Pith. sign in

REVIEW 4 major objections 5 minor 41 references

A Robust Machine Learned Interatomic Potential for Nb: Collision Cascade Simulations with accurate Defect Configurations

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read A machine-learned potential for niobium reproduces the correct self-interstitial defect state and carries it into collision cascade simulations.

desk verdict A solid SNAP potential for Nb with the correct <111> SIA ground state and a genuine out-of-sample cascade check; the main gaps are reproducibility and the unvalidated ZBL blend, both fixable. read the letter →

arxiv 2502.03126 v1 pith:HI6URBYG submitted 2025-02-05 cond-mat.mtrl-sci

Utkarsh Bhardwaj , Vinayak Mishra , Suman Mondal , Manoj Warrier This is my paper · ORCID
classification cond-mat.mtrl-sci
keywords machinelearninginteratomicpotentialSNAPniobiumself-interstitialatomcollisioncascadesimulationradiationdamagemoleculardynamicsdefectformationenergy
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 reports a machine-learned Spectral Neighbor Analysis Potential (SNAP) for niobium, trained on density functional theory data, that reproduces the correct ground-state self-interstitial atom (SIA) configuration. In niobium and other non-magnetic bcc metals, the lowest-energy SIA is a $\langle 111\rangle$ dumbbell, while common embedded-atom and Finnis-Sinclair potentials incorrectly favor $\langle 110\rangle$. The authors show that in 5 keV collision cascade simulations their SNAP potential yields about 98% $\langle 111\rangle$ single dumbbells, matching the DFT-ordered ground state, whereas EAM and FS potentials yield about 99% $\langle 110\rangle$. If true, this means a fitted interatomic potential can carry the correct defect ordering from equilibrium training data into the primary damage state of an irradiation cascade, a prerequisite for reliable predictions of defect evolution in niobium.

What carries the argument

The carrying object is the SNAP (Spectral Neighbor Analysis Potential) framework, which describes each atom's local environment by bispectral descriptors derived from spherical harmonics and fits energy as a linear model in those descriptors, so forces come from analytic derivatives. A ZBL pair potential is blended in at short range to handle high-energy collisions. The training set is curated by Vendi diversity scoring plus D-optimal selection, and includes strained, defective, and liquid configurations, which the authors argue is important for producing complex defect morphologies in cascades. The key output is the relative ordering of SIA formation energies, which then controls dumbbell orientation in the cascade debris.

What would settle it

Rerun the 5 keV cascades with the ZBL short-range repulsion removed or shifted to a different transition radius and measure the fraction of $\langle 111\rangle$ single dumbbells; a significant change would show the result depends on the unvalidated collision model rather than on the trained potential.

Watch

Extended reading notes

Core claim

The central claim is that a SNAP machine-learned potential fitted to a curated DFT dataset reproduces the relative stability of SIA dumbbell configurations in Nb, and that this ordering survives in out-of-sample cascade simulations. The evidence is a series of validation tests on elastic constants, thermal properties, vacancy and SIA formation energies, and ten 5 keV collision cascades. The decisive comparison is in the cascade output: 98% of single SIA dumbbells are $\langle 111\rangle$ with SNAP, versus 99% $\langle 110\rangle$ with the EAM and FS reference potentials, while total defect counts remain comparable. The paper concludes that the SNAP potential is stable under irradiation conditions and resolves a persistent discrepancy in classical potentials.

Load-bearing premise

The cascade predictions assume the short-range repulsive forces that govern violent atomic collisions, which were not part of the training data, are modeled accurately enough that the final defect state is dominated by the correctly ordered equilibrium formation energies.

Editorial extensions

If this is right

  • The SNAP potential matches DFT elastic constants, melting point, and vacancy/SIA formation energies closely, while preserving MD-level speed.
  • In 5 keV cascades, SNAP produces 98% $\langle 111\rangle$ single dumbbells, so primary damage morphology is qualitatively different from EAM and FS, which give 99% $\langle 110\rangle$.
  • Total defect and cluster counts are similar across potentials, so the orientation difference is not an artifact of overall damage production.
  • Because SIA orientation controls diffusion and clustering, the correct ordering is a necessary baseline for modeling void swelling, creep, and embrittlement in Nb.
  • The potential remains stable over ten independent cascade runs, addressing the known risk of machine-learned potential instabilities in long molecular dynamics simulations.

