pith. sign in

arxiv: 2605.05361 · v1 · submitted 2026-05-06 · ❄️ cond-mat.mtrl-sci · cond-mat.dis-nn

From Bulk to Surface: Structure and Dynamics of Amorphous Alumina from Deep Potential Molecular Dynamics

Zheng Yu , Jiayan Xu , Abhirup Patra , Sharan Shetty , Detlef Hohl , Roberto Car This is my paper

Pith reviewed 2026-05-08 16:10 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci cond-mat.dis-nn
keywords amorphous aluminaaluminum oxidesurface structuremolecular dynamicsglass transitionrelaxation dynamicsoxide surfacesdeep potential
0
0 comments X

The pith

Amorphous alumina surfaces show structural differences from the bulk but relax on the same timescale and share a comparable glass transition temperature.

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

This paper uses Deep Potential molecular dynamics to build large models of amorphous aluminum oxide glasses both in the bulk and with free surfaces. It finds that the surface layer is oxygen-rich with more under-coordinated atoms and altered bonding patterns, yet these features do not change how quickly the structure relaxes or where the glass transition occurs. A reader would care because the results indicate the disordered surface stays kinetically stable once formed, which bears on how such materials hold up in coatings, catalysts, or interfaces. The simulations also correct errors in older classical models and give statistical detail on surface sites that experiments cannot easily access.

Core claim

Using Deep Potential molecular dynamics, large-scale models of amorphous Al2O3 are generated that reproduce experimental liquid and glass structure. At the free surface, density recovers to bulk values over roughly 10 angstroms while coordination converges over a slightly wider region. The outermost layer is oxygen-enriched with contracted Al-O bonds and hosts under-coordinated motifs such as AlO3 and OAl2 whose populations depend on glass stability. These motifs pair locally to satisfy bond-valence rules but stay dispersed rather than clustered. Despite the structural heterogeneity, surface relaxation occurs on the same timescale as the bulk and the glass transition temperature is similar,,

What carries the argument

Deep Potential molecular dynamics simulations of melt-quenched bulk glasses and free surfaces that resolve coordination statistics, density profiles, and relaxation dynamics with ab initio-level accuracy on large systems.

If this is right

  • Mass density returns to bulk values within about 10 angstroms of the surface while local coordination needs a wider subsurface zone.
  • The outermost layer contains dispersed reactive under-coordinated sites that remain stable once the surface forms.
  • Polyhedral populations in the bulk glass match experiment better than those from common classical force fields.
  • The surface glass transition temperature is comparable to the bulk, supporting kinetic stability of the disordered layer.
  • Structural heterogeneity at the surface does not produce faster relaxation or lower stability than the interior.

Where Pith is reading between the lines

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

  • The kinetic stability could allow surface reactivity to persist in applications such as thin-film coatings without rapid annealing or restructuring.
  • The same simulation approach could be applied to study how amorphous alumina interfaces with metals or other oxides behave under thermal cycling.
  • Surface-sensitive measurements like certain forms of spectroscopy or atomic-force techniques could be used to check the predicted similarity in relaxation timescales.

Load-bearing premise

The Deep Potential model trained mainly on bulk configurations still gives accurate results for the under-coordinated atoms and dynamics at the free surface without large extrapolation errors.

What would settle it

An experiment that directly measures surface relaxation times or the glass transition temperature at a free amorphous alumina surface and finds them different from the bulk values, or an independent simulation method that yields mismatched surface dynamics.

Figures

Figures reproduced from arXiv: 2605.05361 by Abhirup Patra, Detlef Hohl, Jiayan Xu, Roberto Car, Sharan Shetty, Zheng Yu.

Figure 1
Figure 1. Figure 1: FIG. 1. Structural characterization of liquid Al view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Glass transition of liquid Al view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Atomic structure and depth-dependent profiles of a melt view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Local bonding geometry in surface versus bulk-like regions view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Coordination environments at the a-Al view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Dependence of surface coordination environments on cool view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Comparison of surface and bulk dynamics. (a) Tempera view at source ↗
read the original abstract

