Pith. sign in

REVIEW 5 major objections 5 minor 57 references

Analysis of degradation in perovskite solar cells through physics-based machine learning

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that perovskite solar cell degradation is driven by a coordinated rise in mobile ion density and a fall in ion diffusion, with interface recombination becoming consequential at high ion concentrations.

desk verdict Useful inverse-modeling pipeline, but the headline N0/DI trends rest on broad prior-dominated posteriors and a model degeneracy, so the central claims are less robust than the abstract suggests. read the letter →

arxiv 2608.10691 v1 pith:E243VYLX submitted 2026-08-11 physics.comp-ph

Kjeld O. Jensen , Gemma Giliberti , Aldo Di Carlo , Will Clarke , Giles Richardson , Taylor Blackwell , Petra J. Cameron , Alison B. Walker This is my paper · ORCID
classification physics.comp-ph
keywords perovskitesolarcellsdegradationBayesianparameterestimationionmigrationdrift-diffusionmodelMCMChysteresisinterfacerecombination
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks what happens inside a perovskite solar cell as it degrades over hours of operation. Rather than manually adjusting simulation inputs, the authors fit a drift-diffusion model to measured current-voltage scans at four device ages using Bayesian inference, obtaining probability distributions over material parameters. The result is a specific, coupled mechanism: the mobile ion concentration rises by roughly two orders of magnitude while the ion diffusion constant falls by about a factor of forty, and electron recombination at the hole-transport interface becomes more influential because dense mobile ions screen the built-in field. The paper argues that what is often labelled "ionic loss" is actually this coordinated change, and that physics-based machine learning can replace hand-tuning when interpreting complex device measurements.

What carries the argument

The central machinery is Bayesian parameter estimation with Metropolis-Hastings Markov-Chain Monte Carlo wrapped around a one-dimensional drift-diffusion simulator (IonMonger) that tracks electrons, holes and a single mobile ion species, with the optical generation profile supplied by a transfer-matrix calculation. The inference step evaluates each proposed parameter set by simulating JV scans and comparing nine scan rates worth of $J_{sc}$, forward/reverse $V_{oc}$, and forward/reverse PCE against measurements. The physical identity doing the explanatory work is the near-decoupling of the ionic problem from the charge-carrier problem in the surface-polarisation regime: the cell's hysteresis timescale is set by ionic response, which locks $N_0$ and $D_I$ onto a ridge of constant $\sim N_0 D_I$ behavior. This is why the posterior shows a sharp ridge rather than independent uncertainty in the two ionic parameters.

What would settle it

Measure the same devices' mobile ion density and diffusivity independently of the JV-fit—for example, with low-frequency capacitance and a direct tracer-diffusion experiment at each age—and check whether $N_0$ still rises about 200-fold while $D_I$ falls about 40-fold along the same ridge; if the independent probes show no anti-correlation, or show $N_0 D_I$ varying strongly with age, the paper's central degradation mechanism fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that the measured aging of a triple-cation p-i-n perovskite cell is reproduced not by any single parameter change but by three linked changes: the mobile ion density $N_0$ increases from about $3.76\times10^{21}\,\mathrm{m^{-3}}$ to $8.7\times10^{23}\,\mathrm{m^{-3}}$; the ion diffusion constant $D_I$ falls from about $2.7\times10^{-13}$ to $7.0\times10^{-15}\,\mathrm{m^2\,s^{-1}}$; and the electron interface recombination velocity $v_{nH}$ at the perovskite/HTL interface grows from about $2.5$ to $16.5\,\mathrm{m\,s^{-1}}$. The $N_0$ and $D_I$ estimates are strongly anti-correlated, following a power-law ridge of slope about $-0.8$ in log-log space, so the hysteresis data constrain the ionic conductivity $N_0 D_I$ more tightly than either factor alone. Interface recombination matters only at high $N_0$, because mobile ions screen the internal field and push carriers toward the HTL interface. The same parameter set reproduces the trend of a growing QFLS–$V_{oc}$ gap, though not its magnitude.

Load-bearing premise

The load-bearing premise is that a one-dimensional drift-diffusion model with a single mobile ion species adequately describes the aged device, so the fitted $N_0$, $D_I$ and interface velocities are physical mechanisms rather than lumped effective parameters.

