pith. sign in

arxiv: 2606.17186 · v1 · pith:JJ6PYJHSnew · submitted 2026-06-15 · ❄️ cond-mat.mtrl-sci

Machine learning insights into band gap properties in halide-based perovskites

Chadawan Khamdang , Mengen Wang This is my paper

Pith reviewed 2026-06-27 02:51 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci
keywords halide perovskitesband gap predictionmachine learningfeature importanceensemble methodslead-free materialsoptoelectronicsstructural descriptors
0
0 comments X

The pith

Machine learning models predict band gap energies in halide perovskites from atomic and structural properties with high accuracy.

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

This paper builds machine learning models to forecast band gap energies for halide perovskites across structures such as A2BX6, A2BB'X6, A3B2X9, and A4BX6 using atomic and structural features. Ensemble tree-based methods including random forest regression, gradient boosted regression trees, and extreme gradient boosting achieve strong predictive accuracy on these targets. Feature importance analysis singles out B-site and X-site elemental properties plus the counts of A- and B-site atoms as the dominant influences on the energies. The resulting relationships support direct mapping from composition descriptors to band gaps for material design.

Core claim

Machine learning models based on ensemble tree methods predict band gap energies in multiple types of halide perovskites from atomic and structural descriptors with high accuracy. Feature importance analysis identifies B-site and X-site elemental properties and the number of A- and B-site atoms as the primary drivers of these energies.

What carries the argument

Ensemble tree-based regression models (random forest regression, gradient boosted regression trees, extreme gradient boosting) trained on atomic and structural properties to predict band gaps.

If this is right

  • The models enable rapid screening of new lead-free perovskite compositions for desired band gaps without full electronic structure calculations.
  • The identified key descriptors supply concrete rules for choosing B-site and X-site elements to tune band gap values.
  • The approach extends predictions across the four structure families mentioned, widening the range of accessible materials.
  • Results supply guidance for exploring alternative compositions in low-toxicity optoelectronic applications.

Where Pith is reading between the lines

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

  • Feature rankings could motivate simplified empirical expressions linking specific elemental traits directly to band gap values.
  • The same descriptor set might transfer to predicting related properties such as phase stability in these compounds.
  • Integrating the models into high-throughput screening pipelines would allow faster identification of candidate materials for experimental synthesis.

Load-bearing premise

The training set of halide perovskite compositions and their band gap values is diverse and accurate enough to generalize across the listed structure types.

What would settle it

Prediction errors remain high when the models are tested on band gap measurements for new halide perovskite compositions containing elements or structures absent from the training data.

Figures

Figures reproduced from arXiv: 2606.17186 by Chadawan Khamdang, Mengen Wang.

Figure 1
Figure 1. Figure 1: FIG. 1. Crystal structures of the halide perovskite families in [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Violin plots showing the distribution of calculated [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Pearson correlation matrix showing pairwise correla [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Parity plots from random forest regression (RFR), extreme gradient boosting (XGB), and gradient boosted regression [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
read the original abstract

Halide perovskites show great promise for applications in optoelectronic devices. The lead-free perovskites are attracting increasing interest due to their low toxicity and motivate the exploration of alternative compositions and structures, including A$_2$BX$_6$, A$_2$BB$^\prime$X$_6$, A$_3$B$_2$X$_9$, and A$_4$BX$_6$. Accurate predictions of a wide range of band gap energies are important for designing new materials. It is also desired to generate a direct relationship between the structural and elemental descriptors and the band gap energies. In this work, we develop machine learning models to predict band gap energies across various types of halide perovskites based on atomic and structural properties. Algorithms including ensemble tree-based methods, random forest regression (RFR), gradient boosted regression trees (GBRT), and extreme gradient boosting (XGB) showed strong predictive accuracy. We also analyzed feature importance to identify key descriptors, including B-site and X-site elemental properties, as well as the number of A- and B-site atoms, as primary factors influencing band gap energies. These results improve our understanding of the ML models and provide guidance for designing new halide perovskite materials.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript develops machine learning models using ensemble tree-based methods (random forest regression, gradient boosted regression trees, and extreme gradient boosting) to predict band gap energies in various halide perovskite structures (A2BX6, A2BB'X6, A3B2X9, A4BX6) based on atomic and structural properties. It claims these models achieve strong predictive accuracy and uses feature importance analysis to identify B-site and X-site elemental properties and the number of A- and B-site atoms as primary factors.

