Hydrogen-Bond Trick Tames Evaporated Perovskites for Efficient Near-Infrared LEDs

Perovskite light-emitting diodes made by thermal evaporation have long lagged behind their solution-processed rivals, but a research team now reports a molecular fix that closes much of the gap. By adding a dual-amide molecule that hydrogen-bonds to formamidinium cations, the researchers suppressed premature crystallization during vacuum deposition and produced near-infrared LEDs with a maximum external quantum efficiency of 25.0 percent at an emission wavelength of 805 nanometers, one of the highest values reported so far for thermally evaporated perovskite LEDs. The devices also showed a T50 lifetime of 23 hours at a driving current density of 100 milliamperes per square centimeter, placing them among the most stable evaporated perovskite LEDs reported to date. The work, published in Nature Photonics, offers both a route toward scalable LED manufacturing and a detailed picture of how crystallization unfolds inside a vacuum chamber.

Thermal evaporation is attractive for perovskite optoelectronics because it holds considerable promise for scalable manufacturing and pixel patterning of perovskite LEDs. Yet evaporated perovskite LEDs have consistently underperformed solution-processed ones, and the new study pinpoints why: without effective strategies to regulate perovskite crystallization during vacuum deposition, the perovskite phase forms too quickly and uncontrollably, yielding films with defects that drain energy as heat instead of light. The team, whose authors include Yangyu Liu, Yuhan Zhou and Jianing Chang together with Feng Gao and Chunxiong Bao among many colleagues, set out to understand and then repair that crystallization pathway.

The researchers began with FAPbI3, a formamidinium lead iodide perovskite that emits efficiently in the near-infrared. They first optimized a sequential evaporation process, laying down formamidinium iodide and then lead iodide in separate layers rather than co-evaporating them, which produced larger and more uniform grains. Adding 5-aminovaleric acid, an additive known to work well in solution-processed films, boosted the average external quantum efficiency of evaporated devices from 2.3 percent to around 11.3 percent. That was encouraging, but the same additive lifted solution-processed devices from 4.8 percent to 21.2 percent, revealing a persistent shortfall specific to vacuum deposition.

To find out what was going wrong, the team built an in situ spectroscopy system that could watch the film as it formed. Optical fibers carried light into the evaporation chamber and collected absorption and photoluminescence signals in real time, while grazing-incidence wide-angle X-ray scattering at the Shanghai Synchrotron Radiation Facility tracked crystal structure during annealing. The measurements revealed a surprise: during evaporation, the three-dimensional FAPbI3 phase formed regardless of whether 5-aminovaleric acid was present. In solution processing, by contrast, the additive stabilizes a metastable intermediate phase that later guides orderly crystal growth. The key difference turned out to be the solvent itself. Formamidinium iodide and 5-aminovaleric acid react to form a condensation product that populates the beneficial intermediate phase, but pH strip tests confirmed that this reaction requires a solvent such as N,N-dimethylformamide or dimethyl sulfoxide. In the dry vacuum environment, it simply never happens.

The in situ measurements also mapped three distinct stages during post-deposition annealing of the evaporated films. First, the prematurely formed three-dimensional perovskite partially decomposed, with excess formamidinium intercalating into the lead halide octahedral framework to create a low-dimensional, formamidinium-rich phase identified as FA2PbI4. Second, formamidinium iodide gradually deprotonated and decomposed, a process accelerated by the catalytic effect of the zinc oxide layer beneath the film. Third, continued loss of formamidinium drove the intermediate into the final three-dimensional perovskite. Films containing 5-aminovaleric acid followed essentially the same pathway, delaying three-dimensional phase formation only slightly, from 24 seconds to 29 seconds of annealing. The lesson was clear: what matters is preventing the three-dimensional phase from forming during deposition and then letting it emerge slowly during annealing, guided by a stable intermediate framework.

Armed with that insight, the team designed a molecule that could play the role of the missing solvent. They chose 2,6-pyridine dicarboxamide, or PyDA, a dual-amide compound whose two carbonyl groups form hydrogen bonds with the two amine groups of the formamidinium cation. The molecule had to satisfy two demands: it needed multiple hydrogen-bond donors and acceptors to stabilize formamidinium ions and prevent premature perovskite formation, yet it had to remain stable under vacuum and release the cation during annealing. Sequentially evaporated films containing PyDA showed only lead iodide absorption features during deposition, with no detectable three-dimensional perovskite photoluminescence peak or diffraction rings, confirming that phase formation was effectively suppressed in the chamber.