Editorial extensions

If this is right

  • Fast-versus-slow scan PCE differences cannot be read as a clean "ionic loss" metric, because the same scan-rate signature can arise from correlated changes in $N_0$, $D_I$ and interface recombination.
  • Degradation models for these cells must keep the product $N_0 D_I$ close to the fitted ridge; changing only ion density or only ion mobility is inconsistent with the measured hysteresis.
  • Interface passivation at the hole-transport layer becomes increasingly important as ion density grows, so stability strategies that address ions and interfaces in isolation are likely incomplete.
  • The open-circuit QFLS–$V_{oc}$ gap is reproduced qualitatively but not quantitatively (0.012 eV simulated versus >0.1 eV reported), so current model versions cannot account for the full PL ageing behavior.
  • The same BPE-plus-simulation workflow can be transferred to other device architectures and measurement protocols without manual parameter exploration.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • By extension, the power-law ridge with slope near $-0.8$ suggests that independent measurements of ionic conductivity alone will not separate density from mobility; a probe that directly counts ions, such as low-frequency capacitance or isotope tracer diffusion, would be needed to confirm the individual magnitudes.
  • A testable implication the paper does not draw: if the anti-correlation is physical, devices deliberately grown with different initial defect densities should show a similar drop in effective $D_I$ as $N_0$ rises, a trend that could be checked by controlled stoichiometry experiments.
  • The discrepancy between simulated and measured PL suggests additional physics—halide segregation, interlayer changes, or band-alignment shifts—lies outside the single-ion drift-diffusion description; applying the same Bayesian machinery directly to PL spectra could identify which extension is required.
  • The real-time Bayesian updating described here points toward a practical digital-twin use: inferring a field-deployed cell's degradation state from routine JV scans, without lab characterization.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper applies Bayesian parameter estimation (Metropolis-Hastings MCMC) to the IonMonger drift-diffusion model, augmented with a RayFlare optical model, to infer material parameters from published JV scans of a single p-i-n perovskite solar cell at ages 0, 90, 280, and 480 minutes. The authors report that degradation is accompanied by an increase in mobile ion density N0 (by roughly two orders of magnitude), a decrease in the mobile ion diffusion coefficient DI (by about a factor of forty), a strong N0-DI correlation that they fit with a power-law slope near -0.8, and an increase in the electron interface recombination velocity at the HTL-perovskite interface (vnH). They also simulate photoluminescence through QFLS calculations and acknowledge a quantitative discrepancy with the experimental PL results. The paper positions the work as a demonstration of physics-based machine learning for interpreting device degradation.

Significance. If the central claims were robust, the paper would make a valuable contribution by showing that multi-parameter Bayesian inference can extract degradation mechanisms from routine JV measurements, and the N0-BACE comparison provides an independent check on one inferred quantity. The paper is also honest about the PL failure and includes useful sensitivity analyses. However, the main quantitative conclusions rest on posterior means from distributions whose widths are comparable to the claimed trends, and the N0-DI power-law appears to be a manifestation of a model degeneracy rather than an independently established physical relation. The significance of the work therefore depends on whether the identifiability issues can be resolved or the claims appropriately weakened.

major comments (5)
  1. [Table 5 and Section 3] The posterior standard deviations for log10 N0 and log10 DI at ages 90, 280, and 480 minutes are approximately 1.0-1.2 (Table 5), while the differences between the age-90 and age-480 means are about 0.7 for N0 and 0.5 for DI. Under any conventional significance criterion, these differences are not statistically distinguishable, yet the abstract and conclusions assert a factor-of-200 increase in N0 and a factor-of-40 decrease in DI as the central result. The paper must report credible intervals or posterior probability of the trends, and should temper the causal language if the trends are not statistically supported.
  2. [Section 3, Fig. 4] The power-law relation with slope about -0.8 is obtained by fitting a line to posterior samples in the N0-DI plane. The paper itself states that the JV results are 'much more sensitive to the ion conductivity (proportional to N0 x DI) than to N0 and DI individually', which means the likelihood is nearly flat along the ridge. A line fitted to samples from a ridge whose orientation is set by the model's parameter degeneracy does not constitute evidence for an independent physical power-law relation. The authors should either demonstrate that the ridge orientation is not determined by the prior or by the conductivity timescale, or explicitly reframe the result as a statement about the model's identifiability rather than a physical law.
  3. [Section 2.2 and SI A.2] The likelihood noise variance rho is adjusted to achieve MCMC acceptance rates in the range 0.2-0.31, rather than being set from an actual measurement-error model. This means the posterior widths, and hence all statements about distributions being 'broad' or 'well-defined', are contingent on an arbitrary tuning parameter. The authors should justify rho from the experimental uncertainties in Jsc, Voc, and efficiency, or at least show that the qualitative conclusions are insensitive to the choice of rho.
  4. [Section 3, Fig. 3 and Table 5] For the aged devices (90, 280, 480 minutes), the marginal posteriors for N0 and DI have means near the midpoints of their broad priors (log10 N0 prior 20-26, midpoint 23; log10 DI prior -16 to -11, midpoint -13.5). The age-90 and age-280 means are 23.2 and -13.7, essentially at the prior midpoints, and the age-480 means are 23.9 and -14.2, within one standard deviation of the prior midpoint. This pattern is consistent with the data providing weak information about these parameters individually, and the apparent monotonic trend could be substantially influenced by the prior. A prior-sensitivity analysis (e.g., wider priors or prior predictive checks) is needed before the trends are presented as robust.
  5. [Section 3, PL results] The QFLS simulations fail to reproduce the experimental PL behavior in two respects: the simulated QFLS-Voc gap at age 480 min is 0.012 eV versus more than 0.1 eV reported in Ref. [12], and the simulated QFLS decreases with age while the experimental QFLS increases. The paper acknowledges this and states that no combination of N0, DI, and the four interface recombination velocities can reproduce both features. Since this discrepancy concerns the very interface-recombination mechanism that is one of the paper's key conclusions, the conclusions should be explicitly limited to the JV data, and the PL comparison should be presented as an unresolved tension rather than supportive evidence for the inferred degradation mechanism.
minor comments (5)
  1. [Section 4] There is a typo in the first paragraph: 'changes n the interface recombination velocities' should be 'changes in the interface recombination velocities'.
  2. [Table 2] The parameter Ev is labelled 'Conduction band minimum (eV)' but should be 'Valence band maximum (eV)' to match the symbol and the context.
  3. [Figure A.2 caption] The Beer-Lambert absorption coefficient alpha is given as 6.34 m^-1, which appears to be a typo for 6.34e5 m^-1; please verify the value and units.
  4. [SI A.2] The statement that the MCMC chains have lengths 'always above 200' and numbers 'greater than 20' is vague; the exact values are given in Table 4, but the text should be consistent about whether 400 steps with 40 chains is typical for all analyses.
  5. [Section 3] The paper uses the phrase 'IR V' in Table 1 without defining it in the table caption; the definition 'Interface Recombination Velocity' appears only in the text and should be included in the caption for standalone readability.

