Research News

A decade of New York City fire data shows that citywide trends can mask sharply different realities on the ground, with some neighborhoods moving in the opposite direction of the city as a whole.

A new study published in the Fire Safety Journal breaks down more than 411,000 structural fire incidents recorded by the FDNY between 2013 and 2023 across roughly 2,200 census tracts, small statistical subdivisions of a few thousand residents each, rather than the borough- or citywide figures typically used to track fire trends.

Hannah Bonestroo, a Ph.D. student in Urban Systems at NYU Tandon School of Engineering, conducted the research with Assistant Professor Augustin Guibaud, a faculty member of the Department of Mechanical and Aerospace Engineering affiliated with the NYU Tandon’s Center for Urban Science + Progress.

"If the city is looking at inequalities at the neighborhood scale, or allocating resources at the neighborhood scale, there may be areas with different patterns that get missed," said Bonestroo.

At the citywide level, the story looked simple. Structural fires fell 11.3 percent over the decade, a clear improvement. But that average hid real variation underneath it.

"If you look much more granularly at the census tract level, there were hot spots, areas where fires were actually increasing over the ten years,” said Bonestroo. Eastern Queens and Staten Island were among the places moving against the citywide trend.

Response time showed a different pattern entirely, forming large clusters of similar trends across the city rather than concentrating in pockets. Average response time rose 30 seconds over the decade, from 4 minutes 6 seconds to 4 minutes 36 seconds, now exceeding the National Fire Protection Association's 4-minute travel-time standard.

"In a fire, every second counts," said Guibaud. "Thirty seconds may not sound like much, but it can be enough time for a fire to intensify and for smoke to fill a space, and the first crew on scene can arrive to a situation that's already beyond what they can handle."

"Almost the entire city is following the pattern of it increasing," Bonestroo said, a contrast to fire frequency, which varied sharply block to block. Northwestern Queens saw the sharpest deterioration in response time over the decade.

Recent reviews by the FDNY and the New York State Senate have pointed to traffic congestion and rising call volume – the department responds to more than a million emergency calls a year – rather than neighborhood demographics or distance from firehouses. The researchers' analysis agrees, as fixed socioeconomic and built-environment factors explained only about a quarter of the differences in response time between tracts.

That contrast pointed to a second finding. Fire frequency and response time capture largely different dimensions of risk. Areas with the most fires tended to have faster response times, not slower, something Bonestroo reads as reassuring rather than puzzling.

 "That's a good indicator that FDNY has placed its resources where the fires are," Bonestroo said.

Because the two measures move independently, Bonestroo and Guibaud built a Dual Risk Index, designed to catch tracts that don't top the list on fire count or response time individually, but rank high on both at once. To put it simple, fire frequency reflects the likelihood of a fire. Response time shapes how bad it can get once one starts. The index multiplies the two.

 "If you just looked at the top 50 tracts by number of fires, you might miss a tract that has a fairly high number of fires and fairly slow response time, just not the highest number and the slowest time," Bonestroo said. "That's the kind of area the index is meant to catch."

Mapped across the city, high dual-risk areas cluster in Manhattan, the Bronx, and at the city's periphery, while central Brooklyn and Queens score lowest.

Tested against the locations of multi-alarm fires, the most severe incidents, over the decade, the index performed modestly better than either individual measure at flagging high-priority tracts. None of the measures, including the index, strongly pinpointed where the city's most severe fires occurred.

"It showed that it did capture more, but it wasn't statistically significant once you account for the confidence intervals," Bonestroo said.

The researchers say that shows the need for better public data on fire outcomes. "The paper isn't meant to tell FDNY where they need to reallocate resources," Bonestroo added. "Rather, it’s meant to reveal neighborhoods that may not be following overall citywide trends and are worthy of further inquiry to understand why." The index is a starting point, as the researchers plan to test other ways of accounting for other factors such as overall building safety records.