Understanding the atomic-scale structure and dynamics of amorphous oxide surfaces is essential for interpreting their chemical reactivity, mechanical stability, and interfacial behavior, yet direct experimental characterization remains challenging. We employ Deep Potential (DP) molecular dynamics to generate large-scale, ab initio-quality models of amorphous Al$_2$O$_3$ bulk glasses and melt-quenched free surfaces, enabling a quantitative analysis of both structure and relaxation dynamics with statistical confidence inaccessible to direct ab initio simulation. The trained DP model reproduces experimental liquid and glass structure, captures the cooling-rate dependence of the bulk glass transition, and corrects systematic biases in the polyhedral populations predicted by widely used classical force fields. At the free surface, mass density recovers to bulk values over ~10 $\unicode{x212B}$, while local coordination requires a slightly wider subsurface region to fully converge. The outermost layer is oxygen-enriched, exhibits altered polyhedral connectivity with contracted Al-O bonds, and hosts a broad population of under-coordinated motifs (notably AlO$_3$ and OAl$_2$) whose abundances are governed by glass stability. These reactive Lewis acid and Br$\unicode{x00F8}$nsted base sites are locally paired in a manner consistent with bond-valence compensation, yet remain spatially dispersed rather than aggregating into extended clusters. Despite this pronounced structural heterogeneity, the surface relaxes on the same timescale as the bulk and exhibits a comparable glass transition temperature, suggesting that the disordered surface is kinetically stable once formed. Together, these results establish a molecular-level picture of amorphous alumina surfaces and demonstrate the capability of machine-learned potentials to resolve structure-property relationships in disordered oxide interfaces.

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 employs Deep Potential (DP) molecular dynamics simulations to construct large-scale models of amorphous Al2O3 bulk glasses and melt-quenched free surfaces. It asserts that the DP model reproduces experimental liquid/glass structures, captures cooling-rate dependence of the bulk glass transition, corrects biases in classical force-field polyhedral populations, and reveals surface structural features including oxygen enrichment, contracted Al-O bonds, and dispersed under-coordinated AlO3/OAl2 motifs. The central result is that, despite this heterogeneity, the surface relaxes on the same timescale as the bulk and exhibits a comparable glass transition temperature, implying kinetic stability of the disordered surface.

Significance. If the surface accuracy holds, the work supplies statistically robust atomic-scale details on amorphous alumina interfaces that are difficult to access experimentally, with implications for reactivity, mechanical properties, and applications in coatings or catalysts. The demonstration of DP enabling ab initio-quality large-scale dynamics in disordered oxides, including explicit correction of classical force-field errors, is a methodological strength that could be extended to other oxide systems.

major comments (2)
  1. [Abstract and DP model validation section] Abstract and the section on DP model training/validation: The headline claim that surface relaxation occurs on the same timescale as bulk (and yields comparable Tg) is load-bearing for the conclusion of kinetic stability. The DP model is described as trained primarily on bulk configurations and shown to reproduce bulk experimental structure, but no direct ab initio validation is provided for surface-specific quantities such as coordination populations, Al-O bond lengths, or mean-squared displacements at under-coordinated AlO3/OAl2 sites that dominate the outermost layer. Without this, the reported surface dynamics risk being artifacts of extrapolation rather than physical behavior.
  2. [Results on surface structure and dynamics] Results section on surface structure and dynamics: The statement that mass density recovers to bulk values over ~10 Å while local coordination requires a wider subsurface region is central to interpreting surface heterogeneity. However, the manuscript does not report quantitative convergence metrics or error bars on these length scales, nor does it compare the surface coordination statistics or relaxation functions directly to ab initio reference data, leaving the extent of any extrapolation error unquantified.
minor comments (2)
  1. [Abstract] The unicode symbols for Ångstroms and ø in the abstract reduce readability; standard LaTeX notation (e.g., Å) should be used throughout.
  2. [Methods] Training-set composition, size, and any overlap with prior DP studies on alumina are not detailed, which affects reproducibility and assessment of novelty.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful and constructive review. We address each major comment below, indicating the revisions we plan to incorporate.

read point-by-point responses

  1. Referee: [Abstract and DP model validation section] Abstract and the section on DP model training/validation: The headline claim that surface relaxation occurs on the same timescale as bulk (and yields comparable Tg) is load-bearing for the conclusion of kinetic stability. The DP model is described as trained primarily on bulk configurations and shown to reproduce bulk experimental structure, but no direct ab initio validation is provided for surface-specific quantities such as coordination populations, Al-O bond lengths, or mean-squared displacements at under-coordinated AlO3/OAl2 sites that dominate the outermost layer. Without this, the reported surface dynamics risk being artifacts of extrapolation rather than physical behavior.

