🔬 materials science

High-Throughput Imaging of Degradation-Inducing Microscopic Impurities in Perovskite Solar Cells

This study establishes a practical, high-throughput diagnostic framework that combines nano-Fourier transform infrared spectroscopy with machine learning-enhanced optical imaging to rapidly identify and quantify microscopic impurities in perovskite solar cells, revealing these defects as critical nucleation sites for degradation and enabling early stability screening for scalable device fabrication.

Original authors: Sofiia Kosar, Anil R. Pininti, Vladyslav Hnapovskyi, José P. Jurado, Subhashri Mannar, Lorenzo Mardegan, Anand S. Subbiah, Frédéric Laquai, Stefaan De Wolf

Published 2026-09-29

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Sofiia Kosar, Anil R. Pininti, Vladyslav Hnapovskyi, José P. Jurado, Subhashri Mannar, Lorenzo Mardegan, Anand S. Subbiah, Frédéric Laquai, Stefaan De Wolf

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Solar energy holds the promise of a cleaner future, but for the technology to truly take over, it must be durable enough to last decades in the sun and heat. Among the most promising new materials for capturing sunlight are perovskites, a family of crystals that can be manufactured cheaply and printed onto flexible surfaces. While these materials have already achieved remarkable efficiency in turning light into electricity, their Achilles' heel has been stability. In the real world, these thin films often degrade quickly, losing their power-generating ability long before they are ready for commercial use. The challenge for scientists has been to understand exactly why they fail. Is it the material itself, or is it tiny, invisible flaws introduced during the manufacturing process? If the industry cannot find a way to spot these flaws instantly, scaling up production from a small lab sample to a massive factory line remains a distant dream.

A team of researchers at King Abdullah University of Science and Technology has now uncovered a critical piece of this puzzle, revealing that the stability of these solar cells is often compromised by microscopic impurities left behind during fabrication. These are not large cracks or obvious defects, but rather tiny, invisible clusters of the wrong chemical composition that hide on the surface of the film. The researchers discovered that these impurities act as the starting points for failure. When the solar cell is exposed to light and heat, the degradation process does not begin randomly across the surface; instead, it ignites at the boundary where these impurities meet the healthy crystal. From these small, hidden seeds, the damage spreads outward, eventually destroying the surrounding material and causing the device to fail.

To find these invisible enemies, the scientists first had to map them with extreme precision. They used a sophisticated technique called nano-Fourier transform infrared spectroscopy, which allows researchers to see the chemical makeup of a surface at a scale smaller than a single grain of sand. By scanning their films, they identified two specific types of unwanted residues: leftover lead iodide and a strange, hexagonal version of the perovskite crystal itself. These impurities were chemically distinct from the healthy, light-absorbing material. When the team subjected the films to intense light, mimicking a harsh day in the sun, they watched in real time as the damage began. The healthy crystals remained intact for a while, but the areas around the impurities started to crumble. The healthy crystals nearby eventually broke apart as the decay spread from the impurity sites, confirming that these tiny residues were the triggers for the collapse.

The breakthrough of this work lies not just in identifying the problem, but in solving the practical issue of how to find it quickly. The high-tech methods used to identify the impurities are incredibly slow, taking hours to scan a tiny area, which makes them useless for a factory that needs to check thousands of panels a day. The researchers realized that these impurities, while chemically distinct, also have a different way of reflecting light compared to the healthy material. They developed a system using a standard high-resolution microscope that captures reflected light, allowing them to see these impurities as bright spots against a darker background. To make this fast enough for industrial use, they trained a computer program, a type of artificial intelligence, to recognize these bright spots instantly. This system can scan a film and count the impurities in a matter of seconds, providing a clear, quantitative measure of how many hidden failure points exist before the device is even finished.

The team then tested whether this rapid screening could actually predict how long a solar cell would last. They built solar cells using films that had different numbers of these impurities. The devices made from films with a high density of impurities began to fail almost immediately, showing a rapid drop in performance within the first hour of operation. In contrast, the devices made from films with fewer impurities held their power much longer. This established a direct link: the more impurities present at the start, the faster the device would degrade. Furthermore, the researchers found that this degradation was driven by a combination of light and heat. Even when they heated the films without shining light on them, the same pattern of decay appeared at the impurity sites, proving that temperature alone is enough to activate these failure points.

This work provides a vital new tool for the future of solar energy. By showing that microscopic impurities are the primary cause of early failure, the study shifts the focus of quality control. Instead of waiting for a finished solar cell to fail a long-term test, manufacturers can now inspect the raw film the moment it is made. If the film is too full of these microscopic impurities, it can be discarded or reprocessed immediately, saving time and resources. The ability to detect and quantify these flaws in seconds means that the path to large-scale, stable production of perovskite solar cells is becoming clearer. The researchers have demonstrated that by controlling the cleanliness of the film at the microscopic level, it is possible to build solar cells that are not only efficient but also robust enough to power the world for years to come.

