pith. sign in

arxiv: 2510.04015 · v2 · submitted 2025-10-05 · ❄️ cond-mat.mtrl-sci · math-ph· math.MP

Atomistic Machine Learning with Irreducible Cartesian Natural Tensors

Qun Chen , A. S. L. Subrahmanyam Pattamatta , Boyu Wang , David J. Srolovitz , Mingjian Wen This is my paper

Pith reviewed 2026-05-18 11:02 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci math-phmath.MP
keywords atomistic machine learningCartesian tensorsequivariant networksmachine learning interatomic potentialstensorial propertiesirreducible representationsmaterials sciencemolecular systems
0
0 comments X p. Extension

The pith

A Cartesian tensor framework enables equivariant atomistic models that match spherical-tensor performance for potentials and high-rank tensor properties.

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

The paper develops Cartesian Natural Tensor Networks, or CarNet, to create machine learning models that work directly in Cartesian space while respecting geometric symmetries. It establishes a theory for building irreducible representations from Cartesian natural tensors, along with rules for their products and a method to break down and rebuild high-rank physical tensors. If this holds, models can handle both scalar energies and complex tensor outputs such as dipoles or elastic constants without shifting to spherical coordinates. A sympathetic reader would care because Cartesian approaches keep physical quantities in their native form, which may ease interpretation and implementation for materials and molecular simulations.

Core claim

The central claim is that the theory of irreducible representations using Cartesian natural tensors supplies a systematic, complete scheme for decomposition and reconstruction of high-rank physical tensors. This machinery supports construction of an equivariant Cartesian model that produces machine learning interatomic potentials competitive with leading spherical-tensor methods for both materials and molecules, and that yields accurate structure-property relations for tensorial quantities from dipole moments up to the elastic constant tensor.

What carries the argument

Cartesian natural tensors and their irreducible representations, which supply the decomposition and reconstruction scheme for high-rank tensors and enable equivariant model construction in Cartesian space.

If this is right

  • Machine learning interatomic potentials can be built for both materials and molecular systems at accuracy levels matching current spherical-tensor leaders.
  • Structure-property models become available for tensorial outputs ranging from simple dipoles to high-rank tensors with complex symmetries such as the elastic constant tensor.
  • Equivariant atomistic models can be implemented entirely in Cartesian coordinates while preserving physical tensor symmetries.
  • The same framework supports both scalar potential predictions and vector or tensor property predictions without separate code paths.

Where Pith is reading between the lines

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

  • Cartesian models of this type may integrate more directly with existing molecular dynamics codes that already store coordinates and forces in Cartesian form.
  • The decomposition scheme could be applied to other high-rank tensors that appear in condensed-matter problems, such as piezoelectric or magnetic susceptibility tensors.
  • Hybrid models that combine Cartesian natural tensors with selected spherical components might be tested for further accuracy gains on specific properties.
  • The approach opens a route to parameter-free derivations of symmetry constraints that could be checked against tabulated crystal data.

Load-bearing premise

The assumption that the irreducible Cartesian natural tensors supply a systematic and complete scheme for decomposition and reconstruction that works across the tested atomistic tasks.

What would settle it

Running CarNet and a leading spherical-tensor model on the same benchmark set of elastic constant tensor predictions and checking whether the mean absolute errors remain within a few percent of each other.

Figures

Figures reproduced from arXiv: 2510.04015 by A. S. L. Subrahmanyam Pattamatta, Boyu Wang, David J. Srolovitz, Mingjian Wen, Qun Chen.

Figure 1
Figure 1. Figure 1: FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
read the original abstract

