Pith. sign in

REVIEW 3 major objections 5 minor 123 references

A compact CNN can estimate perovskite cell efficiency retention from multimodal luminescence image pairs without electrical sweeps.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-14 22:04 UTC pith:7EIT2RII

load-bearing objection Solid proof-of-concept pipeline with public code/data and a real but modest spatial gain over a weak intensity baseline; abstract and body disagree on how much spatial features matter. the 3 major comments →

arxiv 2603.12857 v2 pith:7EIT2RII submitted 2026-03-13 cond-mat.mtrl-sci

Quantifying Perovskite Solar Cell Degradation via Machine Learning from Spatially Resolved Multimodal Luminescence Time Series

Giulio Barletta , Simon Ternes , Saif Ali , Zohair Abbas , Chiara Ostendi , Marialucia D'Addio , Erica Magliano , Pietro Asinari
show 2 more authors
Eliodoro Chiavazzo Aldo Di Carlo
This is my paper
classification cond-mat.mtrl-sci
keywords perovskite solar cellsefficiency retentionelectroluminescencephotoluminescenceconvolutional neural networkdegradation monitoringmachine learningstability testing
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Perovskite solar cells reach high efficiencies but still fail too quickly for large-scale use, so labs need a fast, non-invasive way to track how much performance a device has retained. This paper shows that electroluminescence and two photoluminescence images taken at an aged state, stacked with the same device's images from t=0 and simple intensity-ratio maps, contain enough information for a small convolutional network (LumPerNet) to regress efficiency retention RPCE = PCEt/PCE0 inside the operational window 0.8–1.2. Under a device-level hold-out and four-fold cross-validation protocol, the network beats a control that sees only the spatial average of each channel, cutting mean absolute error by about 23 percent and lifting R-squared by 0.42. Modality ablations further show that full multimodality is most stable, though carefully chosen bimodal pairs (especially EL with short-circuit PL) retain most of the accuracy. The result matters because it turns routine luminescence imaging into a quantitative, spatially aware state-of-health readout that can speed stability screening without repeated J–V measurements.

Core claim

Spatially resolved multimodal luminescence time series (EL, open-circuit PL, short-circuit PL at aged time t paired with the same device’s t=0 references and pixel-wise ratios) carry predictive information about device-level efficiency retention beyond global intensity decay; a compact CNN (LumPerNet) regresses RPCE inside [0.8, 1.2] with MAE 0.049, RMSE 0.061 and R2 0.553, improving substantially on an intensity-only MLP baseline trained under identical leakage-aware splits.

What carries the argument

LumPerNet: a lightweight CNN that ingests the nine-channel stacked tensor x(t) = [EL(t), PLoc(t), PLsc(t), EL(0), PLoc(0), PLsc(0), rEL, rPLoc, rPLsc] and maps it to scalar RPCE, learning relative spatial degradation signatures rather than absolute brightness.

Load-bearing premise

The mapping learned from 59 mostly unencapsulated lab cells aged at fixed 30 °C under uncontrolled ambient humidity, labeled only by indoor white-LED J–V averages inside the narrow RPCE window 0.8–1.2, is assumed to transfer to other stacks, encapsulations, stress protocols and full-lifecycle degradation.

What would settle it

Train and test the identical architecture and protocol on an independent set of encapsulated devices aged under controlled humidity or elevated temperature and spanning RPCE well below 0.8; if MAE and R2 collapse to the intensity-only baseline or worse, the claimed spatial predictive advantage does not generalize.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Accelerated stability campaigns can replace frequent full J–V sweeps with periodic luminescence imaging plus a single forward pass to estimate RPCE.
  • Modality ablation shows that EL + short-circuit PL (or open-circuit + short-circuit PL) can retain most accuracy, reducing acquisition burden.
  • Spatially resolved predictions can flag localized failure sites (edges, contacts, shunts) that space-averaged electrical metrics miss.
  • A two-stage classifier-plus-regressor extension could first detect out-of-window samples and then specialize the RPCE estimate inside the operational range.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Because short-circuit PL adds extraction contrast that open-circuit PL and EL largely share, the same multimodal stack could be used to infer which degradation mode (recombination vs. extraction loss) dominates at each pixel.
  • If photometric calibration and humidity control improve, the same network may serve as an early-warning sensor on pilot modules without requiring electrical contact at every inspection.
  • The residual gap between LumPerNet and the intensity baseline suggests that richer temporal context (beyond single reference–current pairs) could further reduce error on rapid early transients.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The manuscript presents a supervised deep-learning pipeline (LumPerNet) that regresses perovskite solar-cell efficiency retention R_PCE = PCE(t)/PCE(0) from multimodal luminescence image tensors (EL, PL_oc, PL_sc at aged and t=0 states, plus ratio channels). Devices are aged in a custom automated chamber with synchronized indoor J–V labels; analysis is restricted to R_PCE ∈ [0.8, 1.2]. Under a device-level held-out split and four-fold CV, LumPerNet reports MAE 0.049±0.007, RMSE 0.061±0.005, R² 0.553±0.077 and is compared to an intensity-only MLP baseline (MAE 0.064±0.015, R² 0.136±0.420). Modality ablations and trajectory reconstructions are used to argue that complementary EL/PL channels and spatial patterns improve prediction and robustness relative to global intensity decay, positioning the approach as a non-invasive route for accelerated stability testing.