Circularity Check

1 steps flagged · score 6.0 of 10

The N0-DI power-law and correlated degradation trends are read off the BPE posterior ridge, i.e. the model's degeneracy between N0 and DI, rather than being independently predicted; N0 is externally validated but DI and the correlation are not.

  1. fitted input called prediction [Section 3 (Results), text around Figures 2-4; repeated in Section 5 (Conclusions)]
    "there is a sharp ridge in the pair-wise joint N0-DI distribution... Figure 4 shows that when lines are fitted to the 90, 280 and 480 min results, they lie close to each other with slopes of ≈-0.8. This result implies that the JV-scan results are much more sensitive to the ion conductivity (∝ N0 × DI) than to N0 and DI individually."

    The N0-DI 'power-law relation' is not an independent measurement: N0 and DI are both fitted parameters in the same BPE analyses of the same JV data. The sharp posterior ridge is the likelihood degeneracy direction, and the paper itself states the JV results are more sensitive to ion conductivity N0×DI than to N0 and DI separately. Fitting a line to the posterior samples therefore recovers the model's identifiability ridge, not a physical law. The age trend (N0 up, DI down) is likewise read off the means of separate fits; for aged devices Table 5 shows these means sit near prior midpoints with log-scale standard deviations of about 1.0-1.2, so the individual DI decrease and the slope are not separately constrained.

full rationale

The paper's central inference chain is: JV scans at four ages -> BPE with IonMonger -> posterior means and correlations -> degradation mechanisms (N0 increase, DI decrease, interface recombination). The N0 increase is given independent support by comparison to BACE measurements from Ref. [12], so that part is not circular. The DI decrease and the claimed N0-DI power law, however, are outputs of the same fits that define them: the posterior ridge in the N0-DI plane is the direction in which the JV likelihood is nearly flat (sensitive mainly to N0×DI), and the reported slope ≈-0.8 is obtained by fitting a line to samples from that ridge. The paper itself explains the ridge by matching the ionic timescale, which is a combination of N0 and DI in the drift-diffusion model; hence the power law is a property of model identifiability, not an independent physical discovery. Reporting it in the abstract and conclusions as a key result is the fitted-input-called-prediction pattern. The DI decrease is not separately validated against external data; the posterior standard deviations in Table 5 (roughly 1.0-1.2 in log10 for aged devices, with means near prior midpoints) show the individual parameters are weakly constrained. The PL comparison is honestly reported as failing quantitatively (0.012 eV vs >0.1 eV gap), which limits the claimed mechanism but is not circular. Self-citations to IonMonger and the Surface Polarisation Model provide background and explanation but are not the sole support for the empirical claim, so they do not raise the score further.

Assumptions & free parameters 9 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical particles, forces, or conserved quantities. The 'single mobile ion species' is an acknowledged modeling simplification, not a posited new entity. The free parameters are dominated by the six material parameters fitted to JV data, plus an ad hoc optical multiplier and likelihood hyperparameters.