Significance. If the predictive models demonstrate robust accuracy with proper validation and the feature rankings prove reliable across structure types, the work could offer practical guidance for designing lead-free halide perovskites by linking specific elemental and structural descriptors to band-gap variation.

major comments (2)
  1. [Abstract] Abstract: the assertion that RFR, GBRT, and XGB 'showed strong predictive accuracy' supplies no quantitative metrics (R², MAE, cross-validation scores), dataset size, or baseline comparisons, leaving the central claim unsupported by visible evidence.
  2. [Abstract] Abstract: no per-structure sample counts, band-gap source (DFT functional or experiment), chemical-space coverage, or stratified validation details are reported for A2BX6, A2BB'X6, A3B2X9, and A4BX6, so the claimed generalization and feature-importance rankings rest on an unverified assumption of dataset representativeness.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We agree that the abstract would be strengthened by the inclusion of quantitative metrics and dataset details, and we will revise it accordingly in the resubmitted manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the assertion that RFR, GBRT, and XGB 'showed strong predictive accuracy' supplies no quantitative metrics (R², MAE, cross-validation scores), dataset size, or baseline comparisons, leaving the central claim unsupported by visible evidence.

    Authors: We acknowledge the point. The results section of the manuscript reports the relevant performance metrics and comparisons; we will condense the key quantitative values (R², MAE, cross-validation scores) and a brief baseline reference into the abstract. revision: yes

  2. Referee: [Abstract] Abstract: no per-structure sample counts, band-gap source (DFT functional or experiment), chemical-space coverage, or stratified validation details are reported for A2BX6, A2BB'X6, A3B2X9, and A4BX6, so the claimed generalization and feature-importance rankings rest on an unverified assumption of dataset representativeness.

    Authors: We will expand the abstract to include a concise statement on total dataset size, the computational source of the band gaps, the distribution across the four structure families, and the stratified validation approach used. revision: yes

Circularity Check

0 steps flagged

No circularity: standard supervised ML on descriptors

full rationale

The paper trains conventional regression models (RFR, GBRT, XGB) on tabulated atomic/structural descriptors to predict band-gap values and reports feature importances as direct model outputs. No equations, self-citations, or ansatzes are described that reduce any claimed prediction or importance ranking to a fitted input by construction. The workflow is self-contained supervised learning whose performance can be externally validated on held-out data; the dataset-representativeness concern raised by the skeptic is a generalization/validity issue, not a circularity in the derivation chain.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The central claim rests on the representativeness of an unspecified training set of halide perovskites and on the assumption that the chosen atomic and structural descriptors capture the dominant physics of the band gap; no explicit free parameters, axioms, or invented entities are stated.

pith-pipeline@v0.9.1-grok · 5742 in / 1146 out tokens · 44080 ms · 2026-06-27T02:51:38.979843+00:00 · methodology

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

75 extracted references

  1. [1]

    Advanced Energy Materials , volume=

    Challenges and perspectives toward future wide-bandgap mixed-halide perovskite photovoltaics , author=. Advanced Energy Materials , volume=. 2023 , publisher=

    2023

  2. [2]

    Nature Reviews Materials , volume=

    Methylammonium-free wide-bandgap metal halide perovskites for tandem photovoltaics , author=. Nature Reviews Materials , volume=. 2023 , publisher=

    2023

  3. [3]

    Materials Chemistry Frontiers , volume=

    Issues of phase segregation in wide-bandgap perovskites , author=. Materials Chemistry Frontiers , volume=. 2023 , publisher=

    2023

  4. [4]

    ACS Energy Letters , volume=

    Wide-bandgap metal halide perovskites for tandem solar cells , author=. ACS Energy Letters , volume=. 2020 , publisher=

    2020

  5. [5]

    Advanced Materials , volume=

    Perovskites for light emission , author=. Advanced Materials , volume=. 2018 , publisher=

    2018

  6. [6]

    Nature Materials , volume=

    Metal halide perovskites for light-emitting diodes , author=. Nature Materials , volume=. 2021 , publisher=

    2021

  7. [7]

    Nature Electronics , volume=

    Perovskite light-emitting diodes , author=. Nature Electronics , volume=. 2022 , publisher=

    2022

  8. [8]

    Advanced Materials , volume=

    Multiple-quantum-well perovskites for high-performance light-emitting diodes , author=. Advanced Materials , volume=. 2020 , publisher=

    2020

  9. [9]

    Journal of Materials Chemistry C , volume=

    Opportunities and challenges in perovskite LED commercialization , author=. Journal of Materials Chemistry C , volume=. 2021 , publisher=

    2021

  10. [10]

    Advanced Science , volume=

    Recent progress on electrical and optical manipulations of perovskite photodetectors , author=. Advanced Science , volume=. 2021 , publisher=

    2021

  11. [11]