Significance. If the spatial and multimodal gains hold under stronger controls, the work supplies a practical, leakage-aware pipeline linking automated EL/PL imaging to a device-level electrical retention metric—something still scarce for PSCs relative to silicon PV. Strengths include public code and processed data, matched device-level splits for all ablations, train-only normalization, and explicit modality physics (EL/PL_oc vs PL_sc). The contribution is primarily methodological and proof-of-concept rather than a definitive performance leap; with only 59 devices and a restricted operational window, the main value is a reproducible baseline for image-based R_PCE monitoring rather than a ready field tool.

major comments (3)
  1. Abstract vs §3.2/Conclusions framing of the spatial claim is inconsistent and load-bearing. The Abstract states that “global luminescence evolution contains most of the predictive signal, while spatial information provides a secondary contribution to robustness,” yet §3.2 and the Conclusions emphasize a substantial spatial improvement (−23% MAE, −26% RMSE, +0.417 R²). The intensity-only baseline is fold-unstable (negative test R² in two of four folds; Table S1), so the headline gain largely reflects a weak control. Please either (i) stabilize the baseline (e.g., stronger regularized MLP, richer global descriptors, or ensemble) and re-report the spatial delta, or (ii) reframe the central claim to match the Abstract’s more cautious wording and quantify how much of the gain is robustness vs absolute accuracy.
  2. §2.4 / Table 2 / §3.5: With N_cells = 59 across only three batches (two architectures), device-level CV variance is high (baseline R² σ = 0.420; LumPerNet R² σ = 0.077). The operational window R_PCE ∈ [0.8, 1.2] and indoor white-LED (not AM1.5G) labels further limit the claim that spatially resolved imaging is a general route for accelerated stability testing. The Limitations section acknowledges this, but the Abstract and Conclusions still over-generalize. Please add a clearer scope statement (lab-scale, unencapsulated, fixed 30 °C ambient, restricted R_PCE window) and, if possible, a leave-one-batch-out or architecture-holdout metric so readers can judge transfer.
  3. §3.3 modality ablation: EL+PL_oc is markedly less stable (R² 0.245±0.170, near-zero/negative in one fold) while EL+PL_sc and PL_oc+PL_sc retain most full-model performance. The physical interpretation (PL_sc as extraction-regime contrast) is plausible, but the paper does not show that the full three-modality model is statistically superior to the best bimodal model under the same splits (EL+PL_sc MAE 0.052±0.002, R² 0.534±0.034 vs full MAE 0.049±0.007, R² 0.553±0.077). Either report a paired significance test across folds or soften the claim that “multimodality is essential” to “complementary bimodal contrast is sufficient and more robust than arbitrary channel stacking.”
minor comments (5)
  1. Notation: Table 1 and Eq. (3) use r_EL(t) = EL(t)/EL(0), but the single-modality example writes r_EL(0); please make ratio indexing consistent.
  2. Figure 4 vs Table 2: ensemble metrics (MAE 0.042, R² 0.676) are better than fold-mean metrics; the caption explains this, but the main text should state which number is the primary reported result to avoid cherry-picking.
  3. §2.2: humidity is uncontrolled and devices are unencapsulated “to increase performance variation”; this design choice should be flagged earlier (Abstract/Introduction) as a deliberate stressor, not only in Limitations.
  4. Supplementary architecture description is clear; a one-sentence parameter count or FLOPs note for LumPerNet vs BaselineMLP in the main text would help readers assess the “compact” claim.
  5. Minor typos: “Flourine-doped” → “Fluorine-doped” (§2.1.1); “V ergata” spacing in affiliations; “Skl’odowska” accent in Acknowledgements.

Circularity Check

0 steps flagged

No circularity: supervised regression of electrically measured R_PCE from independent optical images, not a derivation that redefines the target from its inputs.

full rationale

The paper’s load-bearing claim is empirical ML performance, not a first-principles derivation. R_PCE is defined solely from indoor J–V averages (Eq. 1, Eq. 4: PCE_avg(t)/PCE_avg(0)), while inputs are EL/PLoc/PLsc image stacks plus luminescence ratio channels r_I = I(t)/(I(0)+ε) (Eq. 3). Those ratio channels are optical features, not the electrical label; nothing forces R_PCE to equal mean luminescence decay by construction. Training is standard supervised regression under device-level held-out splits and four-fold CV, with an intensity-only MLP control that discards spatial structure. Ablations (single/bimodal modalities, batch embedding) are comparative experiments, not uniqueness theorems or self-citation chains that forbid alternatives. Self-citations in the bibliography are background (fabrication, stability literature, imaging physics) and do not supply a load-bearing uniqueness or ansatz that the central result reduces to. Correlated response of luminescence and PCE to the same aging history is expected physics and does not make the mapping definitional or fitted-as-prediction. Framing tension between abstract (global signal primary) and body (spatial improvement) is a presentation issue, not circularity. Score 0 with empty steps is the correct honest finding.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 1 invented entities

The work is empirical ML on experimental imaging; load-bearing content is domain measurement choices and modeling restrictions rather than new physics entities. Free parameters are training/window choices that shape the reported metrics. Axioms are standard ML and PSC-domain assumptions. No new physical mediators are postulated.

free parameters (4)
  • R_PCE analysis window [0.8, 1.2]
    Samples outside this band are excluded from training and reported metrics; bounds are chosen for operational relevance and sample density, not derived from first principles, and directly define the claim scope.
  • LumPerNet capacity and training hyperparameters (channels 16→32, dropout, AdamW lr 3e-4, weight decay 1e-4, 100 epochs,
    Architecture and optimization settings are design choices that affect absolute accuracy; paper does not claim uniqueness of these values.
  • Ratio stabilizer ε = 10^{-6}
    Ad hoc constant in r(t)=I(t)/(I(0)+ε) for numerical stability of ratio channels.
  • EL bias (1.5 V), soak/relaxation dwell times, imaging schedule sparsity
    Experimental protocol parameters that define the input distribution; held fixed across batches but not universal standards.
axioms (4)
  • domain assumption Device-level splits prevent temporal leakage and are the right unit for generalization claims.
    Methods §2.4; standard for longitudinal device data but assumes devices are exchangeable within batch strata.
  • domain assumption Indoor white-LED average of forward/reverse PCE is a valid label for relative efficiency retention under consistent lab illumination.
    §2.2–2.3; authors note non-AM1.5G use for relative retention only.
  • domain assumption Spatial EL/PL patterns (with t=0 references) contain degradation-relevant information about device-averaged R_PCE.
    Central modeling hypothesis tested against intensity-only baseline (§3.1–3.2).
  • standard math Standard CNN/MLP supervised regression with MAE/RMSE/R² evaluation is appropriate for this regression task.
    §2.4; conventional ML practice.
invented entities (1)
  • LumPerNet no independent evidence
    purpose: Named compact CNN regressor mapping multimodal image tensors to scalar R_PCE.
    Engineering model name, not a new physical object; independent evidence is the released code and reported metrics only.

