Pith. sign in

REVIEW 3 major objections 5 minor 112 references

DeFecT-FF: a machine learning force field framework for high throughput defect modeling in CdTe-based solar cells

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read A machine-learned force field reproduces hybrid-DFT defect energies across Cd/Zn-Te/Se/S alloys, cutting single-defect optimization from about 8 hours to 1–2 minutes.

desk verdict Strong, useful methods paper—the framework, dataset, and tool are real contributions, but the <0.2 eV defect-formation-energy claim only holds when DFT-derived corrections are added, and the test split is underspecified. read the letter →

arxiv 2510.23514 v3 pith:66RWZMLH submitted 2025-10-27 cond-mat.mtrl-sci cond-mat.dis-nn

Md Habibur Rahman , Maitreyo Biswas , Arun Mannodi-Kanakkithodi This is my paper
classification cond-mat.mtrl-scicond-mat.dis-nn
keywords defectformationenergymachinelearningforcefieldhybridfunctionalHSE06CdTesolarcellsSealloyschargestatesactivegraphneuralnetwork
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 claims that a crystal-graph-based machine learning force field, trained on hybrid-functional (HSE06) data from thousands of charged and neutral defect structures, can predict energies and forces accurately enough that defect formation energies computed on MLFF-optimized geometries stay within 0.20 eV of full hybrid DFT. If correct, this removes the main bottleneck in defect chemistry: a single defect relaxation that normally costs many hours of supercomputer time can be done in one or two minutes, and exhaustive screening across alloy compositions, charge states, and symmetry-broken configurations becomes routine. The authors support the claim with parity plots showing small RMSE values, case studies on As+Cl complexes in CdSeTe, and a public online tool that lets anyone upload a crystal structure and obtain defect formation energy diagrams without running DFT.

What carries the argument

The engine is a machine learning force field built on crystal graphs, trained separately for five charge states (from +2 to −2) on HSE06 energies, forces, and stresses. Because it outputs forces, it can perform gradient-based geometry optimization, replacing the costly hybrid-functional relaxation step. The workflow pairs it with symmetry-breaking initial structure generation (to escape metastable local minima), a final single-point HSE06+SOC calculation, and a standard electrostatic charge correction for periodic charged defects.

What would settle it

Take a defect complex and composition not included in the training set (for instance, a defect in a quaternary Cd-Zn-Se-Te supercell or a double interstitial in an alloy), relax a batch of symmetry-broken configurations with the MLFF, run full HSE06 relaxation on the same inputs, and compare the lowest-energy geometries and formation energies. A systematic RMSE above 0.20 eV for the MLFF-selected ground states would refute the core claim of near-DFT accuracy for new defects.

Watch

Extended reading notes

Core claim

The central discovery is that a graph-based machine learning force field, trained per charge state on energies, forces, and stresses from HSE06 calculations, can serve as a surrogate for hybrid DFT during geometry optimization. Defect structures, including symmetry-broken perturbations and complex multi-defect arrangements, relax to geometries whose total energies, when combined with DFT reference values, electrostatic charge corrections, and chemical potentials, yield defect formation energies with root-mean-square error below 0.20 eV. The framework also identifies new low-energy configurations (for example, a nitrogen interstitial double defect in ZnTe) and supports finite-temperature mole

Load-bearing premise

The load-bearing premise is that the curated HSE06 training subset represents the full diversity of defect configurations, charge states, and alloy compositions; if that subset is unrepresentative or shares relaxation trajectories with the test points, the reported <0.2 eV errors will not generalize to genuinely new structures.

Editorial extensions

If this is right

  • Defect formation energy diagrams for many defects, compositions, and charge states can be produced in hours rather than weeks, making exhaustive screening practical.
  • Charge transition levels predicted with the MLFF (using averaged charge corrections) lie within about 0.2–0.25 eV of hybrid DFT for the four transition types tested.
  • The same force field enables stable finite-temperature molecular dynamics (demonstrated with 100 ps trajectories), a step toward including vibrational entropy in defect thermodynamics.
  • The public tool allows non-specialists to generate defects from an uploaded unit cell and compute formation energies as functions of Fermi level and chemical potentials, bypassing direct DFT.
  • New low-energy defect configurations, e.g., an N interstitial pair in ZnTe, can be found by MLFF-driven scanning before final hybrid-functional validation.