The analysis relies entirely on public data: FDNY's Fire Incident Dispatch Data and NYFIRS, the federal NFIRS system (replaced by a new national reporting system, NERIS, in February 2026), the Mayor's Management Report, and U.S. Census Bureau estimates. “Unfortunately, these sources vary in their reporting, even though records are meant to pass between the systems unchanged, so even the most basic count depends on which record you consulted” said Guibaud.

For 2023, the city's official tally in the Mayor's Management Report lists 35,567 fires, against 30,301 for the federal NFIRS database records, some 5,300 fewer. Individual incidents were sometimes classified differently from one dataset to the next, and Bonestroo carefully mapped the discrepancy by identifying where individual fire event landed in each database.

Bonestroo received a Sheldon Tieszen Student Award in June 2026 at the International Symposium on Fire Safety Science in La Rochelle, France, for the paper.

Bonestroo, H., & Augustin, G. (2026). Structural fires and emergency response in New York City: A spatiotemporal analysis of fire incidence and severity. Fire Safety Journal, 162, 104718. https://doi.org/10.1016/j.firesaf.2026.104718

A new framework for thinking about artificial intelligence at work starts with a question companies have spent relatively little time asking.

What happens if the AI goes away?

Companies are racing to build generative AI into everyday work, from drafting and data analysis to scheduling and decision support. Vedant Das Swain of NYU Tandon and Koustuv Saha of the University of Illinois Urbana-Champaign argue that organizations should also prepare for the opposite possibility: a budget cut, outage, privacy rule, regulatory action or vendor dispute that suddenly makes AI unavailable.

The team describes their Counterfactual Resilience Framework, or CReF, in a Visionary Paper that they presented at the inaugural ACM AI Leadership Summit, which ran August 30 to September 2 in Atlanta. According to Das Swain, the paper is meant to inspire transformative research agendas and identify grand challenges from the engagement of the entire AI ecosystem.

CReF proposes a way for organizations and researchers to expose dependencies that may be difficult to see while AI is working normally. The framework asks users to construct a "counterfactual anchor" with three parts: a workplace setting in which AI is being used, an event that removes access to it, and the aftermath.

The idea is to treat AI's absence as a kind of stress test.

"A lot of research on human-AI teaming and complementarity assumes that LLMs and generative AI tools are like oxygen,” Das Swain said. “ It will be abundant, uniformly accessible, and continually replenished. We wanted to challenge this assumption by urging organizations to ask how they would look if AI was suddenly snatched away. Our framework helps stakeholders confront who looks productive, who retains expertise or whether basic work can continue in the face of a technological barrier, outage, or attack. Existing AI productivity frameworks excite us by focusing on the best case. Instead, we shift attention to pragmatically planning for the worst case."

The paper explores three potential vulnerabilities: whether AI benefits are distributed unevenly across jobs; whether prolonged AI use could erode skills workers need to operate independently; and whether organizations can recover when an AI-dependent workflow is disrupted.

To illustrate them, the authors construct fictional workplaces.

In one, Alex manages complicated relationships and gets relatively little benefit from AI, while Sam works on a data analytics team whose output increases after adopting it. Management begins interpreting their differing output through an AI-influenced notion of productivity. When rising costs temporarily cut off the organization's AI access, Alex's work changes little. The scenario asks whether AI can distort performance expectations when its usefulness differs dramatically among jobs.

Another scenario involves Tanya, a junior compliance analyst who has spent two years reviewing AI-drafted risk assessments rather than regularly writing them herself. When an audit prompts her employer to suspend the system, she struggles to return to manual drafting. The example is fictional, but the concern is not new. The authors connect it to decades of research on automation showing that removing people from routine practice can erode expertise they may later need during unusual or high-stakes situations.

The paper's third scenario imagines an elder-care facility that can no longer afford an AI system woven into care planning and scheduling. Workers trying to reconstruct the previous workflow discover that institutional knowledge once distributed among spreadsheets, handwritten logs and employees has become difficult to recover.

CReF does not predict that these things will happen, and it does not assign organizations a resilience score. The authors describe it instead as a "generative analysis tool" intended to expose hidden dependencies and potential safeguards.