Reading between the lines

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

  • If the equilibrium formation-energy ordering is what determines cascade dumbbell orientation, then formation-energy accuracy could serve as a cheap screening test for radiation-damage potentials in other bcc metals, before running expensive cascades.
  • The training set's liquid configurations may be doing essential work: without them, the potential might not form the experimentally plausible ring-like SIA clusters seen in a fraction of cascades; testing a version trained without liquid frames would isolate that contribution.
  • The short-range ZBL blend is not validated against high-energy collision data in the paper; a dedicated comparison of threshold displacement energies or replacement collision sequences against experiment would test whether the primary damage state remains correct at higher PKA energies.
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 / 5 minor

Summary. The manuscript develops a SNAP machine-learned interatomic potential for Nb using a curated DFT dataset of strained, defective, and liquid configurations. The authors validate elastic constants, thermal properties, vacancy/SIA formation energies, and run ten 5 keV collision cascades, reporting that the SNAP potential yields predominantly <111> single SIA dumbbells in the primary damage state while EAM and FS potentials yield <110>. They argue this resolves a long-standing discrepancy with DFT and supports the use of the potential for irradiation simulations.

Significance. If the cascade result is robust, the paper provides a valuable demonstration that an MLIP trained on near-equilibrium DFT data can transfer to out-of-sample radiation-damage conditions and connect SIA ground-state ordering to primary-damage morphology. The work is strengthened by the use of a diversity-based training-set selection, a broad validation matrix against two classical potentials, and an independent cascade test that was not part of the training data. However, the paper currently neither releases the potential or dataset nor validates the short-range repulsion on which the cascade simulations depend, and it does not quantify run-to-run variability; these gaps prevent full confidence in the headline claim.

major comments (4)
  1. [§2.2] The ZBL/SNAP blending is insufficiently specified and unvalidated at collision-relevant separations. The text states only that SNAP 'smoothly combines' ZBL short-range repulsion, without giving the switching function, inner/outer cutoffs, or any comparison to DFT energies and forces at close separation; no threshold displacement energies or two-body cold curves are reported. Because Table 4's 98% <111> single-dumbbell fraction is the paper's headline out-of-sample claim, and 5 keV cascades sample interatomic separations far below the near-equilibrium training data, the possibility that the cascade orientation statistics are controlled by the repulsive-wall artifact rather than by the fitted SIA energetics is not ruled out. Please add the ZBL transition parameters, a short-range validation, and a sensitivity test of the cascade SIA orientations to the blending parameters.
  2. [§2.1, Table 2] The manuscript does not state whether the defect configurations whose formation energies appear in Table 2 were part of the 550-configuration training set. If those configurations, or nearby AIMD frames from the same trajectories, were included, Table 2 demonstrates fitting accuracy but not independent prediction, and the claim that the potential 'captures' the DFT SIA ordering needs to be qualified accordingly. Please report the train/test split by configuration type and, if feasible, show that the SIA ordering survives exclusion of the corresponding defect configurations from training.
  3. [Tables 3–4, §2.3] No statistical uncertainties are reported for the ten cascade runs. The orientation percentages in Table 4 are given as integers and the defect counts in Table 3 as means, without standard deviations, per-run ranges, or the number of cascades that produced each orientation. Given that the head-to-head comparison between SNAP and the classical potentials turns on these percentages, please report standard errors or the full distribution over cascade directions, together with the random seeds or direction definitions used.
  4. [Data availability] The manuscript does not state where the fitted SNAP potential, the DFT training data, or the cascade input scripts can be obtained. Since the key findings are potential-specific and the paper reports no explicit error bars, independent verification is currently impossible. Please add a Data Availability statement and, if possible, deposit the potential and the training set in a public repository.
minor comments (5)
  1. [Table 1] The relative errors for C44 appear inconsistent with the stated baseline. For FS, 47 GPa is 67% above the experimental 28 GPa but 236% above the DFT 14 GPa; for SNAP, 14 GPa is reported as 0.0% error even though it deviates substantially from the experimental value. Please clarify the reference value used for each error and correct the table accordingly.
  2. [§2.3] The text refers to the 'Csaransh software suite' while Reference [34] is titled 'Saransh'; please unify the name.
  3. [§2.1] The phrase 'We preform ab-initio MD' contains a typo; it should read 'We perform ab-initio MD'.
  4. [Abstract, §4] The abstract and discussion state that the potential captures the ground-state SIA configuration in primary damage, but Table 4 shows this holds for single dumbbells (98% <111>) while clustered dumbbells in SNAP are 63% <110>. Please qualify the wording to avoid overgeneralization.
  5. [§2.2] The paragraph on SNAP and ZBL integration repeats nearly the same sentence twice; please consolidate and use the space to report the actual blending parameters.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the SNAP potential's equilibrium SIA ordering is fitted to DFT data by design, but the headline collision-cascade result is an out-of-sample prediction.