Atomistic machine learning (ML) is a powerful tool for accurate and efficient investigation of material behavior at the atomic scale. While such models have been constructed within Cartesian space to harness geometric information and preserve intuitive physical representations, they face challenges in providing a systematic, irreducible-representation-based formalism analogous to the spherical-tensor machinery widely used in equivariant networks. We address these challenges by proposing Cartesian Natural Tensor Networks (CarNet) as a general framework for atomistic ML. We present the theory of irreducible representations using Cartesian natural tensors, covering their construction and their products, and further develop a systematic scheme for the decomposition and reconstruction of high-rank physical tensors. Leveraging this machinery, we then develop an equivariant Cartesian model and demonstrate its strong performance across diverse atomistic ML tasks. CarNet delivers machine learning interatomic potentials (MLIPs) for both materials and molecular systems, with performance on par with leading spherical-tensor models. Furthermore, it enables the construction of accurate and efficient structure--property relationships for tensorial quantities ranging from simple properties like the dipole moment to high-rank tensors with complex symmetries, such as the elastic constant tensor. This work strengthens Cartesian approaches for advanced atomistic ML in the understanding and design of new materials.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces Cartesian Natural Tensor Networks (CarNet) as a general framework for atomistic machine learning. It develops the theory of irreducible representations via Cartesian natural tensors, including their construction, products, and a systematic scheme for decomposition and reconstruction of high-rank physical tensors. An equivariant Cartesian model is constructed and applied to machine learning interatomic potentials (MLIPs) for both materials and molecular systems, as well as to structure-property predictions for tensorial quantities ranging from dipole moments to the elastic constant tensor, with performance reported on par with leading spherical-tensor models.

Significance. If the central theoretical claims hold, this work would provide a viable Cartesian alternative to spherical-tensor equivariant networks, potentially offering more intuitive representations for physical tensor properties in materials science. The reported benchmarks across MLIPs and high-rank tensor targets constitute a practical strength, and the emphasis on systematic decomposition/reconstruction addresses a known gap in Cartesian approaches.

major comments (2)
  1. [§4.2] §4.2 (Decomposition and reconstruction scheme): The claim of a systematic, complete scheme for arbitrary-rank tensors with complex symmetries is central to both the equivariant model and the elastic-tensor results, yet the manuscript provides no explicit numerical test showing that reconstruction of a rank-4 elastic tensor recovers the original components to machine precision while exactly respecting Voigt symmetries. Without this verification, the completeness and irreducibility of the basis remain unconfirmed for the highest-symmetry case highlighted in the abstract.
  2. [§5.3] §5.3 (Elastic constant tensor results): The reported accuracy for the elastic tensor is presented without direct comparison to spherical-tensor baselines on the same dataset or error bars from multiple independent runs; this weakens the claim that performance is on par with leading models for this specific high-rank target.
minor comments (2)
  1. [§2] The definition of 'natural tensor' and its relation to standard Cartesian tensors could be stated more explicitly in the opening of §2 to aid readers unfamiliar with the construction.
  2. [Figure 3] Figure 3 (tensor decomposition illustration) would benefit from an additional panel showing the reconstruction error for a rank-4 example.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and constructive feedback on our manuscript. We are pleased that the referee recognizes the potential of CarNet as a Cartesian alternative to spherical-tensor methods. Below, we provide point-by-point responses to the major comments and outline the revisions we will make to address them.

read point-by-point responses

  1. Referee: [§4.2] §4.2 (Decomposition and reconstruction scheme): The claim of a systematic, complete scheme for arbitrary-rank tensors with complex symmetries is central to both the equivariant model and the elastic-tensor results, yet the manuscript provides no explicit numerical test showing that reconstruction of a rank-4 elastic tensor recovers the original components to machine precision while exactly respecting Voigt symmetries. Without this verification, the completeness and irreducibility of the basis remain unconfirmed for the highest-symmetry case highlighted in the abstract.

    Authors: We agree that an explicit numerical verification would strengthen the presentation of the decomposition and reconstruction scheme. Although the mathematical construction in §4.2 ensures that the irreducible Cartesian natural tensors form a complete and irreducible basis, and the reconstruction is designed to exactly respect the symmetries (including Voigt for rank-4), we did not include a specific numerical example for the elastic tensor in the original submission. In the revised manuscript, we will add a dedicated numerical test in §4.2. This test will demonstrate the reconstruction of a rank-4 tensor with Voigt symmetries, showing recovery of original components to machine precision (typically 10^{-14} or better) and exact preservation of the symmetry relations. We believe this will confirm the claims for the highest-symmetry case. revision: yes

  2. Referee: [§5.3] §5.3 (Elastic constant tensor results): The reported accuracy for the elastic tensor is presented without direct comparison to spherical-tensor baselines on the same dataset or error bars from multiple independent runs; this weakens the claim that performance is on par with leading models for this specific high-rank target.

