolation.

“It’s been clear that climate change threatens biodiversity, and now this study shows how biodiversity losses can exacerbate climate change,” Fricke says. “Understanding that two-way street helps us understand the connections between these challenges, and how we can address them. These are challenges we need to tackle in tandem, and the contribution of animals to tropical forest carbon shows that there are win-wins possible when supporting biodiversity and fighting climate change at the same time.”

Putting the pieces together

The next time you see a video of a monkey or bird enjoying a piece of fruit, consider that the animals are actually playing an important role in their ecosystems. Research has shown that by digesting the seeds and defecating somewhere else, animals can help with the germination, growth, and long-term survival of the plant.

Fricke has been studying animals that disperse seeds for nearly 15 years. His previous research has shown that without animal seed dispersal, trees have lower survival rates and a harder time keeping up with environmental changes.

“We’re now thinking more about the roles that animals might play in affecting the climate through seed dispersal,” Fricke says. “We know that in tropical forests, where more than three-quarters of trees rely on animals for seed dispersal, the decline of seed dispersal could affect not just the biodiversity of forests, but how they bounce back from deforestation. We also know that all around the world, animal populations are declining.”

Regrowing forests is an often-cited way to mitigate the effects of climate change, but the influence of biodiversity on forests’ ability to absorb carbon has not been fully quantified, especially at larger scales.

For their study, the researchers combined data from thousands of separate studies and used new tools for quantifying disparate but interconnected ecological processes. After analyzing data from more than 17,000 vegetation plots, the researchers decided to focus on tropical regions, looking at data on where seed-dispersing animals live, how many seeds each animal disperses, and how they affect germination.

The researchers then incorporated data showing how human activity impacts different seed-dispersing animals’ presence and movement. They found, for example, that animals move less when they consume seeds in areas with a bigger human footprint.

Combining all that data, the researchers created an index of seed-dispersal disruption that revealed a link between human activities and declines in animal seed dispersal. They then analyzed the relationship between that index and records of carbon accumulation in naturally regrowing tropical forests over time, controlling for factors like drought conditions, the prevalence of fires, and the presence of grazing livestock.

“It was a big task to bring data from thousands of field studies together into a map of the disruption of seed dispersal,” Fricke says. “But it lets us go beyond just asking what animals are there to actually quantifying the ecological roles those animals are playing and understanding how human pressures affect them.”

The researchers acknowledged that the quality of animal biodiversity data could be improved and introduces uncertainty into their findings. They also note that other processes, such as pollination, seed predation, and competition influence seed dispersal and can constrain forest regrowth. Still, the findings were in line with recent estimates.

“What’s particularly new about this study is we’re actually getting the numbers around these effects,” Fricke says. “Finding that seed dispersal disruption explains a fourfold difference in carbon absorption across the thousands of tropical regrowth sites included in the study points to seed dispersers as a major lever on tropical forest carbon.”

Quantifying lost carbon

In forests identified as potential regrowth sites, the researchers found seed-dispersal declines were linked to reductions in carbon absorption each year averaging 1.8 metric tons per hectare, equal to a reduction in regrowth of 57 percent.

The researchers say the results show natural regrowth projects will be more impactful in landscapes where seed-dispersing animals have been less disrupted, including areas that were recently deforested, are near high-integrity forests, or have higher tree cover.

“In the discussion around planting trees versus allowing trees to regrow naturally, regrowth is basically free, whereas planting trees costs money, and it also leads to less diverse forests,” Terrer says. “With these results, now we can understand where natural regrowth can happen effectively because there are animals planting the seeds for free, and we also can identify areas where, because animals are affected, natural regrowth is not going to happen, and therefore planting trees actively is necessary.”

To support seed-dispersing animals, the researchers encourage interventions that protect or improve their habitats and that reduce pressures on species, ranging from wildlife corridors to restrictions on wildlife trade. Restoring the ecological roles of seed dispersers is also possible by reintroducing seed-dispersing species where they’ve been lost or planting certain trees that attract those animals.

The findings could also make modeling the climate impact of naturally regrowing forests more accurate.

