Pith. sign in

REVIEW 3 major objections 5 minor 70 references

Using the lattice bispectrum—a rotation- and convention-invariant reciprocal-space descriptor—as the machine-learning target instead of the six cell parameters reduces powder XRD lattice parameter error by roughly a factor of four.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 06:33 UTC pith:7CYC6XVL

load-bearing objection Novel descriptor with real external-data support, but the headline MP-20 gains are probably inflated because test structures may be in the inversion lookup. the 3 major comments →

arxiv 2607.21829 v1 pith:7CYC6XVL submitted 2026-07-23 physics.comp-ph cond-mat.mtrl-sci

Learning Lattice Parameters from Powder X-Ray Diffraction Data Using Invariants

Elyssa Hofgard , Kyucheol Min , Nofit Segal , David W. Mittan-Moreau , Aria Mansouri Tehrani , Vanessa Oklejas , Jigyasa Nigam , Daniel W. Paley
show 2 more authors
Aaron S. Brewster Tess Smidt
This is my paper
classification physics.comp-ph cond-mat.mtrl-sci
keywords powder X-ray diffractionlattice parametersmachine learningbispectruminvariant representationreciprocal spacetransformerunit cell determination
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that machine learning can determine unit cell parameters from powder X-ray diffraction far more accurately if the model is trained to predict a convention-independent reciprocal-space invariant—the lattice bispectrum—rather than the six conventional cell parameters (a,b,c,alpha,beta,gamma). With identical transformer architecture on a standard 20-atom benchmark set, this swap reduces length error from about 11% to 2.4% mean absolute percentage error and angle error from 12.7% to 3.1%. The bispectrum is differentiable and can be inverted to lattice vectors by a dynamic-programming nearest-neighbor search followed by local optimization. A sympathetic reader would care because it suggests a representation fix, not an architecture fix, may close the accuracy gap that has limited direct parameter prediction, especially for low-symmetry crystals and dominant-zone patterns.

Core claim

The central claim is that the choice of regression target matters more than model architecture for lattice-parameter prediction from powder XRD: the lattice bispectrum—a spherical-harmonic descriptor of the reciprocal lattice that is invariant to rotation, inversion, and primitive-cell reindexing—yields substantially lower lattice parameter error when predicted by a transformer and then inverted, compared with directly predicting (a,b,c,alpha,beta,gamma) with the same transformer. On the standard 20-atom benchmark test set, bispectrum inversion achieves 2.44% length MAPE and 3.07% angle MAPE versus 11.18% and 12.74% for direct prediction; similar trends hold on an augmented full dataset and

What carries the argument

The lattice bispectrum: a descriptor formed by expanding the reciprocal-lattice density rho(k) (Dirac combs at reciprocal lattice points within a cutoff kmax) in spherical harmonics and radial basis functions, then coupling the expansion coefficients via Clebsch-Gordan tensor products to obtain rotation/inversion/translation/permutation invariant scalars. It replaces the piecewise, convention-dependent six cell parameters as the neural network's output, and because the bispectrum calculation is differentiable, it can be inverted: a nearest-neighbor lookup in a database of precomputed bispectra initializes the lattice, and L-BFGS optimization refines it by minimizing the bispectrum residual.

Load-bearing premise

The inversion step seeds its search with the nearest precomputed bispectrum from a database built from the same dataset the model is tested on, and the paper does not say that test structures were excluded from that database; if they were not, the reported accuracy partly reflects retrieval of near-ground-truth cells rather than the learned representation.

What would settle it

Check whether the precomputed bispectrum lookup database contains any test structures from the benchmark sets (e.g., by material id or reduced formula). If it does, rerun the inversion with those entries removed and compare length and angle MAPE; if the gap to direct prediction collapses, the representation's advantage is substantially an artifact of test-set leakage.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Lattice parameters can be recovered from powder XRD with roughly four times lower length error and angle error than direct six-parameter regression, without changing the model architecture.
  • The representation gains concentrate where the six-parameter target is most discontinuous: low-symmetry (triclinic and monoclinic) systems and dominant-zone patterns, where the joint length-and-angle recovery rate at a 5% threshold jumps from 5.0% to 70.7% on dominant zones.
  • The invariant target remains beneficial when training data is augmented with simulated experimental artifacts, and on experimental mineral powder patterns the approach matches a much larger structure-generation model.
  • Being differentiable and invertible, the bispectrum can serve as an intermediate representation for other crystallographic machine-learning tasks, including structure generation.