full rationale

The paper is transparent that the potential is trained on DFT data that explicitly include SIA defects in different orientations (Section 2.1), so Table 2's reproduction of the DFT-predicted ⟨111⟩ ground state is a fit check rather than a derived prediction. The central headline claim, that the potential yields predominantly ⟨111⟩ single dumbbells in 5 keV cascade primary damage (Table 4), is not a training target: the ten cascade runs are an out-of-sample application of the fitted potential. No equation in the paper defines the cascade output in terms of the training labels, and no load-bearing argument reduces to a self-citation chain. The self-citations used to explain C15-like ring clusters (references [35], [38], [39], and [12]) are supporting context for a secondary observation and do not carry the main derivation. The absence of ZBL/SNAP blend details is a correctness or completeness concern, not a circularity concern. Therefore no circular step is exhibited, and the appropriate finding is no significant circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The potential is a fitted model. All predictive power rests on the representativeness of the DFT training data and the expressiveness of the SNAP descriptors. No code, data, or fitted coefficients are provided, so the ledger cannot be independently audited.

free parameters (4)
  • SNAP bispectrum linear coefficients = Not reported (model weights)
    Fitted to DFT energies, forces, and stresses via linear regression. The potential depends entirely on these weights, but they are not released.
  • SNAP hyperparameters jmax and rcut = jmax=6, rcut=5.11 Å
    Chosen via Bayesian optimization (Optuna). These control descriptor dimensionality and locality; different choices would change the fitted surface.
  • ZBL/SNAP blending parameters = Not reported
    The short-range repulsion is switched from SNAP to ZBL with unspecified inner and outer cutoffs and smoothing. Cascade results depend on this transition.
  • Training dataset size and selection threshold = 550 configurations; Vendi saturation point
    Number of training samples and per-category saturation thresholds chosen heuristically. Affects accuracy and overfitting.
assumptions (5)
  • domain assumption DFT/PBE calculations with VASP provide accurate reference energies, forces, and stresses for Nb, including the relative stability of SIA configurations.
    The entire MLIP is fit to this data. If the PBE functional mis-orders SIA ground states, the potential inherits that error. Invoked in Section 2.1.
  • domain assumption The SNAP bispectrum descriptor with jmax=6 and rcut=5.11 Å is sufficiently expressive to represent the Nb potential energy surface, including high-energy configurations encountered in cascades.
    Representational sufficiency is assumed. Only equilibrium and moderately perturbed structures are in the training set. Invoked in Section 2.2.
  • domain assumption The ZBL potential and its smooth blending with SNAP accurately describe short-range repulsion during atomic collisions, with no statement of the blending parameters.
    Cascade simulations rely on this short-range model. No validation against high-energy DFT or experimental stopping data is shown. Invoked in Section 2.2.
  • domain assumption Ten 5 keV cascades in random directions provide statistically sufficient sampling of primary damage morphology.
    The paper reports only means and no error bars. Ten runs may be underpowered for orientation fractions. Invoked in Section 2.3 and Table 4.
  • domain assumption The Vendi diversity score saturation point and D-optimal selection identify a representative subset of the DFT data for training.
    This is a methodological choice that could introduce bias if the saturation heuristic is not robust. Invoked in Section 2.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Robust Machine Learned Interatomic Potential for Nb: Collision Cascade Simulations with accurate Defect Configurations." pith.science (2026). https://pith.science/paper/HI6URBYG

@misc{pith2026250203126,
  author       = {Pith},
  title        = {Pith review of: A Robust Machine Learned Interatomic Potential for Nb: Collision Cascade Simulations with accurate Defect Configurations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HI6URBYG}},
  note         = {Machine review of arXiv:2502.03126}
}
read the original abstract

Niobium (Nb) and its alloys are extensively used in various technological applications owing to their favorable mechanical, thermal and irradiation properties. Accurately modeling Nb under irradiation is essential for predicting microstructural changes, defect evolution, and overall material performance. Traditional interatomic potentials for Nb fail to predict the correct self-interstitial atom (SIA) configuration, a critical factor in radiation damage simulations. We develop a machine learning interatomic potential (MLIP) using the Spectral Neighbor Analysis Potential (SNAP) framework, trained on ab-initio Density Functional Theory (DFT) calculations, which accurately captures the relative stability of different SIA dumbbell configurations. The resulting potential reproduces DFT-level accuracy while maintaining computational efficiency for large-scale Molecular Dynamics (MD) simulations. Through a series of validation tests involving elastic, thermal, and defect properties -- including collision cascade simulations -- we show that our SNAP potential resolves persistent limitations in existing Embedded Atom Method (EAM) and Finnis--Sinclair (FS) potentials and is effective for MD simulations of collision cascades. Notably, it accurately captures the ground-state SIA configuration of Nb in the primary damage of a collision cascade, offering a robust tool for predictive irradiation studies.

