pith. sign in

arxiv: 2604.21222 · v1 · submitted 2026-04-23 · ❄️ cond-mat.mtrl-sci · cs.LG

Neutron and X-ray Diffraction Reveal the Limits of Long-Range Machine Learning Potentials for Medium-Range Order in Silica Glass

Sai Harshit Balantrapu , Atul C. Thakur , Chris Benmore , Ganesh Sivaraman This is my paper

Pith reviewed 2026-05-09 21:58 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci cs.LG
keywords silica glassmedium-range ordermachine learning interatomic potentialsfirst sharp diffraction peakneutron diffractionX-ray diffractionmolecular dynamics
0
0 comments X p. Extension

The pith

Machine learning potentials with long-range interactions still cannot accurately model medium-range order in silica glass.

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

The study tests if long-range extensions to machine learning interatomic potentials can fix the inability of local models to predict medium-range order in glassy silica. Neutron and X-ray diffraction data serve as the benchmark for the first sharp diffraction peak that signals this order. Short-range models create too much order in both liquid and glass states. Adding long-range terms helps the liquid but the quenched glass still shows wrong ring distributions and angle constraints. This points to the need for training data that captures the dynamics of the liquid-to-glass transition.

Core claim

Incorporating explicit long-range interactions via reciprocal-space gated attention in MACE models reduces excess ordering in liquid silica but fails to recover the experimental first sharp diffraction peak and ring statistics in the amorphous glass after quenching, indicating that long-range interactions are necessary but not sufficient for accurate medium-range order.

What carries the argument

Comparison of short-range and long-range MACE machine learning interatomic potentials in large-scale molecular dynamics simulations, benchmarked against neutron and X-ray diffraction for the first sharp diffraction peak and supported by ring-statistics and bond-angle analyses.

If this is right

  • Short-range potentials over-structure the silica network, producing an overly intense first sharp diffraction peak.
  • Long-range potentials improve the liquid structure factor but retain biases in ring populations after quenching.
  • Both models exhibit limited Si-O-Si angle variability, constraining network flexibility.
  • Accurate modeling requires sampling strategies that better represent the vitrification process.

Where Pith is reading between the lines

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

  • Future models may need to incorporate configurations from slower quenching rates or reverse-engineered glass structures in training.
  • The findings could extend to other amorphous network-forming materials beyond silica.
  • Hybrid approaches combining MLIPs with explicit electrostatic long-range terms may address the remaining gaps.

Load-bearing premise

The molecular dynamics quenching protocol used to simulate the liquid-to-glass transition adequately mimics the experimental vitrification process.

What would settle it

A simulation using a training set that includes a diverse set of quenched configurations from multiple rates or direct experimental glass data would produce a first sharp diffraction peak matching experiment if the claim is correct.

read the original abstract

Glassy silica is a foundational material in optics and electronics, yet accurately predicting its medium-range order (MRO) remains a major challenge for machine-learning interatomic potentials (MLIPs). While local MLIPs reproduce the short-range SiO4 tetrahedral network well, it remains unclear whether locality alone is sufficient to recover the first sharp diffraction peak (FSDP), the principal experimental signature of MRO. Here, we combine neutron and X-ray diffraction measurements with large-scale molecular dynamics driven by two MACE-based models: a short-range (SR) potential and a long-range (LR) extension incorporating reciprocal-space gated attention. The SR model systematically over-structures the network, producing an overly intense FSDP in both the liquid and glassy states. Incorporating long-range interactions improves agreement with experiment for the liquid structure by reducing this excess ordering, but the LR model still fails to recover the experimental amorphous MRO after quenching. Ring-statistics and bond-angle analyses reveal that SR model exhibits an artificially narrow distribution dominated by six-membered rings, while the LR model produces a broader but still biased ring population. Despite preserving the correct tetrahedral geometry, both models show limited variability in Si-O-Si angles, indicating constrained network flexibility. These structural signatures demonstrate that both models retain excessive memory of the parent liquid network, leading to kinetically trapped and nonphysical medium-range configurations during vitrification. These results show that explicit long-range interactions are necessary but not sufficient for predictive modelling of disordered silica and suggest that accurate MRO further requires training data and sampling strategies that adequately represent the liquid-to-glass transition.