    Nano Energy , volume=

    Wide-bandgap all-inorganic lead-free perovskites for ultraviolet photodetectors , author=. Nano Energy , volume=. 2023 , publisher=

    2023

  12. [12]

    Journal of Materials Chemistry C , volume=

    Recent progress on highly sensitive perovskite photodetectors , author=. Journal of Materials Chemistry C , volume=. 2019 , publisher=

    2019

  13. [13]

    Applied Materials Today , volume=

    Perovskite photodetectors for flexible electronics: Recent advances and perspectives , author=. Applied Materials Today , volume=. 2022 , publisher=

    2022

  14. [14]

    The Journal of Physical Chemistry C , volume=

    Principles of chemical bonding and band gap engineering in hybrid organic--inorganic halide perovskites , author=. The Journal of Physical Chemistry C , volume=. 2015 , publisher=

    2015

  15. [15]

    Proceedings of the National Academy of Sciences , volume=

    Simultaneous band-gap narrowing and carrier-lifetime prolongation of organic--inorganic trihalide perovskites , author=. Proceedings of the National Academy of Sciences , volume=. 2016 , publisher=

    2016

  16. [16]

    Chemistry of Materials , volume=

    Exploiting ionic radii for rational design of halide perovskites , author=. Chemistry of Materials , volume=. 2019 , publisher=

    2019

  17. [17]

    Chem , volume=

    Long carrier diffusion length in two-dimensional lead halide perovskite single crystals , author=. Chem , volume=. 2022 , publisher=

    2022

  18. [18]

    Prx Energy , volume=

    Relevance of long diffusion lengths for efficient halide perovskite solar cells , author=. Prx Energy , volume=. 2023 , publisher=

    2023

  19. [19]

    Nature , volume=

    High-efficiency and thermally stable FACsPbI3 perovskite photovoltaics , author=. Nature , volume=. 2024 , publisher=

    2024

  20. [20]

    Nature communications , volume=

    Heterojunction formed via 3D-to-2D perovskite conversion for photostable wide-bandgap perovskite solar cells , author=. Nature communications , volume=. 2023 , publisher=

    2023

  21. [21]

    RSC advances , volume=

    Metal halide perovskites for energy applications: recent advances, challenges, and future perspectives , author=. RSC advances , volume=. 2025 , publisher=

    2025

  22. [22]

    Materials Advances , year=

    Lead-free alternatives and toxicity mitigation strategies for sustainable perovskite solar cells: a critical review , author=. Materials Advances , year=

  23. [23]

    Journal of Applied Physics , volume=

    Defect tolerance in halide perovskites: A first-principles perspective , author=. Journal of Applied Physics , volume=. 2022 , publisher=

    2022

  24. [24]

    Energy & Environmental Science , volume=

    Entropic stabilization of mixed A-cation ABX 3 metal halide perovskites for high performance perovskite solar cells , author=. Energy & Environmental Science , volume=. 2016 , publisher=

    2016

  25. [25]

    Energy & environmental science , volume=

    Cesium-containing triple cation perovskite solar cells: improved stability, reproducibility and high efficiency , author=. Energy & environmental science , volume=. 2016 , publisher=

    2016

  26. [26]

    Chemistry of Materials , volume=

    Geometric analysis and formability of the cubic A2BX6 vacancy-ordered double perovskite structure , author=. Chemistry of Materials , volume=. 2020 , publisher=

    2020

  27. [27]

    Chemistry of Materials , volume=

    Computational study of halide perovskite-derived A2BX6 inorganic compounds: chemical trends in electronic structure and structural stability , author=. Chemistry of Materials , volume=. 2017 , publisher=

    2017

  28. [28]

    Advanced Functional Materials , volume=

    Lead-free double perovskite Cs2AgBiBr6: fundamentals, applications, and perspectives , author=. Advanced Functional Materials , volume=. 2021 , publisher=

    2021

  29. [29]

    The Journal of Physical Chemistry C , volume=

    Density functional theory estimate of halide perovskite band gap: When spin orbit coupling helps , author=. The Journal of Physical Chemistry C , volume=. 2022 , publisher=

    2022

  30. [30]

    Chemical Society Reviews , volume=

    Replacing hybrid density functional theory: motivation and recent advances , author=. Chemical Society Reviews , volume=. 2021 , publisher=

    2021

  31. [31]

    ACS Applied Materials & Interfaces , volume=

    Machine Learning-Assisted Prediction and Control of Bandgap for Organic--Inorganic Metal Halide Perovskites , author=. ACS Applied Materials & Interfaces , volume=. 2025 , publisher=

    2025

  32. [32]