Reading between the lines

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

  • If the <0.2 eV accuracy transfers to compositions outside the training set, the same active-learning-plus-force-field recipe could be applied to other alloy families where defect chemistry limits device efficiency, without re-deriving a bespoke force field each time.
  • A testable extension the paper leaves implicit is comparing MLFF-derived charge transition levels against experimental deep-level transient spectroscopy or photoluminescence data for CdSeTe; the prediction that Se alloying deepens the As acceptor level is consistent with one recent experiment, but independent measurements on other defects would sharpen the test.
  • The method's dependence on a curated HSE06 subset means its reliability for exotic non-equilibrium structures beyond the single dislocation core studied remains open; users should validate device-critical cases with targeted HSE06, as the authors themselves advise.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper presents DeFecT-FF, a machine-learned force field framework for defect modeling in Cd/Zn-Te/Se/S compounds. The authors compile a large HSE06 dataset of bulk and defect structures across five charge states, train M3GNet-based MLFF models using active learning and error-aware reweighting, and report CFE RMSE values of 4.8–12 meV/atom on held-out test sets. Defect formation energies computed with MLFF-optimized geometries are reported to have RMSE below 0.20 eV when DFT reference energies and Freysoldt charge corrections are added. The framework is demonstrated on As+Cl complexes in CdSe0.12Te0.88, As/Cl defects across CdSexTe1-x, N-related defects in ZnTe, and finite-temperature MD for AsTe in CdTe. A nanoHUB tool is released that generates defect structures, performs MLFF relaxation, and produces defect formation energy diagrams.

Significance. If fully substantiated, the framework would be a valuable high-throughput screening tool: it provides a large public HSE06-level defect dataset, a charge-state-aware MLFF with gradient-based optimization, a public nanoHUB implementation, and computational speedups of several orders of magnitude for defect relaxation. The paper also demonstrates the MLFF’s ability to find low-energy symmetry-broken configurations and to run stable MD, which are useful beyond the specific CdTe application. However, the central claim that users can compute defect formation energies without DFT is not supported by the reported numbers, and the test-set integrity needs clarification. The magnitude of the contribution after those issues are resolved would still be solid, but the current manuscript overstates the DFT-free capability.

major comments (3)
  1. [Section 3 and Fig. 2(d–f), Table SVII] The headline statement “RMSE in defect formation energies obtained using MLFF-optimized geometries remains below 0.20 eV” is only achieved when DFT-derived reference energies, E_VBM, and Freysoldt charge corrections are added to MLFF total energies. Table SVII shows that without these DFT inputs the RMSE is 0.27–0.30 eV (no correction) or 0.20–0.23 eV (average offsets fitted to the training data). Thus the claimed accuracy is not an MLFF-only result. The abstract and conclusion say users can compute defect formation energies “bypassing expensive DFT calculations,” which is contradicted by Sec. 4 (“final HSE06+SOC single-point calculations must be performed”) and Data Availability (“device-critical cases should be validated with targeted HSE06(+SOC)”). Please specify exactly which quantities still require DFT and revise the bypass claim accordingly.
  2. [SI dataset split; Sec. 2] The SI says the ALIGNN models used a random 60/20/20 train/validation/test split, and Sec. 2 states that all ionic-relaxation snapshots were preserved in the HSE06 dataset. If snapshots from the same relaxation trajectory are split across train and test, the reported CFE RMSE (4.8–12 meV/atom) and the downstream DFE RMSE may be substantially optimistic. No statement is given that snapshots were grouped by parent defect or relaxation trajectory. Please split by trajectory/defect identity and re-report the test errors. The ‘carefully selected subset’ of PBE structures used for HSE06 refinement is also not described by quantitative selection criteria, which matters for generalization claims.
  3. [Section 4.1–4.3, Table SVII] The fully DFT-free workflow is not validated for the new chemistries presented. The average charge-offset corrections (0.10/0.20 eV per charge state) are fitted to the training data, and the case studies in Sec. 4 (AsSe+ClSe in CdSe0.12Te0.88, As/Cl across CdSexTe1-x, and ZnTe N-related defects) are validated by performing HSE06+SOC single-point calculations on MLFF geometries, not by computing defect formation energies from standalone MLFF energies. The transferability of the average offsets to unseen compositions, defect types, or supercell sizes is therefore unsupported, and the claim that the <0.20 eV accuracy holds for a DFT-free user workflow is not established. Please either validate the no-DFT workflow on held-out defect types/compositions or explicitly restrict the accuracy claim to MLFF-optimized geometries followed by DFT energetics.
minor comments (5)
  1. [Section 4.2] The band gap for CdSe0.12Te0.88 is printed as “138 eV”; this should be 1.38 eV.
  2. [Section 3 (text near average offsets)] The list of average offsets is incomplete/erroneous: it reads “0.20 eV for q=+2, 0.10 eV for q=+1, 0.10 eV for q=-1, and 0.20 eV for q=+2 defects,” with q=-2 never listed. Table SVII shows 0.23 eV for q=-2 and 0.20 eV for q=+2; please make the text consistent.
  3. [SI, Eq. (3)] The CFE definition is written for a Cd–Se–Te composition but the dataset spans Cd/Zn and S/Se/Te. Please generalize the expression and define the elemental reference states for all elements.
  4. [Section 4.2; Fig. 5 caption] The use of a modified HSE mixing parameter α=0.31 for CdSexTe1-x band gaps is not justified; the rest of the work uses α=0.25. Please provide a validation or reference for this parameter choice.
  5. [Data Availability vs. reference 74] The nanoHUB URL in Data Availability (nanohub.org/tools/cadetff) differs from the resource URL in the reference list (nanohub.org/resources/cadetff). Please unify the URL.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; MLFF validated on held-out DFT data, DFT corrections cancel in the error metric, and the 'bypass DFT' claim is an overstatement rather than a circular reduction.