Companies should not abandon AI, the researchers argue, but they should develop resilience alongside adoption. The question for an AI-dependent workplace, in other words, may not only be how much better people perform when the technology is available. It may also be what remains when it isn't.

Das Swain, Vedant & Saha, Koustuv. (2026). AI Resilient Future of Work: A Counterfactual Lens on Human-AI Collaboration.

Traffic jams are usually treated as a problem to be measured after they form. Anton Rozhkov and his colleagues are exploring a different approach: giving planners state-of-the-art tools that can anticipate congestion, locate where it is most likely to emerge and translate complex traffic models into plain-language guidance.

In a new study published in the special issue of Transactions in GIS on Ethical and Explainable GeoAI, Rozhkov, Pranav Nitin Motarwar and Rudra Patil, all affiliated with NYU Tandon, developed a geospatial artificial intelligence (GeoAI) framework that combines traffic forecasting, spatial mapping and a locally hosted large language model. Rozhkov led the project and designed the framework, with Motarwar leading the forecasting work and Patil the spatial analysis. The system was tested using 15 years of New York City traffic data.

The motivation came partly from conversations Rozhkov had with urban planning practitioners. Cities increasingly want to incorporate AI into their work, he says, but planning departments may lack the technical expertise to build specialized systems. They may also be reluctant to upload sensitive transportation data to commercial models and chatbots.

“We started with an idea: what if we developed our own AI platform, one that could be hosted locally and would be secure, intuitive, and comfortable for planners to use in their day-to-day work,” Rozhkov said, describing a system that could be hosted locally and used directly by planners.

Traffic offered a useful test case because New York makes extensive transportation data publicly available. The researchers analyzed traffic observations from 2009 through 2024 and tested two methods for predicting future volumes. One was ARIMA, a conventional statistical forecasting technique. The other was a Long Short-Term Memory, or LSTM, neural network, which can identify nonlinear patterns extending across long sequences of data.

Motarwar, who trained the forecasting models, benchmarked the two approaches against each other. “ARIMA gives you seasonality, which is true but not the whole story,” he said. “The LSTM picks up the parts of the pattern that don’t repeat cleanly, and that’s where most of the improvement came from.”

On previously unseen data from 2021 through 2024, the LSTM performed better. Its average prediction error was about 343 vehicles per day, compared with roughly 418 for ARIMA, an improvement of approximately 18 percent. The model then projected average daily traffic increasing from 12,540 vehicles in 2025 to 19,680 in 2029, although those figures come with substantial prediction intervals and should be understood as forecasts rather than certainties.

But forecasting was only one piece of the project. Patil led the spatial component, mapping traffic onto Uber’s H3 system, which divides geographic space into nested hexagonal cells. Unlike citywide averages, this approach allows planners to examine congestion at multiple scales, from borough-level patterns down to localized hotspots. Clustering algorithms then identified areas where heavy traffic repeatedly accumulated. Unsurprisingly, Manhattan emerged as the highest-congestion borough in the analysis, followed by Brooklyn and Queens.

“A citywide average doesn’t help the city planners,” Patil said. “What they need to know is which corridors are badly impacted, and those turn out to be consistent year to year. That is what the hexagonal mapping and clustering were designed to explore.”

The final component turns those analytical results into something closer to a conversation. Building on the forecast and congestion maps, the researchers built a customized portal using Meta’s LLaMA model connected to a project-specific knowledge base containing forecasts, congestion locations and other traffic information. A planner could ask, for example, how a highway expansion near LaGuardia Airport might affect Queens traffic or what could happen if one-way streets in Lower Manhattan were converted to two-way streets.

Crucially, the system is designed to run locally. Agencies could keep sensitive transportation information “behind their own firewall,” Rozhkov said, rather than sending it to an outside AI service, even if sometimes the capabilities of the commercial models might be better. “A general chatbot gives you a reasonable-sounding generic paragraph about congestion,” Motarwar added. “Planners need an answer that comes from their own data. We saw this as much a data problem as a model problem.”