    RSC advances , volume=

    Bandgap tuning strategy by cations and halide ions of lead halide perovskites learned from machine learning , author=. RSC advances , volume=. 2021 , publisher=

    2021

  33. [33]

    Computational Materials Science , volume=

    Python Materials Genomics (pymatgen): A robust, open-source python library for materials analysis , author=. Computational Materials Science , volume=. 2013 , publisher=

    2013

  34. [34]

    Physical Review B , volume=

    Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set , author=. Physical Review B , volume=. 1996 , publisher=

    1996

  35. [35]

    Physical Review B , volume=

    Projector augmented-wave method , author=. Physical Review B , volume=. 1994 , publisher=

    1994

  36. [36]

    The Journal of Chemical Physics , volume=

    Hybrid functionals based on a screened Coulomb potential , author=. The Journal of Chemical Physics , volume=. 2003 , publisher=

    2003

  37. [37]

    Renewable energy , pages=

    Detailed balance limit of efficiency of p--n junction solar cells , author=. Renewable energy , pages=. 2018 , publisher=

    2018

  38. [38]

    Advanced Theory and Simulations , volume=

    Computational Screening of 2D All-Inorganic Lead-Free Halide Perovskites A3B2X9 for Photovoltaic and Photocatalytic Applications , author=. Advanced Theory and Simulations , volume=. 2024 , publisher=

    2024

  39. [39]

    Applied Physics Reviews , volume=

    Lead-free metal halide (halogenidometallate) semiconductors for optoelectronic applications , author=. Applied Physics Reviews , volume=. 2023 , publisher=

    2023

  40. [40]

    Chemistry of Materials , volume=

    Crystal and electronic structures of complex bismuth iodides A 3Bi2I9 (A= K, Rb, Cs) related to perovskite: aiding the rational design of photovoltaics , author=. Chemistry of Materials , volume=. 2015 , publisher=

    2015

  41. [41]

    Nature communications , volume=

    Absolute energy level positions in tin-and lead-based halide perovskites , author=. Nature communications , volume=. 2019 , publisher=

    2019

  42. [42]

    The Journal of Physical Chemistry C , volume=

    Machine learning for predicting the band gaps of ABX3 perovskites from elemental properties , author=. The Journal of Physical Chemistry C , volume=. 2020 , publisher=

    2020

  43. [43]

    Journal of Materials Chemistry C , volume=

    A progressive learning method for predicting the band gap of ABO 3 perovskites using an instrumental variable , author=. Journal of Materials Chemistry C , volume=. 2020 , publisher=

    2020

  44. [44]

    Communications Materials , volume=

    Predicting the formation of fractionally doped perovskite oxides by a function-confined machine learning method , author=. Communications Materials , volume=. 2022 , publisher=

    2022

  45. [45]

    Journal of Materials Chemistry C , volume=

    Defect formation in CsSnI 3 from density functional theory and machine learning , author=. Journal of Materials Chemistry C , volume=. 2025 , publisher=

    2025

  46. [46]

    The Journal of Physical Chemistry Letters , volume=

    Cost-effective high-throughput calculation based on hybrid density functional theory: Application to cubic, double, and vacancy-ordered halide perovskites , author=. The Journal of Physical Chemistry Letters , volume=. 2021 , publisher=

    2021

  47. [47]

    The Journal of Physical Chemistry C , volume=

    Comprehensive and Accurate Prediction of Band Gap for Lead-Free Double Perovskites through Self-Modified Machine Learning Strategy , author=. The Journal of Physical Chemistry C , volume=. 2023 , publisher=

    2023

  48. [48]

    Communications Materials , volume=

    Band gap predictions of double perovskite oxides using machine learning , author=. Communications Materials , volume=. 2023 , publisher=

    2023

  49. [49]

    Chemistry of materials , volume=

    A machine learning approach for the prediction of formability and thermodynamic stability of single and double perovskite oxides , author=. Chemistry of materials , volume=. 2021 , publisher=

    2021

  50. [50]

    Foundations of Crystallography , volume=

    Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides , author=. Foundations of Crystallography , volume=. 1976 , publisher=

    1976

  51. [51]

    Naturwissenschaften , volume=

    Die gesetze der krystallochemie , author=. Naturwissenschaften , volume=. 1926 , publisher=

    1926

  52. [52]

    Acta Crystallographica Section B: Structural Science , volume=

    Formability of abx _3 (x= f, cl, br, i) halide perovskites , author=. Acta Crystallographica Section B: Structural Science , volume=. 2008 , publisher=

    2008

  53. [53]