Where Pith is reading between the lines

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

  • Editorial: the reported end-to-end accuracy depends on excluding test structures from the precomputed bispectrum lookup database used to initialize inversion; the paper does not state this exclusion, so replication should verify it.
  • Editorial: if the smoothness hypothesis is correct, the bispectrum may also improve gradient-based refinement and generative modeling of lattices, where the roughness of the six-parameter landscape is known to cause local optima.
  • Editorial: a natural extension is to include structure factor phases or amplitudes in the descriptor, which would add atomic-basis information and could push the method toward full structure solution rather than lattice parameters only.
  • Editorial: a testable consequence is that prediction error should track the bispectrum's L2 distance rather than raw parameter distance; the paper's error analysis supports this, but a systematic study across crystal systems would sharpen it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes a reciprocal-space invariant representation of the crystal lattice, the lattice bispectrum, and uses it as the target of a transformer trained on powder XRD patterns. The trained model outputs a predicted bispectrum, which is then inverted to lattice parameters via L-BFGS refinement initialized by a nearest-neighbor lookup over precomputed Materials Project bispectra. The central claim is that, with identical architecture, predicting the bispectrum and inverting it yields substantially lower lattice-parameter errors than directly predicting (a,b,c,α,β,γ), with MP-20 length MAPE dropping from 11.18% to 2.44% and angle MAPE from 12.74% to 3.07% (Abstract, Table 1). The paper also benchmarks against Crystalyze and AlphaDiffract on RRUFF experimental data and reports gains on dominant-zone structures.

Significance. If the central claim holds, the bispectrum target would be a valuable contribution to ML-based lattice determination: it is continuous under lattice deformation, independent of primitive-cell convention, differentiable, and the paper provides a practical inversion scheme. The systematic comparison with direct prediction, the use of physically motivated augmentation, and the evaluation on experimental RRUFF data are strengths. However, the headline quantitative claims are currently confounded by the inversion initialization described in Section 4.2, and the reported gains may reflect database retrieval rather than learned representation. The contribution is potentially significant, but the evidence needs to be unconfounded before the claims can be accepted.

major comments (3)
  1. [Section 4.2, Figure 6, Table 1] The inversion step initializes v_guess from "precalculated bispectra from the Materials Project database" via an L2 nearest-neighbor lookup. The MP-20 test set (Section 5.1) and the MP-Full test set (Section 5.2) are subsets of the Materials Project. The paper never states that these test structures were excluded from the lookup database. If a test structure is present in that database, then for any XRD pattern whose predicted bispectrum is even roughly correct, the exact ground-truth bispectrum is a candidate neighbor; the nearest-neighbor initialization then returns essentially the true lattice, and L-BFGS refinement starts at or near the answer. The direct-prediction baseline has no analogous lookup, so the reported improvements in Table 1 (e.g., MP-20 length MAPE 11.18% to 2.44%) are not interpretable as evidence that the bispectrum representation improves learned prediction. This is
  2. [Section 4.2, Section 5.3] The same circularity affects the dominant-zone analysis in Section 5.3, which uses the MP-Full test set and the MP-Full-Aug model. If the inversion database contains the test structures, the 70.7% joint success rate on dominant-zone structures may again reflect retrieval. The paper says the database contains "precalculated bispectra from the Materials Project" without specifying a split; this needs to be clarified and the analysis rerun with an exclusion or no-lookup control.
  3. [Section 4.2, Data Availability] The paper acknowledges that the bispectrum is not mathematically complete ("While the bispectrum is not mathematically complete (i.e. it is not a one-to-one mapping), it can be empirically inverted"). The empirical inversion claim is supported by a sensitivity analysis to Gaussian noise in Figure S1 and a citation to Nigam et al. 2026. However, in the actual pipeline the inversion is initialized from a database that may contain the ground truth. Unless the lookup is removed or restricted, the inversion results do not demonstrate that the bispectrum can be inverted from the model's predicted coefficients in the absence of near-target initialization. Please report inversion accuracy with a lookup-free initialization or with the database restricted to the training split.
minor comments (5)
  1. [Section 4.1] Typo: "the arises" should be "the discontinuity that arises".
  2. [Section 5.1] Typo: "mostly likely" should be "most likely".
  3. [Section 4.2] The term "dynamic programming" is used "loosely" to describe a nearest-neighbor lookup. This is fine, but it should be flagged earlier and more prominently, because the reader may otherwise expect a sequential decomposition algorithm.
  4. [Section 4.1, Figure 7] The descriptor dimension is given as (Nr, sum_l N_allowed) and later as output dimension 350 = 10 × 35. Clarify whether the 35 components include both scalar and pseudoscalar blocks, and whether the pseudoscalar components are identically zero for Bravais lattices and therefore masked in the loss.
  5. [Figure captions] Figure captions use inconsistent notation such as "MP20 Aug" and "MPFull Aug"; make consistent with "MP-20 Aug" and "MP-Full Aug" used in the text.

