pith. sign in

arxiv: 2606.20500 · v1 · pith:D3XIQPHDnew · submitted 2026-06-18 · ❄️ cond-mat.mtrl-sci · physics.chem-ph

A Defect-Free Model of Amorphous Silicon with Pristine Electronic Structure

Louise A. M. Rosset , Chinonso Ugwumadu , Stephen R. Elliott , David A. Drabold , Volker L. Deringer This is my paper

Pith reviewed 2026-06-26 16:04 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci physics.chem-ph
keywords amorphous silicondefect-free modelelectronic bandgapmachine-learning molecular dynamicshybrid density functional theorycontinuous random networkband tails
0
0 comments X

The pith

A defect-free model of amorphous silicon generated by machine-learning molecular dynamics and checked with hybrid DFT reproduces the experimental electronic bandgap.

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

The paper establishes that an 'ideal' continuous random network of a-Si, free of coordination defects, can be produced through machine-learning-driven molecular dynamics and then evaluated at the hybrid density-functional level to yield an electronic bandgap matching experiment. A sympathetic reader would care because previous models of a-Si have struggled to combine structural ideality with accurate electronic properties without introducing artificial defects or relying on empirical adjustments. If correct, the approach supplies a clean platform for examining band tails, optical absorption, and charge transport in the material without the confounding effects of dangling bonds or wrong bonds. The central object carrying the argument is the ML-generated atomic configuration whose electronic structure is recomputed independently at higher theory level.

Core claim

A defect-free ('ideal') model of a-Si from machine-learning-driven molecular-dynamics simulations, subsequently evaluated with hybrid-level density-functional theory computations, can accurately reproduce the experimentally observed electronic bandgap. The authors compare this model with one from the Wooten-Winer-Weaire bond-switching method and with other recent approximants to ideal a-Si, and they position the new structure as a platform for studies of band tails, optical properties, and transport.

What carries the argument

The machine-learning interatomic potential that generates the defect-free a-Si configuration during molecular dynamics, followed by independent hybrid-DFT recomputation of its electronic structure.

If this is right

  • The model supplies a reference structure free of coordination defects for investigating band-tail states.
  • Optical and transport calculations can be performed on the same defect-free network without artificial gap states from dangling bonds.
  • Direct comparison with the WWW bond-switching model isolates the effect of generation method on electronic properties.
  • The workflow can be repeated for other amorphous semiconductors to test transferability of the ML-plus-hybrid-DFT protocol.

Where Pith is reading between the lines

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

  • If the ML potential is unbiased for electronic properties, the same structures could be used to benchmark cheaper semi-local functionals against hybrid results.
  • The approach suggests that structural ideality and correct bandgap can be achieved simultaneously, which may reduce the need for post-hoc defect insertion in device modeling.
  • Extension to larger cells would allow direct simulation of localized tail states and their participation in transport.

Load-bearing premise

The machine-learning potential used to create the atomic arrangement produces a configuration whose electronic bandgap, when recomputed with hybrid DFT, is not biased by the potential's training data or simulation choices.

What would settle it

A direct hybrid-DFT calculation on the ML-generated structure that yields a bandgap differing substantially from the accepted experimental value for a-Si would falsify the central claim.

Figures

Figures reproduced from arXiv: 2606.20500 by Chinonso Ugwumadu, David A. Drabold, Louise A. M. Rosset, Stephen R. Elliott, Volker L. Deringer.

Figure 1
Figure 1. Figure 1: FIG. 1. Electronic structure. (a) Electronic densities of states of the four structural model considered in this study, where [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Chemical bonding. (a) Scatter plots of integrated pairwise crystal orbital bond index (ICOBI [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
read the original abstract

Amorphous silicon (a-Si) is understood to be the canonical continuous random network material, ideally defined by fully fourfold coordination. Here, we show that a defect-free ('ideal') model of a-Si from machine-learning-driven molecular-dynamics simulations [L. A. M. Rosset et al., Nat. Commun. 16, 2360 (2025)], subsequently evaluated with hybrid-level density-functional theory computations, can accurately reproduce the experimentally observed electronic bandgap. We compare this model with one resulting from the Wooten-Winer-Weaire (WWW) bond-switching approach and with other recent approximants to ideal a-Si. More broadly, our work provides a platform for studies of band tails, optical properties, and transport in a-Si.