    Authors: We agree that the absence of direct ab initio validation for surface-specific quantities represents a limitation in the current manuscript. The DP model was trained exclusively on bulk AIMD data, and while it accurately reproduces experimental bulk structures, cooling-rate dependence of Tg, and corrects classical force-field biases, its application to surfaces involves extrapolation. In the revised manuscript we will add an explicit discussion of transferability, including any available cross-checks against smaller ab initio surface calculations from the literature, and we will qualify the surface-dynamics claims with appropriate caveats regarding possible extrapolation error. revision: partial

  2. Referee: [Results on surface structure and dynamics] Results section on surface structure and dynamics: The statement that mass density recovers to bulk values over ~10 Å while local coordination requires a wider subsurface region is central to interpreting surface heterogeneity. However, the manuscript does not report quantitative convergence metrics or error bars on these length scales, nor does it compare the surface coordination statistics or relaxation functions directly to ab initio reference data, leaving the extent of any extrapolation error unquantified.

    Authors: We accept this criticism. In the revised version we will add error bars to the density and coordination profiles (computed from multiple independent runs) and provide quantitative measures of the convergence length scales with statistical uncertainties. Direct ab initio comparisons for surface coordination and relaxation functions remain computationally prohibitive for the system sizes used; we will therefore include a brief discussion of the expected error magnitude based on the bulk validation metrics and note this as an area for future work. revision: partial

Circularity Check

0 steps flagged

No significant circularity in the derivation chain

full rationale

The paper trains a Deep Potential model on ab initio bulk configurations, then performs MD simulations to obtain structure and dynamics for both bulk glass and melt-quenched free surfaces. All reported quantities (density profiles, coordination populations, relaxation timescales, Tg) are direct outputs of these trajectories, compared against external experimental data and classical force-field results. No load-bearing step reduces by construction to a fitted input renamed as prediction, a self-definitional loop, or a self-citation chain; the central claim of comparable surface and bulk relaxation emerges from the simulated trajectories themselves rather than being presupposed by the training procedure or prior author work.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

Central claims rest on the fidelity of the fitted Deep Potential model to the true potential energy surface and on the assumption that melt-quench free surfaces represent physical amorphous alumina.

free parameters (1)
  • Deep Potential neural network parameters
    Weights and biases fitted to ab initio training data; accuracy for surface motifs depends on this fit.
axioms (2)
  • standard math Born-Oppenheimer approximation for atomic interactions in Al2O3
    Standard assumption underlying all classical and ML molecular dynamics.
  • domain assumption Melt-quench protocol generates representative amorphous structures and surfaces
    Common in glass simulations but may not capture all real-world surface formation pathways.

pith-pipeline@v0.9.0 · 5618 in / 1450 out tokens · 83774 ms · 2026-05-08T16:10:46.717947+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

77 extracted references · 77 canonical work pages

  1. [1]

    G. D. Wilk, R. M. Wallace, and J. M. Anthony, High- κ gate dielectrics: Current status and materials properties considera- tions, J. Appl. Phys. 89, 5243 (2001)

    work page 2001

  2. [2]

    Katiyar, C

    P . Katiyar, C. Jin, and R. Narayan, Electrical properties of amor- phous aluminum oxide thin films, Acta Materialia 53, 2617 (2005)

    work page 2005

  3. [3]

    M. D. Groner, F. H. Fabreguette, J. W. Elam, and S. M. George, Low-Temperature Al2O3 Atomic Layer Deposition, Chem. Mater. 16, 639 (2004)

    work page 2004

  4. [4]

    S. M. George, Atomic Layer Deposition: An Overview, Chem. Rev. 110, 111 (2010)

    work page 2010

  5. [5]

    P . O. Oviroh, R. Akbarzadeh, D. Pan, R. A. M. Coet- zee, and T.-C. Jen, New development of atomic layer de- position: processes, methods and applications, Science and Technology of Advanced Materials 20, 465 (2019), _eprint: https://doi.org/10.1080/14686996.2019.1599694

    work page doi:10.1080/14686996.2019.1599694 2019

  6. [6]