Circularity Check

3 steps flagged

Bispectrum inversion initializes from a Materials Project database that, as described, includes the MP-20/MP-Full test population; unless test cells are excluded from the lookup, the headline gains can reduce to database retrieval, and the inversion premise rests on a same-author citation.

specific steps
  1. fitted input called prediction [Section 4.2 (Inversion of Lattice Bispectrum), applied in Table 1]
    "Thus, we initialize v_guess = (a, b, c, α, β, γ)guess using precalculated bispectra from the Materials Project database (with distance between bispectra measured by the L2 norm)."

    MP-20 and MP-Full test sets are described as subsets of the Materials Project (Section 5), and the inversion database is described as precalculated bispectra from the Materials Project database, with no statement that test structures were excluded. For a synthetic test pattern whose predicted bispectrum is only roughly correct, the true bispectrum is therefore a candidate neighbor; L2-nearest lookup yields the true cell as v_guess, and L-BFGS refinement starts essentially at the answer. The direct-prediction baseline has no analogous lookup, so the Table 1 improvements (e.g., length MAPE 11.18% to 2.44%) may measure database retrieval rather than the learned target. An ablation excluding test entries from the lookup is required before the comparison is interpretable.

  2. self citation load bearing [Section 4.2 (Inversion of Lattice Bispectrum)]
    "While the bispectrum is not mathematically complete (i.e. it is not a one-to-one mapping), it can be empirically inverted (Nigam et al., 2026)."

    Nigam et al. 2026 shares authors with this paper (Nigam, Mansouri Tehrani, Smidt) and is cited as the premise that the bispectrum is invertible. The inversion pipeline (database lookup plus gradient refinement) depends on this premise, but the cited work is not machine-checked or independently verified in the present paper, and no theorem or external benchmark is supplied. Thus the feasibility of the recovery step is supported by a same-author citation rather than an independent mathematical fact.

  3. other [Supplementary Section S5.1 (Limitations and Future Directions in Augmentation)]
    "This introduces a degree of circularity: strategies designed by inspecting typical lab-source diffraction data will generalize well to data of that kind, but this is a narrower claim than generalization to experimental data in general."

    The paper itself admits a degree of circularity in its augmentation-driven experimental generalization claim: the augmentations were designed by inspecting lab-source-like data, so RRUFF performance reflects that prior design rather than fully independent generalization. This is an explicit limitation and is honestly narrowed, but it is a circular component of the experimental-data part of the paper and is weighed in the verdict.

full rationale

The core synthetic-data comparison is not fully self-contained as written. The inversion step uses a nearest-neighbor lookup over precalculated Materials Project bispectra, while the MP-20 and MP-Full test sets are Materials Project subsets; the paper never states that test structures were removed from the lookup. If they are present, a modestly accurate predicted bispectrum makes the true lattice the nearest database entry, so the reported bispectrum-vs-direct gains reduce in part to database retrieval rather than to properties of the learned invariant target. The paper's invocation of empirical invertibility via Nigam et al. 2026 (overlapping authors, not independently verified) compounds the issue, although the descriptor itself and the direct-vs-bispectrum learning comparison have substantial independent methodological content. The paper also explicitly acknowledges a 'degree of circularity' in its augmentation strategy (Section S5.1), which primarily limits the experimental-generalization claims. Overall this is a partial circularity in evaluation and inversion, not a fully definitional derivation, so a score of 6 is appropriate rather than 8-10.