    Authors: We thank the referee for this suggestion to improve the robustness of our claims. The elastic constant results are benchmarked against published performances of spherical-tensor models on comparable datasets, as referenced in the manuscript. However, to provide a more direct comparison, we will include results from running a representative spherical-tensor model (such as MACE or NequIP) on the exact same dataset used in our experiments. Additionally, we will conduct multiple independent training runs with different initializations and report the mean and standard deviation of the errors to provide error bars. These updates will be added to §5.3 and the corresponding figures/tables. revision: yes

Circularity Check

0 steps flagged

No significant circularity; theory presented as independent framework

full rationale

The paper introduces a new formalism for irreducible Cartesian natural tensors, including construction, products, and decomposition/reconstruction schemes for physical tensors. The central claims rest on this proposed theory rather than on fitted parameters or self-citations that reduce the result to inputs by construction. No load-bearing step equates a prediction to a fit or renames a known result via self-citation chain. The equivariant model and performance demonstrations follow from the stated machinery without evident self-definitional loops. This yields a minor score reflecting normal self-citation of prior work that is not load-bearing for the core derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the new theoretical construction of Cartesian natural tensors and their use in equivariant models; limited information from abstract yields minimal ledger entries.

axioms (1)
  • domain assumption Cartesian natural tensors admit a systematic construction, product rules, and decomposition/reconstruction scheme for high-rank physical tensors that preserves irreducibility and equivariance.
    This premise is invoked to address challenges in Cartesian space and enable the equivariant model as described in the abstract.
invented entities (1)
  • Cartesian Natural Tensor Networks (CarNet) no independent evidence
    purpose: General framework for atomistic ML using irreducible Cartesian tensors.
    New named framework and associated tensor machinery introduced to provide an alternative to spherical-tensor approaches.

pith-pipeline@v0.9.0 · 5774 in / 1319 out tokens · 53303 ms · 2026-05-18T11:02:01.548490+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Foundation/AlexanderDuality.lean alexander_duality_circle_linking echoes
    ?
    echoes

    ECHOES: this paper passage has the same mathematical shape or conceptual pattern as the Recognition theorem, but is not a direct formal dependency.

    A natural tensor is a fully symmetric Cartesian tensor whose traces vanish on any pair of indices... furnishes a 2n+1-dimensional irreducible representation of the special orthogonal group SO(3)... We propose a systematic approach to decomposing and reconstructing physical tensors of arbitrary rank and symmetry using natural tensors... QR factorization combined with symmetry-informed elimination.

  • IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean absolute_floor_iff_bare_distinguishability echoes
    ?
    echoes

    ECHOES: this paper passage has the same mathematical shape or conceptual pattern as the Recognition theorem, but is not a direct formal dependency.

    natural tensors... intrinsic geometric construction rather than a regular tensor... construction, products, and a systematic scheme for the decomposition and reconstruction of high-rank physical tensors

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

67 extracted references · 67 canonical work pages · 6 internal anchors

  1. [1]

    O. T. Unke, S. Chmiela, H. E. Sauceda, M. Gastegger, I. Poltavsky, K. T. Sch¨ utt, A. Tkatchenko, and K.-R. M¨ uller, Machine learning force fields, Chemical Reviews 121, 10142 (2021)

    work page 2021

  2. [2]

    Musil, A

    F. Musil, A. Grisafi, A. P. Bart´ ok, C. Ortner, G. Cs´ anyi, and M. Ceriotti, Physics-inspired structural representa- tions for molecules and materials, Chemical Reviews121, 9759 (2021)

    work page 2021

  3. [3]

    M. Wen, S. M. Blau, E. W. C. Spotte-Smith, S. Dwarak- nath, and K. A. Persson, Bondnet: a graph neural net- work for the prediction of bond dissociation energies for charged molecules, Chemical Science12, 1858 (2020)

    work page 2020

  4. [4]

    R. Ruff, P. Reiser, J. St¨ uhmer, and P. Friederich, Con- nectivity optimized nested line graph networks for crystal structures, Digital Discovery3, 594 (2024)

    work page 2024

  5. [5]

    M. Veit, D. M. Wilkins, Y. Yang, R. A. DiStasio, and M. Ceriotti, Predicting molecular dipole moments by combining atomic partial charges and atomic dipoles, The Journal of Chemical Physics153, 10.1063/5.0009106 (2020). 12

    work page doi:10.1063/5.0009106 2020

  6. [6]