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 paper compares short-range (SR) and long-range (LR) MACE-based machine-learning interatomic potentials for silica using large-scale MD simulations against experimental neutron and X-ray diffraction data. It reports that the SR model over-structures the network (intense FSDP) in both liquid and glass, while the LR model improves liquid agreement but still fails to recover experimental medium-range order (FSDP, ring statistics, Si-O-Si angles) after quenching, attributing this to both models retaining excessive memory of the parent liquid and becoming kinetically trapped. The central conclusion is that explicit long-range interactions are necessary but not sufficient for predictive MRO modeling and that training data/sampling must better represent the liquid-to-glass transition.

Significance. If the central interpretation holds, the work provides a clear demonstration via direct experimental benchmarks that locality alone is insufficient for MRO in network glasses and that long-range extensions help liquids but not necessarily glasses. The combination of diffraction comparisons with ring and angle analyses is a strength, as is the falsifiable prediction that improved sampling of the vitrification process will be required for accurate MLIPs in disordered materials.

major comments (2)
  1. [Abstract] Abstract: the attribution of MRO discrepancies to the potentials' inability to escape liquid-like configurations (both models 'retain excessive memory of the parent liquid network, leading to kinetically trapped' states) is load-bearing for the 'necessary but not sufficient' claim, yet the manuscript provides no sensitivity tests to cooling rate, ensemble, starting liquid configuration, or fictive temperature. Standard MD quenching rates (10^11–10^13 K/s) are orders of magnitude faster than laboratory vitrification; without such tests it remains possible that the observed ring bias and FSDP mismatch arise from the protocol rather than model limits.
  2. [Abstract] Abstract and results on structural analyses: the statements that the SR model produces 'an artificially narrow distribution dominated by six-membered rings' and that the LR model yields 'a broader but still biased ring population' with 'limited variability in Si-O-Si angles' are central to the kinetic-trapping interpretation, but the manuscript does not report the number of independent quenches, statistical uncertainties, or convergence checks on these distributions, making it difficult to assess whether the differences are robust or protocol-dependent.
minor comments (2)
  1. The abstract would benefit from a concise statement of the specific MACE variants used, the system sizes, and at least one quantitative metric (e.g., FSDP position or intensity deviation) with uncertainty to allow readers to gauge the magnitude of the reported improvements and failures.
  2. Ensure that all figures comparing simulated and experimental structure factors include error bars or shaded regions representing run-to-run variability or experimental uncertainty.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which help clarify the scope and robustness of our claims. We address each major point below and will revise the manuscript to incorporate additional qualifications and statistical details where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the attribution of MRO discrepancies to the potentials' inability to escape liquid-like configurations (both models 'retain excessive memory of the parent liquid network, leading to kinetically trapped' states) is load-bearing for the 'necessary but not sufficient' claim, yet the manuscript provides no sensitivity tests to cooling rate, ensemble, starting liquid configuration, or fictive temperature. Standard MD quenching rates (10^11–10^13 K/s) are orders of magnitude faster than laboratory vitrification; without such tests it remains possible that the observed ring bias and FSDP mismatch arise from the protocol rather than model limits.