Axiom & Free-Parameter Ledger

6 free parameters · 4 axioms · 0 invented entities

No new physical entities are postulated. The lattice bispectrum is a mathematical descriptor, not a new physical quantity.

free parameters (6)
  • kmax = 2/3 Å⁻¹ (Cu Kα, 2θmax=60°)
    Reciprocal-space cutoff; chosen so nearly all MP lattices have bmax below cutoff (Figure 5). Not fitted to errors, but affects descriptor content and is user-set.
  • lmax = 6
    Spherical-harmonic truncation; selected following literature, not benchmarked in this paper.
  • Nr = 10
    Number of Bessel radial basis functions; user-set.
  • strain_max = 0.04 (±4%)
    Augmentation strain bound; hand-selected.
  • texture_alpha = 0.5
    Max intensity reduction in texture augmentation; hand-selected.
  • cctbx_metric_tolerance = 0.1
    Postprocessing tolerance for metric-subgroup search; affects reported errors.
axioms (4)
  • domain assumption The bispectrum is empirically invertible to lattice parameters.
    Section 4.2 and S8: 'While the bispectrum is not mathematically complete ... it can be empirically inverted,' cited to Nigam et al. 2026 (overlapping authors). This is load-bearing and not formally verified.
  • ad hoc to paper The truncated bispectrum (lmax=6, Nr=10) distinguishes the lattices in the evaluation sets.
    The paper relies on this truncation being sufficient (Section 4.1.1) but does not demonstrate completeness for the datasets.
  • domain assumption Powder XRD intensity carries enough information to predict the lattice bispectrum.
    The ML model is trained on simulated intensities; peak positions depend on lattice, but intensities also depend on atomic basis; the mapping is not one-to-one.
  • domain assumption The strain/texture/broadening augmentations approximate experimental variability.
    Section S5; the authors themselves note the 'degree of circularity' in augmentation design.

pith-pipeline@v1.3.0-alltime-deepseek · 26033 in / 12412 out tokens · 117274 ms · 2026-08-01T06:33:11.092244+00:00 · methodology

0 comments
read the original abstract

We present a machine learning (ML) method to determine unit cell parameters from powder X-Ray diffraction (XRD) data using a novel invariant lattice representation. In ML, the data representation used can have a substantial impact on the prediction quality. Previous approaches have directly predicted lattice parameters ($a,b,c,\alpha,\beta,\gamma$) from XRD inputs. However, these parameters depend strongly on the unit cell reduction or convention used. In this work, we construct an invariant representation of the reciprocal lattice that is independent of primitive cell convention, based on the bispectrum--a descriptor built from spherical harmonic projections of lattice points. The calculation of the lattice bispectrum is differentiable, and we demonstrate how to invert it using a dynamic programming approach. We show that when fixing ML model architecture, using the lattice bispectrum as the ML target rather than the unit cell parameters leads to more accurate lattice parameter predictions. For example, using the MP-20 dataset, the bispectrum reduces length mean absolute percentage error (MAPE) from 11.18% to 2.44% and angle MAPE from 12.74% to 3.07% compared to direct prediction with the same model architecture. We additionally benchmark our approach against pre-existing XRD to crystal structure models such as Crystalyze and assess its performance on the experimental RRUFF dataset. Beyond unit cell representation, we anticipate this invariant lattice representation could serve more broadly as a geometry-aware target for other crystallographic machine learning tasks such as structure generation.

Figures

Figures reproduced from arXiv: 2607.21829 by Aaron S. Brewster, Aria Mansouri Tehrani, Daniel W. Paley, David W. Mittan-Moreau, Elyssa Hofgard, Jigyasa Nigam, Kyucheol Min, Nofit Segal, Tess Smidt, Vanessa Oklejas.

