Pith. sign in

REVIEW 4 major objections 6 minor 59 references

Machine Learning Potential-Driven Molecular Dynamics Simulations of Dehydrogenation in Pristine and Doped MgH$_2$

T0 review · 4 major / 6 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Machine-learning dynamics shows MgH2 releases hydrogen by subsurface H2 formation, and a Miedema electron-density volcano puts Ni at the optimum among 22 dopants.

desk verdict Solid screening study with a plausible Ni volcano, but the core quantitative claims rest on an unvalidated proprietary potential and a time-confounded ML analysis. read the letter →

arxiv 2607.18182 v1 pith:NPTBTL6A submitted 2026-07-20 cond-mat.mtrl-sci

Bo Han , Jianchuan Wang , Rui Zhang , Martin Matas , Elias Vigl , Jing Tang , Yong Du , Christian Weiss
show 1 more author
David Holec
This is my paper
classification cond-mat.mtrl-sci
keywords MgH2hydrogenstoragemachine-learningpotentialmoleculardynamicsdehydrogenationMiedemaelectrondensityvolcanorelationshipnickeldoping
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 uses machine-learning molecular dynamics to watch hydrogen leave pristine and doped magnesium hydride, and it makes two linked claims. First, H2 molecules are born in the subsurface of MgH2, not on the surface, and then diffuse outward to desorb, a route that contradicts the usual surface-recombination picture. Second, screening 22 dopants on the most active (100) facet places Ni at the top and shows that catalytic activity follows a volcano in the Miedema electron density, with an optimum between 4.0 and 5.4 x 10^-2 e/bohr^3. If correct, this gives experimentalists a one-number rule for choosing dopants and provides a microscopic explanation for the long-known 'hydrogen pump' role of Ni: the dopant attracts hydrogen into a local reservoir and then lets it couple into H2 and escape. The authors support the mechanism with static DFT desorption energies and with DSC experiments showing that Ni lowers the activation energy of desorption.

What carries the argument

The load-bearing machinery is a universal machine-learned interatomic potential trained on quantum-mechanical (DFT) data, which lets the authors run 50 ps canonical-ensemble trajectories on roughly 2700-atom slabs of MgH2 and its doped variants, scales inaccessible to ab initio dynamics. The descriptor that organizes the results is the Miedema electron density n_ws, a tabulated quantity describing the electron density at the boundary of an element's atomic cell. Time-multiplied versions of elemental descriptors are fed to a machine-learning regression model, and Shapley attribution singles out n_ws as the feature controlling hydrogen release. The paper then maps cumulative H2 release against

What would settle it

A quantum-mechanical reaction-path calculation for subsurface H2 formation in Ni-, W-, and Zr-doped MgH2(100), or a short ab initio molecular-dynamics run on a small doped cell, would settle it: if nickel's barrier is not the lowest, or if H2 forms preferentially on the surface in those calculations, the subsurface mechanism and the volcano ranking collapse.

Watch

Extended reading notes

Core claim

The central discovery is that MgH2 gives up hydrogen through a subsurface route: two hydrogen atoms combine into H2 below the first atomic layer, and the molecule then migrates through the lattice and desorbs into vacuum. The paper's DFT calculation supports this by showing that H2 extracted from the second and third subsurface layers costs 1.59 eV, lower than the 1.96 eV cost from the topmost surface layer. On the doped systems, after ranking 22 elements by cumulative H2 release over 50 ps, the authors find Ni most active and identify the Miedema electron density n_ws as the dominant descriptor. Total hydrogen release plotted against n_ws is a volcano with an optimal window of 4.0 < n_ws <

Load-bearing premise

The whole mechanism and dopant ranking rest on the premise that a machine-learned potential trained on quantum-mechanical data is quantitatively correct for hydrogen pairing, hydrogen migration, and hydrogen escape in MgH2 with 22 different dopants at 800–1000 K, even though the paper checks this directly only for undoped MgH2(100).

Editorial extensions

If this is right

  • MgH2 surface chemistry alone does not explain desorption; rational catalyst design must also target subsurface H2 nucleation and transport of the formed molecule to the surface.
  • Dopants with n_ws in the window 4.0–5.4 x 10^-2 e/bohr^3 should be the first candidates in screening: below it they fail to gather hydrogen, above it they trap the H2 they make.
  • Nickel's experimentally known catalytic superiority in MgH2 gets a microscopic justification: Ni sits at the volcano peak and spontaneously forms Mg2NiH4-like local clusters that act as transient hydrogen pumps.
  • The measured DSC bimodality and lower activation energies (101–135 kJ/mol versus 171 kJ/mol) are consistent with Ni creating a distribution of easy desorption environments rather than a single new channel.
  • The n_ws criterion provides a simple quantitative filter for dopant selection in other hydrogen-storage alloys, consistent with the paper's observation that common storage alloys such as LaNi5-based and FeTi systems combine elements near the window.