Figures

Figures reproduced from arXiv: 2502.03126 by the authors.

Figure 1
Figure 1. The Cumulative diversity scores based on Vendi method [21] for DFT dataset of category SIA defects. We [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The SNAP potential predictions for (a) Energy and (b) Forces against the reference DFT values. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. (a) Energy–volume curve for Nb over a broad volume range, and (b) a focused view around the equilibrium [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Comparison of (a) elastic properties and (b) defect formation energies calculated from di [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Primary damage states in 5 keV collision cascades simulated using di [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Fraction of single and clustered dumbbells present in the primary damage produced using the three poten [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 38 canonical work pages

  1. [1]

    Processing, properties and applications of high-temperature niobium alloys

    C Craig Wojcik. Processing, properties and applications of high-temperature niobium alloys. MRS Online Proceedings Library (OPL), 322:519, 1993

    work page 1993

  2. [2]

    The physical, mechanical, and irradiation behavior of niobium and niobium-base alloys

    C English. The physical, mechanical, and irradiation behavior of niobium and niobium-base alloys. In Niobium-Proceedings of the international symposium, 1984

    work page 1984

  3. [3]

    R.E. Stoller. 1.11 - primary radiation damage formation. In Rudy J.M. Konings, editor, Comprehensive Nuclear Materials, pages 293 – 332. Elsevier, Oxford, 2012

    work page 2012

  4. [4]

    S.J. Zinkle. 1.03 - radiation-induced e ffects on microstructure*. In Rudy J.M. Konings, editor, Comprehensive Nuclear Materials, pages 65 – 98. Elsevier, Oxford, 2012

    work page 2012

  5. [5]

    Molecular dynamics simulation of threshold displacement energy and primary damage state in Niobium

    De-Ye Lin, Haifeng Song, and Xi Dong Hui. Molecular dynamics simulation of threshold displacement energy and primary damage state in niobium. arXiv preprint arXiv:1702.03598, 2017

    work page Pith review arXiv 2017

  6. [6]

    Second nearest-neighbor modified embedded atom method potentials for bcc transition metals

    Byeong-Joo Lee, MI Baskes, Hanchul Kim, and Yang Koo Cho. Second nearest-neighbor modified embedded atom method potentials for bcc transition metals. Physical Review B, 64(18):184102, 2001

    work page 2001

  7. [7]

    Experimental and molecular dynamics simulation study of structure of liquid and amorphous ni62nb38 alloy

    Y Zhang, R Ashcraft, MI Mendelev, CZ Wang, and KF Kelton. Experimental and molecular dynamics simulation study of structure of liquid and amorphous ni62nb38 alloy. The Journal of chemical physics, 145(20), 2016

    work page 2016

  8. [8]

    Force-matched embedded-atom method potential for niobium

    Michael R Fellinger, Hyoungki Park, and John W Wilkins. Force-matched embedded-atom method potential for niobium. Physical Review B—Condensed Matter and Materials Physics, 81(14):144119, 2010

    work page 2010

Show all 41 references
  1. [9]

    Universality of point defect structure in body-centered cubic metals

    Pui-Wai Ma and SL Dudarev. Universality of point defect structure in body-centered cubic metals. Physical Review Materials, 3(1):013605, 2019

    2019

  2. [10]

    Symmetry-broken self-interstitial defects in chromium, molybdenum, and tung- sten

    Pui-Wai Ma and SL Dudarev. Symmetry-broken self-interstitial defects in chromium, molybdenum, and tung- sten. Physical Review Materials, 3(4):043606, 2019

    2019

  3. [11]

    Bhardwaj, Andrea E Sand, and M

    U. Bhardwaj, Andrea E Sand, and M. Warrier. Comparison of sia defect morphologies from di fferent inter- atomic potentials for collision cascades in w. Modelling and Simulation in Materials Science and Engineering, 29(6):065015, jul 2021