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 / 2 minor

Summary. The manuscript claims that a defect-free ('ideal') model of amorphous silicon generated via machine-learning-driven molecular-dynamics simulations (taken from Rosset et al., Nat. Commun. 2025) reproduces the experimentally observed electronic bandgap when its electronic structure is recomputed at the hybrid-DFT level. The work compares this model to the Wooten-Winer-Weaire (WWW) bond-switching model and other recent approximants, positioning the structure as a platform for studies of band tails, optical properties, and transport.

Significance. If the central claim holds after addressing independence concerns, the result would supply a high-quality, large-scale continuous-random-network model of a-Si with a pristine gap, which is valuable for the field because traditional models often contain defects that complicate electronic-property calculations. The ML-MD route enables access to defect-free configurations at scales difficult for direct ab initio methods, and the hybrid-DFT evaluation step is a standard and appropriate choice for gap accuracy.

major comments (2)
  1. [Methods / Results (structure generation and electronic evaluation)] The structure originates from the ML interatomic potential of Rosset et al. (Nat. Commun. 2025). No test is reported in which an identically sized cell generated by an orthogonal protocol (WWW, ab-initio MD, or a different ML potential) is relaxed and evaluated at the same hybrid-DFT level to demonstrate that the reported bandgap match is independent of structural correlations learned during potential training (e.g., ring statistics or bond-angle distributions).
  2. [Abstract and Results section] The abstract states that the hybrid-DFT bandgap 'accurately reproduce[s] the experimentally observed electronic bandgap,' yet the provided text supplies neither the numerical gap value, its uncertainty, nor a direct comparison plot or table against experiment and the WWW reference. Without these quantities the quantitative claim cannot be assessed.
minor comments (2)
  1. [Methods] Notation for the hybrid functional (exact exchange fraction, range-separation parameters) and k-point sampling used in the DFT calculations should be stated explicitly in the methods.
  2. [Figures] Figure captions should include the system size (number of atoms) and the precise definition of 'defect-free' (e.g., coordination cutoff) for each model shown.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed report. The comments identify key areas where additional comparisons and quantitative clarity would strengthen the manuscript. We address each major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Methods / Results (structure generation and electronic evaluation)] The structure originates from the ML interatomic potential of Rosset et al. (Nat. Commun. 2025). No test is reported in which an identically sized cell generated by an orthogonal protocol (WWW, ab-initio MD, or a different ML potential) is relaxed and evaluated at the same hybrid-DFT level to demonstrate that the reported bandgap match is independent of structural correlations learned during potential training (e.g., ring statistics or bond-angle distributions).

    Authors: We agree that an explicit test with an orthogonal generation protocol evaluated at the identical hybrid-DFT level would help demonstrate that the bandgap result is not an artifact of structural features learned by the ML potential. In the revised manuscript we will add such a comparison: a WWW-generated cell of comparable size will be relaxed consistently with the ML potential and then evaluated at hybrid-DFT, with the resulting bandgap reported alongside the ML-MD result and experiment. revision: yes

  2. Referee: [Abstract and Results section] The abstract states that the hybrid-DFT bandgap 'accurately reproduce[s] the experimentally observed electronic bandgap,' yet the provided text supplies neither the numerical gap value, its uncertainty, nor a direct comparison plot or table against experiment and the WWW reference. Without these quantities the quantitative claim cannot be assessed.

    Authors: We acknowledge the omission. The revised manuscript will update the abstract to state the numerical hybrid-DFT bandgap value (with uncertainty) and its relation to experiment. We will also add a dedicated table or figure in the Results section that directly compares the gap values obtained for the ML-MD model, the WWW reference, and the experimental value. revision: yes

Circularity Check

0 steps flagged

No significant circularity; electronic evaluation is independent of structure generation.

full rationale

The paper's chain is: cite prior ML-MD work for a defect-free a-Si model, then recompute its electronic density of states and bandgap at hybrid DFT level, and compare to experiment and to a WWW model. No equation or claim reduces the hybrid-DFT bandgap to the ML potential's training data or loss function by construction. The hybrid-DFT step uses a different Hamiltonian and is externally falsifiable against measured gaps. Self-citation of the structure model is present but not load-bearing for the electronic result, which remains an independent computation. This is the normal case of a self-contained derivation against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No explicit free parameters, axioms or invented entities are stated in the abstract. The machine-learning potential that generated the structure is presumed to contain fitted parameters from the referenced prior work, but none are enumerated here.