Reading between the lines

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

  • An extension the paper leaves implicit: the volcano criterion could be used to design binary or ternary dopant mixtures by averaging n_ws, just as FeTi combines a high-density and a low-density element; the paper hints at this but does not test it.
  • A testable prediction from the subsurface mechanism: pre-opening subsurface channels, through vacancies, interstitials, or strain, should accelerate desorption even without chemical doping.
  • Because the machine-learning descriptor is time-coupled, the ranking may reflect early-time local enrichment around dopants as much as the static n_ws; a coarse-grained sink-plus-escape model could reproduce the volcano with simple radial transport rates, which the paper does not attempt.
  • If the transferability holds, the same n_ws window provides a cheap first-pass filter for dopants in other ionic hydrides before expensive molecular-dynamics screening, a use the authors gesture at in the conclusion.
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

4 major / 6 minor

Summary. The manuscript uses 50 ps NVT molecular dynamics with the proprietary PFP v7.0.0 machine-learning potential to simulate H2 release from MgH2 slabs. It identifies the (100) surface as the most active low-index surface; for pristine (100) at 800 K it observes H2 formation in the subsurface followed by diffusion and desorption, supported by DFT desorption energies (Table 1). It then substitutes three Mg atoms in each of the top and bottom surface layers with each of 22 elements, counts H2 release over five independent runs, and identifies Ni as the best dopant. An XGBoost/SHAP analysis of time-coupled elemental descriptors nominates the Miedema electron density n_ws as the key descriptor, and a volcano in n_ws with an optimal window 4.0–5.4×10^-2 e/bohr^3 is proposed. DSC/Kissinger measurements on pure and Ni-doped MgH2 corroborate the Ni enhancement.

Significance. If correct, the subsurface H2-formation pathway and the n_ws volcano provide a simple, falsifiable screening rule for MgH2 dopants, with plausible links to Sabatier chemistry and to experimental hydrogen-pump phases. The computational screening is ambitious and the experimental DSC data are a useful independent check. However, the central quantitative claims currently rest on two unvalidated pillars: (i) the accuracy of PFP for reactive, doped, high-temperature dynamics, and (ii) an ML descriptor analysis that may be dominated by time. These gaps are addressable with targeted DFT benchmarks and a statistical control, but until then the volcano and Ni ranking are not established. The subsurface mechanism is partially supported by the pristine DFT energies and by agreement with the earlier Morrison et al. MLIP-MD study, which is a genuine strength.

major comments (4)
  1. [§2.2.1/§2.2.2 and §3.3–3.4] The PFP v7.0.0 potential is not validated for the reactive doped dynamics that produce the central volcano and Ni ranking. Table 1 is a static DFT desorption-energy comparison for pristine MgH2(100) only; it does not test H–H recombination barriers in the subsurface, H2 migration through the lattice, or desorption for any of the 22 doped systems at 800–1000 K. Since cumulative H2 counts from PFP are the observable behind Figs. 6 and 7(c), a PFP-vs-DFT benchmark of the key elementary steps (at least for undoped, Ni, W, La, Zr) is required. The statement that PFP is transferable 'without the need for system-specific retraining' is assumed, not demonstrated for this reactive Mg-H-M manifold.
  2. [§2.3 and §3.4] The ML descriptor analysis is circular with respect to time. All input descriptors are multiplied by normalized time, and the model is trained only on these time-coupled descriptors; the target is (or is derived from) time-resolved H2 release. The common time factor makes every descriptor a time proxy, so the test R2=0.998 and the SHAP dominance of n_ws^tau do not independently establish n_ws as causal. The text also inconsistently states the target is 'cumulative H2 release after 50 ps' while using 11,000 time-resolved data points. Please perform a control with static descriptors (or with descriptors and time as separate features) and report cross-validated feature importance.
  3. [§3.4/Fig. 7(c)] No statistical uncertainty is reported for the H2 release counts or for the volcano. Five independent MD runs per dopant are averaged, but without error bars (or per-run points) the volcano shape, the Ni optimum, and the 4.0–5.4 optimal window may be within run-to-run noise. In addition, the optimal window is selected post hoc; if Ag and N are added after the initial 22-element screen (Section 3.4), state explicitly whether they were held out from training and used as a prediction test. At minimum, report mean ± standard deviation for all dopants and mark the 22 vs. additional points in Fig. 7(c).
  4. [§2.2.2] The operational definitions of H2 formation (H–H distance < 0.85 Å) and removal (>30 Å from surface) are ad hoc and directly determine the release counts. The pressure argument in Fig. S1 justifies removal but not the choice of distance threshold or the sensitivity of the counts/volcano to these cutoffs. A short sensitivity analysis (e.g., 0.80/0.90 Å thresholds; 25/35 Å removal distances for at least Ni, W, and undoped) is needed.
minor comments (6)
  1. [§4, conclusion bullet 1] It states 'pristine MgH2(110)' but the simulations and DFT validation in Section 3.2/Table 1/Fig. 3 concern MgH2(100). Fix the mismatch.
  2. [§2.3] Clarify whether the target is the cumulative release after 50 ps or the time-resolved cumulative curve; the current wording supports both readings.
  3. [§2.2.2/§3.3] The doping description says 'randomly substituting three Mg atoms in both the top and bottom surface layers', which means six dopant atoms per slab; clarify this and the definition of the 5.56 at.% concentration.
  4. [§3.5.2] When reporting Ni–H/Ni–Mg coordination numbers, state whether values refer to one dopant or are averaged over the six dopants and five runs.
  5. [Fig. 6] Since only selected curves are shown with full opacity, add a legend for the unlabeled curves or state clearly that the complete set is in Fig. S3; also add axis units.
  6. [Table 1] Specify the sign convention for desorption energy (endothermic positive) and note that these are thermodynamic energies, not activation barriers.