pith-pipeline@v1.1.0-grok45 · 35684 in / 3201 out tokens · 36871 ms · 2026-07-14T22:04:30.060345+00:00 · methodology

0 comments
read the original abstract

Perovskite solar cells achieve remarkable power conversion efficiencies, yet operational stability remains a major barrier to large-scale deployment. Reliable and rapid assessment of device state of health is therefore essential. Conventional electrical diagnostics, such as illuminated current-voltage (J--V) sweeps, provide accurate performance metrics but are time-consuming and do not resolve spatially localized degradation, motivating non-invasive imaging-based alternatives. A deep-learning framework is introduced to estimate PSC efficiency retention, $R_\mathrm{PCE}=\mathrm{PCE}_t/\mathrm{PCE}_0$, directly from multimodal luminescence imaging acquired during device aging. Each sample combines electroluminescence (EL), open-circuit photoluminescence (PLoc), and short-circuit photoluminescence (PLsc) at an aged state with device-specific reference images at $t=0$, enabling learning of degradation-relevant spatial changes. LumPerNet, a compact convolutional neural network, is benchmarked against a spatially homogenized control in which each luminescence channel is replaced by its spatial average while retaining the same learning framework and leakage-aware protocol. The comparison indicates that global luminescence evolution contains most of the predictive signal, while spatial information provides a secondary contribution to robustness. These results establish spatially resolved luminescence imaging as a practical route for accelerated stability testing and non-invasive degradation monitoring in perovskite photovoltaics.

Figures

Figures reproduced from arXiv: 2603.12857 by Aldo Di Carlo, Chiara Ostendi, Eliodoro Chiavazzo, Erica Magliano, Giulio Barletta, Marialucia D'Addio, Pietro Asinari, Saif Ali, Simon Ternes, Zohair Abbas.

Figure 1
Figure 1. Figure 1: Experimental monitoring workflow and acquired data modalities. A custom chamber integrates temperature control and source-measure [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Timing diagram of one automated monitoring cycle executed between two consecutive user-selected aging times, [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: ML framework used in this work. Multimodal luminescence image time series (EL, PL [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: LumPerNet ensemble-mean parity plot and absolute error distribution for [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Cross-validated test performance comparison across models under identical device-level splits. Bars report mean [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Representative time-evolving RPCE(t) trajectories reconstructed from LumPerNet predictions on the held-out test set (point-wise inference at acquisition time, not forecasting). Panels (a)–(c) show three devices sharing the same Architecture (A), selected to represent distinct degradation dynamics within our aging dataset: (a) relatively high stability with slow decay, (b) intermediate stability with modera… view at source ↗
Figure 7
Figure 7. Figure 7: Absolute prediction error versus elapsed time on the held-out test set. Each bin reports the density of test-set samples in the ( [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

123 extracted references · 77 canonical work pages

  1. [1]

    M. A. Green, E. D. Dunlop, M. Yoshita, N. Kopidakis, K. Bothe, G. Siefer et al., Solar Cell Efficiency Tables (Version 66), Progress in Photovoltaics: Research and Applications 33 (7) (2025) 795–810.doi:10.1002/ pip.3919

    2025

  2. [2]

    Kojima, K

    A. Kojima, K. Teshima, Y . Shirai, T. Miyasaka, Organometal Halide Perovskites as Visible-Light Sensitizers for Photovoltaic Cells, J. Am. Chem. Soc. 131 (17) (2009) 6050–6051.doi:10.1021/ja809598r

    doi:10.1021/ja809598r 2009

  3. [3]

    M. A. Green, A. Ho-Baillie, H. J. Snaith, The emergence of perovskite solar cells, Nature Photon 8 (7) (2014) 506–514.doi:10.1038/nphoton.2014.134

    doi:10.1038/nphoton.2014.134 2014

  4. [4]

    H. S. Jung, N.-G. Park, Perovskite Solar Cells: From Materials to Devices, Small 11 (1) (2015) 10–25.doi: 10.1002/smll.201402767

    work page doi:10.1002/smll.201402767 2015

  5. [5]

    J. Y . Kim, J.-W. Lee, H. S. Jung, H. Shin, N.-G. Park, High-Efficiency Perovskite Solar Cells, Chem. Rev. 120 (15) (2020) 7867–7918.doi:10.1021/acs.chemrev.0c00107

    work page doi:10.1021/acs.chemrev.0c00107 2020

  6. [6]

    Correa-Baena, M

    J.-P. Correa-Baena, M. Saliba, T. Buonassisi, M. Grätzel, A. Abate, W. Tress et al., Promises and challenges of perovskite solar cells, Science 358 (6364) (2017) 739–744.doi:10.1126/science.aam6323

    work page doi:10.1126/science.aam6323 2017

  7. [7]

    Xiong, Q

    Z. Xiong, Q. Zhang, K. Cai, H. Zhou, Q. Song, Z. Han et al., Homogenized chlorine distribution for>27% power conversion efficiency in perovskite solar cells, Science 390 (6773) (2025) 638–642.doi:10.1126/ science.adw8780