    M. Wen, M. K. Horton, J. M. Munro, P. Huck, and K. A. Persson, An equivariant graph neural network for the elasticity tensors of all seven crystal systems, Digital Dis- covery3, 869 (2024)

    work page 2024

  7. [7]

    Z. Yao, Y. Lum, A. Johnston, L. M. Mejia-Mendoza, X. Zhou, Y. Wen, A. Aspuru-Guzik, E. H. Sargent, and Z. W. Seh, Machine learning for a sustainable energy future, Nature Reviews Materials8, 202 (2023)

    work page 2023

  8. [8]

    M. Wen, E. W. C. Spotte-Smith, S. M. Blau, M. J. Mc- Dermott, A. S. Krishnapriyan, and K. A. Persson, Chem- ical reaction networks and opportunities for machine learning, Nature Computational Science3, 12 (2023)

    work page 2023

  9. [9]

    H. Li, Y. Jiao, K. Davey, and S.-Z. Qiao, Data-driven machine learning for understanding surface structures of heterogeneous catalysts, Angewandte Chemie Inter- national Edition62, e202216383 (2023)

    work page 2023

  10. [10]

    T. Mou, H. S. Pillai, S. Wang, M. Wan, X. Han, N. M. Schweitzer, F. Che, and H. Xin, Bridging the complexity gap in computational heterogeneous catalysis with ma- chine learning, Nature Catalysis6, 122 (2023)

    work page 2023

  11. [11]

    Vamathevan, D

    J. Vamathevan, D. Clark, P. Czodrowski, I. Dunham, E. Ferran, G. Lee, B. Li, A. Madabhushi, P. Shah, M. Spitzer,et al., Applications of machine learning in drug discovery and development, Nature reviews Drug discovery18, 463 (2019)

    work page 2019

  12. [12]

    Ekins, A

    S. Ekins, A. C. Puhl, K. M. Zorn, T. R. Lane, D. P. Russo, J. J. Klein, A. J. Hickey, and A. M. Clark, Ex- ploiting machine learning for end-to-end drug discovery and development, Nature materials18, 435 (2019)

    work page 2019

  13. [13]

    Behler and M

    J. Behler and M. Parrinello, Generalized neural-network representation of high-dimensional potential-energy sur- faces, Physical review letters98, 146401 (2007)

    work page 2007

  14. [14]

    M. Rupp, A. Tkatchenko, K.-R. M¨ uller, and O. A. Von Lilienfeld, Fast and accurate modeling of molecu- lar atomization energies with machine learning, Physical review letters108, 058301 (2012)

    work page 2012

  15. [15]

    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, Physical Review Letters 120, 143001 (2018)

    work page 2018

  16. [16]

    A. P. Thompson, L. P. Swiler, C. R. Trott, S. M. Foiles, and G. J. Tucker, Spectral neighbor analysis method for automated generation of quantum-accurate interatomic potentials, Journal of Computational Physics285, 316 (2015)

    work page 2015

  17. [17]

    Drautz, Atomic cluster expansion for accurate and transferable interatomic potentials, Physical Review B 99, 014104 (2019)

    R. Drautz, Atomic cluster expansion for accurate and transferable interatomic potentials, Physical Review B 99, 014104 (2019)

    work page 2019

  18. [18]

    Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point clouds

    N. Thomas, T. Smidt, S. Kearnes, L. Yang, L. Li, K. Kohlhoff, and P. Riley, Tensor field networks: Rotation-and translation-equivariant neural networks for 3d point clouds, arXiv preprint arXiv:1802.08219 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  19. [19]

    Batzner, A

    S. Batzner, A. Musaelian, L. Sun, M. Geiger, J. P. Mailoa, M. Kornbluth, N. Molinari, T. E. Smidt, and B. Kozinsky, E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials, Nature Communications13, 1 (2022)

    work page 2022

  20. [20]

    Batatia, D

    I. Batatia, D. P. Kovacs, G. Simm, C. Ortner, and G. Cs´ anyi, Mace: Higher order equivariant message pass- ing neural networks for fast and accurate force fields, Advances in Neural Information Processing Systems35, 11423 (2022)

    work page 2022

  21. [21]