    Authors: We agree that the absence of explicit sensitivity tests limits the strength of the kinetic-trapping interpretation. Our simulations employed standard quenching rates and protocols typical for large-scale silica MD studies, and the consistent observation of trapping in both SR and LR models under identical conditions supports a model-intrinsic component. Nevertheless, without varying cooling rates or initial configurations, we cannot fully exclude protocol dependence. We will revise the abstract and discussion sections to qualify the claim, explicitly noting that the reported trapping occurs under conventional simulation vitrification conditions and that slower laboratory-like rates or enhanced sampling may be required to fully assess MRO recovery. This revision will temper the 'necessary but not sufficient' conclusion accordingly while preserving the comparative evidence between the two models. revision: yes

  2. Referee: [Abstract] Abstract and results on structural analyses: the statements that the SR model produces 'an artificially narrow distribution dominated by six-membered rings' and that the LR model yields 'a broader but still biased ring population' with 'limited variability in Si-O-Si angles' are central to the kinetic-trapping interpretation, but the manuscript does not report the number of independent quenches, statistical uncertainties, or convergence checks on these distributions, making it difficult to assess whether the differences are robust or protocol-dependent.

    Authors: We acknowledge that the manuscript text omits explicit reporting of the number of independent quenches, uncertainties, and convergence checks for the ring statistics and angle distributions. In practice, the presented results were obtained by averaging over multiple independent quenches initiated from distinct equilibrated liquid configurations to improve sampling. We will revise the methods section to state the exact number of independent runs performed for each model, include error estimates or standard deviations on the ring populations and angle distributions, and add a brief convergence check (e.g., comparison of statistics from subsets of the trajectories). These additions will allow readers to evaluate the robustness of the structural signatures used to support the kinetic-trapping argument. revision: yes

Circularity Check

0 steps flagged

No circularity: claims benchmarked directly against external neutron/X-ray data

full rationale

The paper derives its conclusions from direct comparisons of SR and LR MACE-based MD simulations to experimental diffraction patterns, FSDP intensities, ring statistics, and bond-angle distributions. No prediction or structural metric is obtained by fitting to the target MRO observables themselves; the models are evaluated on held-out quenched configurations against independent measurements. The statement that long-range interactions are 'necessary but not sufficient' follows from the observed residual discrepancies with experiment rather than from any self-definition or parameter renaming. No self-citation chain is load-bearing for the central claim, and the derivation remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim depends on the domain assumption that FSDP intensity directly indexes MRO and that MD quenching can isolate model deficiencies from protocol artifacts; no free parameters or invented entities are introduced in the abstract.

axioms (1)
  • domain assumption The first sharp diffraction peak (FSDP) is the principal experimental signature of medium-range order (MRO) in silica glass.
    Invoked to interpret discrepancies between simulated and measured structure factors.

pith-pipeline@v0.9.0 · 5612 in / 1272 out tokens · 33573 ms · 2026-05-09T21:58:54.910172+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

63 extracted references · 4 canonical work pages · 1 internal anchor

  1. [1]

    [1-4] Despite its simple chemical composition and its ability to form a polyhedral network [ 5], amorphous silica exhibits a complex atomic structure across multiple length scales

    Introduction Glassy silica is the archetypal glass forming system, with applications in various fields like optics, microelectronics, geosciences and energy systems. [1-4] Despite its simple chemical composition and its ability to form a polyhedral network [ 5], amorphous silica exhibits a complex atomic structure across multiple length scales. At the sho...

  2. [2]

    model, which includes self -consistent dipole interactions, have demonstrated improved accuracy for certain structural and dynamical properties of silica compared to the BKS potential [20]. Despite these developments, classical force fields, while capable of reproducing key structural features of amorphous silica, including aspects of medium - range order...

  3. [3]

    Methods This section describes the methods used in this study. We first outline the experimental measurements, then summarize the short-range and long-range MACE models, the DFT training dataset, the molecular -dynamics and melt -quench protocol, and the structural analyses used to compare simulated and experimental silica structures. 2.1 Experimental Met...

