🔬 applied physics

Single Best Fits Can Be Misleading: Resolving Common Trapping Signatures across FA--Cs Perovskites

By combining fluence-dependent photoluminescence, Shockley-Read-Hall modeling, and Bayesian inference across FA $_{1-x}$ Cs $_x$ PbI $_3$ perovskites, this study reveals that trap filling resolves parameter degeneracies to identify universal non-radiative recombination signatures ( $T\beta$ and $T\alpha$ ) and composition-specific defects ( $T\epsilon$ ), thereby establishing a robust framework for comparing trapping behaviors in semiconducting materials.

Original authors: Maxim Simmonds, Katarzyna Pydzińska-Białek, Thomas C. Rossi, Mostafa Othman, Aïcha Hessler-Wyser, Christian Wolff, Christophe Ballif, Vincent M. Le Corre, Eva Unger

Published 2026-09-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Maxim Simmonds, Katarzyna Pydzińska-Białek, Thomas C. Rossi, Mostafa Othman, Aïcha Hessler-Wyser, Christian Wolff, Christophe Ballif, Vincent M. Le Corre, Eva Unger

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 cells made from metal-halide perovskites have emerged as a promising alternative to traditional silicon panels, offering the potential for cheaper, lighter, and more flexible energy generation. These materials are semiconductors, meaning they can conduct electricity under specific conditions, and their ability to convert sunlight into power depends heavily on how electrons move through them. However, the journey of these electrons is rarely smooth; they frequently encounter defects—tiny imperfections in the crystal structure of the material. When an electron hits such a defect, it can get stuck or lose its energy as heat instead of contributing to electricity. This process, known as non-radiative recombination, is a primary reason why these solar cells do not reach their theoretical maximum efficiency. For years, scientists have struggled to pinpoint exactly which defects are responsible for these losses, as the signals they leave behind are often complex and difficult to interpret.

A team of researchers at the Helmholtz-Zentrum Berlin and other institutions has now tackled this challenge by developing a new way to look at how these materials behave. Instead of relying on simple models that assume a single, average behavior for all defects, they combined detailed computer simulations with a sophisticated statistical method called Bayesian inference. This approach allowed them to analyze time-resolved photoluminescence, a technique where the material is briefly lit up with a laser and the fading glow is measured over time. By varying the intensity of the laser light, the researchers could observe how the material responded under different conditions, effectively probing the hidden landscape of defects within the crystal. Their work focused on a specific series of perovskite films made by mixing formamidinium lead iodide with varying amounts of cesium, a common strategy used to improve the stability and performance of these solar cells.

The researchers discovered that the way defects behave is far more dynamic than previously thought. When the material is exposed to low levels of light, many of the defect sites are empty and ready to catch passing electrons. In this state, it is nearly impossible to tell the difference between a defect that is very good at catching electrons but is rare, and one that is poor at catching them but is extremely common. The measurements in this regime produce a confusing mix of signals where these two properties are mathematically linked, making it difficult to identify the true nature of the problem. However, as the researchers increased the intensity of the light, they found that the defect sites began to fill up, much like seats in a theater being occupied by an audience. Once these sites were full, the behavior of the remaining electrons changed, and the mathematical confusion cleared up. This "trap filling" effect allowed the researchers to separate the properties of the defects and identify them with much greater precision.

Using this insight, the team mapped out the specific "signatures" of the defects across their different material compositions. They found that one particular type of defect, which they labeled Tβ, was present in all the samples they tested. This defect is characterized by a strong preference for catching electrons over holes, a property that causes the defect sites to fill up quickly under light. This filling process is actually a general feature of how these perovskite materials work, rather than a flaw. In contrast, they identified another defect type, Tα, which acts as a major bottleneck for performance. This defect is present in all the samples and serves as a significant pathway for energy loss, limiting the overall efficiency of the solar cells. Perhaps most notably, they found a third type of defect, Tϵ, which appeared only in the pure formamidinium lead iodide samples without any cesium. This specific defect is shallow and creates the highest rate of energy loss, which aligns with previous observations that these pure materials contain a high density of structural stacking faults, or layers that are slightly misaligned.

The study also highlighted a critical limitation in how scientists have traditionally analyzed this data. For years, researchers have relied on finding a single "best fit" set of numbers to describe a material's defects. The new work demonstrates that this approach can be misleading because many different combinations of defect properties can produce the same experimental results. By using statistical methods to map out the entire range of possible solutions rather than just picking one, the researchers were able to see which defect signatures were truly consistent across different samples and which were unique to specific compositions. This distinction is vital because it allows scientists to compare materials fairly and understand which structural changes actually improve performance. The findings suggest that to truly understand and improve these solar cells, one must account for how the defects fill up during operation, as this state fundamentally changes how the material behaves.

Ultimately, this research provides a clearer roadmap for designing better solar cells. By showing that trap filling is a key factor in how defects are identified, the study offers a more robust framework for comparing different semiconductor materials. The identification of the Tα defect as a universal performance limiter suggests that future efforts should focus on eliminating or mitigating this specific type of imperfection to boost efficiency. Meanwhile, the unique presence of the Tϵ defect in the pure material explains why adding cesium improves the quality of the film, as it reduces the density of these specific structural errors. The work does not claim to have solved the problem of defects entirely, but it establishes a more reliable method for diagnosing them, moving the field away from ambiguous single-number descriptions toward a nuanced understanding of the complex interplay between light, electrons, and the microscopic imperfections that shape the future of solar energy.