    2021

  4. [12]

    Warrier and U

    M. Warrier and U. Bhardwaj. Defect cluster morphology in w from collision cascades: Results comparing five inter-atomic potentials, 2024

    2024

  5. [13]

    Stability and mobility of defect clusters and dislocation loops in metals

    Yu.N Osetsky, D.J Bacon, A Serra, B.N Singh, and S.I Golubov. Stability and mobility of defect clusters and dislocation loops in metals. Journal of Nuclear Materials, 276(1):65 – 77, 2000

    2000

  6. [14]

    Osetsky, D.J

    Yu.N. Osetsky, D.J. Bacon, B.N. Singh, and B. Wirth. Atomistic study of the generation, interaction, accumu- lation and annihilation of cascade-induced defect clusters. Journal of Nuclear Materials, 307-311:852 – 861, 2002

    2002

  7. [15]

    Machine learning interatomic potentials: Keys to first-principles multiscale modeling

    Bohayra Mortazavi, Toshiaki Miyazaki, and Xiaoying Zhuang. Machine learning interatomic potentials: Keys to first-principles multiscale modeling. In Machine Learning for Materials, pages 303–322. Springer, 2023

    2023

  8. [16]

    Construction of machine-learning zr interatomic potentials for identifying the formation process of c-type dislocation loops

    T Okita, S Terayama, K Tsugawa, K Kobayashi, M Okumura, M Itakura, and K Suzuki. Construction of machine-learning zr interatomic potentials for identifying the formation process of c-type dislocation loops. Computational Materials Science, 202:110865, 2022

    2022

  9. [17]

    Gaussian approximation potentials for body- centered-cubic transition metals

    Jesper Byggm ¨astar, Kai Nordlund, and Flyura Djurabekova. Gaussian approximation potentials for body- centered-cubic transition metals. Physical Review Materials, 4(9):093802, 2020

    2020

  10. [18]

    Data efficiency and extrapolation trends in neural network interatomic potentials

    Joshua A Vita and Daniel Schwalbe-Koda. Data efficiency and extrapolation trends in neural network interatomic potentials. Machine Learning: Science and Technology, 4(3):035031, 2023. 8

    2023

  11. [19]

    Spectral neighbor analysis method for automated generation of quantum-accurate interatomic potentials

    Aidan P Thompson, Laura P Swiler, Christian R Trott, Stephen M Foiles, and Garritt J Tucker. Spectral neighbor analysis method for automated generation of quantum-accurate interatomic potentials. Journal of Computational Physics, 285:316–330, 2015

    2015

  12. [20]

    E fficiency of ab-initio total energy calculations for metals and semicon- ductors using a plane-wave basis set

    Georg Kresse and J ¨urgen Furthm¨uller. E fficiency of ab-initio total energy calculations for metals and semicon- ductors using a plane-wave basis set. Computational materials science, 6(1):15–50, 1996

    1996

  13. [21]

    The vendi score: A diversity evaluation metric for machine learning

    Dan Friedman and Adji Bousso Dieng. The vendi score: A diversity evaluation metric for machine learning. Transactions on Machine Learning Research, 2023

    2023

  14. [22]

    Optimal Design of Experiments

    Friedrich Pukelsheim. Optimal Design of Experiments. Society for Industrial and Applied Mathematics, 2006

    2006

  15. [23]

    M. R. Fellinger, H. Park, and J. W. Wilkins. Force-matched embedded-atom method potential for niobium. Physical Review B, 81(14):144119, 2010

    2010

  16. [24]

    M. I. Mendelev. Ni–nb eam fs potential. https://www.ctcms.nist.gov/potentials/Download/ 2022--Mendelev-M-I--Ni-Nb/1/v1_18_NiNb_B2.eam.fs , 2022. Accessed: [Insert date of access]

    2022

  17. [25]

    Identifying sub-cascades from the primary damage state of collision cascades

    Utkarsh Bhardwaj and Manoj Warrier. Identifying sub-cascades from the primary damage state of collision cascades. Modelling and Simulation in Materials Science and Engineering, 32(5):055017, 2024

    2024

  18. [26]

    From ultrasoft pseudopotentials to the projector augmented-wave method

    Georg Kresse and Daniel Joubert. From ultrasoft pseudopotentials to the projector augmented-wave method. Physical review b, 59(3):1758, 1999

    1999

  19. [27]

