Halide perovskites absorb and emit light better than almost any other known solar material — thin enough to coat a phone, flexible enough to wrap a skyscraper. For years, researchers have called them the most promising path to next-generation solar cells.
Yet a fundamental piece of their structure has stubbornly resisted explanation.
That gap has mattered more with every passing year. Global electricity demand — currently around 20% of total energy consumption — is expected to surpass 50% within 25 years, according to the International Energy Agency. The pressure to unlock cleaner, more efficient solar technology has never been greater. So what, exactly, has been hiding inside this material?
A material almost too good to be true
Halide perovskites occupy a rare category: materials that seem almost purpose-built for the job. They absorb and emit light with exceptional efficiency, making them strong candidates not just for solar cells but for LED devices. Their physical properties allow for cells thin and flexible enough to coat a smartphone screen or wrap the entire facade of a skyscraper.
The persistent catch has been durability. These materials degrade quickly, and without a clear understanding of why, engineers have struggled to design around the problem. One compound in particular — formamidinium lead iodide, or FAPbI3 — stood out for its outstanding optoelectronic properties. It also remained poorly understood, and its instability limited how widely it could be deployed.
The missing piece in the research puzzle
The instability problem isn’t unsolvable in principle. Researchers have known that mixing two types of halide perovskites can help stabilize the material, but controlling that mixture requires deep knowledge of both components — and one critical piece of that knowledge was absent.
The low-temperature phase of FAPbI3 had long resisted explanation through experiments alone. Researcher Sangita Dutta at Chalmers put it plainly: “The low-temperature phase of this material has long been a missing piece of the research puzzle and we’ve now settled a fundamental question about the structure of this phase.”
What the team found was subtle but consequential. As FAPbI3 cools, the formamidinium molecules inside get trapped in a semi-stable state — a behavior that shapes the material’s structure in ways that matter significantly for how it performs and how it might eventually be stabilized.
How machine learning changed the simulation game
The Chalmers team’s work centers on building accurate computer models of materials and testing them under varying conditions. Halide perovskites have always been difficult to model, though. Capturing their properties in full requires powerful supercomputers and long simulation run times — resources that, until recently, made comprehensive modeling impractical.
The breakthrough came from combining standard simulation methods with machine learning potentials, and the gain in scale was substantial. “By combining our standard methods with machine learning, we’re now able to run simulations that are thousands of times longer than before,” Dutta explained. “And our models can now contain millions of atoms instead of hundreds, which brings them closer to the real world.” That leap in computational scale was what finally made the hidden low-temperature structure visible.
Simulations meet the real world at –200°C
Computational findings are only as credible as the experimental evidence behind them. To validate their models, the Chalmers team collaborated with researchers at the University of Birmingham, who cooled physical samples of FAPbI3 to –200°C in the lab. The results aligned closely with the simulation models — an agreement that gives the team confidence their structural findings accurately reflect what’s happening inside the material. The study was published in the Journal of the American Chemical Society in August 2025.
What this means for the future of solar energy
The immediate practical value of this work is a clearer roadmap. Engineers now have better information for designing and controlling perovskite mixtures — the approach most likely to address the instability problem that has held this class of materials back. More stable perovskites could broaden the deployment of ultra-thin, flexible solar cells across a far wider range of surfaces and applications than is currently feasible.
The broader significance extends beyond this one compound. The simulation framework itself — machine learning potentials combined with conventional methods — is now a general-purpose tool applicable to other complex halide perovskite materials. As principal investigator Julia Wiktor noted, the team now has “simulation methods that can answer questions that were unresolved just a few years ago.”
With electricity demand expected to more than double its share of global energy consumption within 25 years, the push to develop more efficient solar materials is only intensifying. Resolving a long-standing structural mystery in one of the field’s most promising compounds is a meaningful step forward — and the computational tools that made it possible may prove just as significant as the finding itself.