The researchers are not claiming that the platform can solve congestion on its own. Nor has it yet undergone a large-scale trial with working planners. Rozhkov hopes to test future versions with smaller municipalities and other planning datasets. Instead, the project demonstrates how several emerging tools can be combined into a single planning workflow. Rather than asking AI to make urban decisions, the researchers envision it as an interface between planners and increasingly complicated datasets.

That distinction also explains why the paper appeared in a special issue devoted to ethical and explainable GeoAI. For Rozhkov, the question is not simply whether cities can use increasingly powerful AI. It is how they can use it while keeping the underlying data, assumptions and decisions under local control that will eventually benefit local residents.

Propeller-powered underwater vehicles have long helped scientists explore and monitor aquatic environments. But they're limited by their own mechanics: spinning blades can snag on vegetation, stir up sediment, and startle the wildlife they're often sent to study, making them poorly suited to shallow creeks, dense weeds, or close encounters with fish.

That's one reason roboticists have spent years building machines that swim like fish instead, bending their bodies rather than spinning a propeller. The catch is that most fish-inspired robots are built for one size and one job, so scaling them up or down usually means starting from scratch.

A team of engineers says it's found a way to solve that problem. They've unveiled ScaFi, a robot modeled on fish like cod and mackerel. These fish swim by concentrating most of their body bending toward the tail end, a style that, in nature, spans an unusually wide range of body sizes.

"Right now, if you want to monitor a creek and then monitor a lake, you basically need two different robots, built and tested from the ground up," said NYU Tandon’s Nana Obayashi, currently an assistant professor of mechanical and aerospace engineering and a faculty member of the NYU Center for Robotics and Embodied Intelligence, who led the project while a doctoral researcher at EPFL. "The environments we care about don't come in one size, so we don't think the tools should either."

As described in a paper in npj Robotics, ScaFi has a rigid front section and a flexible tail made of fiberglass rods. A single motor pulls two tendons that cross near the tail's end, producing the "S"-shaped bend required for fish-like swimming motion.

The diameter of the rods forming the tail are the only part that must change with the size of the robot. They grow proportionally thicker as the robot scales up, to preserve similar tail-bending behavior. The underlying motor mechanism and crossed-tendon system stay the same.

That matters because it could cut the engineering effort needed to build fish-like robots for different environments. It also gives researchers a platform for studying how swimming performance changes with scale, a question that's hard to study systematically in animals or custom-made robots alike.

The team built three robots — roughly 0.6, 1.1, and 2.9 meters long — and tested how well each swam. The smallest produced swirling water patterns similar to those left by real fish, and across all three sizes, swimming motion lined up closely once adjusted for body size, evidence the authors say that their scaling approach preserved the fish-like gait even as the robots grew nearly fivefold in length.

They also deployed the robots in the field: the medium-sized one in a Swiss stream, the largest on Lake Geneva, the smallest in creeks only 15–30 centimeters deep. During the stream test, the robot kept swimming even after a GPS dropout.

Energy efficiency proved harder to scale. The two smaller robots performed similarly, but the largest was consistently less efficient and needed a different, more powerful motor. The authors suggest drag and inertia may be to blame, though the exact cause is unresolved, meaning the team scaled the swimming motion itself more cleanly than the energy it takes to produce it.

A similar tradeoff showed up in disturbance tests. The smallest robot was most agile but recovered slowest after being knocked off course, while the larger robots were less nimble but more stable.

The researchers suggest the same approach — scaling around one key structural parameter — could apply to other compliant robots, including ones outside water. Whether energetic performance can be scaled as successfully as the swimming motion remains an open question.

The study adds to broader aquatic robotics efforts at NYU Tandon. Industry Professor Christopher Clark's research includes autonomous underwater robotic systems for exploration and environmental monitoring, while Institute Professor Maurizio Porfiri has pioneered the use of biomimetic robotic fish to study and influence animal behavior.

The co-authors on Obayashi's paper are Josie Hughes, Alexandros Anastasiadis, Karen Mulleners, Kai Junge, and Kyle L. Walker of EPFL, and Jessica Gumowski of Queensland University of Technology. The research was partially funded by the European Union's Horizon 2020 programme under Marie Skłodowska-Curie grant agreement No. 945363.