During annealing, the hydrogen-bonded network did its job gradually. Iodide ions diffused from the formamidinium iodide layer into the lead iodide layer and converted its edge-sharing lattice into corner-sharing octahedra, with the growth of this phase saturating after about 69 seconds as free iodide was depleted. Formamidinium was then released slowly from the PyDA network over roughly 420 seconds to build the three-dimensional perovskite, a formation time almost seven times longer than with 5-aminovaleric acid. Fourier transform infrared spectroscopy and proton solid-state nuclear magnetic resonance verified the hydrogen bonds between the two molecules, and in situ infrared measurements showed the coupled carbonyl and formamidinium vibrational peaks splitting apart as annealing broke the network down. Fitting the temperature-dependent formation rates to the Arrhenius equation yielded apparent activation energies of 1.11, 1.17 and 1.98 electronvolts for the control, 5-aminovaleric acid and PyDA films respectively, with PyDA exhibiting the highest barrier, approaching that of spin-coated films containing 5-aminovaleric acid. Density functional theory calculations independently supported the picture, showing stronger binding for the PyDA-formamidinium complex and a higher barrier to cation release.

The controlled crystallization paid off dramatically in film quality and device performance. Photoluminescence quantum efficiency climbed from 12.6 percent in control films and 38.8 percent with 5-aminovaleric acid to 82.4 percent with PyDA, while the photoluminescence lifetime lengthened from 214 nanoseconds to 809 nanoseconds. Modeling of the carrier dynamics showed that the defect-assisted non-radiative recombination rate fell by roughly two orders of magnitude, from 1.80 times ten to the sixth per second in control films to 1.86 times ten to the fourth per second, while the radiative bimolecular rate stayed comparable, meaning the efficiency gain came primarily from suppressing defect-assisted non-radiative recombination. The resulting LEDs turned on at just 1.3 volts, reached a maximum radiance of about 185 watts per square meter per steradian at 3.1 volts, and delivered a champion external quantum efficiency of 25.0 percent with an average of 22.1 percent, competitive with state-of-the-art solution-processed near-infrared perovskite LEDs. Emission was uniform across a 10-by-10-centimeter film, with device efficiencies consistently exceeding 20 percent across the area, and the team also demonstrated a large-area LED with an active area of 5 by 5 centimeters. Related dual-amide molecules, including 2-butenediamide, succinimide and 1,3-benzene dicarboxamide, also suppressed premature crystallization more effectively than 5-aminovaleric acid and yielded device efficiencies of 18.7, 16.8 and 18.0 percent respectively, suggesting the design principle is transferable, though PyDA remained the strongest performer.

To showcase the practical advantage of vacuum deposition, the researchers replaced lead iodide with lead bromide to tune emission across the near-infrared and deep-red, achieving peak external quantum efficiencies of 12.9 to 17.3 percent at wavelengths from 748 to 771 nanometers, with nearly identical current density-voltage curves across the series. They then integrated LEDs emitting at 805 and 759 nanometers onto the same substrate, creating monolithic dual-color pixels that point toward perovskite pixelated arrays. As a demonstration, they built a wearable pulse oximeter in which the dual-color perovskite emitter and a silicon photodiode exploit the different absorption of oxygenated and deoxygenated hemoglobin at the two wavelengths. The device measured blood oxygen saturation between 96 and 100 percent, with a mean absolute deviation of only 0.95 percent relative to a commercial oximeter used for calibration. Because both light sources share one thin monolithic film rather than discrete LEDs, the architecture offers strong potential for integration into wearable systems.

The study establishes a molecular strategy for intermediate-phase regulation in thermally evaporated formamidinium-based perovskites, a challenge that has limited evaporated devices across the field. By substituting a rationally designed hydrogen-bonding mediator for the solvent coordination that solution chemistry provides, the researchers showed that the intermediate-phase strategy long exploited in spin-coated films can be recreated under high vacuum. The authors note that the hydrogen-bond-mediated interaction is proposed as an important factor contributing to the regulated evolution of the intermediate phase, which subsequently undergoes a gradual conversion into the three-dimensional perovskite during annealing. Their in situ spectroscopic toolkit offers a way to test such mechanisms directly as films grow, and the demonstration of dual-color pixels on a single substrate and a functioning oximeter highlights the potential of evaporated perovskites for multicolor pixelation and wearable sensing applications.

Subject of Research: Crystallization regulation of thermally evaporated perovskite films for efficient near-infrared light-emitting diodes

Article Title: Regulating the crystallization of thermally evaporated perovskites for efficient near-infrared light-emitting diodes

Article References: Liu, Y., Zhou, Y., Chang, J., Yang, J., Yuan, D., Xu, W., Wang, W., Li, Z., Feng, X., Fu, L., Wang, J., Li, Z., He, B., Yu, T., Zou, Z., Gao, F., & Bao, C. (2026). Regulating the crystallization of thermally evaporated perovskites for efficient near-infrared light-emitting diodes. Nature Photonics. https://doi.org/10.1038/s41566-026-02036-5

Image Credits: AI Generated

DOI: 10.1038/s41566-026-02036-5

Keywords: perovskite LEDs, thermal evaporation, crystallization, hydrogen bonding, near-infrared, external quantum efficiency, FAPbI3, pulse oximetry, vacuum deposition, intermediate phase, PyDA, wearable sensors