“Overlooking the impact of seed-dispersal disruption may overestimate natural regrowth potential in many areas and underestimate it in others,” the authors write.

The researchers believe the findings open up new avenues of inquiry for the field.

“Forests provide a huge climate subsidy by sequestering about a third of all human carbon emissions,” Terrer says. “Tropical forests are by far the most important carbon sink globally, but in the last few decades, their ability to sequester carbon has been declining. We will next explore how much of that decline is due to an increase in extreme droughts or fires versus declines in animal seed dispersal.”

Overall, the researchers hope the study helps improves our understanding of the planet’s complex ecological processes.

“When we lose our animals, we’re losing the ecological infrastructure that keeps our tropical forests healthy and resilient,” Fricke says.

The research was supported by the MIT Climate and Sustainability Consortium, the Government of Portugal, and the Bezos Earth Fund.

Scientists often seek new materials derived from polymers. Rather than starting a polymer search from scratch, they save time and money by blending existing polymers to achieve desired properties.

But identifying the best blend is a thorny problem. Not only is there a practically limitless number of potential combinations, but polymers interact in complex ways, so the properties of a new blend are challenging to predict.

To accelerate the discovery of new materials, MIT researchers developed a fully autonomous experimental platform that can efficiently identify optimal polymer blends.

The closed-loop workflow uses a powerful algorithm to explore a wide range of potential polymer blends, feeding a selection of combinations to a robotic system that mixes chemicals and tests each blend.

Based on the results, the algorithm decides which experiments to conduct next, continuing the process until the new polymer meets the user’s goals.

During experiments, the system autonomously identified hundreds of blends that outperformed their constituent polymers. Interestingly, the researchers found that the best-performing blends did not necessarily use the best individual components.

“I found that to be good confirmation of the value of using an optimization algorithm that considers the full design space at the same time,” says Connor Coley, the Class of 1957 Career Development Assistant Professor with shared appointments in the MIT departments of Chemical Engineering and Electrical Engineering and Computer Science and the MIT Schwarzman College of Computing, and senior author of a paper on this new approach. “If you consider the full formulation space, you can potentially find new or better properties. Using a different approach, you could easily overlook the underperforming components that happen to be the important parts of the best blend.”

This workflow could someday facilitate the discovery of polymer blend materials that lead to advancements like improved battery electrolytes, more cost-effective solar panels, or tailored nanoparticles for safer drug delivery.

Coley is joined on the paper by lead author Guangqi Wu, a former MIT postdoc who is now a Marie Skłodowska-Curie Postdoctoral Fellow at Oxford University; Tianyi Jin, an MIT graduate student; and Alfredo Alexander-Katz, the Michael and Sonja Koerner Professor in the MIT Department of Materials Science and Engineering. The work appears today in Matter.

Building better blends

When scientists design new polymer blends, they are faced with a nearly endless number of possible polymers to start with. Once they select a few to mix, they still must choose the composition of each polymer and the concentration of polymers in the blend.

“Having that large of a design space necessitates algorithmic solutions and higher-throughput workflows because you simply couldn’t test all the combinations using brute force,” Coley adds.

While researchers have studied autonomous workflows for single polymers, less work has focused on polymer blends because of the dramatically larger design space.

In this study, the MIT researchers sought new random heteropolymer blends, made by mixing two or more polymers with different structural features. These versatile polymers have shown particularly promising relevance to high-temperature enzymatic catalysis, a process that increases the rate of chemical reactions.

Their closed-loop workflow begins with an algorithm that, based on the user’s desired properties, autonomously identifies a handful of promising polymer blends.

The researchers originally tried a machine-learning model to predict the performance of new blends, but it was difficult to make accurate predictions across the astronomically large space of possibilities. Instead, they utilized a genetic algorithm, which uses biologically inspired operations like selection and mutation to find an optimal solution.

Their system encodes the composition of a polymer blend into what is effectively a digital chromosome, which the genetic algorithm iteratively improves to identify the most promising combinations.