    Bochkarev, Y

    A. Bochkarev, Y. Lysogorskiy, and R. Drautz, Graph atomic cluster expansion for semilocal interactions be- yond equivariant message passing, Physical Review X14, 021036 (2024)

    work page 2024

  22. [22]

    A. V. Shapeev, Moment tensor potentials: A class of sys- tematically improvable interatomic potentials, Multiscale Modeling & Simulation 10.1137/15M1054183 (2016)

    work page doi:10.1137/15m1054183 2016

  23. [23]

    Zhang, J

    Y. Zhang, J. Xia, and B. Jiang, Physically motivated re- cursively embedded atom neural networks: Incorporating local completeness and nonlocality, Physical Review Let- ters127, 156002 (2021)

    work page 2021

  24. [24]

    Cheng, Cartesian atomic cluster expansion for ma- chine learning interatomic potentials, npj Computational Materials10, 1 (2024)

    B. Cheng, Cartesian atomic cluster expansion for ma- chine learning interatomic potentials, npj Computational Materials10, 1 (2024)

    work page 2024

  25. [25]

    Zaverkin, F

    V. Zaverkin, F. Alesiani, T. Maruyama, F. Errica, H. Christiansen, M. Takamoto, N. Weber, and M. Niepert, Higher-rank irreducible cartesian ten- sors for equivariant message passing, ArXiv e-prints 10.48550/arXiv.2405.14253 (2024), 2405.14253

    work page doi:10.48550/arxiv.2405.14253 2024

  26. [26]

    Wen, W.-F

    M. Wen, W.-F. Huang, J. Dai, and S. Adhikari, Cartesian atomic moment machine learning interatomic potentials, npj Computational Materials11, 128 (2025)

    work page 2025

  27. [27]

    Gastegger, K

    M. Gastegger, K. T. Sch¨ utt, and K.-R. M¨ uller, Machine learning of solvent effects on molecular spectra and reac- tions, Chemical Science12, 11473 (2021)

    work page 2021

  28. [28]

    Simeon and G

    G. Simeon and G. De Fabritiis, Tensornet: Cartesian tensor representations for efficient learning of molecu- lar potentials, inAdvances in Neural Information Pro- cessing Systems, Vol. 36, edited by A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine (2023) pp. 37334–37353

    work page 2023

  29. [29]

    Kundu, I

    P. Kundu, I. Cohen, D. Dowling, and J. Capecelatro, Fluid Mechanics(Academic Press, 2024)

    work page 2024

  30. [30]

    Jones, I

    D. Jones, I. Sneddon, S. Ulam, and M. Stark,The Theory of Electromagnetism(Pergamon, 2013)

    work page 2013

  31. [31]

    J. A. R. Coope, R. F. Snider, and F. R. McCourt, Ir- reducible cartesian tensors, Journal of Chemical Physics 43, 2269 (1965)

    work page 1965

  32. [32]

    J. A. R. Coope and R. F. Snider, Irreducible cartesian tensors. ii. general formulation, Journal of Mathematical Physics11, 1003 (1970)

    work page 1970

  33. [33]

    J. A. R. Coope, Irreducible cartesian tensors. iii. clebsch- gordan reduction, Journal of Mathematical Physics11, 1591 (1970)

    work page 1970

  34. [34]

    Zee,Group Theory in a Nutshell for Physicists (Princeton University Press, Princeton, NJ, USA, 2016)

    A. Zee,Group Theory in a Nutshell for Physicists (Princeton University Press, Princeton, NJ, USA, 2016)

    work page 2016

  35. [35]

    Jerphagnon, D

    J. Jerphagnon, D. Chemla, and R. Bonneville, The de- scription of the physical properties of condensed matter using irreducible tensors, Advances in Physics27, 609 (1978)

    work page 1978

  36. [36]

    D. R. Lehman and W. C. Parke, Angular reduction in multiparticle matrix elements, Journal of Mathematical Physics30, 2797 (1989)

    work page 1989

  37. [37]

    Edmonds,Angular Momentum in Quantum Mechan- ics, Investigations in Physics Series (Princeton University Press, 1996)

    A. Edmonds,Angular Momentum in Quantum Mechan- ics, Investigations in Physics Series (Princeton University Press, 1996)

    work page 1996

  38. [38]

    P. R. Morris, Averaging fourth-rank tensors with weight functions, Journal of Applied Physics40, 447 (1969)

    work page 1969

  39. [39]

