Advances in artificial intelligence promise to help chemical engineers discover complex new materials. These materials could be used for reactions such as turning carbon dioxide into fuel, but technical barriers have limited catalysis adoption so far. Researchers at the Ģ����ý are now harnessing the benefits of large language models (LLMs) similar to ChatGPT, Claude, or Gemini to empower more researchers to use AI to discover new materials and accelerate experiment workflows.

In a published in ACS Central Science, a team led by , an associate professor in the , and , visiting associate professor and the cofounder and chief technology officer of , describes an AI-based method they developed that allows users to input natural language prompts about the materials they want to create and suggest optimal procedures for experiments to produce them. As the users run the experiments, they input the results back into the AI model and continue iterating until they reach their goal.

“We’re able to leverage the pre-trained knowledge of large language models and well-established statistical methods for materials discovery to help us as researchers navigate large experimental design spaces more efficiently,” says Porosoff.

Porosoff likens the new AI method to describing a cup of coffee, noting that someone could describe the coffee by its taste, color, and aroma, or by the type of beans, grind size, apparatus, and water temperature used to make the brew. Both representation methods describe the same cup of coffee, but the second approach gives you a recipe to reproduce it that others can easily replicate.

Porosoff and his team are applying the same principle to catalysts for energy applications, using language-based representations to describe materials not just by their properties, but by the steps needed to create them.

To build on their success, the US Department of Energy Advanced Research Projects Agency-Energy (ARPA-E) it will provide nearly $3 million in funding to apply the Ģ����ý team’s method toward creating catalysts for the production of fuel from abundant materials, specifically methanol and ethanol from carbon dioxide and hydrogen. Porosoff will lead a multi-institution project team that includes URochester, Virginia Polytechnic Institute and State University, Stanford University, Northwestern University, A*STAR Institute of Sustainability for Chemicals, Energy and Environment (ISCE2) in Singapore, and OxEon Energy, a small business based in Salt Lake City.

Leveraging the power of LLMs

Traditional AI methods for materials discovery typically use a strategy called Bayesian optimization to identify and design the best candidates. But the result is complex numerical data about a material’s structure, which requires deep expertise to use effectively. The new LLM method instead produces a set of procedures that researchers can easily understand, execute, and verify to determine if the experiment’s output matches the predicted results.

This can be extremely useful for working with complex materials such as trimetallic catalysts, which are made of three metals.

“Our method reduces the technical barrier associated with using Bayesian optimization, which is a well-established method for efficiently exploring large and complicated parameter spaces,” says Shane Michtavy, a URochester chemical engineering PhD student who helped develop the AI method, synthesize materials, and run the chemical reactions described in the paper. “Using pre-trained LLMs allows users to explore using less data than traditional models, as they are deployed in a frozen state with built-in knowledge of the physical world and catalysis.”

The paper shows how the researchers applied the method to several live experiments, including one to identify catalysts for turning carbon dioxide and hydrogen into carbon monoxide and water using trimetallic catalysts made from low-cost metals. Porosoff says that there are about 360,000 possible experiments that could have been run to find the ideal catalyst, but by using procedures produced by the AI model and providing it with the results from the experiments, they were able to find an ideal candidate in just ten experiments.

The study was supported by funding from the National Science Foundation, the National Institutes of Health, and the US Department of Energy. Additional authors included Mayk Caldas, technical staff at Edison Scientific.

Next steps

Now that they have shown the model works as a proof of concept in the lab, Porosoff aims to take the method further using the funding announced through ARPA-E’s Catalytic Application Testing for Accelerated Learning Chemistries via High-throughput Experimentation and Modeling Efficiently (CATALCHEM-E) program.

“Right now, it takes a decade or longer to go from conceptualizing a new catalyst to testing it in a lab to putting it in a real reactor,” says Porosoff. “The CATALCHEM-E program aims to cut that by an order of magnitude to a single year, and we think using AI with text-based representations will be a big factor in shortening the development cycle.”

Porosoff and his collaborators will first demonstrate their workflow on carbon dioxide-to-methanol and then extend the process to higher alcohols such as ethanol, which is a key additive for gasoline and used in pharmaceuticals, cosmetics, and many other applications. Ultimately, they hope to commercially deploy the model for industries to create catalysts to synthesize alcohols for fuel.

The project is scheduled to begin in July and run through 2029. See a full list of CATALCHEM-E programs on the .

The world’s oceans may be quietly amplifying climate change in ways scientists are only beginning to understand.