Obayashi, N., Anastasiadis, A., Gumowski, J. et al. ScaFi: length-scalable, compliant, parametric robotic fish design for operation in multiple environmental niches. npj Robot 4, 38 (2026).

Scientists have long known that ant colonies sometimes seem to move as one. A nest that appears quiet can suddenly erupt into activity, with workers throughout the colony springing into motion almost simultaneously before settling back into stillness.

These synchronized bursts, first documented more than three decades ago, have intrigued biologists because they resemble collective phenomena seen in systems as diverse as neurons, fireflies and even chemical reactions.

Now, new research from engineers and biologists at New York University and the New Jersey Institute of Technology offers a mathematical explanation for how these rhythmic waves of activity emerge.

The study, published in Physical Review X Life, suggests that synchronized bursts arise when a colony balances two competing forces: the ability of a single active ant to rapidly excite its nestmates and the colony’s capacity to fully return to rest before the next wave begins.

Using a computational model grounded in empirical observations of ant behavior, the researchers found that colonies undergo a kind of phase transition — a sudden shift from unsynchronized movement to coordinated collective activity.

“Activity bursts emerge as a balance between the responsiveness of the colony to the first ant that activates and the ability of the colony to completely deactivate before the onset of the next burst,” said lead author Michael Napoli, a doctoral researcher in the Department of Mechanical and Aerospace Engineering at NYU.

The team combined decades of observations of ant movement with established theories of social activation. In the model, ants can occupy one of three states: active, inactive or refractory — a temporary resting period during which they cannot immediately become active again. Active ants move through a virtual nest and interact with others, sometimes triggering them to become active as well.

What emerged from the simulations was a surprisingly powerful role for individual workers. Rather than requiring many ants to coordinate simultaneously, a single ant often acted as the spark that ignited a colony-wide cascade of activity.

The researchers call this worker the “first mover.” Once activated, that ant can stimulate others, which in turn activate additional nestmates, creating a rapid chain reaction that sweeps through the colony. The process resembles a line of falling dominoes or the spread of information through a social network.

“Our results indicate that activity bursts in ant colonies are the result of a first mover that excites the colony in a synchronized regime, thereby favoring the rapid communication of new behaviors throughout the group,” the authors write.

The study also revealed that speed matters. Ants appear to operate in what the researchers describe as a “high-speed interaction regime,” where information spreads through the nest far more quickly than the duration of an activity burst itself. Under these conditions, workers constantly form and break social connections as they move, allowing information to travel efficiently across the colony.

According to senior author Maurizio Porfiri, the findings suggest that synchronized behavior depends not simply on how many interactions occur, but on how rapidly information can propagate through the network of moving individuals.

“The timescale of the motion of individuals through the nest is faster than that of the burst, suggesting that ants operate in a high-speed interaction regime where new behaviors are near-instantaneously transferred through the nest,” Porfiri said.

Although the study focuses on ants, its implications extend beyond insect societies. Simon Garnier, Professor of Biological Sciences at NJIT and coauthor on the paper, suggests similar leader-driven cascades appear in many complex systems, from grazing sheep that suddenly cluster together to neurons firing in coordinated patterns. By identifying the conditions that promote synchronization, the researchers hope to uncover general principles that govern collective behavior across biology.

The work could even inspire new approaches to engineering. Swarms of robots, for example, often rely on local interactions rather than centralized control. Understanding how a single agent can trigger coordinated action across a large population could help designers create more efficient systems for tasks such as warehouse logistics, environmental monitoring or disaster response.

The authors caution that their model simplifies many aspects of real ant colonies, including differences among workers and the complex spatial organization of nests. Future experiments will test whether real colonies operate near the synchronization threshold predicted by the model and whether manipulating density or movement patterns can alter the emergence of activity bursts.

For now, the research offers a compelling explanation for one of social insects’ most mysterious behaviors. What appears to be a colony acting with a single mind may actually begin with one ant taking the first step — and thousands of others rapidly following its lead.