Circularity Check

1 steps flagged · score 6.0 of 10

ML descriptor 'identification' of nws is forced by time-coupling; the volcano/optimal-window claim rests on that circular step.

  1. fitted input called prediction [Section 2.3 (Machine learning-based descriptor analysis) and Section 3.4]
    "Our dataset comprised 11,000 data points derived from time-resolved MD simulations across 22 distinct dopant systems. The cumulative H2 release after 50 ps was used as the primary target variable for kinetic assessment. ... these descriptors were multiplied by normalized time to engineer a suite of time-coupled descriptors. Specifically, to prevent interference from static attributes, the model input was restricted exclusively to these time-coupled descriptors. ... This model achieved an R2 score of 0.998 on the test dataset, demonstrating high predictive accuracy and strong generalization cap"

    Every input feature is d_i(t)=p_i·tau(t), while the target is the cumulative number of H2 molecules released up to time t, a nondecreasing function of t. For an approximately linear release ramp, y(t)≈c·t, so each time-coupled descriptor is proportional to the target by construction: d_i(t)=p_i·t/t_max=(p_i/(c·t_max))·y(t). The near-perfect test R2 is therefore a property of the shared time factor, not of the elemental descriptor p_i. The SHAP dominance of nws^tau is a rank among time-proxy features and does not independently validate nws as the controlling electronic variable; the subsequent volcano and optimal-window claim is anchored to this statistically forced descriptor ranking.

full rationale

The only substantive circularity is in the ML descriptor step. XGBoost is fed time-coupled features (static property × normalized time) and asked to predict time-resolved cumulative H2 release; because cumulative release is a monotone function of time, the high R2 is largely forced by the shared time factor rather than by any elemental descriptor. The paper then uses SHAP to identify nws^tau as the dominant descriptor and builds the volcano/optimal-window relationship on that ranking. The rest of the derivation is not circular: the subsurface recombination mechanism is supported by independent DFT desorption energies (Table 1) and by an external literature result [37]; the Ni ranking comes directly from the PFP-MD release counts and is consistent with external DSC experiments; no load-bearing self-citation chain was found. PFP's accuracy for doped reactive events is a validity risk, but it is an external potential, not a circular input. Overall, partial circularity is present in the descriptor/volcano claim, so the score is 6.

Assumptions & free parameters 7 free parameters · 7 assumptions · 0 invented entities

The paper postulates no new particles, forces, or material phases; the dynamically formed Mg2NiH4-like clusters are emergent features of the simulation, not invented entities. Its quantitative claims rest instead on hand-set thresholds, post-hoc fitted windows, and a proprietary ML potential whose reactive accuracy for doped systems is not established.

free parameters (7)
  • H2 formation distance threshold = 0.85 Å
    Defines an H2 molecule; hand-set above equilibrium 0.74 Å to account for thermal vibration; directly sets H2 counts.
  • H2 removal distance = 30 Å
    H2 molecules are removed beyond 30 Å from the surface to keep partial pressure low; affects computed cumulative release by preventing re-adsorption.
  • MD temperature = 1000 K (800 K for mechanism run)
    Chosen to accelerate events to observable rates; far above experimental desorption (~683 K), so kinetics are extrapolated.
  • Doping concentration = 5.56 at.% relative to substituted Mg surface layers
    Three Mg atoms replaced per surface; arbitrary choice affecting activity ranking.
  • XGBoost hyperparameters = max_depth=6, learning_rate=0.2, n_estimators=100
    Tuned by 10-fold cross-validation; model is fit on the same data used for interpretation.
  • Optimal Miedema window boundaries = 4.0 < nws < 5.4 x10^-2 e/bohr^3
    Drawn post hoc from the 22-dopant volcano; no independent validation is provided.
  • Time-coupling normalization of descriptors = descriptor x normalized time
    Transforms static elemental properties into time-dependent inputs; introduces a time proxy that confounds SHAP importance.
assumptions (7)
  • domain assumption PFP (v7.0.0) accurately describes Mg-H-M interactions and reactive dynamics without retraining.
    Central simulations rely on this; validation is limited to static DFT desorption energies for undoped (100) in Section 3.2 and Table 1.
  • domain assumption DFT/PBE energies and forces are a sufficient ground truth for training PFP and for the desorption-energy check.
    Used as the reference for the ML potential and for the Ed calculation in Section 2.4/3.2.
  • domain assumption NVT Nosé-Hoover MD at 800-1000 K for 50 ps samples representative dehydrogenation events.
    The entire kinetic analysis uses these short, high-temperature trajectories.
  • domain assumption An H-H distance below 0.85 Å identifies a formed H2 molecule, and tracking its diffusion/release is meaningful.
    Counts of H2 formation and release depend on this threshold; no sensitivity analysis is given.
  • domain assumption Miedema electron density is a valid intrinsic descriptor for dopant catalytic activity.
    Used as the central descriptor without derivation from the MD physics.
  • domain assumption The Sabatier principle applies and explains the observed volcano.
    Invoked in Section 3.4 to interpret the nws trend.
  • domain assumption Slab models with ~37 Å thickness, 100 Å vacuum, and fixed bottom layers emulate a real surface.
    Standard surface-modeling assumption used in Section 2.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine Learning Potential-Driven Molecular Dynamics Simulations of Dehydrogenation in Pristine and Doped MgH$_2$." pith.science (2026). https://pith.science/paper/NPTBTL6A