In a published in the journal Proceedings of the National Academy of Sciences, ��������Գپ����ٲ�—i�Գ����ܻ徱�Բ� , an associate professor in the , as well as graduate student Shengyu Wang and postdoctoral research associate Hairong Xu in Weber’s lab—uncovered a key mechanism behind methane production in the open ocean. Their research indicates that this mechanism could intensify as the planet warms, providing an alarming feedback loop for global warming.

Methane is a powerful greenhouse gas, and for decades scientists have puzzled over a paradox: surface ocean waters consistently release methane into the atmosphere, even though surface water is rich in oxygen. Traditionally, methane production has been associated with oxygen-free environments, such as wetlands or deep sediments.

Weber’s team set out to solve this puzzle using a global dataset and computer modeling. Their findings point to a specific microbial process that is responsible for methane production in the ocean environment: certain bacteria generate methane as a byproduct when they break down organic compounds, but they only do this when the nutrient phosphate is scarce.

“This means that phosphate scarcity is the primary control knob for methane production and emissions in the open ocean,” Weber says.

The findings reframe how scientists understand methane production in the ocean. Rather than being a rare or unusual process, methane production in oxygen-rich environments may be widespread in regions where phosphate is limited.

But the study extends further than explaining marine methane production in the present—it also offers a troubling glimpse into the future.

“Climate change is warming the ocean from the top down, increasing the density difference between surface and deep waters,” Weber says. “This is expected to slow the vertical mixing that carries nutrients like phosphate up from depth.”

According to the team’s model, with less vertical mixing, surface waters could become increasingly nutrient-starved, creating ideal conditions for methane-producing microbes to thrive.

The result, Weber warns, would be more methane released from the ocean into the atmosphere. Because methane is such a potent greenhouse gas, this creates the potential for a harmful feedback loop: warming oceans lead to more methane emissions, which in turn drive further warming.

The findings highlight how even processes occurring at the microscopic level in the ocean can have global consequences.

Crucially, this feedback is not currently included in major climate projection models. As researchers continue to refine climate models, incorporating feedbacks such as this may be essential for accurately predicting the pace and scale of future climate change.

“Our work will help fill a key gap in climate predictions, which often overlook interactions between the changing environment and natural greenhouse gas sources to the atmosphere,” Weber says.

When lasers were invented in the 1960s, they opened new avenues for scientific discovery and everyday applications from scanners at the grocery store to corrective eye surgery. Conventional lasers control photons—individual particles of light—but over the past 20 years, scientists have invented lasers that control other fundamental particles, including phonons—individual particles of vibration or sound. Controlling phonons could open even more possibilities with lasers, such as taking advantage of unique quantum properties like entanglement.

A new squeezed phonon laser developed by researchers at the and Rochester Institute of Technology provides precise control over phonons at the nanoscale level. This could give new insights into the nature of gravity, particle acceleration, and quantum physics. In in Nature Communications, the researchers describe how they coax these individual particles of mechanical motion to behave like a laser.

, the Marie C. Wilson and Joseph C. Wilson Professor of Optical Physics with the Ģ����ý , and his collaborators first demonstrated a phonon laser by trapping and levitating phonons with an optical tweezer in a vacuum in 2019. But to make this technology useful for extremely accurate measurements, they had to overcome a key obstacle fundamental to both photon and phonon lasers: noise, or unwanted disturbances that make a signal difficult to accurately read.

“While a laser looks to the naked eye like a steady beam, there’s actually a lot of fluctuation, which causes noise when you’re using lasers for measurement,” says Vamivakas. “By pushing and pulling on a phonon laser with light in the right way, we can reduce that phonon laser fluctuation significantly.”

Specifically, the researchers were able to squeeze or reduce the thermal noise intrinsic to the phonon laser. Vamivakas says that noise reduction provides the ability to measure acceleration more accurately than techniques that use photon lasers or radio frequency waves.

Vamivakas envisions researchers using the phonon laser to obtain pinpoint accurate measurements of gravity and other forces, which could be important in applications such as navigation. Scientists have envisioned quantum compasses as more accurate, “unjammable” alternatives to GPS navigation that do not require the use of satellites, and Vamivakas is intrigued by seeing if the phonon laser could be a step toward such systems.

The research was supported by the National Science Foundation. Vamivakas’ collaborators on the paper include Ģ����ý optics PhD student Kai Zhang, RIT postdoctoral researcher Kewen Xiao, and Mishkat Bhattacharya ’05, a professor of physics at RIT.