free parameters (9)
  • N0 (mobile ion density) = 3.76e21 to 8.70e23 m^-3 (per age)
    Fitted to JV scans; central claim is its increase with age.
  • DI (mobile ion diffusion coefficient) = 2.68e-13 to 7.01e-15 m^2/s
    Fitted to JV scans; central claim is its decrease with age.
  • vnH (electron interface recombination velocity at HTL) = 2.5 to 16.5 m/s
    Fitted; central to claim of interface recombination at high N0.
  • vnE (electron interface recombination velocity at ETL) = 3.02 to 17.54 m/s
    Fitted but posterior is broad; not well constrained.
  • vpE (hole interface recombination velocity at ETL) = 8.0 to 14.2 m/s
    Fitted but posterior is broad; not well constrained.
  • vpH (hole interface recombination velocity at HTL) = 2.4 to 8.8 m/s
    Fitted; posterior becomes peaked at later ages.
  • G_adj (optical generation multiplier) = 1.184
    Fitted to pristine-device Jsc in a preliminary BPE, then held constant; an ad hoc correction for optical model mismatch.
  • rho (likelihood noise variance) = 0.0001 (age 0), 0.001 (other ages)
    Ad hoc variance parameter tuned to achieve acceptance rates 0.20-0.31.
  • Likelihood weights = w=2 for Jsc and Vocf; w=1 otherwise
    Ad hoc weighting to emphasize sensitive outputs in the WMSE.
assumptions (5)
  • domain assumption IonMonger's 1D drift-diffusion model with one mobile ionic species is an adequate description of the perovskite solar cell.
    Invoked throughout Section 2.1; if false (e.g., multiple ion species or 2D effects), the inferred parameters are effective rather than physical.
  • ad hoc to paper The optical model with data from a similar but not identical stack, plus a constant G_adj, accurately represents generation in the measured device.
    Section 2.1 and SI A.1; the authors themselves note refractive indices and layer thicknesses may differ and apply an ad hoc multiplier.
  • standard math Uniform priors and the tuned diagonal-covariance likelihood yield a Bayesian posterior that can be interpreted as parameter uncertainty.
    Section 2.2; the likelihood is not derived from measurement error but set ad hoc.
  • domain assumption The six fitted parameters capture the relevant degradation mechanisms; carrier lifetimes and doping densities do not change with age.
    Section 3; supported only by the finding that their posteriors are flat in a preliminary 10-parameter BPE.
  • domain assumption The published JV and BACE measurements from Ref [12] are accurate and reproducible.
    Used as ground truth; no independent verification of the data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Analysis of degradation in perovskite solar cells through physics-based machine learning." pith.science (2026). https://pith.science/paper/E243VYLX

@misc{pith2026260810691,
  author       = {Pith},
  title        = {Pith review of: Analysis of degradation in perovskite solar cells through physics-based machine learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E243VYLX}},
  note         = {Machine review of arXiv:2608.10691}
}
read the original abstract

Degradation in lead halide perovskite solar cells is analysed by inverse modelling of published measurements of characteristics of a single solar cell at ages 0, 90, 280, 480 minutes. We employ machine learning to deduce distributions of material parameter values and hence the physics linked to measured changes. Bayesian parameter estimation is coupled with drift diffusion simulations using the IonMonger code combined with an optical model. We accurately replicated measured changes in device performance with age through variations in model input parameters. Our key result is that degradation is influenced by correlated changes in the concentrations and diffusion coefficients of mobile ions and by interface recombination at large mobile ion concentrations. This study demonstrates the power of machine learning combined with simulations to reliably interpret experimental results, a task which is problematic if using simulation models with only manual exploration of the input parameter space.

Figures

Figures reproduced from arXiv: 2608.10691 by the authors.

Figure 1
Figure 1. Posterior distributions (counts of samples from the MCMC chains) from the BPE analysis [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Posterior distributions from the BPE analysis of the age-480-min device. See caption to [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Posterior distributions from BPE for the four ages for selected input parameters, showing [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Scatter plot of log10(DI ) vs log10(N0) from BPE samples for the age-480-min device (blue), linear fit to these points (orange) and the linear fits to similar scatter plots for the ages 90 min and 280 min devices (green and purple lines, respectively). The diamonds sho…
Figure 5
Figure 5. Figure 5: IonMonger JV-scan results (orange) with inputs set to the means from the BPE (Table [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Panel (a): Comparison of Mobile Ion Densities: BPE means from the current analysis [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 57 canonical work pages

  1. [12]

    Ion-induced field screening as a dominant factor in perovskite solar cell operational stability.Nature Energy, 9:1–13, 2024

    Jarla Thiesbrummel, Sahil Shah, Emilio Gutierrez-Partida, Fengshuo Zu, Francisco Pe˜ na- Camargo, Stefan Zeiske, Jonas Diekmann, Fangyuan Ye, Karol P Peters, Kai O Brinkmann, et al. Ion-induced field screening as a dominant factor in perovskite solar cell operational stability.Nature Energy, 9:1–13, 2024. 17

    work page 2024

  2. [1]

    De rerum natura: How do halide perovskites self-heal from damage?Advanced Materials, 38(21):e18808, 2026

    Davide Raffaele Ceratti, Gary Hodes, and David Cahen. De rerum natura: How do halide perovskites self-heal from damage?Advanced Materials, 38(21):e18808, 2026

    work page 2026

  3. [2]

    Roadmap on established and emerging photovoltaics for sustainable energy conversion.J

    J C et al Blakesley. Roadmap on established and emerging photovoltaics for sustainable energy conversion.J. Phys. Energy, 6:041501, 2024

    work page 2024

  4. [3]

    Eperon, Alessandro Virtuani, Quentin Jeangros, Dana B

    Sara Baumann, Giles E. Eperon, Alessandro Virtuani, Quentin Jeangros, Dana B. Kern, Dounya Barrit, Jackson Schall, Wanyi Nie, Gernot Oreski, Mark Khenkin, Carolin Ulbrich, Robby Peibst, Joshua S. Stein, and Marc K¨ ontges. Stability and reliability of perovskite containing solar cells and modules: degradation mechanisms and mitigation strategies.Energy En...

    work page 2024

  5. [4]

    Ion migration in perovskite solar cells.Nature Reviews Chemistry, pages 1–17, 2026