pith-pipeline@v0.9.1-grok · 5683 in / 1143 out tokens · 27915 ms · 2026-06-26T16:04:36.770304+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

73 extracted references

  1. [1]

    W. H. Zachariasen, J. Am. Chem. Soc.54, 3841 (1932)

    1932

  2. [2]

    Laaziri, S

    K. Laaziri, S. Kycia, S. Roorda, M. Chicoine, J. L. Robertson, J. Wang, and S. C. Moss, Phys. Rev. B60, 13520 (1999)

    1999

  3. [3]

    Birney, J

    R. Birney, J. Steinlechner, Z. Tornasi, S. MacFoy, D. Vine, A. S. Bell, D. Gibson, J. Hough, S. Rowan, P. Sortais, S. Sproules, S. Tait, I. W. Martin, and S. Reid, Phys. Rev. Lett.121, 191101 (2018)

    2018

  4. [4]

    Wooten, K

    F. Wooten, K. Winer, and D. Weaire, Phys. Rev. Lett. 54, 1392 (1985)

    1985

  5. [5]

    Zongo, H

    K. Zongo, H. Sun, C. Ouellet-Plamondon, N. Mousseau, and L. K. B´ eland, Phys. Rev. B111, 214209 (2025)

    2025

  6. [6]

    R. H. Klazes, M. H. L. M. van den Broek, J. Bezemer, and S. Radelaar, Philos. Mag. B45, 377 (1982)

    1982

  7. [7]

    W. Y. Ching, D. J. Lam, and C. C. Lin, Phys. Rev. B 21, 2378 (1980)

    1980

  8. [8]

    Allan, C

    G. Allan, C. Delerue, and M. Lannoo, Phys. Rev. B57, 6933 (1998)

    1998

  9. [9]

    R. V. Meidanshahi, S. Bowden, and S. M. Goodnick, Phys. Chem. Chem. Phys.21, 13248 (2019)

    2019

  10. [10]

    L. A. M. Rosset, D. A. Drabold, and V. L. Deringer, Nat. Commun.16, 2360 (2025)

    2025

  11. [11]

    J. D. Morrow and V. L. Deringer, J. Chem. Phys.157, 104105 (2022)

    2022

  12. [12]

    11 was distilled from the Si- GAP-18 potential introduced in A

    The MLIP model of Ref. 11 was distilled from the Si- GAP-18 potential introduced in A. P. Bart´ ok, J. Ker- mode, N. Bernstein, and G. Cs´ anyi, Phys. Rev. X8, 041048 (2018)

    2018

  13. [13]

    Further details are given in the Supplementary Material

  14. [14]

    A. P. Bart´ ok, R. Kondor, and G. Cs´ anyi, Phys. Rev. B 87, 184115 (2013)

    2013

  15. [15]

    B. R. Djordjevi´ c, M. F. Thorpe, and F. Wooten, Phys. Rev. B52, 5685 (1995)

    1995

  16. [16]

    Pedersen, L

    A. Pedersen, L. Pizzagalli, and H. J´ onsson, New J. Phys. 19, 063018 (2017)

    2017

  17. [17]

    Mousseau, L

    N. Mousseau, L. K. B´ eland, P. Brommer, J.-F. Joly, F. El-Mellouhi, E. Machado-Charry, M.-C. Marinica, and P. Pochet, J. Phys. B: At. Mol. Phys.2012, 925278 (2012)

    2012

  18. [18]

    P. N. Keating, Phys. Rev.145, 637 (1966)

    1966

  19. [19]

    Y. Pan, M. Zhang, and D. Drabold, J. Non-Cryst. Solids 354, 3480 (2008)

    2008

  20. [20]

    Opletal, T

    G. Opletal, T. C. Petersen, I. K. Snook, and D. G. Mc- Culloch, J. Chem. Phys.126, 214705 (2007)

    2007

  21. [21]

    M. J. Cliffe, M. T. Dove, D. A. Drabold, and A. L. Good- win, Phys. Rev. Lett.104, 125501 (2010)

    2010

  22. [22]

    Pandey, P

    A. Pandey, P. Biswas, and D. A. Drabold, Sci. Rep.6, 33731 (2016)

    2016

  23. [23]

    Igram, B

    D. Igram, B. Bhattarai, P. Biswas, and D. Drabold, J. Non-Cryst. Solids492, 27 (2018)

    2018

  24. [24]