full rationale

The derivation chain is a standard supervised ML pipeline: M3GNet MLFFs are trained on HSE06 total energies, forces, and stresses and evaluated on held-out DFT data (Fig. 2a–c; Figs. S14–S15). The defect formation energy validation (Fig. 2d–f; Table SVII) uses the thermodynamic expression ΔE_f(D^q) = E_tot(D^q) − E_tot(bulk) + Σn_iμ_i + q(E_F + E_VBM) + E_corr, with μ_i, E_VBM, and E_corr explicitly 'evaluated from DFT and added directly to the MLFF-derived bulk and defect energies.' In the RMSE comparison, these DFT reference terms cancel between the DFT and MLFF channels, so the reported <0.20 eV RMSE measures the MLFF's reproduction of DFT total-energy differences—not a self-defined quantity. The alternative 'average offset' correction is a single per-charge-state constant (0.10–0.20 eV) approximating the physical Freysoldt correction; the paper transparently reports all three correction schemes and shows the average-offset RMSE is slightly worse (0.20–0.23 eV), so it is not a concealed fit to the target. Self-citations (refs 40, 55, 62, 75 for prior PBE data; ref 74 for the nanoHUB tool) are data/software provenance, not load-bearing arguments; no uniqueness theorem or ansatz is imported from the authors' prior work. The abstract's claim that users can 'bypass expensive DFT calculations' is contradicted by Sec. 4 ('final HSE06+SOC single-point calculations must be performed') and the Data Availability note that 'device-critical cases should be validated with targeted HSE06(+SOC) calculations'—this is an overstatement/limitation, but not a circular reduction. Potential concerns about HSE06 subset selection and trajectory leakage in the train/test split are data-quality risks, not demonstrated circularity. No step in the paper's derivation reduces to its own inputs; score 1 reflects minor non-load-bearing self-citation but no circularity.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The framework introduces no new physical entities or forces; it is a surrogate model plus corrections. The main ledger entries are the fitted charge-correction offsets and the HSE06 mixing parameter, plus the standard DFT/correction/sampling assumptions that any defect-DFT study relies on. The MLFF model weights themselves are not enumerated here because they are conventional trained parameters, but the generalization benchmark is the key unproven quantity.

free parameters (2)
  • Average charge-offset corrections = 0.20 eV (q=+2), 0.10 eV (q=+1), 0.10 eV (q=-1), 0.20 eV (q=-2)
    Fitted to reduce the RMSE between MLFF and DFT defect formation energies; used in the reported DFE and charge-transition-level accuracy. Listing in Table SVII and Section 3.
  • HSE06 mixing parameter alpha for CdSexTe1-x band gaps = 0.31 (modified from default 0.25)
    Chosen to obtain band gaps of 1.5-1.30 eV for CdTe to CdSe0.25Te0.75; these band gaps set the Fermi-level bounds in the defect formation energy diagrams. Section 4.2.
assumptions (5)
  • domain assumption HSE06 hybrid DFT is an accurate reference ground truth for defect formation energies and transition levels in these semiconductors.
    All MLFF labels and final validation energies are HSE06; the paper's accuracy claims are relative to this functional.
  • domain assumption Gamma-point-only sampling and a 400 eV plane-wave cutoff are sufficient for HSE06 calculations on 216-atom defect supercells.
    Stated in Section 3; no convergence tests for k-points or cutoff are shown in the main text.
  • domain assumption The Freysoldt charge-correction scheme correctly removes finite-size electrostatic errors for charged defects.
    Used throughout for charged defect formation energies; the SI describes the scheme but provides limited convergence testing.
  • domain assumption SQS supercells and ShakeNBreak perturbations adequately represent the configurational space of alloy disorder and defect metastability.
    The claim that the lowest-energy configuration is found depends on the completeness of symmetry-broken sampling in Section 4 and the SQS construction of alloys.
  • domain assumption The curated HSE06 training subset is representative enough that the MLFF generalizes to new compositions, charge states, and defect complexes.
    This is the load-bearing generalization assumption; the paper does not specify the selection criteria for the HSE06 subset or the test-split leakage control.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DeFecT-FF: a machine learning force field framework for high throughput defect modeling in CdTe-based solar cells." pith.science (2026). https://pith.science/paper/66RWZMLH

@misc{pith2026251023514,
  author       = {Pith},
  title        = {Pith review of: DeFecT-FF: a machine learning force field framework for high throughput defect modeling in CdTe-based solar cells},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/66RWZMLH}},
  note         = {Machine review of arXiv:2510.23514}
}
read the original abstract