    Jarla Thiesbrummel, Jovana V Mili´ c, Carsten Deibel, Erik C Garnett, Shuxia Tao, Thomas Kirchartz, Antonio Guerrero, Petra Cameron, Wolfgang Tress, M Saiful Islam, et al. Ion migration in perovskite solar cells.Nature Reviews Chemistry, pages 1–17, 2026

    work page 2026

  6. [5]

    Nonradiative re- combination in perovskite solar cells: the role of interfaces.Advanced Materials, 31(52):1902762, 2019

    Christian M Wolff, Pietro Caprioglio, Martin Stolterfoht, and Dieter Neher. Nonradiative re- combination in perovskite solar cells: the role of interfaces.Advanced Materials, 31(52):1902762, 2019

    work page 2019

  7. [6]

    Will Clarke, Petra Cameron, and Giles Richardson. Predicting long-term stability from short- term measurement: Insights from modeling degradation in perovskite solar cells during voltage scans and impedance spectroscopy.The Journal of Physical Chemistry Letters, 15(47):11730– 11736, 2024. PMID: 39547710

    work page 2024

  8. [7]

    The impact of interfacial quality and nanoscale performance disorder on the stability of alloyed perovskite solar cells.Nature Energy, 10(1):66–76, 2025

    Kyle Frohna, Cullen Chosy, Amran Al-Ashouri, Florian Scheler, Yu-Hsien Chiang, Milos Duba- jic, Julia E Parker, Jessica M Walker, Lea Zimmermann, Thomas A Selby, et al. The impact of interfacial quality and nanoscale performance disorder on the stability of alloyed perovskite solar cells.Nature Energy, 10(1):66–76, 2025

    work page 2025

Show all 57 references
  1. [8]

    Challenges and opportunities for the characterization of electronic properties in halide perovskite solar cells.Chemical Science, 16(19):8153–8195, 2025

    Thomas Kirchartz. Challenges and opportunities for the characterization of electronic properties in halide perovskite solar cells.Chemical Science, 16(19):8153–8195, 2025

    2025

  2. [9]

    How halide segregation governs the ion density evolution and ionic performance losses: From degradation to recovery.Advanced Energy Materials, page e03866, 2026

    Nikhil Kalasariya, Paria Forozi Sowmeeh, Francisco Pena-Camargo, Francesco Vanin, Tino Lukas, Yuxin Dong, Qifan Feng, Ziwei Liu, Waqar Ali Memon, Danpeng Gao, et al. How halide segregation governs the ion density evolution and ionic performance losses: From degradation to reco...

    2026

  3. [10]

    Will Clarke, Laurence J Bennett, Yoana Grudeva, Jamie M Foster, Giles Richardson, and Nicola E Courtier. Ionmonger 2.0: software for free, fast and versatile simulation of current, voltage and impedance response of planar perovskite solar cells.Journal of Computational Electro...

    2023

  4. [11]

    O’Regan, and Piers R.F

    Philip Calado, Andrew M Telford, Daniel Bryant, Xiaoe Li, Jenny Nelson, Brian C. O’Regan, and Piers R.F. Barnes. Evidence for ion migration in hybrid perovskite solar cells with minimal hysteresis.Nat. Commun., 7:13831, dec 2016

    2016

  5. [13]

    Lucy J. F. Hart, Fraser J. Angus, Yin Li, Abdul Khaleed, Philip Calado, James R. Durrant, Aleksandra B. Djuriˇ si´ c, Pablo Docampo, and Piers R. F. Barnes. More is different: mobile ions improve the design tolerances of perovskite solar cells.Energy Environ. Sci., 17:7107–7118, 2024

    2024

  6. [14]

    Torre Cachafeiro and W Tress

    Miguel A. Torre Cachafeiro and W Tress. Ionic Losses and Gains in Perovskite Solar Cells: Impact on Efficiency and Stability.ACS Energy Lett., 10:4849–4855, 2025

    2025

  7. [15]

    The curse(s) of dimensionality.Nat Methods, 15(15):399–400, 2018

    Naomi Altman and Martin Krzywinski. The curse(s) of dimensionality.Nat Methods, 15(15):399–400, 2018

    2018

  8. [16]

    Bayesian parameter estimation for characterising mobile ion vacancies in perovskite solar cells.Journal of Physics: Energy, 6:015005, 2023

    Samuel G McCallum, Oliver Nicholls, Kjeld O Jensen, Matthew V Cowley, James E Lerpini` ere, and Alison B Walker. Bayesian parameter estimation for characterising mobile ion vacancies in perovskite solar cells.Journal of Physics: Energy, 6:015005, 2023

    2023

  9. [17]

    Taking control of ion transport in halide perovskite solar cells.ACS Energy Letters, 3(8):1983–1990, 2018

    Aron Walsh and Samuel D Stranks. Taking control of ion transport in halide perovskite solar cells.ACS Energy Letters, 3(8):1983–1990, 2018

    1983

  10. [18]

    De Souza

    Denis Barboni and Roger A. De Souza. The thermodynamics and kinetics of iodine vacancies in the hybrid perovskite methylammonium lead iodide.Energy Environ. Sci., 11:3266–3274, 2018