@misc{pith2026260718182,
  author       = {Pith},
  title        = {Pith review of: Machine Learning Potential-Driven Molecular Dynamics Simulations of Dehydrogenation in Pristine and Doped MgH$_2$},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NPTBTL6A}},
  note         = {Machine review of arXiv:2607.18182}
}
read the original abstract

Machine learning potential-driven molecular dynamics simulations (ML-MD) were employed to provide atomistic insights into the dehydrogenation kinetics of pristine and doped MgH2. Through systematic investigation of distinct surface orientations, the MgH2 (100) surface was identified as the most active low-index surface for hydrogen release. For pristine MgH2, our simulations revealed a novel H2 formation mechanism characterized by H2 generation in the subsurface region followed by diffusion to the surface for desorption, highlighting the critical role of subsurface processes beyond conventional surface-driven pathways. Comprehensive screening of 22 doping elements identified Ni as the most effective dopant. Among several descriptors, machine learning analysis identified the time-coupled Miedema electron density as the critical descriptor, underscoring the role of electronic properties. Consequently, a volcano-shaped relationship was uncovered between the intrinsic Miedema electron density ( nws ) and total hydrogen release (optimal window: 4.0 < nws < 5.4x10-2 e/bohr3). Dopants within this range serve a dual function: acting as thermodynamic sinks for H attraction while maintaining a balanced interaction strength to facilitate H-H coupling and H2 release. This atomistic-level validation provides strong theoretical support for the experimentally observed "hydrogen pump" effect of catalytic phases. The present study demonstrates the strong capability of ML- MD in navigating through complex catalytic mechanisms and establishing quantitative property-activity relationships, providing a robust framework for rational design of high-performance catalysts for MgH2 and other hydrogen storage materials.

Figures

Figures reproduced from arXiv: 2607.18182 by the authors.

Figure 1
Figure 1. Atomic structures of MgH2 (110), (100), (101), and (001) surfaces. 2.2 Calculation details 2.2.1 Matlantis potential In this work, we employed the universal neural network potential known as the ‘PreFerred Potential’ (PFP, version 7.0.0), which incorporates van der Waals (vdW) corrections. This potential was developed by Preferred Computational Chemistry (PFCC) and is accessible through the cloud-based simulation pl… view at source ↗
Figure 2
Figure 2. Dehydrogenation curves of MgH2(100), (001), (101), and (110) surfaces. 3.2 Dehydrogenation pathway Elucidating the atomic-scale pathway of hydrogen release is the first step in understanding dehydrogenation kinetics. To observe the hydrogen desorption process, MD simulations of the pristine MgH2(100) surface were performed at 800 K. This temperature, which is higher than the experimental desorption temperature for M… view at source ↗
Figure 3
Figure 3. Schematic illustration of an atomic-scale pathway of hydrogen release from MgH2(100) surface revealed by MD snapshots from the simulation at 800 K. Orange and blue spheres represent Mg and H atoms, respectively. The sequential desorption process shows: (1) H2 formation at the subsurface, (2) migration to the surface, and (3) desorption into vacuum. The formed H2 molecule is highlighted in dark blue with H-H bonds vi… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: (a) Selected doping elements for surface doping study; (b) Atomic structures of the doped MgH2(100) surface viewed from the top (left) and bottom (right). The large orange, small light blue, and large dark blue spheres represent Mg, H, and dopants, respectively [PITH_…
Figure 5
Figure 5. Figure 5: Depth evolution of dopants during MD simulations. (a) Time-dependent mean depth of Ni dopants. (b) Mean depths of all screened doping elements at 50 ps [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

59 extracted references · 42 canonical work pages

  1. [1]

    Amirkhiz, B

    B.S. Amirkhiz, B. Zahiri, P. Kalisvaart, D. Mitlin, Synergy of elemental Fe and Ti promoting low temperature hydrogen sorption cycling of magnesium, Int. J. Hydrogen Energy 36 (2011) 6711 -6722, https://doi.org/10.1016/j.ijhydene.2011.02.141

    work page doi:10.1016/j.ijhydene.2011.02.141 2011

  2. [2]

    Zhang, Y

    X. Zhang, Y . Sun, S. Ju, J. Ye, X. Hu, W . Chen, L. Yao, G. Xia, F. Fang, D. Sun, X. Y u, Solar-Driven Reversible Hydrogen Storage, Adv. Mater. 35 (2023) 2206946, https://doi.org/10.1002/adma.202206946

    work page doi:10.1002/adma.202206946 2023

  3. [3]