We developed a framework for predicting the energies and ground state configurations of native point defects, extrinsic dopants and impurities, and defect complexes across zinc blende-phase Cd/Zn-Te/Se/S compounds, important for CdTe-based solar cells. This framework, named DeFecT-FF, is powered by high-throughput density functional theory (DFT) computations and crystal graph-based machine learning force field (MLFF) models trained on the DFT data. The Cd/Zn-Te/Se/S chemical space is chosen because alloying at Cd or Te sites is a promising avenue to tailor the electronic and defect properties of the CdTe absorber layer to potentially improve solar cell performance. The sheer number of defect configurations achievable when considering all possible singular defects and their combinations, symmetry-breaking operations, and defect charge states, as well as the expense of running large supercell calculations, makes this an ideal problem for developing accurate and widely-applicable force field models. Here, we introduce our dataset of structures and energies from HSE06 geometry optimization, including bulk and alloyed supercells with and without defects. Data were gradually expanded using active learning and accurate MLFF models were trained to predict energies and atomic forces across different charge states. Via accelerated prediction and screening, we identified many new low energy defect configurations and obtained high-fidelity defect formation energy diagrams using HSE06 calculations with spin-orbit coupling. The DeFecT-FF framework has been released publicly as an online tool on the nanoHUB platform, allowing users to upload any crystallographic information file, generate defects of interest, and compute defect formation energies as a function of Fermi level and chemical potential conditions, thus bypassing expensive DFT calculations.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

112 extracted references · 3 canonical work pages

  1. [1]

    Rojsatien, A

    S. Rojsatien, A. Mannodi-Kanakkithodi, T. Walker, T. Nietzold, E. Colegrove, B. Lai, Z. Cai, M. Holt, M. K. Chan and M. I. Bertoni, Radiation Physics and Chemistry, 2023, 202, 110548

    2023

  2. [2]

    Rojsatien, A

    S. Rojsatien, A. Mannodi-Kanakkithodi, T. Walker, N. Mohan Kumar, T. Nietzold, E. Colegrove, D. Mao, M. E. Stuckelberger, B. Lai, Z. Cai, M. K. Y. Chan and M. I. Bertoni, Chemistry of Materials, 2023, 35, 9935--9944

    2023

  3. [3]

    Gloeckler, I

    M. Gloeckler, I. Sankin and Z. Zhao, IEEE Journal of Photovoltaics, 2013, 3, 1389--1393

    2013

  4. [4]

    Nideep, M

    T. Nideep, M. Ramya and M. Kailasnath, Superlattices and Microstructures, 2020, 141, 106477

    2020

  5. [5]

    X. Yang, Y. Long, Y. Zheng, J. Wang, B. Zhou, S. Xie, B. Li, J. Zhang, X. Hao, S. Karazhanov, G. Zeng and L. Feng, Materials Science in Semiconductor Processing, 2023, 156, 107267

    2023

  6. [6]

    Freysoldt, B

    C. Freysoldt, B. Grabowski, T. Hickel, J. Neugebauer, G. Kresse, A. Janotti and C. G. Van de Walle, Rev. Mod. Phys., 2014, 86, 253--305

    2014

  7. [7]

    A. M. Ganose, D. O. Scanlon, A. Walsh and R. L. Z. Hoye, Nature Communications, 2022, 13, 4715

    2022

  8. [8]

    Mannodi-Kanakkithodi, The devil is in the defects - Nature Physics, 2023, https://www.nature.com/articles/s41567-023-02049-9

    A. Mannodi-Kanakkithodi, The devil is in the defects - Nature Physics, 2023, https://www.nature.com/articles/s41567-023-02049-9

    2023

Show all 112 references
  1. [9]

    S. R. Kavanagh, A. Walsh and D. O. Scanlon, ACS Energy Lett., 2021, 6, 1392--1398

    2021

  2. [10]

    M. E. Turiansky, A. Alkauskas, M. Engel, G. Kresse, D. Wickramaratne, J.-X. Shen, C. E. Dreyer and C. G. Van de Walle, Computer Physics Communications, 2021, 267, 108056

    2021

  3. [11]

    Yang, W.-J

    J.-H. Yang, W.-J. Yin, J.-S. Park, W. Metzger and S.-H. Wei, Journal of Applied Physics, 2016, 119, 045104

    2016

  4. [12]

    D. N. Krasikov, A. V. Scherbinin, A. A. Knizhnik, A. N. Vasiliev, B. V. Potapkin and T. J. Sommerer, Journal of Applied Physics, 2016, 119, 085706

    2016

  5. [13]

    Krasikov, A

    D. Krasikov, A. Knizhnik, B. Potapkin, S. Selezneva and T. Sommerer, Thin Solid Films, 2013, 535, 322--325

    2013

  6. [14]

    Krasikov, D

    D. Krasikov, D. Guo, S. Demtsu and I. Sankin, Solar Energy Materials and Solar Cells, 2021, 224, 111012

    2021

  7. [15]

    Krasikov and I

    D. Krasikov and I. Sankin, Physical Review Materials, 2018, 2, 103803

    2018

  8. [16]

    Krasikov, Nat

    D. Krasikov, Nat. Energy, 2019, 4, 442--443

    2019

  9. [17]

    Ablekim, S

    T. Ablekim, S. K. Swain, W.-J. Yin, K. Zaunbrecher, J. Burst, T. M. Barnes, D. Kuciauskas, S.-H. Wei and K. G. Lynn, Scientific Reports, 2017, 7, 4563

    2017

  10. [18]