    2018

  11. [19]

    Mocanu, and M

    Allison Nicole Arber, Vikram, Felix C. Mocanu, and M. Saiful Islam. Ion migration and dopant effects in the gamma-cspbi3 perovskite photovoltaic material: Atomistic insights through ab initio and machine learning methods.Chemistry of Materials, 37:4416–4424, 2025

    2025

  12. [20]

    American Chemical Society, 2022

    Keith T Butler, Felipe Oviedo, and Pieremanuele Canepa.Machine learning in materials science, volume 29. American Chemical Society, 2022

    2022

  13. [21]

    Rombach, David P

    Manuel Kober-Czerny, Akash Dasgupta, Seongrok Seo, Florine M. Rombach, David P. McMeekin, Heon Jin, and Henry J. Snaith. Determining parameters of metal-halide perovskites using photoluminescence with bayesian inference.PRX Energy, 4:013001, Jan 2025

    2025

  14. [22]

    Rapid parameter prediction in perovskite solar cells via ai-assisted performance analysis.ACS Energy Letters, 2026

    Jonas Diekmann, Zitong Yang, Nurlan Tokmoldin, Ziwei Liu, Francisco Pe˜ na-Camargo, Zhuofan Xiong, Paria Forozi Sowmeeh, Hongsheng Li, Qiuqiang Kong, Xiaoliang Ju, et al. Rapid parameter prediction in perovskite solar cells via ai-assisted performance analysis.ACS Energy Letters, 2026

    2026

  15. [23]

    Inversion of the impedance response towards physical parameter extraction using interpretable machine learning.Advanced Energy Materials, page e06352, 2026

    Mahmoud Nabil, Isel Grau, Ricardo Grau-Crespo, Said Hamad, and Juan A Anta. Inversion of the impedance response towards physical parameter extraction using interpretable machine learning.Advanced Energy Materials, page e06352, 2026

    2026

  16. [24]

    Bayesian optimization approach to quantify the effect of input parameter uncertainty on predictions of numerical physics simulations.APL Machine Learning, 1(4), 2023

    Samuel G McCallum, James E Lerpini` ere, Kjeld O Jensen, Pascal Friederich, and Alison B Walker. Bayesian optimization approach to quantify the effect of input parameter uncertainty on predictions of numerical physics simulations.APL Machine Learning, 1(4), 2023

    2023

  17. [25]

    Rayflare: flexible optical modelling of solar cells.Journal of Open Source Software, 6(65):3460, 2021

    Phoebe M Pearce. Rayflare: flexible optical modelling of solar cells.Journal of Open Source Software, 6(65):3460, 2021

    2021

  18. [26]

    Giles Richardson, Simon E. J. O’Kane, Ralf Gregor Niemann, Timo A. Peltola, Jamie Michael Foster, Petra Jane Cameron, and Alison B. Walker. Can slow-moving ions explain hysteresis in the current-voltage curves of perovskite solar cells?Energy Environ. Sci., 9(4):1476–1485, 2016. 18

    2016

  19. [27]

    Nicola E Courtier, James M Cave, Jamie M Foster, Alison B Walker, and Giles Richardson. How transport layer properties affect perovskite solar cell performance: insights from a coupled charge transport/ion migration model.Energy & Environmental Science, 12(1):396–409, 2019

    2019

  20. [28]

    Inverted hysteresis as a diagnostic tool for perovskite solar cells: Insights from the drift-diffusion model.Journal of Applied Physics, 133(9), 2023

    Will Clarke, Matthew V Cowley, Matthew J Wolf, Petra Cameron, Alison Walker, and Giles Richardson. Inverted hysteresis as a diagnostic tool for perovskite solar cells: Insights from the drift-diffusion model.Journal of Applied Physics, 133(9), 2023

    2023

  21. [29]

    Laurence J Bennett, Antonio J Riquelme, Juan A Anta, Nicola E Courtier, and Giles Richard- son. Avoiding ionic interference in computing the ideality factor for perovskite solar cells and an analytical theory of their impedance-spectroscopy response.Physical Review Applied, 19...

    2023

  22. [30]

    Understanding the full zoo of perovskite so- lar cell impedance spectra with the standard drift-diffusion model.Advanced Energy Materials, page 2400955, 2024

    Will Clarke, Giles Richardson, and Petra Cameron. Understanding the full zoo of perovskite so- lar cell impedance spectra with the standard drift-diffusion model.Advanced Energy Materials, page 2400955, 2024

    2024

  23. [31]

    Daniel A Jacobs, Heping Shen, Florian Pfeffer, Jun Peng, Thomas P White, Fiona J Beck, and Kylie R Catchpole. The two faces of capacitance: New interpretations for electrical impedance measurements of perovskite solar cells and their relation to hysteresis.Journal of Applied P...