    M. J. Y oung, N. M. Bedford, A. Y anguas-Gil, S. Letourneau, M. Coile, D. J. Mandia, B. Aoun, A. S. Cavanagh, S. M. George, and J. W. Elam, Probing the Atomic-Scale Structure of Amorphous Aluminum Oxide Grown by Atomic Layer De- position, ACS Appl. Mater. Interfaces 12, 22804 (2020)

    work page 2020

  7. [7]

    B. R. Goldsmith, B. Peters, J. K. Johnson, B. C. Gates, and S. L. Scott, Beyond Ordered Materials: Understanding Cat- alytic Sites on Amorphous Solids, ACS Catal. 7, 7543 (2017)

    work page 2017

  8. [8]

    Lamparter and R

    P . Lamparter and R. Kniep, Structure of amorphous Al2O3, Physica B: Condensed Matter 234–236, 405 (1997)

    work page 1997

  9. [9]

    L. B. Skinner, A. C. Barnes, P . S. Salmon, L. Hennet, H. E. Fis- cher, C. J. Benmore, S. Kohara, J. K. R. Weber, A. Bytchkov, M. C. Wilding, J. B. Parise, T. O. Farmer, I. Pozdnyakova, S. K. Tumber, and K. Ohara, Joint diffraction and modeling approach to the structure of liquid alumina, Phys. Rev. B 87, 024201 (2013)

    work page 2013

  10. [11]

    Hashimoto, Y

    H. Hashimoto, Y . Onodera, S. Tahara, S. Kohara, K. Y azawa, H. Segawa, M. Murakami, and K. Ohara, Structure of alumina glass, Sci Rep 12, 516 (2022)

    work page 2022

  11. [12]

    A. F. Harper, S. P . Emge, P . C. Magusin, C. P . Grey, and A. J. Morris, Modelling amorphous materials via a joint solid-state nmr and x-ray absorption spectroscopy and dft approach: ap- plication to alumina, Chem. Sci. 14, 1155 (2023)

    work page 2023

  12. [13]

    A. F. Harper, K. Iwanowski, W. C. Witt, M. C. Payne, and M. Simoncelli, Vibrational and thermal properties of amor- phous alumina from first principles, Phys. Rev. Mater. 8, 043601 (2024)

    work page 2024

  13. [14]

    A. H. Tavakoli, P . S. Maram, S. J. Widgeon, J. Rufner, K. V an Benthem, S. Ushakov, S. Sen, and A. Navrotsky, Amor- phous alumina nanoparticles: structure, surface energy, and thermodynamic phase stability, The Journal of Physical Chem- istry C 117, 17123 (2013)

    work page 2013

  14. [15]

    S. K. Lee, S. B. Lee, S. Y . Park, Y . S. Yi, and C. W. Ahn, Struc- ture of Amorphous Aluminum Oxide, Phys. Rev. Lett. 103, 095501 (2009)

    work page 2009

  15. [16]

    S. K. Lee, S. Y . Park, Y . S. Yi, and J. Moon, Structure and Dis- order in Amorphous Alumina Thin Films: Insights from High- Resolution Solid-State NMR, J. Phys. Chem. C 114, 13890 (2010)

    work page 2010

  16. [17]

    Sarou-Kanian, A

    V . Sarou-Kanian, A. N. Gleizes, P . Florian, D. Samélor, D. Mas- siot, and C. V ahlas, Temperature dependent 4-, 5- and 6-fold coordination of aluminum in MOCVD-grown amorphous alu- mina films: A very high field 27Al-NMR study, The Journal of Physical Chemistry C 117, 21965 (2013)

    work page 2013

  17. [18]

    Hashimoto, K

    H. Hashimoto, K. Y azawa, H. Asoh, and S. Ono, NMR Spectro- scopic Analysis of the Local Structure of Porous-Type Amor- phous Alumina Prepared by Anodization, J. Phys. Chem. C 121, 12300 (2017)

    work page 2017

  18. [19]

    Baggetto, V

    L. Baggetto, V . Sarou-Kanian, P . Florian, A. N. Gleizes, D. Massiot, and C. V ahlas, Atomic scale structure of amorphous aluminum oxyhydroxide, oxide and oxycarbide films probed by very high field 27Al nuclear magnetic resonance, Physical Chemistry Chemical Physics 19, 7648 (2017)

    work page 2017

  19. [20]