    2025

  8. [8]

    URLhttps://www.longi.com/eu/news/2730-hbc-world-record/

    LONGi sets new world record with 27.30% for heterojunction back contact cells and renames next generation silicon tera wafer to TaiRay Wafer (2024). URLhttps://www.longi.com/eu/news/2730-hbc-world-record/

    2024

  9. [9]

    URLhttps://www.longi.com/en/news/silicon-perovskite-tandem-solar-cells-new-world- efficiency/ 17

    34.85%! LONGi Breaks World Record for Crystalline Silicon-Perovskite Tandem Solar Cell Efficiency Again (2025). URLhttps://www.longi.com/en/news/silicon-perovskite-tandem-solar-cells-new-world- efficiency/ 17

    2025

  10. [10]

    R. Lin, H. Gao, J. Lou, J. Xu, M. Yin, P. Wu et al., All-perovskite tandem solar cells with dipolar passivation, Nature 648 (8094) (2025) 600–606.doi:10.1038/s41586-025-09773-7

    work page doi:10.1038/s41586-025-09773-7 2025

  11. [11]

    M. V . Khenkin, E. A. Katz, A. Abate, G. Bardizza, J. J. Berry, C. Brabec et al., Consensus statement for stability assessment and reporting for perovskite photovoltaics based on ISOS procedures, Nat Energy 5 (1) (2020) 35–49.doi:10.1038/s41560-019-0529-5

    work page doi:10.1038/s41560-019-0529-5 2020

  12. [12]

    R. Wang, M. Mujahid, Y . Duan, Z.-K. Wang, J. Xue, Y . Yang, A Review of Perovskites Solar Cell Stability, Advanced Functional Materials 29 (47) (2019) 1808843.doi:10.1002/adfm.201808843

    work page doi:10.1002/adfm.201808843 2019

  13. [13]

    P. T. Mokabane, V . T. Lukong, T.-C. Jen, A review of the effect of stability issues and wide-bandgap in the application of perovskite solar cells, Mater Renew Sustain Energy 14 (2) (2025) 34.doi:10.1007/s40243- 025-00307-9

    work page doi:10.1007/s40243- 2025

  14. [14]

    Yaghoobi Nia, M

    N. Yaghoobi Nia, M. Zendehdel, B. Paci, J. Xu, M. Di Giovannantonio, A. Generosi et al., Co-crystal engineer- ing of a two-dimensional perovskite phase for perovskite solar modules with improved efficiency and stability, Nat Energy 11 (1) (2026) 135–149.doi:10.1038/s41560-025-01903-9

    doi:10.1038/s41560-025-01903-9 2026

  15. [15]

    M. A. Fazal, S. Rubaiee, Progress of PV cell technology: Feasibility of building materials, cost, performance, and stability, Solar Energy 258 (2023) 203–219.doi:10.1016/j.solener.2023.04.066

    work page doi:10.1016/j.solener.2023.04.066 2023

  16. [16]

    H. Cui, G. Heath, T. Remo, D. Ravikumar, T. Silverman, M. Deceglie et al., Technoeconomic analysis of high- value, crystalline silicon photovoltaic module recycling processes, Solar Energy Materials and Solar Cells 238 (2022) 111592.doi:10.1016/j.solmat.2022.111592

    work page doi:10.1016/j.solmat.2022.111592 2022

  17. [17]

    Y . Chen, D. Chen, P. P. Altermatt, S. Zhang, L. Wang, X. Zhang et al., Technology evolution of the photovoltaic industry: Learning from history and recent progress, Progress in Photovoltaics: Research and Applications 31 (12) (2023) 1194–1204.doi:10.1002/pip.3626

    work page doi:10.1002/pip.3626 2023

  18. [18]

    J. Bian, L. Li, L. Chen, W. Zhu, T. Zhang, Y . Liu et al., Unraveling of the reversible light soaking mech- anisms in intrinsic halide perovskite film, Journal of Power Sources 653 (2025) 237704.doi:10.1016/ j.jpowsour.2025.237704

    arXiv 2025

  19. [19]

    L. Lin, L. Yang, G. Du, X. Li, Y .-n. Li, J. Deng et al., Light Soaking Effects in Perovskite Solar Cells: Mechanism, Impacts, and Elimination, ACS Appl. Energy Mater. 6 (20) (2023) 10303–10318.doi:10.1021/ acsaem.2c04120

    2023

  20. [20]

    Ahmad, M

    Z. Ahmad, M. A. Najeeb, R. A. Shakoor, A. Alashraf, S. A. Al-Muhtaseb, A. Soliman et al., Instability in CH3NH3PbI3 perovskite solar cells due to elemental migration and chemical composition changes, Sci Rep 7 (1) (2017) 15406.doi:10.1038/s41598-017-15841-4

    work page doi:10.1038/s41598-017-15841-4 2017

  21. [21]

    Kalasariya, P

    N. Kalasariya, P. F. Sowmeeh, F. Pena-Camargo, F. Vanin, T. Lukas, Y . Dong et al., How Halide Segregation Governs the Ion Density Evolution and Ionic Performance Losses: From Degradation to Recovery, Advanced Energy Materials (2026) e03866doi:10.1002/aenm.202503866

    work page doi:10.1002/aenm.202503866 2026

  22. [22]

    J. T. DuBose, P. V . Kamat, Hole Trapping in Halide Perovskites Induces Phase Segregation, Acc. Mater. Res. 3 (7) (2022) 761–771.doi:10.1021/accountsmr.2c00076

    work page doi:10.1021/accountsmr.2c00076 2022

  23. [23]