    Song, S.N

    M.Y . Song, S.N. Kwon, H.R. Park, S. -H. Hong, Improvement in the hydrogen storage properties of Mg by mechanical grinding with Ni, Fe and V under H 2 atmosphere, Int. J. Hydrogen Energy 36 (2011) 13587 -13594, https://doi.org/10.1016/j.ijhydene.2011.07.107

    work page doi:10.1016/j.ijhydene.2011.07.107 2011

  4. [4]

    J. Qin, X. Zhou, Y . Fu, J. Liu, H. Wang, L. Ouyang, M. Zeng, Y .-J. Zhao, M. Zhu, Construction of Mg/Zr superlattice structure to achieve efficient hydrogen storage via atomic-scale interaction in Mg -Zr modulation films, Acta Mater. 263 (2024) 119470, https://doi.org/10.1016/j.actamat.2023.119470

    arXiv 2024

  5. [5]

    Hongo, K

    T. Hongo, K. Edalati, M. Arita, J. Matsuda, E. Akiba, Z. Horita, Significance of grain boundaries and stacking faults on hydrogen storage properties of Mg 2Ni intermetallics processed by high -pressure torsion, Acta Mater. 92 (2015) 46 -54, https://doi.org/10.1016/j.actamat.2015.03.036

    work page doi:10.1016/j.actamat.2015.03.036 2015

  6. [6]

    Yartys, M.V

    V .A. Yartys, M.V . Lototskyy, E. Akiba, R. Albert, V .E. Antonov, J.R. Ares, M. Baricco, N. Bourgeois, C.E. Buckley, J.M. Bellosta von Colbe, J.C. Crivello, F. Cuevas, R.V . Denys, M. Dornheim, M. Felderhoff, D.M. Grant, B.C. Hauback, T.D. Humphries, I. Jacob, T.R. Jensen, P.E. de Jongh, J.M. Joubert, M.A. Kuzovnikov, M. Latroche, M. Paskevicius, L. Pasq...

    2019

  7. [7]

    Q. Hou, X. Yang, J. Zhang, Review on Hydrogen Storage Performance of MgH 2: Development and Trends, Chemistryselect 6 (2021) 1589 -1606, https://doi.org/10.1002/slct.202004476

    work page doi:10.1002/slct.202004476 2021

  8. [8]

    Z. Ding, Y . Li, H. Yang, Y . Lu, J. Tan, J. Li, Q. Li, Y .a. Chen, L.L. Shaw, F. Pan, Tailoring MgH 2 for hydrogen storage through nanoengineering and catalysis, J. Magnesium Alloys 10 (2022) 2946-2967, https://doi.org/10.1016/j.jma.2022.09.028

    work page doi:10.1016/j.jma.2022.09.028 2022

Show all 59 references
  1. [9]

    X. Tan, B. Zahiri, C.M.B. Holt, A. Kubis, D. Mitlin, A TEM based study of the microstructure during room temperature and low temperature hydrogen storage 31 cycling in MgH 2 promoted by Nb –V , Acta Mater. 60 (2012) 5646 -5661, https://doi.org/10.1016/j.actamat.2012.06.009

    2012 doi

  2. [10]

    Smirnova, S.V

    D.E. Smirnova, S.V . Starikov, A.M. Vlasova, New interatomic potential for simulation of pure magnesium and magnesium hydrides, Comput. Mater. Sci. 154 (2018) 295-302, https://doi.org/10.1016/j.commatsci.2018.07.051

    2018 doi

  3. [11]

    X. Zhou, S. Kang, T.W . Heo, B.C. Wood, V . Stavila, M.D. Allendor f, An Analytical Bond Order Potential for Mg−H Systems, Chemphyschem 20 (2019) 1404- 1411, https://doi.org/10.1002/cphc.201800991

    2019 doi

  4. [12]

    Jacobs, D

    R. Jacobs, D. Morgan, S. Attarian, J. Meng, C. Shen, Z. Wu, C.Y . Xie, J.H. Yang, N. Artrith, B. Blaiszik, G. Ceder, K. Choudhary, G. Csanyi, E.D. Cubuk, B. Deng, R. Drautz, X. Fu, J. Godwin, V . Honavar, O. Isayev, A. Johansson, B. Kozinsky, S. Martiniani, S.P. Ong, I. Poltav...

    2025

  5. [13]

    Takamoto, C

    S. Takamoto, C. Shinagawa, D. Motoki, K. Nakago, W . Li, I. Kurata, T. Watanabe, Y . Yayama, H. Iriguchi, Y . Asano, T. Onodera, T. Ishii, T. Kudo, H. Ono, R. Sawada, R. Ishitani, M. Ong, T. Yamaguchi, T. Kataoka, A. Hayashi, N. Charoenphakdee, T. Ibuka, Towards universal neur...