“This algorithm is not new, but we had to modify the algorithm to fit into our system. For instance, we had to limit the number of polymers that could be in one material to make discovery more efficient,” Wu adds.

In addition, because the search space is so large, they tuned the algorithm to balance its choice of exploration (searching for random polymers) versus exploitation (optimizing the best polymers from the last experiment).

The algorithm sends 96 polymer blends at a time to the autonomous robotic platform, which mixes the chemicals and measures the properties of each.

The experiments were focused on improving the thermal stability of enzymes by optimizing the retained enzymatic activity (REA), a measure of how stable an enzyme is after mixing with the polymer blends and being exposed to high temperatures.

These results are sent back to the algorithm, which uses them to generate a new set of polymers until the system finds the optimal blend.

Accelerating discovery

Building the robotic system involved numerous challenges, such as developing a technique to evenly heat polymers and optimizing the speed at which the pipette tip moves up and down.

“In autonomous discovery platforms, we emphasize algorithmic innovations, but there are many detailed and subtle aspects of the procedure you have to validate before you can trust the information coming out of it,” Coley says.

When tested, the optimal blends their system identified often outperformed the polymers that formed them. The best overall blend performed 18 percent better than any of its individual components, achieving an REA of 73 percent.

“This indicates that, instead of developing new polymers, we could sometimes blend existing polymers to design new materials that perform even better than individual polymers do,” Wu says.

Moreover, their autonomous platform can generate and test 700 new polymer blends per day and only requires human intervention for refilling and replacing chemicals.

While this research focused on polymers for protein stabilization, their platform could be modified for other uses, like the development or new plastics or battery electrolytes.

In addition to exploring additional polymer properties, the researchers want to use experimental data to improve the efficiency of their algorithm and develop new algorithms to streamline the operations of the autonomous liquid handler.

“Technologically, there are urgent needs to enhance thermal stability of proteins and enzymes. The results demonstrated here are quite impressive. Being a platform technology and given the rapid advancement in machine learning and AI for material science, one can envision the possibility for this team to further enhance random heteropolymer performances or to optimize design based on end needs and usages,” says Ting Xu, a professor at the University of California at Berkeley, who was not involved with this work.

This work is funded, in part, by the U.S. Department of Energy, the National Science Foundation, and the Class of 1947 Career Development Chair.

MIT physicists have performed an idealized version of one of the most famous experiments in quantum physics. Their findings demonstrate, with atomic-level precision, the dual yet evasive nature of light. They also happen to confirm that Albert Einstein was wrong about this particular quantum scenario.

The experiment in question is the double-slit experiment, which was first performed in 1801 by the British scholar Thomas Young to show how light behaves as a wave. Today, with the formulation of quantum mechanics, the double-slit experiment is now known for its surprisingly simple demonstration of a head-scratching reality: that light exists as both a particle and a wave. Stranger still, this duality cannot be simultaneously observed. Seeing light in the form of particles instantly obscures its wave-like nature, and vice versa.

The original experiment involved shining a beam of light through two parallel slits in a screen and observing the pattern that formed on a second, faraway screen. One might expect to see two overlapping spots of light, which would imply that light exists as particles, a.k.a. photons, like paintballs that follow a direct path. But instead, the light produces alternating bright and dark stripes on the screen, in an interference pattern similar to what happens when two ripples in a pond meet. This suggests light behaves as a wave. Even weirder, when one tries to measure which slit the light is traveling through, the light suddenly behaves as particles and the interference pattern disappears.

The double-slit experiment is taught today in most high school physics classes as a simple way to illustrate the fundamental principle of quantum mechanics: that all physical objects, including light, are simultaneously particles and waves.

Nearly a century ago, the experiment was at the center of a friendly debate between physicists Albert Einstein and Niels Bohr. In 1927, Einstein argued that a photon particle should pass through just one of the two slits and in the process generate a slight force on that slit, like a bird rustling a leaf as it flies by. He proposed that one could detect such a force while also observing an interference pattern, thereby catching light’s particle and wave nature at the same time. In response, Bohr applied the quantum mechanical uncertainty principle and showed that the detection of the photon’s path would wash out the interference pattern.