This research was supported by a grant from the National Science Foundation.

Napoli, M., Garnier, S., & Porfiri, M. (2026). Nest-Level Phase Transition Drives Synchronized Activity Bursts in Ant Colonies. PRX Life, 4, 033010. doi:10.1103/ghnl-p5c1

As demand for electricity surges from artificial intelligence, data centers, and other new technologies, companies are increasingly pledging to procure carbon-free electricity to mitigate emissions associated with their electricity use. While quantifying system-wide emissions is relatively straightforward, the complex nature of power grid operations makes it more challenging to ascertain the emissions impacts of a single consumer that also procures its own clean energy. This challenge is further compounded by clean energy sources like solar and wind, whose hour to hour generation can vary drastically, depending on weather patterns.

A new study by researchers at NYU Tandon and the Massachusetts Institute of Technology evaluated the system impacts of so-called 24/7 or hourly matching, involving matching grid electricity consumption with generation from procured clean electricity, a strategy of growing corporate and regulatory interest. By incorporating multiple years of weather variability into sophisticated power-system models, the researchers found that procuring clean electricity to match consumption every hour of the day can deliver emissions benefits, but at higher cost than previous studies suggested. The work also indicates that in regions already pursuing aggressive renewable energy policies, less strict approaches may achieve nearly the same climate benefits for considerably less money.

The study, published in Environmental Science & Technology, focuses on electricity-based production of hydrogen, a pathway of growing interest to decarbonize difficult-to-electrify segments of the economy like heavy industry. Recently, governments in both the United States and Europe have proposed or implemented rules requiring producers to match their electricity use with generation from newly built renewable energy resources, initially on an annual basis and eventually on an hourly basis. Beyond hydrogen, the findings apply broadly to other large electricity consumers, including data centers, and speak directly to ongoing revisions of the Greenhouse Gas Protocol's Scope 2 rules, the standard many companies use to account for and report the emissions tied to their electricity use, where hourly matching is a central proposal under debate.

Until now, however, most analyses on hourly matching have relied on a single year of weather data to characterize wind and solar resource variability. In reality, wind and solar resources fluctuate significantly from year to year.

"We wanted to understand how real-world, inter-annual weather variability changes the economics and emissions of clean electricity procurement," said senior author Dharik S. Mallapragada, associate professor of chemical and biomolecular engineering at NYU Tandon. "When projects are expected to operate for decades, designing them around a single weather year can paint an overly optimistic picture of both cost and performance."

To explore that question, the researchers modeled a grid-connected hydrogen facility operating in the Texas power grid, one of the nation's largest renewable energy markets. They optimized investments across nine representative years of historical wind and solar conditions, both individually and collectively, and then tested those designs for their robustness against additional weather scenarios.

The team's model consistently showed that hourly matching costs more than annual matching and that the cost premium is sensitive to the weather year, varying between ranging from $0.68 to $1.18/kg of H2 produced. For context, the average cost of fossil H2 production is around $1/kg. The team also showed that such single weather-year based investment plans can lead to shortfalls in clean electricity supply across multiple hours of the year when tested against additional weather scenarios not included in the planning analysis.

The cost of clean electricity procurement to be robust to inter-annual weather variations was found to lead to a still higher cost premium, of $1.29 per kilogram compared with annual matching. This cost premium stems from oversizing the capacity of wind and solar farms to manage shortfalls in supply as well as installing larger electrolyzers capable of ramping production up and down, along with extensive hydrogen storage to buffer periods when renewable generation falls short.

The researchers also examined ways to reduce costs of 24/7 matching without sacrificing most of the environmental benefits. Allowing hydrogen producers to match80 to 90 percent of their hourly electricity needs with clean power lowered costs while retaining much of the emissions benefits achieved under full compliance. Likewise, if hydrogen production operates within electricity grids already governed by strong renewable portfolio standards, annual matching performed nearly as well as hourly matching from an emissions standpoint, but at much lower costs. A further option, with similar emissions outcomes at a comparable or lower cost, is to fold new electricity demand directly into those existing renewable portfolio standards rather than layering separate, granular, matching requirements on top. These findings speak directly to ongoing debates among governments and corporations over the role of "24/7 matching" in voluntary and regulatory emissions reporting.