  4. [4]

    The structure factor S(Q) was computed for direct comparison with experimental data

    definition of rings to quantify medium -range order through network topology. The structure factor S(Q) was computed for direct comparison with experimental data. These structural descriptors provide a direct link between atomistic simulations and experimental measurements and enabling us to assess how interaction range influences medium - range order in silica

  5. [5]

    Results and Discussion 3.1- Model Validation 8 Before analysing structural observables, we first verify that both models reproduce the DFT reference energies and forces. Fig. 2 shows that both the models provide good agreement with DFT, indicating that both provide a sufficiently accurate description of the reference potential -energy surface for molecula...

    1973

  6. [6]

    While both SR and LR models reproduce the local SiO₄ tetrahedral geometry, they diverge in their description of the FSDP and associa ted network descriptors

    Discussion The results presented in this work reveal a clear separation between local and medium -range structure in disordered silica. While both SR and LR models reproduce the local SiO₄ tetrahedral geometry, they diverge in their description of the FSDP and associa ted network descriptors. The SR model over-structures the network, whereas the LR 14 mod...

  7. [7]

    While both models accurately reproduce short-range tetrahedral order, significant differences emerge at the level of medium -range order (MRO)

    Conclusions In this work, we examined the role of locality in machine-learned interatomic potentials for disordered silica by comparing short -range (SR) and long -range (LR) MACE models against experimental high energy X-ray and neutron diffraction data. While both models accurately reproduce short-range tetrahedral order, significant differences emerge ...

    work page doi:10.46936/aps-182067/60010593

  8. [8]

    Formation and applications of nanoparticles in silica optical fibers

    Blanc, Wilfried, and Bernard Dussardier. "Formation and applications of nanoparticles in silica optical fibers." Journal of Optics 45, no. 3 (2016): 247–254

    2016

  9. [9]

    Silicate glasses and their impact on humanity

    Youngman, Randall E. "Silicate glasses and their impact on humanity." Reviews in Mineralogy and Geochemistry 87, no. 1 (2022): 1015–1038

    2022

  10. [10]

    Principles of the development of a silica dielectric for microelectronics packaging

    Gupta, Tapan K., and Jau -Ho Jean. "Principles of the development of a silica dielectric for microelectronics packaging." Journal of Materials Research 11, no. 1 (1996): 243–263

    1996

  11. [11]

    A review on dielectric breakdown in thin dielectrics: silicon dioxide, high -k, and layered dielectrics

    Palumbo, Felix, Chao Wen, Salvatore Lombardo, Sebastian Pazos, Fernando Aguirre, Moshe Eizenberg, Fei Hui, and Mario Lanza. "A review on dielectric breakdown in thin dielectrics: silicon dioxide, high -k, and layered dielectrics." Advanced Functional Materials 30, no. 18 (2020): 1900657

    2020

  12. [12]

    Devine, Roderick A. B. The Physics and Technology of Amorphous SiO₂. Springer Science & Business Media, 2012

    2012

  13. [13]

    Neutron and X -ray diffraction of glass

    Cormier, Laurent. "Neutron and X -ray diffraction of glass." In Springer Handbook of Glass, pp. 1047 –1094. Cham: Springer International Publishing, 2019

    2019

  14. [14]

    Force fields for silicas and alumino phosphates based on ab initio calculations

    Van Beest, B. W. H., Gert Jan Kramer, and R. A. Van Santen. "Force fields for silicas and alumino phosphates based on ab initio calculations." Physical Review Letters 64, no. 16 (1990): 1955

    1990

  15. [15]

    Cooling-rate effects in amorphous silica: A computer-simulation study

    Vollmayr, Katharina, Walter Kob, and Kurt Binder. "Cooling-rate effects in amorphous silica: A computer-simulation study." Physical Review B 54, no. 22 (1996): 15808

    1996

  16. [16]

    Short- and medium -range orders in as -quenched and deformed SiO₂ glasses: An atomistic study

    Koziatek, P., J. L. Barrat, and D. Rodney. "Short- and medium -range orders in as -quenched and deformed SiO₂ glasses: An atomistic study." Journal of Non -Crystalline Solids 414 (2015): 7 – 15

    2015

  17. [17]