    G. Busca, The surface of transitional aluminas: A critical re- view, Catalysis Today Acid-Base Catalysis Advanced Sciences and Spreading Applications to Solutions of Environmental, Re- sources and Energy Issues: ABC-7, 7th International Sympo- sium on Acid-Base Catalysis, Tokyo, May 12-15, 2013, 226, 2 (2014)

    work page 2013

  20. [21]

    Z. Zhao, D. Xiao, K. Chen, R. Wang, L. Liang, Z. Liu, I. Hung, Z. Gan, and G. Hou, Nature of Five-Coordinated Al in γ-Al2O3 Revealed by Ultra-High-Field Solid-State NMR, ACS Cent. Sci. 8, 795 (2022)

    work page 2022

  21. [22]

    Giacomazzi, N

    L. Giacomazzi, N. S. Shcheblanov, M. E. Povarnitsyn, Y . Li, A. Mavri, B. Zupani, J. Grdadolnik, and A. Pasquarello, In- frared spectra in amorphous alumina: A combined ab initio and experimental study, Phys. Rev. Mater. 7, 045604 (2023)

    work page 2023

  22. [23]

    Rakita, S

    Y . Rakita, S. Nikolov, K. Guzzetta, C. S. Thieu, J. Brgoch, S. Banerjee, and K. W. Chapman, Cryo-ePDF: Overcoming electron beam damage to study the local atomic structure of amorphous ALD aluminum oxide thin films within a TEM, ACS Omega 6, 8479 (2021)

    work page 2021

  23. [24]

    Friederich, F

    P . Friederich, F. Häse, J. Proppe, and A. Aspuru-Guzik, Machine-learned potentials for next-generation matter simula- tions, Nature Materials 20, 750 (2021)

    work page 2021

  24. [25]

    V ollmayr, W

    K. V ollmayr, W. Kob, and K. Binder, Cooling-rate effects in amorphous silica: A computer-simulation study, Physical Re- view B 54, 15808 (1996)

    work page 1996

  25. [26]

    Z. Y u, Q. Liu, I. Szlufarska, and B. Wang, Structural signatures for thermodynamic stability in vitreous silica: Insight from ma- chine learning and molecular dynamics simulations, Physical Review Materials 5, 015602 (2021)

    work page 2021

  26. [27]

    Gutiérrez and B

    G. Gutiérrez and B. Johansson, Molecular dynam- ics study of structural properties of amorphous ${\mathrm{Al}}_{2}{\mathrm{O}}_{3}$, Phys. Rev. B 65, 104202 (2002)

    work page 2002

  27. [28]

    S. P . Adiga, P . Zapol, and L. A. Curtiss, Atomistic simula- tions of amorphous alumina surfaces, Phys. Rev. B 74, 064204 (2006)

    work page 2006

  28. [29]

    H. M. Carruzzo, A. Bilmes, J. Lisenfeld, Z. Y u, B. Wang, Z. Wan, J. Schmidt, and C. C. Y u, Distribution of two-level system couplings to strain and electric fields in glasses at low temperatures, Physical Review B 104, 134203 (2021)

    work page 2021

  29. [30]

    Zhang, M

    J. Zhang, M. De Meulder, E. J. Frankberg, and A. Kuronen, De- pendence between glass transition and plasticity in amorphous aluminum oxide: A molecular dynamics study, Journal of Non- Crystalline Solids 628, 122840 (2024)

    work page 2024

  30. [31]

    J.-S. Lee, J. Ji, U. Jeong, and B.-J. Lee, An atomistic simulation study on ductility of amorphous aluminum oxide, Acta Materi- alia 274, 119985 (2024)

    work page 2024

  31. [32]

    Matsui, A transferable interatomic potential model for crys- tals and melts in the system CaO-MgO-Al2O3-SiO2, Miner- 13 alogical Magazine 58, 571 (1994)

    M. Matsui, A transferable interatomic potential model for crys- tals and melts in the system CaO-MgO-Al2O3-SiO2, Miner- 13 alogical Magazine 58, 571 (1994)

    work page 1994

  32. [33]

    Car and M

    R. Car and M. Parrinello, Unified approach for molecular dy- namics and density-functional theory, Physical review letters 55, 2471 (1985)

    work page 1985

  33. [34]

