Imagine a display that harvests ambient light when it is not actively in use, offsetting some of its own energy consumption. The materials physics shows that this is possible, the same semiconductor material can, in principle, emit and absorb light efficiently. What has been missing is a device architecture that allows it to do both without reductions in efficiency of either application. A new study reports a perovskite diode that converts sunlight to electricity at 26.7% efficiency (a world record at the time of publication) and emits light at 31% efficiency, figures that would be high for a device designed to do only one of those things.
Metal-halide perovskites are a class of materials named for their distinctive crystal structure, that have emerged over the past decade as some of the most promising candidates for next-generation solar cells and light-emitting diodes (LEDs). They are relatively inexpensive to produce, can be tuned to absorb or emit different wavelengths of light, and have shown efficiency levels that rival far more costly semiconductor materials. Yet despite sharing the same underlying material, perovskite solar cells and perovskite LEDs have largely been developed as separate technologies, because the physical requirements of each push device design in opposite directions. A collaborative study published in Joule by a team led by Michael McGehee at the University of Colorado Boulder, and Jixian Xu at the University of Science and Technology of China, now demonstrates that this conflict can be resolved, and that resolving it improves both devices at once.
The challenge of doing two things at once
The tension between perovskite LEDs and solar cells comes down to a question of thickness. An effective LED needs an extremely thin, discontinuous layer of perovskite, typically around 50 nanometers (roughly one thousandth the width of a human hair), because thin, slightly uneven films naturally scatter light outward, helping photons escape the device. A solar cell, by contrast, needs a layer roughly sixteen times thicker to absorb enough incoming sunlight and convert it into electricity efficiently. For years, this meant that researchers optimizing a perovskite LED were building something poorly suited to harvesting solar energy, and vice versa. Thanks to these different needs the two applications have followed separate architectural paths, and devices that attempted to do both tended to do neither particularly well.
There is a further complication. Even in a well-made perovskite LED device, much of the light generated inside never escapes. When a photon (a particle of light) is produced inside the material, it travels outward and hits the surface. If it arrives at too steep an angle, it is reflected back inside rather than escaping, a phenomenon governed by the physics of how light moves between materials with different optical properties. Once trapped, that photon bounces around until it is absorbed by a microscopic defect in the material and converted to heat, essentially wasted energy. Reducing these losses requires both giving trapped photons a better route out and patching the defects that absorb them along the way. These have typically been treated as separate engineering problems.
A useful way to think about what the team describe in this research is to consider what a texture does to a pane of glass. Smooth, flat glass transmits light reasonably well in one direction, but offers little control over what happens to light approaching from awkward angles. Some passes through, some reflects, and the behavior is largely determined by the geometry. A textured or patterned surface changes this: by introducing deliberate variations in the surface structure, light arriving from many different angles can be redirected more usefully, whether that means bending it inward toward an internal target (for a solar cell) or redirecting it outward toward an observer (for an LED). The same surface feature serves both directions of travel. The team's approach works on a closely related principle, applied to structures far smaller than any surface texture visible to the naked eye, and with the added benefit that the material forming those structures also repairs the defects that were previously wasting energy as heat.
Building porous textured sponges
Building on earlier collaborative work published in Science in 2023, by McGehee and Xu, which demonstrated that porous alumina nanoplates (a form of aluminum oxide) could reduce energy losses at perovskite interfaces, the team set out to extend that principle into a more sophisticated architecture. The key advance was developing a method to assemble alumina nanoparticles into micrometer-sized islands (each around five micrometers across and half a micrometer tall) embedded within the perovskite device. The assembly process uses electrostatic attraction: two populations of alumina nanoparticles are given opposite surface charges, and when mixed, they cluster together naturally into porous, sponge-like islands. One population is treated with a negatively charged molecule (Me-4PACz) and the other population treated with a positively charged molecule (ODA). The team refer to these as e-Al₂O₃, where the "e" denotes "electrostatic" assembly.
The porous sponge-like structure is critical. Earlier approaches to introducing low-refractive-index materials (materials that are less optically dense than the surrounding perovskite) into LED devices tended to block the flow of electrical charge, undermining device performance. Because the e-Al₂O₃ islands are porous, the perovskite material can grow through them, maintaining electrical contact with the electrode beneath. The islands therefore redirect light without interrupting the charge transport the device depends on.
The surface treatments applied to the alumina nanoparticles were designed to serve a second, equally important function. The molecules used to give the particles their opposite charges are the same molecules known to passivate perovskite surfaces, essentially chemically neutralizing the defects where energy can be lost as heat. The surface recombination velocity, a measure of how quickly electrical charges are lost at interfaces, dropped from 20.2 cm/s in a flat control device to 1.4 cm/s in the e-Al₂O₃ device. This brings the rate of energy loss at the interface close to levels seen in high-performance silicon solar cells.
With defect losses suppressed to this degree, a useful secondary effect called photon recycling becomes significant. When a photon is generated inside the perovskite and would otherwise be trapped and lost, it now has a reasonable chance of being reabsorbed by the material and re-emitted, effectively getting a second, or third, attempt to find an exit. This would be counterproductive in a defect-rich material, because each reabsorption event would risk the photon being lost to heat. However, with defects minimized, photon recycling amplifies the benefit of the improved light routing, pushing external efficiency higher than the geometry of the device alone would predict.
Operated as a solar cell, the e-Al₂O₃ device achieved an externally certified stabilized power-conversion efficiency of 26.7%. At the time this work was submitted for publication this cell held the world record for the power conversion efficiency for perovskite devices (held between 05/2024 – 02/2025). Operated as an LED with the same 800 nm thick perovskite layer, the device reached an external quantum efficiency of approximately 31%, meaning roughly 31 out of every 100 injected electrons produced a photon that successfully escaped the device. Radiance (a measure of light output intensity) was nearly ten times higher than the flat control device. Across both operating modes, the e-Al₂O₃ devices also showed meaningfully improved long-term stability, retaining 95% of their initial solar cell efficiency after 1,200 hours of continuous operation, compared with 67% for the flat control.
The authors note that this combination of greater than 26% solar cell efficiency and greater than 30% LED efficiency in a single polycrystalline device is, across all photovoltaic materials, only the second time this has been demonstrated, the first being single-crystal gallium arsenide, a material that is substantially more expensive and more difficult to manufacture at scale.
The practical implication of a device that converts sunlight to electricity efficiently and emits light efficiently is not merely academic. Displays that harvest ambient light to extend battery life, or lighting systems that recover energy when not actively in use, become more plausible when the same device architecture serves both functions without meaningful compromise in either. More fundamentally, the work demonstrates that the long-standing separation between emissive and photovoltaic device design is not a physical inevitability but an engineering problem, one that careful co-optimization of optical and electronic properties can address.