    Y . Guo, C. Zhang, L. Wang, X. Yin, B. Sun, C. Wei et al., Unveiling the impact of photoinduced halide segre- gation on performance degradation in wide-bandgap perovskite solar cells, Energy & Environmental Science 18 (5) (2025) 2308–2317.doi:10.1039/D4EE05604C

    work page doi:10.1039/d4ee05604c 2025

  24. [24]

    H. Zhu, S. Teale, M. N. Lintangpradipto, S. Mahesh, B. Chen, M. D. McGehee et al., Long-term operating sta- bility in perovskite photovoltaics, Nat Rev Mater 8 (9) (2023) 569–586.doi:10.1038/s41578-023-00582- w. 18

    work page doi:10.1038/s41578-023-00582- 2023

  25. [25]

    Ahmed Chowdhury, M

    T. Ahmed Chowdhury, M. A. B. Zafar, M. S.-U. Islam, M. Shahinuzzaman, M. Aminul Islam, M. Uddin Khan- daker, Stability of perovskite solar cells: issues and prospects, RSC Advances 13 (3) (2023) 1787–1810. doi:10.1039/D2RA05903G

    work page doi:10.1039/d2ra05903g 2023

  26. [26]

    Kim, J.-Y

    H.-S. Kim, J.-Y . Seo, N.-G. Park, Material and Device Stability in Perovskite Solar Cells, ChemSusChem 9 (18) (2016) 2528–2540.doi:10.1002/cssc.201600915

    work page doi:10.1002/cssc.201600915 2016

  27. [27]

    T. J. Silverman, M. G. Deceglie, I. Repins, M. Owen-Bellini, J. J. Berry, J. S. Stein et al., Durability research is pivotal for perovskite photovoltaics, Nat Energy 10 (8) (2025) 934–940.doi:10.1038/s41560-025-01786- w

    work page doi:10.1038/s41560-025-01786- 2025

  28. [28]

    Jiang, R

    Q. Jiang, R. Tirawat, R. A. Kerner, E. A. Gaulding, Y . Xian, X. Wang et al., Towards linking lab and field lifetimes of perovskite solar cells, Nature 623 (7986) (2023) 313–318.doi:10.1038/s41586-023-06610-7

    doi:10.1038/s41586-023-06610-7 2023

  29. [29]

    M. Jošt, B. Lipovšek, B. Glažar, A. Al-Ashouri, K. Brecl, G. Matiˇc et al., Perovskite Solar Cells go Outdoors: Field Testing and Temperature Effects on Energy Yield, Advanced Energy Materials 10 (25) (2020) 2000454. doi:10.1002/aenm.202000454

    work page doi:10.1002/aenm.202000454 2020

  30. [30]

    G. M. Meheretu, A. K. Worku, M. T. Yihunie, R. K. Koech, G. A. Wubetu, The recent advancement of outdoor performance of perovskite photovoltaic cells technology, Heliyon 10 (17) (Sep. 2024).doi: 10.1016/j.heliyon.2024.e36710

    work page doi:10.1016/j.heliyon.2024.e36710 2024

  31. [31]

    T. J. Jacobsson, A. Hultqvist, A. García-Fernández, A. Anand, A. Al-Ashouri, A. Hagfeldt et al., An open- access database and analysis tool for perovskite solar cells based on the FAIR data principles, Nat Energy 7 (1) (2022) 107–115.doi:10.1038/s41560-021-00941-3

    work page doi:10.1038/s41560-021-00941-3 2022

  32. [32]

    Nur-E-Alam, M

    M. Nur-E-Alam, M. S. Islam, T. Abedin, M. A. Islam, B. K. Yap, T. S. Kiong et al., Current scenario and future trends on stability issues of perovskite solar cells: A mini review, Current Opinion in Colloid & Interface Science 76 (2025) 101895.doi:10.1016/j.cocis.2025.101895

    work page doi:10.1016/j.cocis.2025.101895 2025

  33. [33]

    R. Azmi, E. Ugur, A. Seitkhan, F. Aljamaan, A. S. Subbiah, J. Liu et al., Damp heat–stable perovskite solar cells with tailored-dimensionality 2D/3D heterojunctions, Science 376 (6588) (2022) 73–77.doi:10.1126/ science.abm5784

    2022

  34. [34]

    Mavlonov, Y

    A. Mavlonov, Y . Hishikawa, Y . Kawano, T. Negami, A. Hayakawa, S. Tsujimura et al., Thermal stability test on flexible perovskite solar cell modules to estimate activation energy of degradation on temperature, Solar Energy Materials and Solar Cells 277 (2024) 113148.doi:10.1016/j.solmat.2024.113148

    work page doi:10.1016/j.solmat.2024.113148 2024

  35. [35]

    M. N. Lintangpradipto, H. Zhu, B. Shao, W. J. Mir, L. Gutiérrez-Arzaluz, B. Turedi et al., Single-Crystal Methylammonium-Free Perovskite Solar Cells with Efficiencies Exceeding 24% and High Thermal Stability, ACS Energy Lett. 8 (11) (2023) 4915–4922.doi:10.1021/acsenergylett.3c01935

    work page doi:10.1021/acsenergylett.3c01935 2023

  36. [36]

    M. O. Reese, S. A. Gevorgyan, M. Jørgensen, E. Bundgaard, S. R. Kurtz, D. S. Ginley et al., Consensus stability testing protocols for organic photovoltaic materials and devices, Solar Energy Materials and Solar Cells 95 (5) (2011) 1253–1267.doi:10.1016/j.solmat.2011.01.036

    work page doi:10.1016/j.solmat.2011.01.036 2011

  37. [37]

    M. V . Khenkin, A. K. M, I. Visoly-Fisher, Y . Galagan, F. D. Giacomo, B. Ramesh Patil et al., Reconsidering figures of merit for performance and stability of perovskite photovoltaics, Energy & Environmental Science 11 (4) (2018) 739–743.doi:10.1039/C7EE02956J

    work page doi:10.1039/c7ee02956j 2018

  38. [38]