    Noise Reduction in Speech Processing , pages=

    Pearson correlation coefficient , author=. Noise Reduction in Speech Processing , pages=. 2009 , publisher=

    2009

  54. [54]

    Journal of Machine Learning Research , volume=

    Scikit-learn: Machine learning in python , author=. Journal of Machine Learning Research , volume=. 2011 , url=

    2011

  55. [55]

    Journal of Materials Science , volume=

    Accelerated screening of functional atomic impurities in halide perovskites using high-throughput computations and machine learning , author=. Journal of Materials Science , volume=. 2022 , publisher=

    2022

  56. [56]

    Journal of Applied Physics , volume=

    Machine learning substitutional defect formation energies in ABO _3 perovskites , author=. Journal of Applied Physics , volume=. 2020 , publisher=

    2020

  57. [57]

    Chemistry of Materials , volume=

    A-site cation chemistry in halide perovskites , author=. Chemistry of Materials , volume=. 2024 , publisher=

    2024

  58. [58]

    Chinese Physics B , volume=

    Nature of the band gap of halide perovskites ABX3 (A= CH3NH3, Cs; B= Sn, Pb; X= Cl, Br, I): First-principles calculations , author=. Chinese Physics B , volume=. 2015 , publisher=

    2015

  59. [59]

    Biometrika , volume=

    A new measure of rank correlation , author=. Biometrika , volume=. 1938 , publisher=

    1938

  60. [60]

    The American journal of psychology , volume=

    The proof and measurement of association between two things , author=. The American journal of psychology , volume=. 1987 , publisher=

    1987

  61. [61]

    Machine Learning , volume=

    Random forests , author=. Machine Learning , volume=. 2001 , publisher=

    2001

  62. [62]

    Empirical inference: Festschrift in honor of vladimir n

    Kernel ridge regression , author=. Empirical inference: Festschrift in honor of vladimir n. vapnik , pages=. 2013 , publisher=

    2013

  63. [63]

    Automatic model construction with Gaussian processes , author=

  64. [64]

    R package version 0.4-2 , volume=

    Xgboost: extreme gradient boosting , author=. R package version 0.4-2 , volume=. 2015 , url=

    2015

  65. [65]

    Frontiers in neurorobotics , volume=

    Gradient boosting machines, a tutorial , author=. Frontiers in neurorobotics , volume=. 2013 , publisher=

    2013

  66. [66]

    Chemistry of Materials , volume=

    Perspectives and design principles of vacancy-ordered double perovskite halide semiconductors , author=. Chemistry of Materials , volume=. 2019 , publisher=

    2019

  67. [67]

    Nano Research , volume=

    Unraveling the triplet excited-state dynamics of Bi3+ in vacancy-ordered double perovskite Cs2SnCl6 nanocrystals , author=. Nano Research , volume=. 2022 , publisher=

    2022

  68. [68]

    Acs Photonics , volume=

    Zero-dimensional Cs2TeI6 perovskite: solution-processed thick films with high X-ray sensitivity , author=. Acs Photonics , volume=. 2018 , publisher=

    2018

  69. [69]

    EcoMat , volume=

    Pt and Pt-group transition metal 0D vacancy ordered halide perovskites: A review , author=. EcoMat , volume=. 2024 , publisher=

    2024

  70. [70]

    Angewandte Chemie International Edition , volume=

    Vacancy Effect on the Luminescent and Water Responsive Properties of Vacancy-Ordered Double Perovskite Derivatives , author=. Angewandte Chemie International Edition , volume=. 2024 , publisher=

    2024

  71. [71]

    Advanced Optical Materials , volume=

    Vacancy-Ordered Double Perovskite Rb2ZrCl6- xBrx: Facile Synthesis and Insight into Efficient Intrinsic Self-Trapped Emission , author=. Advanced Optical Materials , volume=. 2022 , publisher=

    2022

  72. [72]

    npj Computational Materials , volume=

    Efficiently charting the space of mixed vacancy-ordered perovskites by machine-learning encoded atomic-site information , author=. npj Computational Materials , volume=. 2025 , publisher=

    2025

  73. [73]

    Octahedral tilting in perovskites. II. Structure stabilizing forces , author=. Structural Science , volume=. 1997 , publisher=

    1997

  74. [74]

    Nano letters , volume=

    Cation-induced band-gap tuning in organohalide perovskites: interplay of spin--orbit coupling and octahedra tilting , author=. Nano letters , volume=. 2014 , publisher=

    2014

  75. [75]

    Nature communications , volume=

    Steric engineering of metal-halide perovskites with tunable optical band gaps , author=. Nature communications , volume=. 2014 , publisher=

    2014