    Car and M

    R. Car and M. Parrinello, Structural, dymanical, and elec- tronic properties of amorphous silicon: an ab initio molecular- dynamics study, Physical review letters 60, 204 (1988)

    work page 1988

  34. [35]

    Sarnthein, A

    J. Sarnthein, A. Pasquarello, and R. Car, Structural and elec- tronic properties of liquid and amorphous si o 2: An ab ini- tio molecular dynamics study, Physical review letters 74, 4682 (1995)

    work page 1995

  35. [36]

    Lizárraga, E

    R. Lizárraga, E. Holmström, S. C. Parker, and C. Arrouvel, Structural characterization of amorphous alumina and its poly- morphs from first-principles XPS and NMR calculations, Phys- ical Review B 83, 094201 (2011)

    work page 2011

  36. [37]

    Sivaraman, A

    G. Sivaraman, A. N. Krishnamoorthy, M. Baur, C. Holm, M. Stan, G. Csányi, C. Benmore, and A. Vázquez-Mayagoitia, Machine-learned interatomic potentials by active learning: amorphous and liquid hafnium dioxide, npj Comput. Mater. 6, 104 (2020)

    work page 2020

  37. [38]

    Davis and G

    S. Davis and G. Gutiérrez, Structural, elastic, vibrational and electronic properties of amorphous al2o3 from ab initio cal- culations, Journal of Physics: Condensed Matter 23, 495401 (2011)

    work page 2011

  38. [39]

    Zhang, W

    Z. Zhang, W. Kob, and S. Ispas, First-principles study of the surface of silica and sodium silicate glasses, Physical Review B 103, 184201 (2021)

    work page 2021

  39. [40]

    Behler, Four Generations of High-Dimensional Neural Net- work Potentials, Chem

    J. Behler, Four Generations of High-Dimensional Neural Net- work Potentials, Chem. Rev. 121, 10037 (2021)

    work page 2021

  40. [41]

    V . L. Deringer, M. A. Caro, and G. Csányi, Machine Learning Interatomic Potentials as Emerging Tools for Materials Science, Adv. Mater. 31, 1902765 (2019)

    work page 2019

  41. [42]

    Xu, X.-M

    J. Xu, X.-M. Cao, and P . Hu, Perspective on computational re- action prediction using machine learning methods in heteroge- neous catalysis, Phys. Chem. Chem. Phys. 23, 11155 (2021)

    work page 2021

  42. [43]

    Cheng, C

    X. Cheng, C. Wu, J. Xu, Y . Han, W. Xie, and P . Hu, Leveraging machine learning potentials for in-situ searching of active sites in heterogeneous catalysis, Precis. Chem. 2, 570 (2024)

    work page 2024

  43. [44]

    L. C. Erhard, J. Rohrer, K. Albe, and V . L. Deringer, A machine- learned interatomic potential for silica and its relation to empir- ical models, npj Computational Materials 8, 90 (2022)

    work page 2022

  44. [45]

    Rodríguez-Martínez, T

    C. Rodríguez-Martínez, T. Schwedek, E. Salazar, and X. Bokhimi, Utilizing Wyckoff Sites to Construct Machine- Learning-Driven Interatomic Potentials for Crystalline Materi- als: A Case Study on α-Alumina, J. Phys. Chem. C 128, 1746 (2024)

    work page 2024

  45. [46]

    Zhang, W

    L. Zhang, W. Luo, R. Liu, M. Chen, Z. Y an, and K. Cao, Explor- ing the energy landscape of aluminas through machine learning interatomic potential, Phys. Rev. Mater. 9, 023801 (2025)

    work page 2025

  46. [47]

    Y ang, C

    X. Y ang, C. Shang, and Z.-P . Liu, Resolving the atomic struc- ture of γ-alumina: a non-spinel phase with a distorted anion lattice and three adjacent long channels, J. Mater. Chem. A 10.1039/D5TA01715G (2025)

    work page doi:10.1039/d5ta01715g 2025

  47. [48]

    X. Du, W. Shao, C. Bao, L. Zhang, J. Cheng, and F. Tang, Re- vealing the molecular structures of -Al2O3(0001)water inter- face by machine learning based computational vibrational spec- troscopy, The Journal of Chemical Physics 161, 124702 (2024)

    work page 2024

  48. [49]