Figure 1
Figure 1. Figure 1: Workflow for predicting the bispectrum and then inverting to obtain lattice parameters. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Plot of ρ(k) for a cubic lattice with a = 3 ˚A, lmax = 6, 10 Bessel function radial basis functions. One can see that the ρ(k) attains its maximum values at reciprocal lattice points. P m |cnlm| 2 for each allowed l. However, this loses information pertaining to angular correlations between different l channels. To retain more information, we combine three sets of coefficients. The bispectrum bl1l2l3 is co… view at source ↗
Figure 3
Figure 3. Figure 3: Sample bispectra for crystal systems with varying levels of symmetry. Each column [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Example interpolation between a cubic and triclinic lattice and the corresponding change [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Distribution of the magnitude of the maximum reciprocal lattice vector (using the crys [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Algorithm to invert the lattice bispectrum. Given an initial bispectrum, the starting [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Transformer model architecture used for predicting the bispectrum from XRD input. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: MAE for bispectrum + inversion compared to direct predictions per bravais lattice for [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: MAE for bispectrum + inversion compared to direct predictions per bravais lattice for [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Predicted vs. true for a, b, c and volume for bispec + inversion and directly predicting parameters using the MP20 augmented dataset. than length predictions with more diversity in unit cells, although this trend is not uniform across Bravais lattices, see Figure S13. See Section S9.4 for additional per-Bravais lattice MAE and par￾ity plots comparing bispectrum and direct prediction. Section S9.3 addition… view at source ↗
Figure 11
Figure 11. Figure 11: MAE for bispectrum + inversion compared to direct predictions per bravais lattice for [PITH_FULL_IMAGE:figures/full_fig_p016_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Predicted vs. true for a, b, c and volume for bispec + inversion and directly predicting parameters using the MPFull augmented dataset. Note that we use a diffraction-based definition rather than considering real space lattice anisotropy as lattice recovery from XRD patterns depends on reciprocal-space indexing. Using this criteria, 1,100 out of 30,000 structures in the test set were identified as dominan… view at source ↗
Figure 13
Figure 13. Figure 13: Parity plot of bispectrum prediction and direct lattice parameter predictions for the [PITH_FULL_IMAGE:figures/full_fig_p018_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Cumulative distribution function for length and angle MAE for the RRUFF Crystalyze [PITH_FULL_IMAGE:figures/full_fig_p019_14.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

70 extracted references · 25 canonical work pages

  1. [1]

    and Downs, R

    Lafuente, B. and Downs, R. T. and Yang, H. and Stone, N. , title =. Highlights in Mineralogical Crystallography , editor =. 2015 , publisher =

    2015

  2. [2]

    Nuclear Instruments , volume =

    Choice of collimators for a crystal spectrometer for neutron diffraction , author =. Nuclear Instruments , volume =. 1958 , doi =

    1958

  3. [3]

    2026 , langid =

    Reconstructing Local Environments from Concise Atomistic Representations , author =. 2026 , langid =

    2026

  4. [4]

    Chierotti and Carlo Nervi and Martin U

    Carina Schlesinger and Arnd Fitterer and Christian Buchsbaum and Stefan Habermehl and Michele R. Chierotti and Carlo Nervi and Martin U. Schmidt , title =. IUCrJ , year =. doi:10.1107/S2052252522004237 , issn =

    doi:10.1107/s2052252522004237

  5. [5]

    Kenneth D. M. Harris , title =. Acta Crystallographica Section B: Structural Science, Crystal Engineering and Materials , year =. doi:10.1107/S2052520622003717 , url =

    work page doi:10.1107/s2052520622003717

  6. [6]

    Acta Crystallographica Section A: Foundations and Advances , volume=

    Structure determination from powder diffraction data , author=. Acta Crystallographica Section A: Foundations and Advances , volume=. 2008 , publisher=. doi:10.1107/S0108767307064252 , url=

    work page doi:10.1107/s0108767307064252 2008

  7. [7]

    Journal of Applied Crystallography , author =

    Automated prediction of lattice parameters from. Journal of Applied Crystallography , author =. 2021 , note =. doi:10.1107/S1600576721010840 , abstract =

    work page doi:10.1107/s1600576721010840 2021

  8. [8]