    Aristidou, I

    N. Aristidou, I. Sanchez-Molina, T. Chotchuangchutchaval, M. Brown, L. Martinez, T. Rath et al., The Role of Oxygen in the Degradation of Methylammonium Lead Trihalide Perovskite Photoactive Layers, Angewandte Chemie International Edition 54 (28) (2015) 8208–8212.doi:10.1002/anie.201503153

    work page doi:10.1002/anie.201503153 2015

  39. [39]

    Thiesbrummel, S

    J. Thiesbrummel, S. Shah, E. Gutierrez-Partida, F. Zu, F. Peña-Camargo, S. Zeiske et al., Ion-induced field screening as a dominant factor in perovskite solar cell operational stability, Nat Energy 9 (6) (2024) 664–676. doi:10.1038/s41560-024-01487-w. 19

    doi:10.1038/s41560-024-01487-w 2024

  40. [40]

    J. A. Kress, C. Quarti, Q. An, S. Bitton, N. Tessler, D. Beljonne et al., Persistent Ion Accumulation at In- terfaces Improves the Performance of Perovskite Solar Cells, ACS Energy Lett. 7 (10) (2022) 3302–3310. doi:10.1021/acsenergylett.2c01636

    work page doi:10.1021/acsenergylett.2c01636 2022

  41. [41]

    Moghadamzadeh, I

    S. Moghadamzadeh, I. M. Hossain, M. Jakoby, B. A. Nejand, D. Rueda-Delgado, J. A. Schwenzer et al., Spon- taneous enhancement of the stable power conversion efficiency in perovskite solar cells, Journal of Materials Chemistry A 8 (2) (2020) 670–682.doi:10.1039/C9TA09584E

    work page doi:10.1039/c9ta09584e 2020

  42. [42]

    S. Shao, M. Abdu-Aguye, L. Qiu, L.-H. Lai, J. Liu, S. Adjokatse et al., Elimination of the light soaking effect and performance enhancement in perovskite solar cells using a fullerene derivative, Energy & Environmental Science 9 (7) (2016) 2444–2452.doi:10.1039/C6EE01337F

    work page doi:10.1039/c6ee01337f 2016

  43. [43]

    B. P. Finkenauer, Akriti, K. Ma, L. Dou, Degradation and Self-Healing in Perovskite Solar Cells, ACS Appl. Mater. Interfaces 14 (21) (2022) 24073–24088.doi:10.1021/acsami.2c01925

    work page doi:10.1021/acsami.2c01925 2022

  44. [44]

    Prete, M

    M. Prete, M. V . Khenkin, D. Glowienka, B. R. Patil, J. S. Lissau, I. Dogan et al., Bias-Dependent Dynamics of Degradation and Recovery in Perovskite Solar Cells, ACS Appl. Energy Mater. 4 (7) (2021) 6562–6573. doi:10.1021/acsaem.1c00588

    work page doi:10.1021/acsaem.1c00588 2021

  45. [45]

    Y . Song, Z. Liu, X. Cai, H. Ge, X. Liu, X. Wang et al., Efficient and Moisture Resistant Wide-Bandgap Per- ovskite Solar Cells with Phosphinate-Based Iodine Defect Passivation, Advanced Energy Materials 15 (30) (2025) 2500650.doi:10.1002/aenm.202500650

    work page doi:10.1002/aenm.202500650 2025

  46. [46]

    T. Chen, J. Xie, P. Gao, Ultraviolet Photocatalytic Degradation of Perovskite Solar Cells: Progress, Chal- lenges, and Strategies, Advanced Energy and Sustainability Research 3 (6) (2022) 2100218.doi:10.1002/ aesr.202100218

    2022

  47. [47]

    S.-W. Lee, S. Kim, S. Bae, K. Cho, T. Chung, L. E. Mundt et al., UV Degradation and Recovery of Perovskite Solar Cells, Sci Rep 6 (1) (2016) 38150.doi:10.1038/srep38150

    work page doi:10.1038/srep38150 2016

  48. [48]

    Farooq, I

    A. Farooq, I. M. Hossain, S. Moghadamzadeh, J. A. Schwenzer, T. Abzieher, B. S. Richards et al., Spectral Dependence of Degradation under Ultraviolet Light in Perovskite Solar Cells, ACS Appl. Mater. Interfaces 10 (26) (2018) 21985–21990.doi:10.1021/acsami.8b03024

    work page doi:10.1021/acsami.8b03024 2018

  49. [49]

    J. Li, Q. Dong, N. Li, L. Wang, Direct Evidence of Ion Diffusion for the Silver-Electrode-Induced Thermal Degradation of Inverted Perovskite Solar Cells, Advanced Energy Materials 7 (14) (2017) 1602922.doi: 10.1002/aenm.201602922

    doi:10.1002/aenm.201602922 2017

  50. [50]

    M. Qi, L. Jin, H. Yao, Z. Xu, T. Cheng, Q. Chen et al., Recent progress on electrical failure and stability of perovskite solar cells under reverse bias, Acta Physico-Chimica Sinica 41 (8) (2025) 100088.doi:10.1016/ j.actphy.2025.100088

    arXiv 2025

  51. [51]

    S. Bae, S. Kim, S.-W. Lee, K. J. Cho, S. Park, S. Lee et al., Electric-Field-Induced Degradation of Methy- lammonium Lead Iodide Perovskite Solar Cells, J. Phys. Chem. Lett. 7 (16) (2016) 3091–3096.doi: 10.1021/acs.jpclett.6b01176

    work page doi:10.1021/acs.jpclett.6b01176 2016

  52. [52]

    Jowett, Oxford PV starts commercial distribution of perovskite solar modules (Sep