    T. A. M. Fiducia, B. G. Mendis, K. Li, C. R. M. Grovenor, A. H. Munshi, K. Barth, W. S. Sampath, L. D. Wright, A. Abbas, J. W. Bowers and J. M. Walls, Nat. Energy, 2019, 4, 504--511

    2019

  11. [19]

    Gorai, D

    P. Gorai, D. Krasikov, S. Grover, G. Xiong, W. K. Metzger and V. Stevanović, Science Advances, 2023, 9, eade3761

    2023

  12. [20]

    De Souza and G

    R. De Souza and G. Harrington, Nature Materials, 2023, 22, 794–797

    2023

  13. [21]

    D. V. Lang, Journal of Applied Physics, 2003, 45, 3023--3032

    2003

  14. [22]

    Wickramaratne, C

    D. Wickramaratne, C. E. Dreyer, B. Monserrat, J.-X. Shen, J. L. Lyons, A. Alkauskas and C. G. Van de Walle, Applied Physics Letters, 2018, 113, 192106

    2018

  15. [23]

    J. Y. Kim, L. Gelczuk, M. P. Polak, D. Hlushchenko, D. Morgan, R. Kudrawiec and I. Szlufarska, Npj 2D Materials and Applications, 2022, 6, 75

    2022

  16. [24]

    Broberg, K

    D. Broberg, K. Bystrom and et al, npj Computational Materials, 2023, 9, 72

    2023

  17. [25]

    M. Y. Toriyama, J. Qu, G. J. Snyder and P. Gorai, J. Mater. Chem. A, 2021, 9, 20685--20694

    2021

  18. [26]

    C.-W. Lee, N. U. Din, K. Yazawa, W. Nemeth, R. W. Smaha, N. M. Haegel and P. Gorai, Journal of Applied Physics, 2024, 135, 155101

    2024

  19. [27]

    Grill and A

    R. Grill and A. Zappettini, Prog. Cryst. Growth Charact. Mater., 2004, 48-49, 209--244

    2004

  20. [28]

    Buckeridge, Computer Physics Communications, 2019, 244, 329–342

    J. Buckeridge, Computer Physics Communications, 2019, 244, 329–342

    2019

  21. [29]

    Mannodi-Kanakkithodi, X

    A. Mannodi-Kanakkithodi, X. Xiang, L. Jacoby, R. Biegaj, S. T. Dunham, D. R. Gamelin and M. K. Y. Chan, Patterns (N. Y.), 2022, 3, 100450

    2022

  22. [30]

    M. H. Rahman and A. Mannodi-Kanakkithodi, Defect Generation in Graph Neural Networks, 2023, https://github.com/msehabibur/defect_GNN_gen_1, GitHub repository

    2023

  23. [31]

    Mannodi-Kanakkithodi, J.-S

    A. Mannodi-Kanakkithodi, J.-S. Park, A. B. F. Martinson and M. K. Y. Chan, The Journal of Physical Chemistry C, 2020, 124, 16729--16738

    2020

  24. [32]

    Kim, J.-S

    S. Kim, J.-S. Park, S. Hood and A. Walsh, Journal of Materials Chemistry A, 2019, 7, 2686–2693

    2019

  25. [33]

    J. L. Lyons and C. G. Van de Walle, Npj Comput. Mater., 2017, 3, 1--10

    2017

  26. [34]

    Machín and F

    A. Machín and F. Márquez, Materials, 2024, 17, 1165

    2024

  27. [35]

    Mosquera-Lois, J

    I. Mosquera-Lois, J. Klarbring and A. Walsh, Chemical Science, 2025

    2025

  28. [36]

    Mosquera-Lois, S

    I. Mosquera-Lois, S. R. Kavanagh, J. Klarbring, K. Tolborg and A. Walsh, Chemical Society Reviews, 2023, 52, 5812–5826

    2023

  29. [37]

    M. P. Polak, R. Jacobs, A. Mannodi-Kanakkithodi, M. K. Y. Chan and D. Morgan, The Journal of Chemical Physics, 2022, 156, 114110

    2022

  30. [38]

    Mannodi-Kanakkithodi and M

    A. Mannodi-Kanakkithodi and M. K. Y. Chan, Journal of Materials Science, 2022, 57, 10736--10754

    2022

  31. [39]

    Xie and J

    T. Xie and J. C. Grossman, Phys. Rev. Lett., 2018, 120, 145301

    2018

  32. [40]

    Choudhary and B

    K. Choudhary and B. G. Sumpter, AIP Advances, 2023, 13, 095109

    2023

  33. [41]

    Z. Chen, X. Li and J. Bruna, Supervised Community Detection with Line Graph Neural Networks, 2017, https://arxiv.org/abs/1705.08415v6

    2017 arXiv

  34. [42]

    T. N. Kipf and M. Welling, Semi-Supervised Classification with Graph Convolutional Networks, 2016, https://arxiv.org/abs/1609.02907v4

    2016 arXiv

  35. [43]

    M. D. Witman, A. Goyal, T. Ogitsu, A. H. McDaniel and S. Lany, Nature Computational Science, 2023

    2023

  36. [44]