The Ģ����ý’s (LLE) and have established a $6.9 million partnership, the largest single industrial-sponsored research agreement awarded to LLE, to address fundamental challenges in  and accelerate progress toward practical, sustainable fusion power.

When you get better at a skill—recognizing a familiar face in a crowd, spotting a typo at a glance, or anticipating the next move in a game—sensory neurons in your brain become more coordinated, sharing information rather than acting more independently. That’s the conclusion of a by researchers at the and its , published in Science, which challenges a long-held assumption in neuroscience that learning improves efficiency by minimizing repetition across neural signals.

Led by Shizhao Liu, a graduate student in the labs of and , both faculty members in the , the study shows that learning instead increases shared activity among neurons. The findings could provide insights into learning disorders and inspire more flexible, human-like artificial intelligence tools.

“The dominant view in neuroscience has been that learning makes the brain more efficient by pushing neurons to act more independently, so information can be read out more cleanly,” Liu says. “Our results support a different idea, that sensory areas of the brain aren’t just passively encoding the world. They’re actively performing inference by combining what’s coming in with what the brain has learned to expect.”

How learning reshapes neural teamwork

For decades, researchers believed that learning streamlined how the brain processes information by reducing shared activity among neurons, allowing information to be read out more efficiently. The idea shaped how researchers thought about everything from perception to decision-making.

But the research from Liu, Haefner, Snyder, and their team suggests a different mechanism. Rather than becoming more independent, neurons become more coordinated as learning unfolds, increasing the amount of information they share, particularly when the brain is actively engaged in a task and making decisions.

This coordination reflects the brain’s growing reliance on internal expectations. As learning progresses, feedback from higher-level brain areas appears to shape how sensory neurons respond, allowing perception to incorporate both incoming information and what the brain has learned from past experiences.

Tracking neurons as learning unfolds

The researchers tracked the activity of the same small networks of neurons in the visual cortex over several weeks as subjects learned to tell apart different visual patterns. The team measured whether neurons were increasingly acting on their own or sharing more information as learning progressed.

The researchers discovered that before learning, neurons mostly worked independently. But as subjects honed their visual skills, the neurons started to behave more like a well-trained sports team, communicating and working together in a coordinated way.

“It’s a bit like a group of people solving a problem,” Snyder says. “Instead of everyone working in isolation as efficiently as possible, learning makes them communicate more. That shared information makes each individual better informed and potentially makes the group more flexible and adaptive.”

Importantly, this coordinated effect only appeared when subjects were actively performing a task and making decisions based on what they saw. When they passively looked at the same images without needing to respond, the effect disappeared.

The neurons most important for the task showed the biggest boost in coordination, especially at the moments when decisions were made.

But these are flexible, not permanent, changes. The researchers believe these shifts are guided by feedback signals from higher-level brain areas, allowing neurons to adjust their behavior on the fly, depending on the task.

The results support a growing idea in neuroscience that the brain isn’t a simple conveyor belt that passes information forward. Instead, it constantly blends what we see with what we expect to see, creating a richer, more informed picture of the world. And that blending requires groups of neurons to act together, not separately.

Insights for health and AI

Understanding how the brain coordinates neurons during learning could provide new insights into learning disorders and conditions that affect perception. It could also help scientists design artificial intelligence systems that generalize better by taking inspiration from the way the brain flexibly blends prior expectations with new sensory information.

“Most current artificial intelligence systems are built on discriminative architectures that map sensory inputs directly to outputs,” Haefner says. “Our new research suggests that incorporating generative feedback loops—in which internal models shape sensory representations—may lead to systems that learn faster from limited data, are more robust to uncertainty, and adapt more flexibly to changing tasks.”

When you go for a walk, how does your brain know the difference between a parked car and a moving car? This seemingly simple distinction is challenging because eye movements, such as the ones we make when watching a car pass by, make even stationary objects move across the retina—motion that has long been thought of as visual “noise” the brain must subtract out.

Now, researchers at the have discovered that instead of being meaningless interference, the visual motion of an image caused by eye movements helps us understand the world. The specific patterns of visual motion created by eye movements are useful to the brain for figuring out how objects move and where they are located in 3D space.

“The conventional idea has been that the brain needs to somehow discount, or subtract off, the image motion that is produced by eye movements, as this motion has been thought to be a nuisance,” says , George Eastman Professor; professor in the Departments of , , and and the ; member of the ; and lead author of the new research, published in . “But we found that the visual motion produced by our eye movements is not just a nuisance variable to be subtracted off; rather, our brains analyze these global patterns of image motion and use this to infer how our eyes have moved relative to the world.”