Scientists have since carried out multiple versions of the double-slit experiment, and they have all, to various degrees, confirmed the validity of the quantum theory formulated by Bohr. Now, MIT physicists have performed the most “idealized” version of the double-slit experiment to date. Their version strips down the experiment to its quantum essentials. They used individual atoms as slits, and used weak beams of light so that each atom scattered at most one photon. By preparing the atoms in different quantum states, they were able to modify what information the atoms obtained about the path of the photons. The researchers thus confirmed the predictions of quantum theory: The more information was obtained about the path (i.e. the particle nature) of light, the lower the visibility of the interference pattern was. 

They demonstrated what Einstein got wrong. Whenever an atom is “rustled” by a passing photon, the wave interference is diminished.

“Einstein and Bohr would have never thought that this is possible, to perform such an experiment with single atoms and single photons,” says Wolfgang Ketterle, the John D. MacArthur Professor of Physics and leader of the MIT team. “What we have done is an idealized Gedanken experiment.”

Their results appear in the journal Physical Review Letters. Ketterle’s MIT co-authors include first author Vitaly Fedoseev, Hanzhen Lin, Yu-Kun Lu, Yoo Kyung Lee, and Jiahao Lyu, who all are affiliated with MIT’s Department of Physics, the Research Laboratory of Electronics, and the MIT-Harvard Center for Ultracold Atoms.

Cold confinement

Ketterle’s group at MIT experiments with atoms and molecules that they super-cool to temperatures just above absolute zero and arrange in configurations that they confine with laser light. Within these ultracold, carefully tuned clouds, exotic phenomena that only occur at the quantum, single-atom scale can emerge.

In a recent experiment, the team was investigating a seemingly unrelated question, studying how light scattering can reveal the properties of materials built from ultracold atoms.

“We realized we can quantify the degree to which this scattering process is like a particle or a wave, and we quickly realized we can apply this new method to realize this famous experiment in a very idealized way,” Fedoseev says.

In their new study, the team worked with more than 10,000 atoms, which they cooled to microkelvin temperatures. They used an array of laser beams to arrange the frozen atoms into an evenly spaced, crystal-like lattice configuration. In this arrangement, each atom is far enough away from any other atom that each can effectively be considered a single, isolated and identical atom. And 10,000 such atoms can produce a signal that is more easily detected, compared to a single atom or two.

The group reasoned that with this arrangement, they might shine a weak beam of light through the atoms and observe how a single photon scatters off two adjacent atoms, as a wave or a particle. This would be similar to how, in the original double-slit experiment, light passes through two slits.

“What we have done can be regarded as a new variant to the double-slit experiment,” Ketterle says. “These single atoms are like the smallest slits you could possibly build.”

Tuning fuzz

Working at the level of single photons required repeating the experiment many times and using an ultrasensitive detector to record the pattern of light scattered off the atoms. From the intensity of the detected light, the researchers could directly infer whether the light behaved as a particle or a wave.

They were particularly interested in the situation where half the photons they sent in behaved as waves, and half behaved as particles. They achieved this by using a method to tune the probability that a photon will appear as a wave versus a particle, by adjusting an atom’s “fuzziness,” or the certainty of its location. In their experiment, each of the 10,000 atoms is held in place by laser light that can be adjusted to tighten or loosen the light’s hold. The more loosely an atom is held, the fuzzier, or more “spatially extensive,” it appears. The fuzzier atom rustles more easily and records the path of the photon. Therefore, in tuning up an atom’s fuzziness, researchers can increase the probability that a photon will exhibit particle-like behavior. Their observations were in full agreement with the theoretical description.

Springs away

In their experiment, the group tested Einstein’s idea about how to detect the path of the photon. Conceptually, if each slit were cut into an extremely thin sheet of paper that was suspended in the air by a spring, a photon passing through one slit should shake the corresponding spring by a certain degree that would be a signal of the photon’s particle nature. In previous realizations of the double slit experiment, physicists have incorporated such a spring-like ingredient, and the spring played a major role in describing the photon’s dual nature.