    Chen and S

    C. Chen and S. P. Ong, Nature Computational Science, 2022, 2, 718--728

    2022

  37. [45]

    M. H. Rahman, P. Gollapalli, P. Manganaris, S. K. Yadav, G. Pilania, B. DeCost, K. Choudhary and A. Mannodi-Kanakkithodi, APL Machine Learning, 2024, 2, 016122

    2024

  38. [46]

    M. H. Rahman, S. Rojsatien, D. Krasikov, M. K. Chan, M. Bertoni and A. Mannodi-Kanakkithodi, Solar Energy Materials and Solar Cells, 2025, 293, 113857

    2025

  39. [47]

    M. H. Rahman and A. Mannodi-Kanakkithodi, Journal of Physics Materials, 2025, 8, 022001

    2025

  40. [48]

    J. Pan, W. K. Metzger and S. Lany, Phys. Rev. B, 2018, 98, 054108

    2018

  41. [49]

    Borlido, J

    P. Borlido, J. Schmidt, A. W. Huran, F. Tran, M. A. L. Marques and S. Botti, npj Computational Materials, 2020, 6,

    2020

  42. [50]

    C. Vona, D. Nabok and C. Draxl, Advanced Theory and Simulations, 2022, 5, 2100496

    2022

  43. [51]

    J. L. Lyons and C. G. Van de Walle, npj Computational Materials, 2017, 3, 12

    2017

  44. [52]

    Menéndez‐Proupin, M

    E. Menéndez‐Proupin, M. Casanova‐Páez, A. L. Montero‐Alejo, M. A. Flores and W. Orellana, Physica B Condensed Matter, 2019, 568, 81--87

    2019

  45. [53]

    F. K. Alfadhili, A. B. Phillips, G. K. Liyanage, J. M. Gibbs, M. K. Jamarkattel and M. J. Heben, MRS Advances, 2019, 4, 913--919

    2019

  46. [54]

    de Melo, M

    O. de Melo, M. Behar, J. F. Dias, R. Ribeiro-Andrade, M. da Silva, A. G. de Oliveira and J. C. González, Materials Science in Semiconductor Processing, 2019, 97, 17--20

    2019

  47. [55]

    K. Luo, W. Wu, S. Xie, Y. Jiang, S. Liao and D. Qin, Applied Sciences, 2019, 9, 1885--1885

    2019

  48. [56]

    Chen, C.-Y

    W.-C. Chen, C.-Y. Chen, Y.-R. Lin, J.-K. Chang, C.-H. Chen, Y.-P. Chiu, N.-I. Wu, K.-H. Chen and L.-C. Chen, Interface engineering of CdS/CZTSSe heterojunctions for enhancing the Cu2ZnSn(S,Se)4 solar cell efficiency, 2019, https://www.sciencedirect.com/science/article/pii/S246...

    2019

  49. [57]

    K. Shen, X. Wang, Y. Zhang, H. Zhu, Z. Chen, C. Huang and Y. Mai, Solar Energy, 2020, 201, 55--62

    2020

  50. [58]

    J. Miao, X. Liu, K. Jo, K. He, R. Saxena, B. Song, H. Zhang, J. He, M. Han, W. Hu and D. Jariwala, Nano Letters, 2020, 20, 2907--2915

    2020

  51. [59]

    A. G. García and S. Zarate, Microscopy and Microanalysis, 2020, 26, 2804--2805

    2020

  52. [60]

    Zheng, E

    X. Zheng, E. Colegrove, J. N. Duenow, J. Moseley and W. K. Metzger, Journal of Applied Physics, 2020, 128, 053102

    2020

  53. [61]

    Wardak, W

    A. Wardak, W. Chromiński, A. Reszka, D. Kochanowska, M. Witkowska‐Baran, M. Lewandowska and A. Mycielski, Journal of Alloys and Compounds, 2021, 874, 159941--159941

    2021

  54. [62]

    P. D. Hatton, M. J. Watts, Y. Zhou, R. Smith and P. Goddard, Journal of Physics Condensed Matter, 2022, 35, 75702--75702

    2022

  55. [63]

    M. A. Scarpulla, B. E. McCandless, A. B. Phillips, Y. Yan, M. J. Heben, C. A. Wolden, G. Xiong, W. K. Metzger, D. Mao, D. Krasikov, I. Sankin, S. Grover, A. Munshi, W. Sampath, J. R. Sites, A. Bothwell, D. S. Albin, M. O. Reese, A. Romeo, M. Nardone, R. F. Klie, J. M. Walls, T...