    On force fields for molecular dynamics simulations of crystalline silica

    Cowen, Benjamin J., and Mohamed S. El - Genk. "On force fields for molecular dynamics simulations of crystalline silica." Computational Materials Science 107 (2015): 88–101

    2015

  18. [18]

    New fitting scheme to obtain effective potential from Car -Parrinello molecular-dynamics simulations: Application to silica

    Carré, Antoine, Juergen Horbach, Simona Ispas, and Walter Kob. "New fitting scheme to obtain effective potential from Car -Parrinello molecular-dynamics simulations: Application to silica." EPL (Europhysics Letters) 82, no. 1 (2008): 17001

    2008

  19. [19]

    An ab initio parametrized interatomic force field for 16 silica

    Tangney, Paul, and Sandro Scandolo. "An ab initio parametrized interatomic force field for 16 silica." The Journal of Chemical Physics 117, no. 19 (2002): 8898-8904

    2002

  20. [20]

    Using an extended Tersoff interatomic potential to analyze the static -fatigue strength of SiO₂ under atmospheric influence

    Yasukawa, Akio. "Using an extended Tersoff interatomic potential to analyze the static -fatigue strength of SiO₂ under atmospheric influence." JSME International Journal Series A 39, no. 3 (1996): 313–320

    1996

  21. [21]

    Charge optimized many -body potential for the Si/SiO₂ system

    Yu, Jianguo, Susan B. Sinnott, and Simon R. Phillpot. "Charge optimized many -body potential for the Si/SiO₂ system." Physical Review B 75, no. 8 (2007): 085311

    2007

  22. [22]

    Second -generation charge- optimized many -body potential for Si/SiO₂ and amorphous silica

    Shan, Tzu-Ray, Bryce D. Devine, Jeffery M. Hawkins, Aravind Asthagiri, Simon R. Phillpot, and Susan R. Sinnott. "Second -generation charge- optimized many -body potential for Si/SiO₂ and amorphous silica." Physical Review B 82, no. 23 (2010): 235302

    2010

  23. [23]

    Interaction potential for SiO₂: A molecular-dynamics study of structural correlations

    Vashishta, P., Rajiv K. Kalia, José P. Rino, and Ingvar Ebbsjö. "Interaction potential for SiO₂: A molecular-dynamics study of structural correlations." Physical Review B 41, no. 17 (1990): 12197

    1990

  24. [24]

    First sharp diffraction peak and intermediate-range order in amorphous silica: finite-size effects in molecular dynamics simulations

    Nakano, Aiichiro, Rajiv K. Kalia, and Priya Vashishta. "First sharp diffraction peak and intermediate-range order in amorphous silica: finite-size effects in molecular dynamics simulations." Journal of Non -Crystalline Solids 171, no. 2 (1994): 157–163

    1994

  25. [25]

    Direct atomistic simulation of quartz crystal oscillators: Bulk properties and nanoscale devices

    Broughton, Jeremy Q., Christopher A. Meli, Priya Vashishta, and Rajiv K. Kalia. "Direct atomistic simulation of quartz crystal oscillators: Bulk properties and nanoscale devices." Physical Review B 56, no. 2 (1997): 611

    1997

  26. [26]

    Interatomic potential for Si –O systems using Tersoff parameterization

    Munetoh, Shinji, Teruaki Motooka, Koji Moriguchi, and Akira Shintani. "Interatomic potential for Si –O systems using Tersoff parameterization." Computational Materials Science 39, no. 2 (2007): 334–339

    2007

  27. [27]

    Comparison of model potentials for molecular-dynamics simulations of silica

    Herzbach, Daniel, Kurt Binder, and Martin H. Müser. "Comparison of model potentials for molecular-dynamics simulations of silica." The Journal of Chemical Physics 123, no. 12 (2005)

    2005

  28. [28]