The research team developed a new theoretical framework to predict how humans should perceive an object’s motion and depth during different types of eye movements. They tested these predictions by having participants view 3D virtual environments in which a target object moved through a scene while the participants kept their eyes focused on a single point. In one task, participants estimated the direction the target object was moving by using a dial to match its motion with a second object. In a second task that measured depth perception, participants reported whether the target object appeared nearer or farther than the fixation spot. Across both tasks, the researchers found consistent, predictable patterns of errors that closely matched the theoretical predictions.

“We show that the brain considers many pieces of information to understand the 3D structure of the world through vision, including the patterns of image motion caused by eye movements,” says DeAngelis. “Contrary to conventional ideas, the brain doesn’t ignore or suppress image motion produced by eye movement. Instead, it uses this image motion to understand a scene and accurately estimate an object’s motion and depth.”

This research has important implications for understanding visual perception, which informs how the brain interprets everyday activities like reading and recognizing faces. But it could also provide insight and new applications for visual technologies, such as virtual reality headsets.

“VR headsets don’t factor in how the eyes are moving relative to the scene when they compute the images to show to each eye. There may be a stark mismatch between the image motion that is shown to the observer in VR and what the brain is expecting to receive based on the eye movements that the observer is making,” says DeAngelis. This could be what causes some people to experience motion sickness while using a VR headset.

Additional authors include first author Zhe-Xin Xu ’25 (PhD), a former graduate student in the DeAngelis lab who is now a postdoctoral fellow at Harvard University; Jiayi Pang ’25 (BS), who is now a graduate student at Brown University; and Akiyuki Anzai, a research associate at the URochester. The National Institutes of Health supported this research.

The has awarded the Ģ����ý a $1.3 million grant for research at the forefront of how light and matter interact. The project, titled “Quantum Electrodynamics for Selective Transformations,” aims to create new chemistry using quantum light. The ambitious project has the potential to unlock new opportunities for chemical and material synthesis.

“We are thrilled to receive support from the W. M. Keck Foundation that will allow us to pursue high-risk, high-reward research that we hope will open up new frontiers at the intersection of chemistry, photonics, and quantum science,” says , the Jay Last Professor in Arts, Sciences & Engineering in the and the .

Krauss leads a team of researchers that includes , the Dean and Laura Marvin Endowed Professor in Physical Chemistry and an associate professor of optics; Dan Weix at the University of Wisconsin–Madison (and former faculty member at Ģ����ý), and Rachel Bangle at North Carolina A&T State University.

“The work of Professor Krauss and his team is an example of Rochester’s long tradition of working across cutting-edge disciplines to advance science and improve our understanding of the physical world,” says University President Sarah Mangelsdorf. “We’re grateful for the support of the W. M. Keck Foundation in recognizing the enormous potential in this research.”

Using quantum light to create new chemistry

A grand challenge in the field of chemistry is controlling chemical bond formation at any stage in a reaction.

Chemistry is governed by an established set of rules that dictate how simple molecules react with each other to form new, more complex molecules. These rules are related to how electrons are distributed in the molecules and underpin the field of synthetic chemistry. The constraints imposed by these rules have a direct impact on society because they can limit access to potential new drugs or materials. In the past, chemists have used temperature, pressure, light, and other ways to control and perform chemistry.

For the newly funded project, Ģ����ý researchers and their colleagues at other institutions seek to discover if it is possible to use the quantum light of an optical cavity to bend or break these fundamental rules of reactivity by changing how electrons are distributed. To test the idea, researchers will couple light inside an optical cavity to the electronic states of molecules, forming a hybrid light-matter state called an electron-polariton.

While polariton chemistry has the potential to alter the fundamental rules of chemical reactivity, verifying this new concept experimentally has been challenging because of the varied expertise required. To overcome that hurdle, Krauss has assembled just such a diverse team, including synthetic organic chemists, materials scientists, spectroscopists, and theoreticians, who will work to help establish this new field of research.

Krauss notes, “It isn’t often that one has the chance to discover a new set of rules that govern the makeup of matter in the universe.”

Ģ����ý the Keck Foundation

The W. M. Keck Foundation was established in 1954 in Los Angeles by William Myron Keck, founder of The Superior Oil Company. One of the nation’s largest philanthropic organizations, the W. M. Keck Foundation supports outstanding science, engineering, and medical research. The foundation also supports undergraduate education and maintains a program within Southern California to support arts and culture, education, health, and community service projects.