    2023

  56. [64]

    Huang, S

    Y. Huang, S. R. Kavanagh, D. O. Scanlon, A. Walsh and R. L. Z. Hoye, Nanotechnology, 2020, 32, 132004--132004

    2020

  57. [65]

    Bidaud, J

    T. Bidaud, J. Moseley, M. Amarasinghe, M. Al-Jassim, W. K. Metzger and S. Collin, Imaging CdCl2 defect passivation and formation in polycrystalline CdTe films by cathodoluminescence, 2021, https://journals.aps.org/prmaterials/abstract/10.1103/PhysRevMaterials.5.064601

    2021 doi

  58. [66]

    Shi and M

    J. Shi and M. Zikry, Materials Science and Engineering A, 2009, 520, 121--133

    2009

  59. [67]

    Y. Zhao, E. M. D. Siriwardane, Z. Wu, N. Fu, M. Al‐Fahdi, M. Hu and J. Hu, npj Computational Materials, 2023, 9, 38

    2023

  60. [68]

    Manna, H

    S. Manna, H. Chan, A. Ghosh, T. Chakrabarti and S. K. R. S. Sankaranarayanan, Computational Materials Science, 2023, 229, 112384--112384

    2023

  61. [69]

    Cheng, C.-L

    M. Cheng, C.-L. Fu, B. Yu, E. Rha, A. Chotrattanapituk, D. L. Abernathy, Y. Cheng and M. Li, 2025

    2025

  62. [70]

    M. Jin, J. Miao, M. Khafizov, B. Chen, Y. Zhang and D. H. Hurley, 2025

    2025

  63. [71]

    D. Xue, P. V. Balachandran, J. Hogden, J. Theiler, D. Xue and T. Lookman, Nature Communications, 2016, 7, 11241

    2016

  64. [72]

    A. G. Kusne, T. Gao, A. Mehta, L. Ke, M. C. Nguyen, K.-M. Ho, V. Antropov, C.-Z. Wang, M. J. Kramer, C. Long and I. Takeuchi, Scientific Reports, 2014, 4, 6367

    2014

  65. [73]

    F. Ren, L. Ward, T. Williams, K. J. Laws, C. Wolverton, J. Hattrick-Simpers and A. Mehta, Science Advances, 2018, 4, eaaq1566

    2018

  66. [74]

    C. Kim, A. Chandrasekaran, A. Jha and R. Ramprasad, MRS Communications, 2019, 9, 860–866

    2019

  67. [75]

    J. C. Verduzco, E. E. Marinero and A. Strachan, Integrating Materials and Manufacturing Innovation, 2021, 10, 299–310

    2021

  68. [76]

    J. E. Gentle, Computational Statistics, 2010, https://doi.org/10.1016/b978-0-08-044894-7.01316-6

    2010 doi

  69. [77]

    D. E. Farache, J. C. Verduzco, Z. D. McClure, S. Desai and A. Strachan, Computational Materials Science, 2022, 209, 111386

    2022

  70. [78]

    J. Heyd, G. E. Scuseria and M. Ernzerhof, The Journal of Chemical Physics, 2003, 118, 8207--8215

    2003

  71. [79]

    Bapst, T

    V. Bapst, T. Keck, A. Grabska-Barwińska, C. Donner, E. D. Cubuk, S. S. Schoenholz, A. Obika, A. W. R. Nelson, T. Back, D. Hassabis and P. Kohli, Nature Physics, 2020, 16, 448–454

    2020

  72. [80]

    C. Li, J. Poplawsky, Y. Yan and S. J. Pennycook, Mater. Sci. Semicond. Process., 2017, 65, 64--76

    2017

  73. [81]

    Battaglia, A

    C. Battaglia, A. Cuevas and S. De Wolf, Energy Environ. Sci., 2016, 9, 1552--1576

    2016

  74. [82]

    Mannodi-Kanakkithodi, Modelling and Simulation in Materials Science and Engineering, 2022, 30, 044001

    A. Mannodi-Kanakkithodi, Modelling and Simulation in Materials Science and Engineering, 2022, 30, 044001

    2022

  75. [83]

    Mannodi-Kanakkithodi, M

    A. Mannodi-Kanakkithodi, M. Y. Toriyama, F. G. Sen, M. J. Davis, R. F. Klie and M. K. Y. Chan, npj Computational Materials, 2020, 6, 39

    2020

  76. [84]

    Zunger, S.-h

    A. Zunger, S.-h. Wei, L. G. Ferreira and J. E. Bernard, Physical Review Letters, 1990, 65, 353–356

    1990

  77. [85]

    F. G. Sen, A. Mannodi-Kanakkithodi, T. Paulauskas, J. Guo, L. Wang, A. Rockett, M. J. Kim, R. F. Klie and M. K. Chan, Solar Energy Materials and Solar Cells, 2021, 232, 111279

    2021

  78. [86]

    S. R. Kavanagh, A. G. Squires, A. Nicolson, I. Mosquera-Lois, A. M. Ganose, B. Zhu, K. Brlec, A. Walsh and D. O. Scanlon, The Journal of Open Source Software, 2024, 9, 6433

    2024

  79. [87]

    Choudhary and B

    K. Choudhary and B. DeCost, npj Computational Materials, 2022, 8, 221

    2022

  80. [88]

    Mosquera-Lois, S

    I. Mosquera-Lois, S. R. Kavanagh, A. M. Ganose and A. Walsh, Npj Computational Materials, 2024, 10, 121

    2024

  81. [89]