    Neural network potentials

    Zeng, Jinzhe, Liqun Cao, and Tong Zhu. "Neural network potentials." In Quantum Chemistry in the Age of Machine Learning , pp. 279–294. Elsevier, 2023

    2023

  29. [29]

    Gaussian approximation potentials: The accuracy of quantum mechanics, without the electrons

    Bartók, Albert P., Mike C. Payne, Risi Kondor, and Gábor Csányi. "Gaussian approximation potentials: The accuracy of quantum mechanics, without the electrons." Physical Review Letters 104, no. 13 (2010): 136403

    2010

  30. [30]

    Moment tensor potentials: A class of systematically improvable interatomic potentials

    Shapeev, Alexander V. "Moment tensor potentials: A class of systematically improvable interatomic potentials." Multiscale Modeling & Simulation 14, no. 3 (2016): 1153–1173

    2016

  31. [31]

    Deep potential molecular dynamics: a scalable model with the accuracy of quantum mechanics

    Zhang, Linfeng, Jiequn Han, Han Wang, Roberto Car, and Weinan E. "Deep potential molecular dynamics: a scalable model with the accuracy of quantum mechanics." Physical Review Letters 120, no. 14 (2018): 143001

    2018

  32. [32]

    MACE: Higher order equivariant message passing neural 17 networks for fast and accurate force fields

    Batatia, Ilyes, David P. Kovacs, Gregor Simm, Christoph Ortner, and Gábor Csányi. "MACE: Higher order equivariant message passing neural 17 networks for fast and accurate force fields." Advances in Neural Information Processing Systems 35 (2022): 11423–11436

    2022

  33. [33]

    Long -Range Machine Learning Interatomic Potentials via Reciprocal - Space Gated Attention

    Thakur, Atul C., Alvaro Vazquez-Mayagoitia, and Ganesh Sivaraman. "Long -Range Machine Learning Interatomic Potentials via Reciprocal - Space Gated Attention." (In Preparation)

  34. [34]

    On the Atomic Cluster Expansion: interatomic potentials and beyond

    Ortner, Christoph. "On the Atomic Cluster Expansion: interatomic potentials and beyond." arXiv preprint arXiv:2308.06462 (2023)

    work page arXiv 2023

  35. [35]

    Evaluation of the MACE force field architecture: From medicinal chemistry to materials science

    Kovács, Dávid Péter, Ilyes Batatia, Eszter Sára Arany, and Gábor Csányi. "Evaluation of the MACE force field architecture: From medicinal chemistry to materials science." The Journal of Chemical Physics 159, no. 4 (2023)

    2023

  36. [36]

    A foundation model for atomistic materials chemistry

    Batatia, I., P. Benner, Y. Chiang, A. M. Elena, D. P. Kovács, J. Riebesell, X. R. Advincula et al. "A foundation model for atomistic materials chemistry." arXiv preprint arXiv:2401.00096 (2024)

    work page internal anchor Pith review arXiv 2024

  37. [37]

    A machine -learned interatomic potential for silica and its relation to empirical models

    Erhard, Linus C., Jochen Rohrer, Karsten Albe, and Volker L. Deringer. "A machine -learned interatomic potential for silica and its relation to empirical models." npj Computational Materials 8, no. 1 (2022): 90

    2022

  38. [38]

    Strongly constrained and appropriately normed semilocal density functional

    Sun, Jianwei, Adrienn Ruzsinszky, and John P. Perdew. "Strongly constrained and appropriately normed semilocal density functional." Physical Review Letters 115, no. 3 (2015): 036402

    2015

  39. [39]

    The atomic simulation environment-a Python library for working with atoms

    Hjorth Larsen, Ask, Jens Jørgen Mortensen, Jakob Blomqvist, Ivano E. Castelli, Rune Christensen, Marcin Dułak, Jesper Friis et al. "The atomic simulation environment-a Python library for working with atoms." Journal of Physics: Condensed Matter 29, no. 27 (2017): 273002

    2017

  40. [40]