    P. Jowett, Oxford PV starts commercial distribution of perovskite solar modules (Sep. 2024). URLhttps://www.pv-magazine.com/2024/09/05/oxford-pv-starts-commercial-distribution- of-perovskite-solar-modules/

    2024

  53. [53]

    Kennedy, Plenitude tests perovskite-silicon tandem solar modules in utility-scale installation pilot (Nov

    R. Kennedy, Plenitude tests perovskite-silicon tandem solar modules in utility-scale installation pilot (Nov. 2025). URLhttps://www.pv-magazine.com/2025/11/26/plenitude-tests-perovskite-silicon- tandem-solar-modules-in-utility-scale-installation-pilot/ 20

    2025

  54. [54]

    Shaw, Chinese PV Industry Brief: PowerChina signs 1.05 GW Terra solar deal (Dec

    V . Shaw, Chinese PV Industry Brief: PowerChina signs 1.05 GW Terra solar deal (Dec. 2024). URLhttps://www.pv-magazine.com/2024/12/03/chinese-pv-industry-brief-powerchina- signs-1-05-gw-terra-solar-deal/

    2024

  55. [55]

    V . E. Puranik, R. Kumar, R. Gupta, Progress in module level quantitative electroluminescence imag- ing of crystalline silicon PV module: A review, Solar Energy 264 (2023) 111994.doi:10.1016/ j.solener.2023.111994

    arXiv 2023

  56. [56]

    Z. Wang, P. Zheng, B. Bahadir Kocer, M. Kovac, Drone-Based Solar Cell Inspection With Autonomous Deep Learning, in: Infrastructure Robotics, John Wiley & Sons, Ltd, 2024, pp. 337–365.doi:10.1002/ 9781394162871.ch16

    2024

  57. [57]

    B. Doll, J. Hepp, M. Hoffmann, R. Schüler, C. Buerhop-Lutz, I. M. Peters et al., Photoluminescence for Defect Detection on Full-Sized Photovoltaic Modules, IEEE Journal of Photovoltaics 11 (6) (2021) 1419–1429.doi: 10.1109/JPHOTOV.2021.3099739

    work page doi:10.1109/jphotov.2021.3099739 2021

  58. [58]

    B. Chen, J. Peng, H. Shen, T. Duong, D. Walter, S. Johnston et al., Imaging Spatial Variations of Opti- cal Bandgaps in Perovskite Solar Cells, Advanced Energy Materials 9 (7) (2019) 1802790.doi:10.1002/ aenm.201802790

    2019

  59. [59]

    A. M. Soufiani, M. J. Y . Tayebjee, S. Meyer, A. Ho-Baillie, J. Sung Yun, R. W. MacQueen et al., Electro- and photoluminescence imaging as fast screening technique of the layer uniformity and device degradation in planar perovskite solar cells, J. Appl. Phys. 120 (3) (2016) 035702.doi:10.1063/1.4956436

    work page doi:10.1063/1.4956436 2016

  60. [60]

    Mariam, Z

    T. Mariam, Z. S. Almutawah, A. B. Phillips, S. Fu, J. Chung, Y . Li et al., Environmentally Controlled Electro- luminescence/Photoluminescence Imaging System with Current Density-V oltage Capabilities for Quantitative Degradation Analysis of Perovskite Thin Film Solar Cells, in: 2023 IEEE 50th Photovoltaic Specialists Con- ference (PVSC), 2023, pp. 1–6.do...

    doi:10.1109/pvsc48320.2023.10359933 2023

  61. [61]

    J. W. Schall, A. Glaws, N. Y . Doumon, T. J. Silverman, M. Owen-Bellini, K. Terwilliger et al., Accelerated Stress Testing of Perovskite Photovoltaic Modules: Differentiating Degradation Modes with Electrolumines- cence Imaging, Solar RRL 7 (14) (2023) 2300229.doi:10.1002/solr.202300229

    work page doi:10.1002/solr.202300229 2023

  62. [62]

    Campanari, F

    V . Campanari, F. Martelli, A. Agresti, S. Pescetelli, N. Y . Nia, F. Di Giacomo et al., Reevaluation of Photolu- minescence Intensity as an Indicator of Efficiency in Perovskite Solar Cells, Solar RRL 6 (8) (2022) 2200049. doi:10.1002/solr.202200049

    work page doi:10.1002/solr.202200049 2022

  63. [63]

    X. Geng, Y . Liu, X. Zou, E. M. J. Johansson, J. Sá, Can photoluminescence quenching be a predictor for perovskite solar cell efficiencies?, Physical Chemistry Chemical Physics 25 (34) (2023) 22607–22613.doi: 10.1039/D3CP02190D

    work page doi:10.1039/d3cp02190d 2023

  64. [64]

    S. Kim, J. H. Jang, Z. Wu, M. J. Lee, H. Y . Woo, I. Hwang, Interfacial Defects Change the Correlation between Photoluminescence, Ideality Factor, and Open-Circuit V oltage in Perovskite Solar Cells, Small 17 (33) (2021) 2101839.doi:10.1002/smll.202101839

    work page doi:10.1002/smll.202101839 2021

  65. [65]

    Krückemeier, B

    L. Krückemeier, B. Krogmeier, Z. Liu, U. Rau, T. Kirchartz, Understanding Transient Photoluminescence in Halide Perovskite Layer Stacks and Solar Cells, Advanced Energy Materials 11 (19) (2021) 2003489.doi: 10.1002/aenm.202003489

    work page doi:10.1002/aenm.202003489 2021

  66. [66]

    Dileep K, S

    R. Dileep K, S. Mandati, E. Ramasamy, S. Mallick, T. N. Rao, G. Veerappan, Rapid assessment of photovoltaic activity of perovskite solar cells by photoluminescence spectroscopy, Materials Letters 299 (2021) 130056. doi:10.1016/j.matlet.2021.130056

    work page doi:10.1016/j.matlet.2021.130056 2021

  67. [67]