Michael A. Giovanniello, Dharik S. Mallapragada; Emissions and Cost Trade-Offs of Time-Matched Clean Electricity Procurement under Interannual Weather Variability: A Case Study of Hydrogen Production. Environ. Sci. Technol. 21 July 2026; 60 (28): 19854–19865. https://doi.org/10.1021/acs.est.6c00988

Solar panels have become the most iconic symbol of the clean energy transition. They cover rooftops, stretch across deserts, and quietly convert sunlight into electricity without emitting carbon dioxide. But hidden behind this success story is a growing environmental challenge that few people think about: what happens when millions of those panels reach the end of their lives?

By 2050, discarded photovoltaic modules could generate more than 80 million metric tons of waste. Unless new recycling methods are developed, many of those panels could end up in landfills, taking with them valuable metals that required enormous amounts of energy and mining to produce in the first place.

"Solar panels are the clean energy infrastructure of the future," says Juanita Hidalgo, Assistant Professor of Chemical and Biomolecular Engineering at NYU Tandon. "But to make solar truly sustainable, we also need to think about what happens after these technologies reach the end of their lifetime."

A new perspective paper, coauthored by Hidalgo and post-doctoral researcher Sara Hamilton, argues that one of the most promising solutions lies not in hotter furnaces or more intensive manufacturing, but in chemistry. Researchers are increasingly turning to hydrometallurgy, a family of recycling techniques that uses liquids to selectively dissolve and recover valuable metals. Instead of heating an entire solar panel at temperatures approaching 2,000 degrees Celsius, hydrometallurgy carefully separates individual components at low temperatures so they can be reused in new devices.

"It's a much more selective approach," explains Hamilton. "Rather than treating a solar panel as waste, we're treating it as a source of valuable materials that can be recovered and put back into the supply chain."

Today's recycling systems recover relatively simple materials such as aluminum frames, glass, and copper wiring. But the heart of every solar panel contains metals that are both economically valuable and strategically important. Silver, indium, gallium, tellurium, and lead all play critical roles in different kinds of solar cells, yet recovering them remains technically difficult and often too expensive to justify. Current recycling methods frequently rely on pyrometallurgy, which uses extremely high temperatures to melt materials apart. While effective, the process consumes large amounts of energy and can make it difficult to separate individual metals cleanly.

Hydrometallurgy offers a more tunable alternative. Carefully selected solvents dissolve specific metals, which can then be purified and recovered for reuse. In principle, the approach requires less energy and can recover materials with much greater precision.

But the researchers found that not all solar technologies are equally easy to recycle. Conventional crystalline silicon panels, which account for roughly 95 percent of the global solar market, typically require strong acids such as nitric acid to extract valuable silver. Those acids work well, but they are corrosive, hazardous to handle, and difficult to recycle themselves. Thin-film solar cells face similar challenges. Although they contain smaller amounts of material overall, they rely on critical metals such as indium, gallium, and tellurium that are usually recovered using equally aggressive chemical treatments.

One surprise was the recyclability of the hottest new solar technology. Perovskite solar cells have generated enormous excitement because they can be manufactured at lower cost than conventional silicon while achieving record-setting efficiencies in the laboratory. The new perspective suggests they are also remarkably well suited for environmentally friendly recycling.

Unlike conventional solar cells, perovskites are built from layers connected by relatively weak chemical interactions. As a result, several recent studies have shown that researchers can recover one of their most important ingredients — lead — using something unexpectedly simple: hot water. As the water cools, the dissolved lead crystallizes back into a compound that can be used to manufacture new perovskite solar cells.

The authors argue that recyclability should become a design goal rather than an afterthought. Instead of maximizing efficiency first and worrying about disposal decades later, engineers could build future solar cells with disassembly and material recovery in mind from the very beginning.