    J. Heyd, J. E. Peralta, G. E. Scuseria, and R. L. Martin, J. Chem. Phys.123, 174101 (2005)

    2005

  25. [25]

    H. J. Xiang, B. Huang, E. Kan, S.-H. Wei, and X. G. Gong, Phys. Rev. Lett.110, 118702 (2013)

    2013

  26. [26]

    Jarolimek, E

    K. Jarolimek, E. Hazrati, R. De Groot, and G. De Wijs, Phys. Rev. Appl.8, 014026 (2017)

    2017

  27. [27]

    J. D. Morrow, C. Ugwumadu, D. A. Drabold, S. R. Elliott, A. L. Goodwin, and V. L. Deringer, Angew. Chem. Int. Ed.63, e202403842 (2024)

    2024

  28. [28]

    P. E. Bl¨ ochl, Phys. Rev. B50, 17953 (1994)

    1994

  29. [29]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Phys. Rev. B54, 11169 (1996)

    1996

  30. [30]

    Kresse and D

    G. Kresse and D. Joubert, Phys. Rev. B59, 1758 (1999)

    1999

  31. [31]

    J. Heyd, G. E. Scuseria, and M. Ernzerhof, J. Chem. Phys.118, 8207 (2003)

    2003

  32. [32]

    J. Heyd, G. E. Scuseria, and M. Ernzerhof, J. Chem. Phys.124, 219906 (2006)

    2006

  33. [33]

    J. S. Custer, M. O. Thompson, D. C. Jacobson, J. M. Poate, S. Roorda, W. C. Sinke, and F. Spaepen, Appl. Phys. Lett.64, 437 (1994). 6

    1994

  34. [34]

    Laaziri, S

    K. Laaziri, S. Roorda, and J. M. Baribeau, J. Non- Cryst. Solids191, 193 (1995)

    1995

  35. [35]

    Y. Liu, Y. Zhou, R. Ademuwagun, L. Walter- bos, J. George, S. R. Elliott, and V. L. Deringer, J. Am. Chem. Soc.148, 9400 (2026)

    2026

  36. [36]

    S. C. Moss and J. F. Graczyk, Phys. Rev. Lett.23, 1167 (1969)

    1969

  37. [37]

    L. J. Lewis, J. Non-Cryst. Solids580, 121383 (2022)

    2022

  38. [38]

    P. M. Larsen, S. Schmidt, and J. Schiøtz, Model. Simul. Mater. Sci. Eng.24, 055007 (2016)

    2016

  39. [39]

    J. D. Honeycutt and H. C. Andersen, J. Phys. Chem.91, 4950 (1987)

    1987

  40. [40]

    A. S. Clarke and H. J´ onsson, Phys. Rev. E47, 3975 (1993)

    1993

  41. [41]

    Stukowski, Model

    A. Stukowski, Model. Simul. Mater. Sci. Eng.20, 045021 (2012)

    2012

  42. [42]

    Maras, O

    E. Maras, O. Trushin, A. Stukowski, T. Ala-Nissila, and H. J´ onsson, Comput. Phys. Commun.205, 13 (2016)

    2016

  43. [43]

    Bernstein, B

    N. Bernstein, B. Bhattarai, G. Cs´ anyi, D. A. Drabold, S. R. Elliott, and V. L. Deringer, Angew. Chem. Int. Ed. 58, 7057 (2019)

    2019

  44. [44]

    Roorda and L

    S. Roorda and L. J. Lewis, Science338, 1539 (2012)

    2012

  45. [45]

    D. A. Drabold, Phys. Status Solidi Rapid Res. Lett.5, 359 (2011)

    2011

  46. [46]

    Stukowski, Model

    A. Stukowski, Model. Simul. Mater. Sci. Eng18, 015012 (2010)

    2010

  47. [47]

    M. F. Thorpe and D. Weaire, Phys. Rev. Lett.27, 1581 (1971)

    1971

  48. [48]

    Dong and D

    J. Dong and D. A. Drabold, Phys. Rev. Lett.80, 1928 (1998)

    1928

  49. [49]

    Urbach, Phys

    F. Urbach, Phys. Rev.92, 1324 (1953)

    1953

  50. [50]

    Aljishi, J

    S. Aljishi, J. D. Cohen, S. Jin, and L. Ley, Phys. Rev. Lett.64, 2811 (1990)

    1990

  51. [51]

    D. A. Drabold, Y. Li, B. Cai, and M. Zhang, Phys. Rev. B 83, 045201 (2011)

    2011

  52. [52]