    2022 doi

  6. [14]

    Takamoto, D

    S. Takamoto, D. Okanohara, Q. -J. Li, J. Li, Towards universal neural network interatomic potential, Journal of Materiomics 9 (2023) 447 -454, https://doi.org/10.1016/j.jmat.2022.12.007

    2023 doi

  7. [15]

    Ishikawa, Y

    T. Ishikawa, Y . Tanaka, S. Tsuneyuki, Evolutionary search for superconducting phases in the lanthanum -nitrogen-hydrogen system with universal neural network potential, Phys. Rev. B 109 (2024) 094106, https://doi.org/10.1103/PhysRevB.109.094106

    2024 doi

  8. [16]

    T. Kato, F. Lodesani, S. Urata, Boron coordination and three -membered ring formation in sodium borate glasses: a machine -learning molecular dynamics study, J. Am. Ceram. Soc. 107 (2024) 2888-2900, https://doi.org/10.1111/jace.19629

    2024 doi

  9. [17]

    M. Kim, S. Kim, B. Shong, Adsorption of dimethylaluminum isopropoxide (DMAI) on the Al2O3 surface: A machine-learning potential study, J. Sci.:Adv. Mater. Devices 9 (2024) 100754, https://doi.org/10.1016/j.jsamd.2024.100754

    2024

  10. [18]

    Z. Mao, W . Li, J. Tan, Dielectric tensor prediction for inorganic materials using latent information from preferred potential, Npj Comput. Mat er. 10 (2024) 265, https://doi.org/10.1038/s41524-024-01450-z

    2024 doi

  11. [19]

    Leiner, D

    T. Leiner, D. Holec, Revealing trends in catalytic activity of adatoms for hydrogen adsorption on carbon: A case study of graphen e and carbon nanotube, Carbon Trends 20 (2025) 100535, https://doi.org/10.1016/j.cartre.2025.100535

    2025

  12. [20]

    Park, C.I

    J.C. Park, C.I. Choi, W .P. Jeon, T. Thi Ngoc V an, W .-H. Kim, J. -H. Ahn, B. Shong, T.J. Pa rk, High -temperature atomic layer deposition of HfO 2 film with low impurity using a novel Hf precursor, J. Mater. Chem. C 13 (2025) 1637 -1645, https://doi.org/10.1039/D4TC01256A. 32

    2025 doi

  13. [21]

    J. Moon, U. Jeon, S. Choung, J.W . Han, CatBench framework for benchmarking machine learning interatomic potentials in adsorption energy predictions for heterogeneous catalysis, Cell Rep. Phys. Sci. 6 (2025) 102968, https://doi.org/10.1016/j.xcrp.2025.102968

    2025

  14. [22]

    B. Han, Y . Jia, J. Wang, X. Xiao, L. Chen, L. Sun, Y . Du, The structural, energetic and dehydrogenation properties of pure and Ti-doped Mg(0001)/MgH2(110) interfaces, J. Mater. Chem. A 11 (2023) 26602 -26616, https://doi.org/10.1039/D3TA06177A

    2023 doi

  15. [23]

    Bortz, B

    M. Bortz, B. Bertheville, G. Böttger, K. Yvon, Structure of the high pressure phase γ-MgH2 by neutron powder diffraction, J. Alloys Compd. 287 (1999) L4 -L6, https://doi.org/10.1016/S0925-8388(99)00028-6

    1999 doi

  16. [24]

    Kresse, J

    G. Kresse, J. Furthmüller, Efficiency of ab -initio total energy calculations for metals and semiconductors using a plane -wave basis set, Comput. Mater. Sci. 6 (1996) 15-50, https://doi.org/10.1016/0927-0256(96)00008-0

    1996 doi

  17. [25]

    Kresse, J

    G. Kresse, J. Furthmüller, Efficient iterative schemes for ab initio total -energy calculations using a plane -wave basis set, Phys. Rev. B 54 (1996) 11169 -11186, https://doi.org/10.1103/PhysRevB.54.11169

    1996 doi

  18. [26]

    Blöchl, Projector augmented-wave method, Phys

    P .E. Blöchl, Projector augmented-wave method, Phys. Rev. B 50 (1994) 17953 - 17979, https://doi.org/10.1103/PhysRevB.50.17953

    1994 doi

  19. [27]

    Kresse, D

    G. Kresse, D. Jou bert, From ultrasoft pseudopotentials to the projector augmented-wave method, Phys. Rev. B 59 (1999) 1758 -1775, https://doi.org/10.1103/PhysRevB.59.1758

    1999 doi

  20. [28]

    Perdew, K

    J.P . Perdew, K. Burke, M. Ernzerhof, Generalized Gradient Approximation Made Simple, Phys. Rev. Lett. 77 (1996) 3865 -3868, https://doi.org/10.1103/PhysRevLett.77.3865

    1996 doi

  21. [29]

    Bogdanović, K

    B. Bogdanović, K. Bohmhammel, B. Christ, A. Reiser, K. Schlicht e, R. V ehlen, U. Wolf, Thermodynamic investigation of the magnesium–hydrogen system, J. Alloys Compd. 282 (1999) 84-92, https://doi.org/10.1016/S0925-8388(98)00829-9

    1999 doi

  22. [30]

    Nosé, A unified for mulation of the constant temperature molecular dynamics methods, J

    S. Nosé, A unified for mulation of the constant temperature molecular dynamics methods, J. Chem. Phys. 81 (1984) 511-519, https://doi.org/10.1063/1.447334

    1984 doi

  23. [31]

    Hjorth Larsen, J

    A. Hjorth Larsen, J. Jørgen 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. Bjerre Jensen, J. Kermode, J.R. Kitchin, E. Leonhard Kolsbjerg, J. Kubal, K. Kaasbjerg, S. Lysgaard, J. Be...