But Ketterle and his colleagues were able to perform the experiment without the proverbial springs. The team’s cloud of atoms is initially held in place by laser light, similar to Einstein’s conception of a slit suspended by a spring. The researchers reasoned that if they were to do away with their “spring,” and observe exactly the same phenomenon, then it would show that the spring has no effect on a photon’s wave/particle duality.

This, too, was what they found. Over multiple runs, they turned off the spring-like laser holding the atoms in place and then quickly took a measurement in a millionth of a second,  before the atoms became more fuzzy and eventually fell down due to gravity. In this tiny amount of time, the atoms were effectively floating in free space. In this spring-free scenario, the team observed the same phenomenon: A photon’s wave and particle nature could not be observed simultaneously.

“In many descriptions, the springs play a major role. But we show, no, the springs do not matter here; what matters is only the fuzziness of the atoms,” Fedoseev says. “Therefore,  one has to use a more profound description, which uses quantum correlations between photons and atoms.”

The researchers note that the year 2025 has been declared by the United Nations as the International Year of Quantum Science and Technology, celebrating the formulation of quantum mechanics 100 years ago. The discussion between Bohr and Einstein about the double-slit experiment took place only two years later.

“It’s a wonderful coincidence that we could help clarify this historic controversy in the same year we celebrate quantum physics,” says co-author Lee.

This work was supported, in part, by the National Science Foundation, the U.S. Department of Defense, and the Gordon and Betty Moore Foundation.

In an office at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), a soft robotic hand carefully curls its fingers to grasp a small object. The intriguing part isn’t the mechanical design or embedded sensors — in fact, the hand contains none. Instead, the entire system relies on a single camera that watches the robot’s movements and uses that visual data to control it.

This capability comes from a new system CSAIL scientists developed, offering a different perspective on robotic control. Rather than using hand-designed models or complex sensor arrays, it allows robots to learn how their bodies respond to control commands, solely through vision. The approach, called Neural Jacobian Fields (NJF), gives robots a kind of bodily self-awareness. An open-access paper about the work was published in Nature on June 25.

“This work points to a shift from programming robots to teaching robots,” says Sizhe Lester Li, MIT PhD student in electrical engineering and computer science, CSAIL affiliate, and lead researcher on the work. “Today, many robotics tasks require extensive engineering and coding. In the future, we envision showing a robot what to do, and letting it learn how to achieve the goal autonomously.”

The motivation stems from a simple but powerful reframing: The main barrier to affordable, flexible robotics isn't hardware — it’s control of capability, which could be achieved in multiple ways. Traditional robots are built to be rigid and sensor-rich, making it easier to construct a digital twin, a precise mathematical replica used for control. But when a robot is soft, deformable, or irregularly shaped, those assumptions fall apart. Rather than forcing robots to match our models, NJF flips the script — giving robots the ability to learn their own internal model from observation.

Look and learn

This decoupling of modeling and hardware design could significantly expand the design space for robotics. In soft and bio-inspired robots, designers often embed sensors or reinforce parts of the structure just to make modeling feasible. NJF lifts that constraint. The system doesn’t need onboard sensors or design tweaks to make control possible. Designers are freer to explore unconventional, unconstrained morphologies without worrying about whether they’ll be able to model or control them later.

“Think about how you learn to control your fingers: you wiggle, you observe, you adapt,” says Li. “That’s what our system does. It experiments with random actions and figures out which controls move which parts of the robot.”

The system has proven robust across a range of robot types. The team tested NJF on a pneumatic soft robotic hand capable of pinching and grasping, a rigid Allegro hand, a 3D-printed robotic arm, and even a rotating platform with no embedded sensors. In every case, the system learned both the robot’s shape and how it responded to control signals, just from vision and random motion.

The researchers see potential far beyond the lab. Robots equipped with NJF could one day perform agricultural tasks with centimeter-level localization accuracy, operate on construction sites without elaborate sensor arrays, or navigate dynamic environments where traditional methods break down.