    D. Y. Kim, S. Stefanoski, O. O. Kurakevych, and T. A. Strobel, Nat. Mater.14, 169 (2015)

    2015

  53. [53]

    Durandurdu, D

    M. Durandurdu, D. A. Drabold, and N. Mousseau, Phys. Rev. B62, 15307 (2000)

    2000

  54. [54]

    G. T. Barkema and N. Mousseau, Phys. Rev. B62, 4985 (2000)

    2000

  55. [55]

    G. K. M. Thutupalli and S. G. Tomlin, J. Phys. C: Solid State Phys.10, 467 (1977)

    1977

  56. [56]

    Park, C.-J

    N.-M. Park, C.-J. Choi, T.-Y. Seong, and S.-J. Park, Phys. Rev. Lett.86, 1355 (2001)

    2001

  57. [57]

    Huang, Y

    C.-C. Huang, Y. Tang, M. van der Laan, J. van de Groep, A. F. Koenderink, and K. Dohnalov´ a, ACS Appl. Nano Mater.4, 288 (2021)

    2021

  58. [58]

    D. A. Drabold, P. A. Fedders, O. F. Sankey, and J. D. Dow, Phys. Rev. B42, 5135 (1990)

    1990

  59. [59]

    Prasai, P

    K. Prasai, P. Biswas, and D. A. Drabold, Semi- cond. Sci. Technol.31, 073002 (2016)

    2016

  60. [60]

    D. J. Thouless, Phys. Rep.13, 93 (1974)

    1974

  61. [61]

    Herath, P

    U. Herath, P. Tavadze, X. He, E. Bousquet, S. Singh, F. Mu˜ noz, and A. H. Romero, Comput. Phys. Commun. 251, 107080 (2020)

    2020

  62. [62]

    L. Lang, P. Tavadze, A. Tellez, E. Bousquet, H. Xu, F. Mu˜ noz, N. Vasquez, U. Herath, and A. H. Romero, Comput. Phys. Commun.297, 109063 (2024)

    2024

  63. [63]

    Y. Pan, F. Inam, M. Zhang, and D. A. Drabold, Phys. Rev. Lett.100, 206403 (2008)

    2008

  64. [64]

    P. A. Fedders, D. A. Drabold, and S. Nakhmanson, Phys. Rev. B58, 15624 (1998)

    1998

  65. [65]

    F. Inam, J. P. Lewis, and D. A. Drabold, Phys. Status Solidi A207, 599 (2010)

    2010

  66. [66]

    Maintz, V

    S. Maintz, V. L. Deringer, A. L. Tchougr´ eeff, and R. Dronskowski, J. Comput. Chem.34, 2557 (2013)

    2013

  67. [67]

    Nelson, C

    R. Nelson, C. Ertural, J. George, V. L. Deringer, G. Hau- tier, and R. Dronskowski, J. Comput. Chem.41, 1931 (2020)

    1931

  68. [68]

    P. C. M¨ uller, C. Ertural, J. Hempelmann, and R. Dron- skowski, J. Phys. Chem. C125, 7959 (2021)

    2021

  69. [69]

    Pauling, J

    L. Pauling, J. Am. Chem. Soc.69, 542 (1947)

    1947

  70. [70]

    Winer, I

    K. Winer, I. Hirabayashi, and L. Ley, Phys. Rev. Lett. 60, 2697 (1988)

    1988

  71. [71]

    Y. Lee, Y. Hu, D. Kim, S. Datta, and K. Cho, Phys. Rev. B105, 085201 (2022)

    2022

  72. [72]

    H. Lin, M. Yang, X. Ru, G. Wang, S. Yin, F. Peng, C. Hong, M. Qu, J. Lu, L. Fang, C. Han, P. Procel, O. Isabella, P. Gao, Z. Li, and X. Xu, Nat. Energy8, 789 (2023)

    2023

  73. [73]

    Oliphant, V

    E. Oliphant, V. Mantena, M. Brod, G. J. Snyder, and W. Sun, Mater. Horiz.12, 3073 (2025). Supplemental Material for ‘A Defect-Free Model of Amorphous Silicon with Pristine Electronic Structure’ Louise A. M. Rosset1, Chinonso Ugwumadu2, Stephen R. Elliott3, David A. Drabold4, and V olker L. Deringer*1 1Inorganic Chemistry Laboratory, Department of Chemistr...

    2025