    W. Li, Y . Ando, and S. Watanabe, Effects of density and com- position on the properties of amorphous alumina: A high- dimensional neural network potential study, J. Chem. Phys. 153, 164119 (2020)

    work page 2020

  49. [50]

    Gramatte, O

    S. Gramatte, O. Politano, N. Jakse, C. Cancellieri, I. Utke, L. P . H. Jeurgens, and V . Turlo, Unveiling hydrogen chemi- cal states in supersaturated amorphous alumina via machine learning-driven atomistic modeling, npj Comput Mater 11, 170 (2025)

    work page 2025

  50. [51]

    Zhang, J

    L. Zhang, J. Han, H. Wang, R. Car, and W. E, Deep Potential Molecular Dynamics: A Scalable Model with the Accuracy of Quantum Mechanics, Phys. Rev. Lett. 120, 143001 (2018)

    work page 2018

  51. [52]

    C. S. Gorham, J. T. Gaskins, G. N. Parsons, M. D. Losego, and P . E. Hopkins, Density dependence of the room temperature thermal conductivity of atomic layer deposition-grown amor- phous alumina (al2o3), Applied Physics Letters 104 (2014)

    work page 2014

  52. [53]

    M. W. Chase et al. , Nist-janaf thermochemical tables, Journal of physical and chemical reference data 28, 1951 (1998)

    work page 1951

  53. [54]

    F. H. Stillinger and T. A. Weber, Computer simulation of local order in condensed phases of silicon, Physical Review B 31, 5262 (1985)

    work page 1985

  54. [55]

    P . G. Debenedetti and F. H. Stillinger, Supercooled liquids and the glass transition, Nature 410, 259 (2001)

    work page 2001

  55. [56]

    J. He, D. Avnir, and L. Zhang, Sol-gel derived alumina glass: Mechanistic study of its structural evolution, Acta Materialia 174, 418 (2019)

    work page 2019

  56. [57]

    Digne, P

    M. Digne, P . Sautet, P . Raybaud, P . Euzen, and H. Toulhoat, Use of dft to achieve a rational understanding of acid–basic proper- ties of γ-alumina surfaces, J. Catal. 226, 54 (2004)

    work page 2004

  57. [58]

    Gramatte, V

    S. Gramatte, V . Turlo, and O. Politano, Do we really need ma- chine learning interatomic potentials for modeling amorphous metal oxides? case study on amorphous alumina by recycling an existing ab initio database, Modelling and Simulation in Ma- terials Science and Engineering 32, 045010 (2024)

    work page 2024

  58. [59]

    Mavri, M

    A. Mavri, M. V alant, C. Cui, and Z. M. Wang, Advanced appli- cations of amorphous alumina: From nano to bulk, Journal of Non-Crystalline Solids 521, 119493 (2019)

    work page 2019

  59. [60]

    Y . Li, A. Annamareddy, D. Morgan, Z. Y u, B. Wang, C. Cao, J. H. Perepezko, M. Ediger, P . M. V oyles, and L. Y u, Surface diffusion is controlled by bulk fragility across all glass types, Physical Review Letters 128, 075501 (2022)

    work page 2022

  60. [61]

    S. F. Swallen, K. L. Kearns, M. K. Mapes, Y . S. Kim, R. J. McMahon, M. D. Ediger, T. Wu, L. Y u, and S. Satija, Organic glasses with exceptional thermodynamic and kinetic stability, Science 315, 353 (2007)

    work page 2007

  61. [62]

    Berthier, P

    L. Berthier, P . Charbonneau, E. Flenner, and F. Zamponi, Ori- gin of ultrastability in vapor-deposited glasses, Physical review letters 119, 188002 (2017)

    work page 2017

  62. [63]

    Annamareddy, P

    A. Annamareddy, P . M. V oyles, J. Perepezko, and D. Morgan, Mechanisms of bulk and surface diffusion in metallic glasses determined from molecular dynamics simulations, Acta Mate- rialia 209, 116794 (2021)

    work page 2021

  63. [64]

    H. Wang, L. Zhang, J. Han, and E. Weinan, DeePMD-kit: A deep learning package for many-body potential energy repre- sentation and molecular dynamics, Comput. Phys. Commun. 228, 178 (2018)

    work page 2018

  64. [65]