    ACS Omega , author =

    Mlatticeabc:. ACS Omega , author =. 2021 , note =. doi:10.1021/acsomega.1c00781 , abstract =

    work page doi:10.1021/acsomega.1c00781 2021

  9. [9]

    The Journal of Machine Learning Research , author =

    Multiple output regression with latent noise , url =. The Journal of Machine Learning Research , author =. 2016 , note =. doi:10.5555/2946645.3007075 , abstract =

    arXiv 2016

  10. [10]

    Convolutional neural networks for time series classification , url =

  11. [11]

    Acta Crystallographica Section A: Foundations and Advances , author =

    A space for lattice representation and clustering , volume =. Acta Crystallographica Section A: Foundations and Advances , author =. 2019 , note =. doi:10.1107/S2053273319002729 , abstract =

    work page doi:10.1107/s2053273319002729 2019

  12. [12]

    Acta Crystallographica Section A: Foundations and Advances , author =

    Selling reduction versus. Acta Crystallographica Section A: Foundations and Advances , author =. 2019 , note =. doi:10.1107/S2053273318015413 , abstract =

    doi:10.1107/s2053273318015413 2019

  13. [13]

    arXiv.org , author =

    Measuring. arXiv.org , author =. 2023 , file =

    2023

  14. [14]

    Acta Crystallographica Section A: Foundations of Crystallography , author =

    Classification of. Acta Crystallographica Section A: Foundations of Crystallography , author =. 1997 , note =. doi:10.1107/S010876739700411X , abstract =

    work page doi:10.1107/s010876739700411x 1997

  15. [15]

    JOSA A , author =

    Phase retrieval by iterated projections , volume =. JOSA A , author =. 2003 , note =. doi:10.1364/JOSAA.20.000040 , abstract =

    work page doi:10.1364/josaa.20.000040 2003

  16. [16]

    Applied Optics , author =

    Phase retrieval algorithms: a personal tour [. Applied Optics , author =. 2013 , note =. doi:10.1364/AO.52.000045 , abstract =

    doi:10.1364/ao.52.000045 2013

  17. [17]

    An indexing algorithm independent of peak position extraction for

  18. [18]

    2023 , pages =

    Journal of Applied Crystallography , author =. 2023 , pages =. doi:10.1107/S1600576723000596 , abstract =

    work page doi:10.1107/s1600576723000596 2023

  19. [19]

    Journal of Materials Science , author =

    Exploring supervised machine learning for multi-phase identification and quantification from powder. Journal of Materials Science , author =. 2023 , note =. doi:10.1007/s10853-023-08343-4 , abstract =

    work page doi:10.1007/s10853-023-08343-4 2023

  20. [20]

    Advanced Intelligent Systems , author =

    A. Advanced Intelligent Systems , author =. 2023 , note =. doi:10.1002/aisy.202300140 , abstract =

    work page doi:10.1002/aisy.202300140 2023

  21. [21]

    Settembre, Gaetano and Corriero, Nicola and Del Buono, Nicoletta and Esposito, Flavia and Rizzi, Rosanna , year =. Machine. Machine. doi:10.1007/978-3-031-25599-1_8 , abstract =

    work page doi:10.1007/978-3-031-25599-1_8

  22. [22]

    Chemical Physics Reviews , author =

    Machine learning on neutron and x-ray scattering and spectroscopies , volume =. Chemical Physics Reviews , author =. 2021 , note =. doi:10.1063/5.0049111 , abstract =

    doi:10.1063/5.0049111 2021

  23. [23]

    Handbook of

    X-ray and. Handbook of. 2003 , doi =

    2003

  24. [24]

    Journal of Applied Crystallography , author =

    Machine learning for scattering data: strategies, perspectives and applications to surface scattering , volume =. Journal of Applied Crystallography , author =. 2023 , note =. doi:10.1107/S1600576722011566 , abstract =

    work page doi:10.1107/s1600576722011566 2023

  25. [25]