Technical Summary: High-Throughput Imaging of Degradation-Inducing Microscopic Impurities in Perovskite Solar Cells

Problem Statement
The scalable fabrication of high-quality, large-area perovskite solar cells (PSCs) is currently hindered by microscopic inhomogeneities, specifically residual compositional impurities formed during processing. While high-resolution techniques like scanning probe microscopies and nano-spectroscopies can identify these impurities, they are inherently slow and incompatible with the high-throughput, in-line diagnostics required for industrial upscaling. Conversely, rapid camera-based imaging lacks the spatial resolution and chemical specificity to detect stability-relevant microscopic impurities. This creates a critical gap: microscopic impurity phases may escape detection during fabrication yet act as deterministic seeds for catastrophic device failure under operational stress. There is an urgent need for a chemically validated, non-invasive framework to rapidly screen and quantify these impurities prior to device completion.

Methodology
The authors investigated blade-coated, wide-bandgap (~1.68 eV) CsFA-rich mixed-halide perovskite films relevant to tandem solar cell applications. Their approach involved a multi-step workflow:

Correlative Microscopy for Identification: The team utilized a suite of correlative techniques to map local morphology and chemistry. They employed nano-Fourier transform infrared spectroscopy (nano-FTIR) based on scattering-type scanning near-field optical microscopy (s-SNOM) to obtain nanoscale chemical information. This was complemented by atomic force microscopy (AFM), Kelvin probe force microscopy (KPFM), confocal photoluminescence (PL) microscopy, X-ray diffraction (XRD), and confocal Raman spectroscopy.
Photo-Stress Experiments: Films were subjected to white light illumination at 3-Sun equivalent intensity in an inert atmosphere, as well as thermal stress at 85 °C, to observe the progression of degradation.
High-Throughput Imaging Framework: To translate fundamental insights into a rapid screening tool, the authors developed a reflected light microscopy (RLM) system. They demonstrated that compositional impurities exhibit distinct optical contrast (enhanced reflectance) compared to pristine grains.
Machine Learning Analysis: A supervised convolutional neural network (CNN) with a U-Net architecture was trained on RLM images to automatically identify and segment compositional impurities. The model was validated against ground-truth data from nano-FTIR and AFM.
Quantitative Metrics: The framework enabled the calculation of quantitative metrics, including impurity coverage ( $CI$ ), impurity size ( $SI$ ), and nearest-neighbor distances among impurities ( $NND_{I-I}$ ).

Key Results

Impurity Identification: The correlative microscopy revealed two primary types of remnant impurities on the film surface: hexagonally-shaped clusters identified as residual $PbI_2$ and isolated grain-like features identified as the 4H hexagonal perovskite polytype. These impurities showed a lack of FA+ cation vibrational resonances (C=N bond) in nano-FTIR spectra, confirming their photo-inactive nature.
Degradation Mechanism: Photo-stress experiments demonstrated that degradation initiates specifically at the interfaces between impurities and pristine perovskite grains. The process involves the decomposition of the perovskite lattice, evidenced by the depletion of FA+ and the formation of $PbI_2$ and intermediate Cs-rich phases.
Coupled Photo-Thermo-Chemical Instability: The study revealed that impurity-mediated instability arises from coupled photo- and thermo-chemical processes. While photo-chemical reactions (e.g., defect-assisted photolysis) occur under illumination, thermal effects (even under illumination) play a critical role in accelerating halide/cation migration and chemical reactivity. Thermal stress alone (85 °C) produced degradation patterns nearly identical to those observed under illumination, confirming that heating cannot be neglected in degradation pathways.
High-Throughput Detection: The RLM combined with CNN analysis successfully identified impurities with high accuracy (Dice-Sørensen coefficient ~0.91) within seconds. Films with dense impurity distributions (shorter $NND_{I-I}$ ) exhibited accelerated early degradation in single-junction device tests, whereas films with moderate impurity density showed improved stability.
Tracking Degradation: The RLM framework was extended to monitor degradation in real-time. Local darkening in RLM contrast was observed at impurity-perovskite junctions, allowing for the mapping of degradation propagation without invasive techniques.

Significance and Claims
The paper establishes a practical, chemically-validated diagnostic imaging framework for the rapid pre-screening of perovskite films. By linking microscopic impurity density to device stability, the authors propose impurity density (specifically $NND_{I-I}$ ) as an early-warning metric for stability screening.

The work claims to bridge the gap between fundamental nanoscale characterization and industrial quality control. It demonstrates that controlling remnant microscopic impurities is essential for achieving stable PSCs, particularly for wide-bandgap compositions requiring high annealing temperatures. The developed RLM-CNN framework offers a non-destructive, high-throughput tool capable of identifying detrimental impurities in as-prepared films, enabling the optimization of fabrication processes before subsequent device layers are applied. The authors suggest this approach could be integrated into in-line and robotic coating platforms to accelerate the upscaling of perovskite photovoltaic technologies.

More like this