    2018

  24. [32]

    CreateSpace, United States, 3rd ed edition, 2013

    Andrew Gelman,{John B.}Carlin,{Hal S.}Stern,{David B.}Dunson, Aki Vehtari, and {Donald B.}Rubin.Bayesian Data Analysis. CreateSpace, United States, 3rd ed edition, 2013

    2013

  25. [33]

    Multifunctional dual-interface layer enables efficient and stable inverted perovskite solar cells.Physical Chemistry Chemical Physics, 26(10):8299–8307, 2024

    Chaofeng Wang, Yi Guo, Shuang Liu, Jiajia Huang, Xiaohui Liu, Jing Zhang, Ziyang Hu, Yuejin Zhu, and Like Huang. Multifunctional dual-interface layer enables efficient and stable inverted perovskite solar cells.Physical Chemistry Chemical Physics, 26(10):8299–8307, 2024

    2024

  26. [34]

    Solution-processed metal-oxide nanoparticles to prevent the sputtering damage in perovskite/silicon tandem solar cells.ACS Applied Materials & Interfaces, 17(11):17599–17610, 2025

    Erica Magliano, Francesco Di Giacomo, Harshavardhan Reddy Sathy, Shirin M Pourmotlagh, Gemma Giliberti, David Becerril Rodriguez, Giuseppe Ammirati, Paolo Mariani, Francesca Zarotti, Fabio Matteocci, et al. Solution-processed metal-oxide nanoparticles to prevent the sputtering...

    2025

  27. [35]

    Kauf- mann, and Aldo Di Carlo

    Gemma Giliberti, Guillermo Farias-Basulto, Klaus J¨ ager, Thede Mehlhop, Christian A. Kauf- mann, and Aldo Di Carlo. Assessment and optimization of 2t perovskite/cigs tandems via data-driven and optoelectronic modelling.Nano-Micro Letters, 18:312, 2026

    2026

  28. [36]

    Hill, Matthew V

    Nathan S. Hill, Matthew V. Cowley, Nadja Gluck, Miriam H. Fsadni, Will Clarke, Yinghong Hu, Matthew J. Wolf, Noel Healy, Marina Freitag, Thomas J. Penfold, Giles Richardson, Alison B. Walker, Petra J. Cameron, and Pablo Docampo. Ionic accumulation as a diagnostic tool in perov...

    2023

  29. [37]

    N. E. Courtier, J. M. Foster, S. E. J. O’Kane, A. B. Walker, and G. Richardson. Systematic derivation of a surface polarisation model for planar perovskite solar cells.Eur. J. Appl. Math., page doi:10.1017/S0956792518000207, 2018

    doi

  30. [38]

    Jonas Diekmann, Francisco Pe˜ na-Camargo, Nurlan Tokmoldin, Jarla Thiesbrummel, Jonathan Warby, Emilio Gutierrez-Partida, Sahil Shah, Dieter Neher, and Martin Stolterfoht. Determi- nation of mobile ion densities in halide perovskites via low-frequency capacitance and charge ex...

    2023

  31. [39]

    Cave, Nicola E

    James M. Cave, Nicola E. Courtier, Isabelle A. Blakborn, Timothy W. Jones, Dibyajyoti Ghosh, Kenrick F. Anderson, Liangyou Lin, Andrew A. Dijkhoff, Gregory J. Wilson, Krishna Feron, M. Saiful Islam, Jamie M. Foster, Giles Richardson, and Alison B. Walker. Deducing transport pr...

    2020

  32. [40]

    Understanding performance limiting interfacial recombination in pin perovskite solar cells.Advanced Energy Materials, 13:230313, 2023

    Jonathan Warby, Fengshuo Zu, Stefan Zeiske, Emilio Gutierrez-Partida, Lennart Frohloff, Si- mon Kahmann, Kyle Frohna, Edoardo Mosconi, Eros Radicchi, Felix Lang, et al. Understanding performance limiting interfacial recombination in pin perovskite solar cells.Advanced Energy M...

    2023

  33. [41]

    Rombach, Akash Dasgupta, Manuel Kober-Czerny, et al

    Florian M. Rombach, Akash Dasgupta, Manuel Kober-Czerny, et al. Disentangling degrada- tion pathways of narrow bandgap lead-tin perovskite material and photovoltaic devices.Nat Commun, 16:5450, 2025

    2025

  34. [42]

    Preventing phase segregation in mixed-halide per- ovskites: a perspective.Energy & Environmental Science, 13(7):2024–2046, 2020

    Alexander J Knight and Laura M Herz. Preventing phase segregation in mixed-halide per- ovskites: a perspective.Energy & Environmental Science, 13(7):2024–2046, 2020

    2024

  35. [43]

    De Souza and Denis Barboni

    Roger A. De Souza and Denis Barboni. Iodide-ion conduction in methylammonium lead iodide perovskite: some extraordinary aspects.Chem. Commun., 55:1108–1111, 2019

    2019

  36. [44]

    Defect chemistry of mixed ionic–electronic conductors under light: halide perovskites as a master example.Mater

    Davide Moia and Joachim Maier. Defect chemistry of mixed ionic–electronic conductors under light: halide perovskites as a master example.Mater. Horiz, 13:763–778, 2026

    2026

  37. [45]

    Matthias Diethelm, Tino Lukas, Joel Smith, Akash Dasgupta, Pietro Caprioglio, Moritz Futscher, Roland Hany, and Henry J Snaith. Probing ionic conductivity and electric field screening in perovskite solar cells: a novel exploration through ion drift currents.Energy & Environmen...