    R. A. Harris, W. M. McClain, and C. F. Sloane, On the theory of polarized light scattering from dilute polymer solutions, Molecular Physics28, 381 (1974)

    work page 1974

  40. [40]

    M. C. Russ and C. A. Hollingsworth, Properties of three- dimensional cartesian tensors. i. some properties of ir- reducible tensors, Journal of Mathematical Physics23, 1025 (1982). 13

    work page 1982

  41. [41]

    Golub and C

    G. Golub and C. Van Loan,Matrix Computations, Johns Hopkins Studies in the Mathematical Sciences (Johns Hopkins University Press, 1996)

    work page 1996

  42. [42]

    L. B. Skinner, C. J. Benmore, J. C. Neuefeind, and J. B. Parise, The structure of water around the compressibil- ity minimum, Journal of Chemical Physics141, 214507 (2014)

    work page 2014

  43. [43]

    A blender-powered toolkit for scientific visualization of atoms, molecules, crystals, and beyond,https:// github.com/spattamatta/atomviz(2025)

    work page 2025

  44. [44]

    Cheng, E

    B. Cheng, E. A. Engel, J. Behler, C. Dellago, and M. Ce- riotti, Ab initio thermodynamics of liquid and solid wa- ter, Proceedings of the National Academy of Sciences 116, 1110 (2019)

    work page 2019

  45. [45]

    X. Fu, Z. Wu, W. Wang, T. Xie, S. Keten, R. Gomez- Bombarelli, and T. S. Jaakkola, Forces are not enough: Benchmark and critical evaluation for machine learning force fields with molecular simulations, Transactions on Machine Learning Research (2023)

    work page 2023

  46. [46]

    Soper, F

    A. Soper, F. Bruni, and M. Ricci, Site–site pair correla- tion functions of water from 25 to 400 c: Revised analysis of new and old diffraction data, The Journal of chemical physics106, 247 (1997)

    work page 1997

  47. [47]

    Sch¨ utt, O

    K. Sch¨ utt, O. Unke, and M. Gastegger, Equivariant mes- sage passing for the prediction of tensorial properties and molecular spectra, inInternational Conference on Ma- chine Learning, Vol. 139 (PMLR, 2021) pp. 9377–9388

    work page 2021

  48. [48]

    Voigt,Lehrbuch der kristallphysik (mit Ausschluss der Kristalloptik)(Vieweg+Teubner Verlag Wiesbade, 1966)

    W. Voigt,Lehrbuch der kristallphysik (mit Ausschluss der Kristalloptik)(Vieweg+Teubner Verlag Wiesbade, 1966)

    work page 1966

  49. [49]

    Hill, The elastic behaviour of a crystalline aggregate, Proceedings of the Physical Society

    R. Hill, The elastic behaviour of a crystalline aggregate, Proceedings of the Physical Society. Section A65, 349 (1952)

    work page 1952

  50. [50]

    Wen, Irreducible Cartesian tensors: Theory, compu- tation, and applications, In Preparation (2025)

    M. Wen, Irreducible Cartesian tensors: Theory, compu- tation, and applications, In Preparation (2025)

    work page 2025

  51. [51]

    B. Deng, P. Zhong, K. Jun, J. Riebesell, K. Han, C. J. Bartel, and G. Ceder, Chgnet as a pretrained universal neural network potential for charge-informed atomistic modelling, Nature Machine Intelligence5, 1031 (2023)

    work page 2023

  52. [52]

    Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models

    L. Barroso-Luque, M. Shuaibi, X. Fu, B. M. Wood, M. Dzamba, M. Gao, A. Rizvi, C. L. Zitnick, and Z. W. Ulissi, Open materials 2024 (omat24) inorganic materi- als dataset and models, arXiv preprint arXiv:2410.12771 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  53. [53]

    A graph neural network for the era of large atomistic models.arXiv:2506.01686, 2025

    D. Zhang, A. Peng, C. Cai, W. Li, Y. Zhou, J. Zeng, M. Guo, C. Zhang, B. Li, H. Jiang,et al., Graph neu- ral network model for the era of large atomistic models, arXiv preprint arXiv:2506.01686 (2025)

    work page arXiv 2025

  54. [54]