    J. Zeng, D. Zhang, D. Lu, P . Mo, Z. Li, Y . Chen, M. Rynik, L. Huang, Z. Li, S. Shi, et al. , DeePMD-kit v2: A software package for deep potential models, J. Chem. Phys. 159, 054801 (2023)

    work page 2023

  65. [66]

    Xu, GDPy: Generating deep potential with Python (2026 (Accessed Feburary 28, 2026))

    J. Xu, GDPy: Generating deep potential with Python (2026 (Accessed Feburary 28, 2026))

    work page 2026

  66. [67]

    Zhang, D.-Y

    L. Zhang, D.-Y . Lin, H. Wang, R. Car, and E. Weinan, Active learning of uniformly accurate interatomic potentials for mate- rials simulation, Phys. Rev. Mater. 3, 023804 (2019)

    work page 2019

  67. [68]

    Kresse and J

    G. Kresse and J. Furthmüller, Efficiency of ab-initio total en- ergy calculations for metals and semiconductors using a plane- wave basis set, Computational Materials Science 6, 15 (1996)

    work page 1996

  68. [69]

    H. J. Monkhorst and J. D. Pack, Special points for brillouin- zone integrations, Phys. Rev. B 13, 5188 (1976). 14

    work page 1976

  69. [70]

    A. P . Thompson, H. M. Aktulga, R. Berger, D. S. Bolintineanu, W. M. Brown, P . S. Crozier, P . J. i. t. V eld, A. Kohlmeyer, S. G. Moore, T. D. Nguyen, R. Shan, M. J. Stevens, J. Tranchida, C. Trott, and S. J. Plimpton, LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales, Computer Physics Com...

    work page 2022

  70. [71]

    G. J. Martyna, M. L. Klein, and M. Tuckerman, Nosé–hoover chains: The canonical ensemble via continuous dynamics, The Journal of chemical physics 97, 2635 (1992)

    work page 1992

  71. [72]

    A. P . Willard and D. Chandler, Instantaneous Liquid Interfaces, J. Phys. Chem. B 114, 1954 (2010)

    work page 1954

  72. [73]

    From Bulk to Surface: Structure and Dynamics of Amorphous Alumina from Deep Potential Molecular Dynamics

    T. Faber and J. Ziman, A theory of the electrical properties of liquid metals: Iii. the resistivity of binary alloys, Philosophical Magazine 11, 153 (1965). Supporting Information for “From Bulk to Surface: Structure and Dynamics of Amorphous Alumina from Deep Potential Molecular Dynamics” Zheng Yu, 1, ∗ Jiayan Xu, 1, ∗ Abhirup Patra, 2, † Sharan Shetty,3...

    work page 1965

  73. [74]

    stronger

    surface was considered. Melt-quenched bulk structures were generated through melt–quench simulations starting from cubic α-Al2O3 bulk structures with two sizes: p(1 × 3 × 1) (180 atoms) and p(2 × 3 × 1) (360 atoms). The simulations were performed under NPT conditions at 1 bar following a three-stage protocol: the systems were first equilibrated at 3800 K ...

    work page

  74. [75]

    Digne, P

    M. Digne, P. Sautet, P. Raybaud, P. Euzen, and H. Toulhoat, Use of dft to achieve a rational understanding of acid–basic properties of γ-alumina surfaces, J. Catal. 226, 54 (2004)

    work page 2004

  75. [76]

    C. Shi, O. L. G. Alderman, D. Berman, J. Du, J. Neuefeind, A. Tamalonis, J. K. R. Weber, J. You, and C. J. Benmore, The Structure of Amorphous and Deeply Supercooled Liquid Alumina, Front. Mater. 6, 10.3389/fmats.2019.00038 (2019)

    work page doi:10.3389/fmats.2019.00038 2019

  76. [77]

    J. He, D. A vnir, and L. Zhang, Sol-gel derived alumina glass: Mechanistic study of its structural evolution, Acta Materialia 174, 418 (2019)

    work page 2019

  77. [78]

    Hashimoto, Y

    H. Hashimoto, Y. Onodera, S. Tahara, S. Kohara, K. Yazawa, H. Segawa, M. Murakami, and K. Ohara, Structure of alumina glass, Sci Rep 12, 516 (2022)

    work page 2022