    2025

  38. [46]

    Electrical conductivity of halide perovskites follows expectations from classical defect chemistry.European Journal of Inorganic Chemistry, 2021(28):2882–2889, 2021

    Nico Leupold, Anna Lena Seibel, Ralf Moos, and Fabian Panzer. Electrical conductivity of halide perovskites follows expectations from classical defect chemistry.European Journal of Inorganic Chemistry, 2021(28):2882–2889, 2021

    2021

  39. [47]

    Toolsets for assessing ionic migration in halide per- ovskites.Joule, 8(5):1239–1273, 2024

    Natalia Yantara and Nripan Mathews. Toolsets for assessing ionic migration in halide per- ovskites.Joule, 8(5):1239–1273, 2024

    2024

  40. [48]

    Effect of thermal stress on the ion density and mobility distribution in perovskite solar cells.The Journal of Physical Chemistry Letters, 2026

    Manjeet Kumar, Changzeng Ding, Jiansheng Yang, Oskar J Sandberg, Mathias Nyman, Chang- Qi Ma, and Ronald ¨Osterbacka. Effect of thermal stress on the ion density and mobility distribution in perovskite solar cells.The Journal of Physical Chemistry Letters, 2026

    2026

  41. [49]

    Ionic-defect distribution revealed by improved evaluation of deep-level transient spectroscopy on perovskite solar cells.Physical Review Applied, 13(3):034018, 2020

    Sebastian Reichert, Jens Flemming, Qingzhi An, Yana Vaynzof, Jan-Frederik Pietschmann, and Carsten Deibel. Ionic-defect distribution revealed by improved evaluation of deep-level transient spectroscopy on perovskite solar cells.Physical Review Applied, 13(3):034018, 2020

    2020

  42. [50]

    How to tell the difference between a model and a digital twin.Advanced Modeling and Simulation in Engineering Sciences, 7:7–13, 2020

    Louise Wright and Stuart Davidson. How to tell the difference between a model and a digital twin.Advanced Modeling and Simulation in Engineering Sciences, 7:7–13, 2020

    2020

  43. [51]

    A fast and robust numerical scheme for solving models of charge carrier transport and ion vacancy motion in perovskite solar cells

    NE Courtier, Giles Richardson, and Jamie M Foster. A fast and robust numerical scheme for solving models of charge carrier transport and ion vacancy motion in perovskite solar cells. Applied Mathematical Modelling, 63:329–348, 2018. 20

    2018

  44. [52]

    Substitution of lead with tin suppresses ionic transport in halide perovskite optoelectronics

    Krishanu Dey, Dibyajyoti Ghosh, Matthew Pilot, Samuel R Pering, Bart Roose, Priyanka Deswal, Satyaprasad P Senanayak, Petra J Cameron, M Saiful Islam, and Samuel D Stranks. Substitution of lead with tin suppresses ionic transport in halide perovskite optoelectronics. Energy & ...

    2024

  45. [53]

    Avoid pitfalls in identi- fying perovskite grain size.The journal of physical chemistry letters, 13(31):7236–7242, 2022

    Bowen Jin, Yidong Ming, Zihui Liang, Kai Wang, and Congcong Wu. Avoid pitfalls in identi- fying perovskite grain size.The journal of physical chemistry letters, 13(31):7236–7242, 2022

    2022

  46. [54]

    Light management in perovskite photovoltaic solar cells: A perspective

    Florian Berry, Rapha¨ el Mermet-Lyaudoz, Jose Maria Cuevas Davila, Djihad Amina Djemmah, Hai Son Nguyen, Christian Seassal, Erwann Fourmond, C´ eline Chevalier, Mohamed Amara, and Emmanuel Drouard. Light management in perovskite photovoltaic solar cells: A perspective. Adv Ene...

    2022

  47. [55]

    Martin Stolterfoht, Pietro Caprioglio, Christian M Wolff, Jos´ e A M´ arquez, Joleik Nordmann, Shanshan Zhang, Daniel Rothhardt, Ulrich H¨ ormann, Yohai Amir, Alex Redinger, et al. The impact of energy alignment and interfacial recombination on the internal and external open- ...

    2019

  48. [56]

    Functionalized substrates for reduced nonradiative recombination in metal-halide per- ovskites.The Journal of Physical Chemistry Letters, 16(1):372–377, 2025

    Guus JW Aalbers, Willemijn HM Remmerswaal, Ralph HC van den Heuvel, Laura Bellini, Lana M Kessels, Christ HL Weijtens, Nick RM Schipper, Martijn M Wienk, and Ren´ e AJ Janssen. Functionalized substrates for reduced nonradiative recombination in metal-halide per- ovskites.The J...

    2025

  49. [57]

    30 (a) (b) (c) Figure A.6: QFLS,qV oc and their difference, in eV, from open-circuit IonMonger runs for combina- tions of interface recombination velocity values

    who showed that for mixed-halide perovskites phase segregation, where halide atoms segregate into regions dominated by one or other of the halide constituents, can occur on similar time scales to those analysed in this paper, i.e.<1000 minutes, and that this can have a marked ...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.