    Nosé –Hoover chains: The canonical ensemble via continuous dynamics

    Martyna, Glenn J., Michael L. Klein, and Mark Tuckerman. "Nosé –Hoover chains: The canonical ensemble via continuous dynamics." The Journal of Chemical Physics 97, no. 4 (1992): 2635–2643

    1992

  41. [41]

    Polymorphic transitions in single crystals: A new molecular dynamics method

    Parrinello, Michele, and Aneesur Rahman. "Polymorphic transitions in single crystals: A new molecular dynamics method." Journal of Applied Physics 52, no. 12 (1981): 7182–7190

    1981

  42. [42]

    Ring statistics analysis of topological networks: New approach and application to amorphous GeS₂ and SiO₂ systems

    Le Roux, Sébastien, and Philippe Jund. "Ring statistics analysis of topological networks: New approach and application to amorphous GeS₂ and SiO₂ systems." Computational Materials Science 49, no. 1 (2010): 70–83

    2010

  43. [43]

    Ring structure of the crystalline and amorphous forms of silicon dioxide

    Guttman, Lester. "Ring structure of the crystalline and amorphous forms of silicon dioxide." Journal of Non -Crystalline Solids 116 (1990)

    1990

  44. [44]

    First -principles simulations of liquid silica: Structural and dynamical behavior at high pressure

    Karki, Bijaya B., Dipesh Bhattarai, and Lars Stixrude. "First -principles simulations of liquid silica: Structural and dynamical behavior at high pressure." Physical Review B 76, no. 10 (2007): 104205

    2007

  45. [45]

    Revealing hidden medium-range order in silicate glass formers using many -body correlation functions

    Zhang, Zhen, and Walter Kob. "Revealing hidden medium-range order in silicate glass formers using many -body correlation functions." Physical Review B 110, no. 10 (2024): 104203

    2024

  46. [46]

    Machine 18 learning molecular dynamics reveals the structural origin of the first sharp diffraction peak in high - density silica glasses

    Kobayashi, Keita, Masahiko Okumura, Hiroki Nakamura, Mitsuhiro Itakura, Masahiko Machida, Shingo Urata, and Kentaro Suzuya. "Machine 18 learning molecular dynamics reveals the structural origin of the first sharp diffraction peak in high - density silica glasses." Scientific Reports 13, no. 1 (2023): 18721

    2023

  47. [47]

    Revealing the medium-range structure of glassy silica using force -enhanced atomic refinement

    Zhou, Qi, Ying Shi, Binghui Deng, Tao Du, Lijie Guo, Morten M. Smedskjaer, and Mathieu Bauchy. "Revealing the medium-range structure of glassy silica using force -enhanced atomic refinement." Journal of Non-Crystalline Solids 573 (2021): 121138

    2021

  48. [48]

    A time resolved high energy X -ray diffraction study of cooling liquid SiO 2

    Skinner, L. B., C. J. Benmore, J. K. R. Weber, M. C. Wilding, S. K. Tumber, and J. B. Parise. "A time resolved high energy X -ray diffraction study of cooling liquid SiO 2." Physical Chemistry Chemical Physics 15, no. 22 (2013): 8566-8572

    2013

  49. [49]

    Machine learning interatomic potentials and long- range physics

    Anstine, Dylan M., and Olexandr Isayev. "Machine learning interatomic potentials and long- range physics." The Journal of Physical Chemistry A 127, no. 11 (2023): 2417-2431

    2023

  50. [50]

    Machine -learned interatomic potentials by active learning: amorphous and liquid hafnium dioxide

    Sivaraman, Ganesh, et al. "Machine -learned interatomic potentials by active learning: amorphous and liquid hafnium dioxide." npj Computational Materials 6.1 (2020): 104

    2020

  51. [51]

    A combined machine learning and high-energy X-ray diffraction approach to understanding liquid and amorphous metal oxides