    Hacene, F

    B. Hacene, F. Laufer, S. Ternes, A. Farag, R. Pappenberger, P. Fassl et al., Intensity Dependent Photolumines- cence Imaging for In-Line Quality Control of Perovskite Thin Film Processing, Advanced Materials Technolo- gies 9 (11) (2024) 2301279.doi:10.1002/admt.202301279. 21

    work page doi:10.1002/admt.202301279 2024

  68. [68]

    Basumatary, J

    P. Basumatary, J. Kumari, P. Agarwal, Probing the defects states in MAPbI3 perovskite thin films through photoluminescence and photoluminescence excitation spectroscopy studies, Optik 266 (2022) 169586.doi: 10.1016/j.ijleo.2022.169586

    work page doi:10.1016/j.ijleo.2022.169586 2022

  69. [69]

    J. Chen, C. Zhang, X. Liu, L. Peng, J. Lin, X. Chen, Carrier dynamic process in all-inorganic halide perovskites explored by photoluminescence spectra, Photon. Res., PRJ 9 (2) (2021) 151–170.doi:10.1364/PRJ.410290

    work page doi:10.1364/prj.410290 2021

  70. [70]

    H. Wu, J. Zhang, Y . Zhang, F. Cao, Z. Qiu, L. Zhang et al., Anomalous Electroluminescence Character- istics of Perovskite Modules, ACS Appl. Mater. Interfaces 16 (32) (2024) 41986–41995.doi:10.1021/ acsami.4c03397

    2024

  71. [71]

    M. A. Torre Cachafeiro, E. L. Comi, S. Parayil Shaji, S. Narbey, S. Jenatsch, E. Knapp et al., Ion Migration in Mesoscopic Perovskite Solar Cells: Effects on Electroluminescence, Open Circuit V oltage, and Photovoltaic Quantum Efficiency, Advanced Energy Materials 15 (5) (2025) 2403850.doi:10.1002/aenm.202403850

    work page doi:10.1002/aenm.202403850 2025

  72. [72]

    M. Gao, Y . Xie, P. Song, J. Qian, X. Sun, J. Liu, A Definition Rule for Defect Classification and Grading of Solar Cells Photoluminescence Feature Images and Estimation of CNN-Based Automatic Defect Detection Method, Crystals 13 (5) (2023) 819.doi:10.3390/cryst13050819

    work page doi:10.3390/cryst13050819 2023

  73. [73]

    Louis, S

    B. Louis, S. Seth, Q. An, R. Ji, Y . Vaynzof, J. Hofkens et al., In Operando Locally-Resolved Photophysics in Perovskite Solar Cells by Correlation Clustering Imaging, Advanced Materials 37 (7) (2025) 2413126.doi: 10.1002/adma.202413126

    work page doi:10.1002/adma.202413126 2025

  74. [74]

    Taherimakhsousi, B

    N. Taherimakhsousi, B. P. MacLeod, F. G. L. Parlane, T. D. Morrissey, E. P. Booker, K. E. Dettelbach et al., Quantifying defects in thin films using machine vision, npj Comput Mater 6 (1) (2020) 111.doi:10.1038/ s41524-020-00380-w

    2020

  75. [75]

    Ternes, F

    S. Ternes, F. Laufer, P. Scharfer, W. Schabel, B. S. Richards, I. A. Howard et al., Correlative In Situ Multichan- nel Imaging for Large-Area Monitoring of Morphology Formation in Solution-Processed Perovskite Layers, Solar RRL 6 (3) (2022) 2100353.doi:10.1002/solr.202100353

    work page doi:10.1002/solr.202100353 2022

  76. [76]

    Laufer, S

    F. Laufer, S. Ziegler, F. Schackmar, E. A. Moreno Viteri, M. Götz, C. Debus et al., Process Insights into Perovskite Thin-Film Photovoltaics from Machine Learning with In Situ Luminescence Data, Solar RRL 7 (7) (2023) 2201114.doi:10.1002/solr.202201114

    work page doi:10.1002/solr.202201114 2023

  77. [77]

    Laufer, M

    F. Laufer, M. Götz, U. W. Paetzold, Deep learning for augmented process monitoring of scalable perovskite thin-film fabrication, Energy & Environmental Science 18 (4) (2025) 1767–1782.doi:10.1039/D4EE03445G

    work page doi:10.1039/d4ee03445g 2025

  78. [78]

    Kumar, B

    J. Kumar, B. Nath, T. Bhattacharjee, P. C. Ramamurthy, Machine learning aided optoelectric characterization modelling and prediction of the IV parameters of perovskite solar cells with>90% accuracy, Sci Rep 15 (1) (2025) 30660.doi:10.1038/s41598-025-93004-6

    work page doi:10.1038/s41598-025-93004-6 2025

  79. [79]

    Zhang, B

    R. Zhang, B. Motes, S. Tan, Y . Lu, M.-C. Shih, Y . Hao et al., Machine Learning Prediction of Organic–Inorganic Halide Perovskite Solar Cell Performance from Optical Properties, ACS Energy Lett. 10 (4) (2025) 1714–1724. doi:10.1021/acsenergylett.4c03592

    work page doi:10.1021/acsenergylett.4c03592 2025

  80. [80]

    Glaws, J

    A. Glaws, J. W. Schall, A. Ballen, A. E. Louks, K. O. Davis, A. F. Palmstrom et al., Explainable artificial intel- ligence relates perovskite luminescence images to current-voltage metrics, Energy and AI 22 (2025) 100640. doi:10.1016/j.egyai.2025.100640

    work page doi:10.1016/j.egyai.2025.100640 2025

Showing first 80 references.