    A sub-sampling algorithm preventing outliers

    L Deldossi, E Pesce, and C Tommasi. A sub-sampling algorithm preventing outliers. arXiv preprint arXiv:2208.06218, 2022

    2022 arXiv

  20. [28]

    A performance and cost assessment of machine learning interatomic potentials

    Yunxing Zuo, Chi Chen, Xiangguo Li, Zhi Deng, Yiming Chen, J ¨org Behler, et al. A performance and cost assessment of machine learning interatomic potentials. arXiv:1906.08888, 2019. Preprint

    1906 arXiv

  21. [29]

    Fitsnap: Atomistic machine learning with lammps

    Andrew Rohskopf, C Sievers, Nicholas Lubbers, MA Cusentino, James Go ff, Jan Janssen, Megan McCarthy, D Montes Oca de Zapiain, Svetoslav Nikolov, Khachik Sargsyan, et al. Fitsnap: Atomistic machine learning with lammps. Journal of Open Source Software, 8(84):5118, 2023

    2023

  22. [30]

    J. F. Ziegler, J. P. Biersack, and U. Littmark. The Stopping and Range of Ions in Solids. Pergamon, New York, 1985

    1985

  23. [31]

    Polyvalent machine-learned potential for cobalt: From bulk to nanoparticles

    Marthe Bideault, J ´erˆome Creuze, Ryoji Asahi, and Erich Wimmer. Polyvalent machine-learned potential for cobalt: From bulk to nanoparticles. Physical Review Materials, 8(12):123803, 2024

    2024

  24. [32]

    Optuna: A next-generation hyperparameter optimization framework

    Takuya Akiba, Shotaro Sano, Toshihiko Yanase, Takeru Ohta, and Masanori Koyama. Optuna: A next-generation hyperparameter optimization framework. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2019

    2019

  25. [33]

    A. P. Thompson et al. Lammps—a flexible simulation tool for particle-based materials modeling at the atomic, meso and continuum scales. Computer Physics Communications, 2022. Preprint; see https://doi.org/10.1016/j.cpc.2021.108171

    2022

  26. [34]

    Bhardwaj, Harsh Hemanu, M

    U. Bhardwaj, Harsh Hemanu, M. Warrier, N. Semwal, K. Ali, and A. Arya. Saransh : Software suite to study molecular dynamics simulations of collision cascades. Journal of Open Source Software, 4(41):1461, 2019

    2019

  27. [35]

    Bhardwaj, Andrea E

    U. Bhardwaj, Andrea E. Sand, and M. Warrier. Graph theory based approach to characterize self interstitial defect morphology. Computational Materials Science, 195:110474, 2021

    2021

  28. [36]

    Self-interstitial atom defects in bcc transition metals: Group- specific trends

    D Nguyen-Manh, AP Horsfield, and SL Dudarev. Self-interstitial atom defects in bcc transition metals: Group- specific trends. Physical Review B—Condensed Matter and Materials Physics, 73(2):020101, 2006

    2006

  29. [37]

    Crc handbook of chemistry and physics, 100th edn., 2019–2020, 2019

    JR Rumble. Crc handbook of chemistry and physics, 100th edn., 2019–2020, 2019

    2019

  30. [38]

    Bhardwaj, Andrea E

    U. Bhardwaj, Andrea E. Sand, and M. Warrier. Classification of clusters in collision cascades. Computational Materials Science, 172:109364, 2020. 9

    2020

  31. [39]

    Molecular dynamics simulations of the defect evolution in tungsten on successive collision cascades, 2024

    Utkarsh Bhardwaj and Manoj Warrier. Molecular dynamics simulations of the defect evolution in tungsten on successive collision cascades, 2024

    2024

  32. [40]

    Formation mechanism of⟨111⟩ interstitial dislocation loops from irradiation-induced c15 clusters in tungsten

    Lixia Liu, Ning Gao, Yangchun Chen, Rongyang Qiu, Wangyu Hu, Fei Gao, and Huiqiu Deng. Formation mechanism of⟨111⟩ interstitial dislocation loops from irradiation-induced c15 clusters in tungsten. Phys. Rev. Mater., 5:093605, Sep 2021

    2021

  33. [41]

    D ´ezerald, M.-C

    L. D ´ezerald, M.-C. Marinica, Lisa Ventelon, D. Rodney, and F. Willaime. Stability of self-interstitial clusters with c15 laves phase structure in iron. Journal of Nuclear Materials, 449(1):219 – 224, 2014. Credit Authorship Statement • Utkarsh Bhardwaj: Machine Learning pote...

    2014

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.