    2017

  24. [32]

    T. Chen, C. Guestrin, XGBoost: A Scalable Tree Boosting System, Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Association for Computing Machinery, San Francisco, California, USA, 2016, pp. 785-794

    2016

  25. [33]

    Lundberg, G

    S.M. Lundberg, G. Erion, H. Chen, A. DeGrave, J.M. Prutkin, B. Nair, R. Katz, J. Himmelfarb, N. Bansal, S. -I. Lee, From local explanations to global understanding 33 with explainable AI for trees, Nat. Mach. Intell. 2 (2020) 56 -67, https://doi.org/10.1038/s42256-019-0138-9

    2020 doi

  26. [34]

    Pedregosa, G

    F. Pedregosa, G. V aroquaux, A. Gramfort, V . Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V . Dubourg, Scikit -learn: Machine learning in Python, the Journal of machine Learning research 12 (2011) 2825 -2830, https://doi.org/10.6009/jjrt.2023-2266

    2011 doi

  27. [35]

    Monkhorst, J.D

    H.J. Monkhorst, J.D. Pack, Special points for Brillouin -zone integrations, Phys. Rev. B 13 (1976) 5188-5192, https://doi.org/10.1103/PhysRevB.13.5188

    1976 doi

  28. [36]

    Kissinger, Reaction kinetics in differential thermal analysis, Anal

    H.E. Kissinger, Reaction kinetics in differential thermal analysis, Anal. Chem. 29 (1957) 1702-1706, https://doi.org/10.1021/ac60131a045

    1957 doi

  29. [37]

    Morrison, E

    O. Morrison, E. Uteva, G.S. Walker, D.M. Grant, S. Ling, Long Time Scale Molecular Dynamics Simulation of Magnesium Hydride Dehydrogenation Enabled by Machine Learning Interatomic Potentials, Acs Appl. Energy Mater. 8 (2025) 492 -502, https://doi.org/10.1021/acsaem.4c02627

    2025 doi

  30. [38]

    Hanada, T

    N. Hanada, T. Ichikawa , H. Fujii, Catalytic Effect of Nanoparticle 3d -Transition Metals on Hydrogen Storage Properties in Magnesium Hydride MgH 2 Prepared by Mechanical Milling, J. Phys. Chem. B 109 (2005) 7188 -7194, https://doi.org/10.1021/jp044576c

    2005 doi

  31. [39]

    H. Y u, S. Bennici, A. Auroux, Hydrogen storage and release: Kinetic and thermodynamic studies of MgH 2 activated by transition metal nanoparticles, Int. J. Hydrogen Energy 39 (2014) 11633 -11641, https://doi.org/10.1016/j.ijhydene.2014.05.069

    2014 doi

  32. [40]

    N. Xu, Z. Y uan, Z. Ma, X. Guo, Y . Zhu, Y . Zou, Y . Zhang, Effects of highly dispersed Ni nanoparticles on the hydrogen storage performance of MgH 2, Int. J. Miner., Metall. Mater. 30 (2023) 54-62, https://doi.org/10.1007/s12613-022-2510-8

    2023 doi

  33. [41]

    Tsuda, W .A

    M. Tsuda, W .A. Diño, H. Kasai, H. Nakanishi, H. Aikawa, Mg–H dissociation of magnesium hydride MgH 2 catalyzed by 3 d transition metals, Thin Solid Films 509 (2006) 157-159, https://doi.org/10.1016/j.tsf.2005.09.139

    2006 doi

  34. [42]

    Lakhal, M

    M. Lakhal, M. Bhihi, A. Benyoussef, A. El Kenz, M. Loulidi, S. Naji, The hydrogen ab/desorpt ion kinetic properties of doped magnesium hydride MgH 2 systems by first principles calculations and kinetic Monte Carlo simulations, Int. J. Hydrogen Energy 40 (2015) 6137 -6144, http...

    2015 doi

  35. [43]

    Cermak, B

    J. Cermak, B. David, Catalytic effect of Ni, Mg 2Ni and Mg2NiH4 upon hydrogen desorption from MgH 2, Int. J. Hydrogen Energy 36 (2011) 13614 -13620, https://doi.org/10.1016/j.ijhydene.2011.07.133

    2011 doi

  36. [44]

    Y . Fu, Z. Ding, S. Ren, X. Li, S. Zhou, L. Zhang, W . Wang, L. Wu, Y . Li, S. Han, Effect of in -situ formed Mg 2Ni/Mg2NiH4 compounds on hydrogen storage performance of MgH 2, Int. J. Hydrogen Energy 45 (2020) 28154 -28162, https://doi.org/10.1016/j.ijhydene.2020.03.089

    2020 doi

  37. [45]

    S. Ding, Y . Qiao, X. Cai, C. Du, Y . Wen, X. Shen, L. Xu, S. Guo, W . Gao, T. Shen, Catalytic mechanisms of nickel nanoparticles for the improved dehydriding kinetics of magnesium hydride, J. Magnesium Alloys 12 (2024) 4278 -4288, https://doi.org/10.1016/j.jma.2023.07.002. 34

    2024 doi

  38. [46]