    Dusson, M

    G. Dusson, M. Bachmayr, G. Cs´ anyi, R. Drautz, S. Etter, C. van der Oord, and C. Ortner, Atomic cluster expan- sion: Completeness, efficiency and stability, Journal of Computational Physics454, 110946 (2022)

    work page 2022

  55. [55]

    cuequivariance library,https://github.com/NVIDIA/ cuEquivariance(2025)

    work page 2025

  56. [56]

    Bharadwaj, A

    V. Bharadwaj, A. Glover, A. Buluc, and J. Demmel, An efficient sparse kernel generator for o(3)-equivariant deep networks, inSIAM Conference on Applied and Compu- tational Discrete Algorithms (ACDA25)(Society for In- dustrial and Applied Mathematics, 2025)

    work page 2025

  57. [57]

    X. Chen, N. Ravikumar, Y. Xia, R. Attar, A. Diaz- Pinto, S. K. Piechnik, S. Neubauer, S. E. Petersen, and A. F. Frangi, Shape registration with learned deforma- tions for 3d shape reconstruction from sparse and incom- plete point clouds, Medical image analysis74, 102228 (2021)

    work page 2021

  58. [58]

    Hierarchical Feature Learning for Medical Point Clouds via State Space Model

    G. Zhang, J. Yang, and Y. Li, Hierarchical feature learn- ing for medical point clouds via state space model, arXiv preprint arXiv:2504.13015 10.48550/arXiv.2504.13015 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2504.13015 2025

  59. [59]

    C. R. Qi, H. Su, K. Mo, and L. J. Guibas, Pointnet: Deep learning on point sets for 3d classification and segmenta- tion, inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)(2017)

    work page 2017

  60. [60]

    Shi and R

    W. Shi and R. Rajkumar, Point-gnn: Graph neural net- work for 3d object detection in a point cloud, inProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)(2020)

    work page 2020

  61. [61]

    M. K. Horton, P. Huck, R. X. Yang, J. M. Munro, S. Dwaraknath, A. M. Ganose, R. S. Kingsbury, M. Wen, J. X. Shen, T. S. Mathis, A. D. Kaplan, K. Berket, J. Riebesell, J. George, A. S. Rosen, E. W. C. Spotte- Smith, M. J. McDermott, O. A. Cohen, A. Dunn, M. C. Kuner, G.-M. Rignanese, G. Petretto, D. Waroquiers, S. M. Griffin, J. B. Neaton, D. C. Chrzan, M....

    work page doi:10.1038/s41563-025-02272-0 2025

  62. [62]

    I. S. Novikov, K. Gubaev, E. V. Podryabinkin, and A. V. Shapeev, The mlip package: moment tensor potentials with mpi and active learning, Machine Learning: Science and Technology2, 025002 (2020)

    work page 2020

  63. [63]

    K. He, X. Zhang, S. Ren, and J. Sun, Deep resid- ual learning for image recognition, ArXiv e-prints 10.48550/arXiv.1512.03385 (2015), 1512.03385

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1512.03385 2015

  64. [64]

    PyTorch: An Imperative Style, High-Performance Deep Learning Library

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. K¨ opf, E. Yang, Z. DeVito, M. Rai- son, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, and S. Chintala, Pytorch: An imperative style, high-performance deep learning library, ArXiv e-prints 10.48550/arXiv.1912.01703 (20...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1912.01703 1912

  65. [65]

    pytorch-lightning (2025), [Online; accessed 2. Sept. 2025]

    work page 2025

  66. [66]

    D. P. Kingma and J. Ba, Adam: A method for stochastic optimization, ArXiv e-prints 10.48550/arXiv.1412.6980 (2014), 1412.6980

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1412.6980 2014

  67. [67]

    A. H. Larsen, J. J. Mortensen, J. Blomqvist, I. E. Castelli, R. Christensen, M. Du lak, J. Friis, M. N. Groves, B. Hammer, C. Hargus, E. D. Hermes, P. C. Jennings, P. B. Jensen, J. Kermode, J. R. Kitchin, E. L. Kolsbjerg, J. Kubal, K. Kaasbjerg, S. Lysgaard, J. B. Maronsson, T. Maxson, T. Olsen, L. Pastewka, A. Pe- terson, C. Rostgaard, J. Schiøtz, O. Sch...

    work page 2017