    Sivaraman, Ganesh, et al. "A combined machine learning and high-energy X-ray diffraction approach to understanding liquid and amorphous metal oxides." Journal of the Physical Society of Japan 91.9 (2022): 091009

    2022

  52. [52]

    Revisiting machine learning potentials for silicate glasses: The missing role of dispersion interactions

    Pedone, Alfonso, Marco Bertani, and Matilde Benassi. "Revisiting machine learning potentials for silicate glasses: The missing role of dispersion interactions." Journal of Chemical Theory and Computation 21, no. 9 (2025): 4769-4778

    2025

  53. [53]

    Hierarchical structures of amorphous solids characterized by persistent homology

    Hiraoka, Yasuaki, Takenobu Nakamura, Akihiko Hirata, Emerson G. Escolar, Kaname Matsue, and Yasumasa Nishiura. "Hierarchical structures of amorphous solids characterized by persistent homology." Proceedings of the National Academy of Sciences 113, no. 26 (2016): 7035 - 7040

    2016

  54. [54]

    Persistent homology analysis for materials research and persistent homology software: HomCloud

    Obayashi, Ippei, Takenobu Nakamura, and Yasuaki Hiraoka. "Persistent homology analysis for materials research and persistent homology software: HomCloud." Journal of the physical society of japan 91, no. 9 (2022): 091013

    2022

  55. [55]

    Intermediate-range order in vitreous SiO2 and GeO2

    Kohara, Shinji, and Kentaro Suzuya. "Intermediate-range order in vitreous SiO2 and GeO2." Journal of Physics: Condensed Matter 17, no. 5 (2005): S77-S86

    2005

  56. [56]

    Area detector corrections for high quality synchrotron X -ray structure factor measurements

    Skinner, Lawrie B., Chris J. Benmore, and John B. Parise. "Area detector corrections for high quality synchrotron X -ray structure factor measurements." Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 662, no. 1 (2012): 61-70

    2012

  57. [57]

    The nanoscale ordered materials diffractometer NOMAD at the spallation neutron source 19 SNS

    Neuefeind, Jörg, Mikhail Feygenson, John Carruth, Ron Hoffmann, and Kenneth K. Chipley. "The nanoscale ordered materials diffractometer NOMAD at the spallation neutron source 19 SNS." Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms 287 (2012): 68-75

    2012

  58. [58]

    Structure of liquid sio 2: a measurement by high energy X-ray diffraction

    Mei, Q., C. J. Benmore, and J. K. R. Weber. "Structure of liquid sio 2: a measurement by high energy X-ray diffraction." Physical Review Letters 98, no. 5 (2007): 057802

    2007

  59. [59]

    Origin of the first sharp diffraction peak in the structure factor of covalent glasses

    Elliott, S. R. "Origin of the first sharp diffraction peak in the structure factor of covalent glasses." Physical Review Letters 67, no. 6 (1991): 711

    1991

  60. [60]

    Formation of glasses from liquids and biopolymers

    Angell, C. Austen. "Formation of glasses from liquids and biopolymers." Science 267, no. 5206 (1995): 1924-1935

    1995

  61. [61]

    Intermediate range order in vitreous silica from a partial structure factor analysis

    Mei, Q., C. J. Benmore, S. Sen, R. Sharma, and J. L. Yarger. "Intermediate range order in vitreous silica from a partial structure factor analysis." Physical Review B —Condensed Matter and Materials Physics 78, no. 14 (2008): 144204

    2008

  62. [62]

    RSGA: Reciprocal Space Gated Attention

    Thakur, Atul C., Alvaro Vazquez -Mayagoitia, and Ganesh Sivaraman. "RSGA: Reciprocal Space Gated Attention" . Zenodo, April 20, 2026. https://doi.org/10.5281/zenodo.19673766

    work page doi:10.5281/zenodo.19673766 2026

  63. [63]

    Center for Computational Research, University at Buffalo, http://hdl.handle.net/10477/79221