    J. Cui, J. Liu, H. Wang, L. Ouyang, D. Sun, M. Zhu, X. Y ao, Mg–TM (TM: Ti, Nb, V , Co, Mo or Ni) core –shell like nanostructures: synthesis, hydrogen storage performance and catalytic mechanism, J. Mater. Chem. A 2 (2014) 9645 -9655, https://doi.org/10.1039/C4TA00221K

    2014 doi

  39. [47]

    Kudiiarov, A

    V .N. Kudiiarov, A. Kenzhiyev, R.R. Elman, N. Kurdyumov, I.A. Ushakov, A.V . Tereshchenko, R.S. Laptev, M.A. Kruglyakov, P.I. Khomidzoda, The Defect Structure Evolution in MgH2-EEWNi Composites in Hydrogen Sorption–Desorption Processes, Metals Basel. 15 (2025) 72, https://doi....

    2025 doi

  40. [48]

    Y . Jia, X. Yao, Carbon scaffold modified by metal (Ni) or non -metal (N) t o enhance hydrogen storage of MgH 2 through nanoconfinement, Int. J. Hydrogen Energy 42 (2017) 22933-22941, https://doi.org/10.1016/j.ijhydene.2017.07.106

    2017 doi

  41. [49]

    Z. Ma, J. Zhang, Y . Zhu, H. Lin, Y . Liu, Y . Zhang, D. Zhu, L. Li, Facile Synthesis of Carbon Supported Nano -Ni Particles with Superior Catalytic Effect on Hydrogen Storage Kinetics of MgH 2, Acs Appl. Energy Mater. 1 (2018) 1158 -1165, https://doi.org/10.1021/acsaem.7b00266

    2018 doi

  42. [50]

    C. Duan, Y . Tian, X. Wang, M. Wu, D. Fu, Y . Zhang, W . Lv, Z. Su, Z. Xue, Y . Wu, Ni -CNTs as an efficient confining framework and catalyst for improving dehydriding/rehydriding properties of MgH 2, Renew. Energy 18 7 (2022) 417 -427, https://doi.org/10.1016/j.renene.2022.01.048

    2022 doi

  43. [51]

    Zhang, W

    J. Zhang, W . Wang, X. Chen, J. Jin, X. Yan, J. Huang, Single-Atom Ni Supported on TiO2 for Catalyzing Hydrogen Storage in MgH 2, J. Am. Chem. Soc. 146 (2024) 10432-10442, https://doi.org/10.1021/jacs.3c13970

    2024 doi

  44. [52]

    Sabatier, La catalyse en chimie organique, C

    P. Sabatier, La catalyse en chimie organique, C. Béranger1920

  45. [53]

    Che, Nobel Prize in chemistry 1912 to Sabatier: Organic chemistry or catalysis?, Catal

    M. Che, Nobel Prize in chemistry 1912 to Sabatier: Organic chemistry or catalysis?, Catal. Today 218 -219 (2013) 162 -171, https://doi.org/10.1016/j.cattod.2013.07.006

    1912 doi

  46. [54]

    Pototschnig, M

    U. Pototschnig, M. Matas, D. Scheiblehner, D. Neuschitzer, R. Obenaus -Emler, H. Antrekowitsch, D. Holec, Predictive Model for Catalytic Methane Pyrolysis, The Journal of Physical Chemistry C 128 (2024) 9034 -9040, https://doi.org/10.1021/acs.jpcc.4c01690

    2024 doi

  47. [55]

    S. Wang, M. Gao, Z. Yao, K. Xian, M. Wu, Y . Liu, W . Sun, H. Pan, High-loading, ultrafine Ni nanoparticles dispersed on porous hollow carbon nanospheres for fast (de)hydrogenation kinetics of MgH 2, J. Magnesium All oys 10 (2022) 3354 -3366, https://doi.org/10.1016/j.jma.2021.05.004

    2022 doi

  48. [56]

    Brostow, Radial distribution function peaks and coordination numbers in liquids and in amorphous solids, Chem

    W . Brostow, Radial distribution function peaks and coordination numbers in liquids and in amorphous solids, Chem. Phys. Lett. 49 (1977) 285 -288, https://doi.org/10.1016/0009-2614(77)80588-5

    1977 doi

  49. [57]

    García, J.P

    G.N. García, J.P. Abriata, J.O. Sofo, Calculation of the electronic and structural properties of cubic Mg 2NiH4, Phys. Rev. B 59 (1999) 11746 -11754, https://doi.org/10.1103/PhysRevB.59.11746

    1999 doi

  50. [58]

    Myers, L

    W .R. Myers, L. -W . Wang, T.J. Richardson, M.D. Rubin, Calculation of thermodynamic, electronic, and optical properties of monoclinic Mg 2NiH4, J. Appl. Phys. 91 (2002) 4879-4885, https://doi.org/10.1063/1.1454206. 35

    2002 doi

  51. [59]

    K. Yvon, J. Schefer, F. Stucki, Structural studies of the hydrogen storage material Mg2NiH4. 1. Cubic high -temperature structure, Inorg. Chem. 20 (1981) 2776 -2778, https://doi.org/10.1021/ic50223a006

    1981 doi

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.