Technical Summary: Single Best Fits Can Be Misleading: Resolving Common Trapping Signatures across FA–Cs Perovskites

Problem Statement
Identifying the specific defects responsible for non-radiative recombination in metal-halide perovskites remains a significant challenge. While time-resolved photoluminescence (trPL) is a standard diagnostic tool, the characteristically strong non-monoexponential decay behavior of perovskite films has proven difficult to describe using conventional models that reduce Shockley–Read–Hall (SRH) recombination to a single characteristic lifetime. Simplified models (e.g., mono- or stretched-exponential) often fail to capture kinetics across broad carrier density ranges.

More physically complete models based on the full SRH formalism, which explicitly treat electron/hole capture, emission, and defect occupation, have shown promise. However, these models introduce a large number of fitted parameters ( $N_t$ , $\beta_n$ , $\beta_p$ , $E_t$ ) that are often strongly correlated. Consequently, a single "best-fit" parameter set does not guarantee a unique physical description; different parameter combinations can reproduce experimental transients with similar accuracy. This degeneracy makes it difficult to reliably compare defect signatures across different samples or compositions, as apparent differences may simply reflect different positions within correlated regions of parameter space rather than distinct physical defects.

Methodology
The authors address these limitations by combining fluence-dependent time-resolved photoluminescence (trPL) measurements with full SRH modeling and Bayesian posterior inference.

Samples: The study investigates an $FA_{1-x}Cs_xPbI_3$ compositional series ( $x = 0, 0.15, 0.30$ ), selected to represent distinct microstructural regimes: pure FAPbI $_3$ (high stacking fault density), an intermediate composition with minimized defects, and a Cs-rich composition containing secondary $\delta$ -CsPbI $_3$ phases.
Modeling Framework: A numerical model based on the full SRH formalism is employed, explicitly treating carrier capture, emission, and time-dependent trap occupation. The model accounts for carrier diffusion, cyclic excitation effects (due to the 5 kHz repetition rate), and effective bulk trapping populations that encompass both bulk and surface recombination.
Inference Strategy: Instead of relying on single best-fit solutions obtained via Bayesian optimization, the authors utilize Markov Chain Monte Carlo (MCMC) sampling to map the posterior distributions of the model parameters. This approach quantifies the uncertainty and correlation structures of the inferred parameters (trap density $N_t$ , capture coefficients $\beta_n, \beta_p$ , and energy $E_t$ ).
Analysis: The authors analyze the topology of these posterior distributions to identify "degenerate ridges" (where parameters are correlated) and regions of overlap across different compositions to distinguish between sample-specific and common trapping signatures.

Key Results

Parameter Degeneracy and Trap Occupation: The study confirms that parameter identifiability is governed by trap occupation. In regimes of weak trap occupation, a strong degeneracy exists between the electron capture coefficient ( $\beta_n$ ) and trap density ( $N_t$ ), as well as between these parameters and trap energy ( $E_t$ ). The measured dynamics primarily constrain the products of these parameters rather than their individual values.
Lifting Degeneracy via Trap Filling: The authors demonstrate that significant trap filling (where the number of occupied traps becomes comparable to the total trap density) lifts the $\beta_n$ – $N_t$ degeneracy. In this regime, the capture rate depends explicitly on the number of unoccupied states, allowing the capture coefficient and trap density to be independently constrained.
Identification of Common Signatures: By comparing posterior distributions across the compositional series, the authors identify four distinct effective trapping signatures:
- $T_\beta$ : A non-degenerate signature present across all compositions ( $x=0, 0.15, 0.30$ ). It is characterized by a strongly asymmetric capture coefficient ratio ( $\beta_p/\beta_n \ll 1$ ), leading to substantial trap filling. This signature is associated with the power-law decays observed in perovskite trPL.
- $T_\alpha$ : A signature present in all compositions but exhibiting pronounced parameter degeneracies ( $\beta_n$ – $N_t$ ). It represents a major non-radiative recombination channel that limits device performance.
- $T_\gamma$ : A signature overlapping between the intermediate ( $x=0.15$ ) and Cs-rich ( $x=0.30$ ) samples, likely originating from homogeneous Cs-containing grains.
- $T_\epsilon$ : A composition-specific signature unique to pure FAPbI $_3$ ( $x=0$ ). It is characterized by a shallow trap energy and produces the highest steady-state non-radiative recombination rate, consistent with the high density of stacking faults observed in this composition.
Recombination Rates: While $T_\beta$ is ubiquitous, it is $T_\alpha$ and the unique $T_\epsilon$ that drive the highest non-radiative recombination rates under steady-state conditions, directly correlating with the lower quasi-Fermi-level splitting (QFLS) observed in the Cs0 and Cs30 samples compared to Cs15.

Significance and Claims
The paper argues that relying on individual best-fit parameter values is insufficient and potentially misleading for constructing a physically meaningful defect landscape from trPL data. The authors claim that:

Posterior inference provides a robust framework: Quantifying the joint uncertainty and correlation structure of parameters is essential for reliably comparing trapping signatures across different semiconducting materials.
Trap filling is a key determinant: The degree of trap filling dictates whether defect parameters are identifiable or degenerate. The presence of a non-degenerate signature ( $T_\beta$ ) with strong asymmetry in capture coefficients appears to be a general feature of perovskite carrier dynamics.
Methodological transferability: The identified parameter degeneracies originate from the structure of the SRH equations and are not specific to the FA–Cs system. The approach offers a pathway to identifying and comparing trapping signatures across various semiconductor systems using room-temperature measurements, provided that complementary observables or conditions (like varying fluence) are used to break degeneracies.

The study concludes that this framework enables the identification of recurring trapping signatures and their correlation with structural features (such as stacking faults or secondary phases), offering a more rigorous basis for understanding and mitigating defect-mediated recombination losses in perovskite optoelectronics.

More like this