    Choudhary, B

    K. Choudhary, B. DeCost, L. Major, K. Butler, J. Thiyagalingam and F. Tavazza, Digital Discovery, 2023, 2, 346–355

    2023

  82. [90]

    C. Chen, W. Ye, Y. Zuo, C. Zheng and S. P. Ong, Chemistry of Materials, 2019, 31, 3564--3572

    2019

  83. [91]

    Cheng, C

    J. Cheng, C. Zhang and L. Dong, Communications Materials, 2021, 2, 92

    2021

  84. [92]

    Lee and R

    J. Lee and R. Asahi, Computational Materials Science, 2021, 190, 110314

    2021

  85. [93]

    V. Fung, J. Zhang, E. Juarez and B. G. Sumpter, npj Computational Materials, 2021, 7, 84

    2021

  86. [94]

    S. R. Bahn and K. W. Jacobsen, Comput. Sci. Eng., 2002, 4, 56--66

    2002

  87. [95]

    A. H. Larsen, J. J. Mortensen, J. Blomqvist, I. E. Castelli, R. Christensen, M. Dułak, J. Friis, M. N. Groves, B. Hammer, C. Hargus, E. D. Hermes, P. C. Jennings, P. B. Jensen, J. Kermode, J. R. Kitchin, E. L. Kolsbjerg, J. Kubal, K. Kaasbjerg, S. Lysgaard, J. B. Maronsson, T....

    2017

  88. [96]

    M. H. Rahman and A. K. M. Kanakkithodi, Introducing DeFecT-FF for Accelerated Modeling of Defect Thermodynamics in CdSeTe Solar Cells, 2025, https://nanohub.org/resources/cadetff

    2025

  89. [97]

    Ščajev, M

    P. Ščajev, M. Nardone, C. Reich, R. Farshchi, K. McReynolds, D. Krasikov and D. Kuciauskas, Advanced Energy Materials, 2024, 2403902

    2024

  90. [98]

    F. A. Stevie and C. L. Donley, Journal of Vacuum Science & Technology A Vacuum Surfaces and Films, 2020, 38, 063204

    2020

  91. [99]

    Mahoney, C

    J. Mahoney, C. A. Monroe, A. M. Swartley, M. G. Ucak-Astarlioglu and C. A. Zoto, Spectroscopy Letters, 2020, 53, 726--736

    2020

  92. [100]

    A. Born, F. O. L. Johansson, T. Leitner, D. Kühn, A. Lindblad, N. Mårtensson and A. Föhlisch, Scientific Reports, 2021, 11, 16596

    2021

  93. [101]

    Lanza, M

    G. Lanza, M. J. Jimenez, F. Alvarez, J. Pérez and A. Ávila, ACS Omega, 2022, 7, 34521--34527

    2022

  94. [102]

    H. Chen, D. T. L. Alexander and C. Hébert, Nano Letters, 2024, 24, 10177--10185

    2024

  95. [103]

    H. Xie, X. Cheng and H. Huang, Investigation on the Interfaces in Organic Devices by Photoemission Spectroscopy, 2025, https://doi.org/10.3390/nano15090680

    2025 doi

  96. [104]

    J. H. Lee, J. H. Lee, S. H. Jung, T. K. Hyun, M. Feng, J.-Y. Kim, J. Lee, H.-Y. Lee, J. S. Kim, C. Kang, K.-Y. Kwon and J. H. Jung, Chemical Communications, 2015, 51, 7463--7465

    2015

  97. [105]

    L. Zhao, C. Sun, G. Tian and Q. Pang, Journal of Colloid and Interface Science, 2017, 502, 1--7

    2017

  98. [106]

    Y. Li, G. Zha, D. Wei, F. Yang, J. Dong, S. Xi, L. Xu and W. Jie, Sensors, 2020, 20, 2032--2032

    2020

  99. [107]

    T. Li, Y. Zhu, X. Ji, W. Zheng, Z. Lin, X. Lu and F. Huang, The Journal of Physical Chemistry Letters, 2020, 11, 8901--8907

    2020

  100. [108]

    Dragoni, T

    D. Dragoni, T. D. Daff, G. Csányi and N. Marzari, Physical Review Materials, 2018, 2, 013808

    2018

  101. [109]

    Berger, M

    E. Berger, M. Bagheri and H. Komsa, Small, 2025

    2025

  102. [110]

    Mosquera-Lois, S

    I. Mosquera-Lois, S. R. Kavanagh, A. Walsh and D. O. Scanlon, Npj Computational Materials, 2023, 9, 25

    2023

  103. [111]

    Mosquera-Lois, S

    I. Mosquera-Lois, S. R. Kavanagh, A. Walsh and D. O. Scanlon, The Journal of Open Source Software, 2022, 7, 4817

    2022

  104. [112]

    crystal formation energy

    G. Cheng, X.-G. Gong and W.-J. Yin, Nature Communications, 2022, 13, 1492 mcitethebibliography main.tex0000664000000000000000000032430215077721123011235 0ustar rootroot [twoside, twocolumn] article [utf8] inputenc [T1] fontenc graphicx [switch] lineno longtable wrapfig rotatin...

    2022

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.