At the core of NJF is a neural network that captures two intertwined aspects of a robot’s embodiment: its three-dimensional geometry and its sensitivity to control inputs. The system builds on neural radiance fields (NeRF), a technique that reconstructs 3D scenes from images by mapping spatial coordinates to color and density values. NJF extends this approach by learning not only the robot’s shape, but also a Jacobian field, a function that predicts how any point on the robot’s body moves in response to motor commands.

To train the model, the robot performs random motions while multiple cameras record the outcomes. No human supervision or prior knowledge of the robot’s structure is required — the system simply infers the relationship between control signals and motion by watching.

Once training is complete, the robot only needs a single monocular camera for real-time closed-loop control, running at about 12 Hertz. This allows it to continuously observe itself, plan, and act responsively. That speed makes NJF more viable than many physics-based simulators for soft robots, which are often too computationally intensive for real-time use.

In early simulations, even simple 2D fingers and sliders were able to learn this mapping using just a few examples. By modeling how specific points deform or shift in response to action, NJF builds a dense map of controllability. That internal model allows it to generalize motion across the robot’s body, even when the data are noisy or incomplete.

“What’s really interesting is that the system figures out on its own which motors control which parts of the robot,” says Li. “This isn’t programmed — it emerges naturally through learning, much like a person discovering the buttons on a new device.”

The future is soft

For decades, robotics has favored rigid, easily modeled machines — like the industrial arms found in factories — because their properties simplify control. But the field has been moving toward soft, bio-inspired robots that can adapt to the real world more fluidly. The trade-off? These robots are harder to model.

“Robotics today often feels out of reach because of costly sensors and complex programming. Our goal with Neural Jacobian Fields is to lower the barrier, making robotics affordable, adaptable, and accessible to more people. Vision is a resilient, reliable sensor,” says senior author and MIT Assistant Professor Vincent Sitzmann, who leads the Scene Representation group. “It opens the door to robots that can operate in messy, unstructured environments, from farms to construction sites, without expensive infrastructure.”

“Vision alone can provide the cues needed for localization and control — eliminating the need for GPS, external tracking systems, or complex onboard sensors. This opens the door to robust, adaptive behavior in unstructured environments, from drones navigating indoors or underground without maps to mobile manipulators working in cluttered homes or warehouses, and even legged robots traversing uneven terrain,” says co-author Daniela Rus, MIT professor of electrical engineering and computer science and director of CSAIL. “By learning from visual feedback, these systems develop internal models of their own motion and dynamics, enabling flexible, self-supervised operation where traditional localization methods would fail.”

While training NJF currently requires multiple cameras and must be redone for each robot, the researchers are already imagining a more accessible version. In the future, hobbyists could record a robot’s random movements with their phone, much like you’d take a video of a rental car before driving off, and use that footage to create a control model, with no prior knowledge or special equipment required.

The system doesn’t yet generalize across different robots, and it lacks force or tactile sensing, limiting its effectiveness on contact-rich tasks. But the team is exploring new ways to address these limitations: improving generalization, handling occlusions, and extending the model’s ability to reason over longer spatial and temporal horizons.

“Just as humans develop an intuitive understanding of how their bodies move and respond to commands, NJF gives robots that kind of embodied self-awareness through vision alone,” says Li. “This understanding is a foundation for flexible manipulation and control in real-world environments. Our work, essentially, reflects a broader trend in robotics: moving away from manually programming detailed models toward teaching robots through observation and interaction.”

This paper brought together the computer vision and self-supervised learning work from the Sitzmann lab and the expertise in soft robots from the Rus lab. Li, Sitzmann, and Rus co-authored the paper with CSAIL affiliates Annan Zhang SM ’22, a PhD student in electrical engineering and computer science (EECS); Boyuan Chen, a PhD student in EECS; Hanna Matusik, an undergraduate researcher in mechanical engineering; and Chao Liu, a postdoc in the Senseable City Lab at MIT. 

The research was supported by the Solomon Buchsbaum Research Fund through MIT’s Research Support Committee, an MIT Presidential Fellowship, the National Science Foundation, and the Gwangju Institute of Science and Technology.