    Journal of Applied Physics , author =

    A novel experimental procedure for removing ambiguity from the interpretation of neutron and x‐ray reflectivity measurements: ‘‘. Journal of Applied Physics , author =. 1991 , note =. doi:10.1063/1.349629 , abstract =

    work page doi:10.1063/1.349629 1991

  26. [26]

    npj Computational Materials , author =

    Deep learning for visualization and novelty detection in large. npj Computational Materials , author =. 2021 , note =. doi:10.1038/s41524-021-00575-9 , abstract =

    work page doi:10.1038/s41524-021-00575-9 2021

  27. [27]

    Machine Learning: Science and Technology , author =

    Direct prediction of inelastic neutron scattering spectra from the crystal structure* , volume =. Machine Learning: Science and Technology , author =. 2023 , note =. doi:10.1088/2632-2153/acb315 , abstract =

    work page doi:10.1088/2632-2153/acb315 2023

  28. [28]

    Advanced Science , author =

    Direct. Advanced Science , author =. 2021 , note =. doi:10.1002/advs.202004214 , abstract =

    doi:10.1002/advs.202004214 2021

  29. [29]

    Journal of applied crystallography , author =

    Recent developments in the. Journal of applied crystallography , author =. 2019 , pmid =. doi:10.1107/S160057671900997X , abstract =

    doi:10.1107/s160057671900997x 2019

  30. [30]

    Acta Crystallographica Section B: Structural Science, Crystal Engineering and Materials , author =

    The. Acta Crystallographica Section B: Structural Science, Crystal Engineering and Materials , author =. 2016 , note =. doi:10.1107/S2052520616003954 , language =

    work page doi:10.1107/s2052520616003954 2016

  31. [31]

    arXiv.org , author =

    A. arXiv.org , author =. 2021 , file =

    2021

  32. [32]

    arXiv.org , author =

    Completeness of. arXiv.org , author =. 2023 , file =

    2023

  33. [33]

    Physical Review B , author =

    Through the eyes of a descriptor:. Physical Review B , author =. 2021 , note =. doi:10.1103/PhysRevB.104.144110 , abstract =

    work page doi:10.1103/physrevb.104.144110 2021

  34. [34]

    arXiv.org , author =

    On the. arXiv.org , author =. 2020 , doi =

    2020

  35. [35]

    arXiv.org , author =

    Neural networks trained on synthetically generated crystals can extract structural information from. arXiv.org , author =. 2023 , doi =

    2023

  36. [36]

    2023 , doi =

    Automated. 2023 , doi =

    2023

  37. [37]

    Scientific Reports , author =

    Symmetry prediction and knowledge discovery from. Scientific Reports , author =. 2020 , note =. doi:10.1038/s41598-020-77474-4 , abstract =

    work page doi:10.1038/s41598-020-77474-4 2020

  38. [38]

    An indexing algorithm independent of peak position extraction for X-ray powder diffraction patterns

    Coelho, Alan A. An indexing algorithm independent of peak position extraction for X-ray powder diffraction patterns. Journal of Applied Crystallography. 2017. doi:10.1107/S1600576717011359 , url =

    work page doi:10.1107/s1600576717011359 2017

  39. [39]

    and Mackey, Tsach and Nilforoshan, Hamed and Xu, Minkai and Badding, Catherine K

    Riesel, Eric A. and Mackey, Tsach and Nilforoshan, Hamed and Xu, Minkai and Badding, Catherine K. and Altman, Alison B. and Leskovec, Jure and Freedman, Danna E. , year =. Crystal. Journal of the American Chemical Society , publisher =. doi:10.1021/jacs.4c10244 , urldate =

    doi:10.1021/jacs.4c10244

  40. [40]

    and Bartel, Christopher J

    Szymanski, Nathan J. and Bartel, Christopher J. and Zeng, Yan and Tu, Qingsong and Ceder, Gerbrand , year =. Probabilistic. Chemistry of Materials , publisher =. doi:10.1021/acs.chemmater.1c01071 , urldate =

    work page doi:10.1021/acs.chemmater.1c01071

  41. [41]

    2013 , month = may, journal =

    On Representing Chemical Environments , author =. 2013 , month = may, journal =. doi:10.1103/PhysRevB.87.184115 , urldate =

    doi:10.1103/physrevb.87.184115 2013

  42. [42]

    Mario Geiger and Tess Smidt and Alby M. and Benjamin Kurt Miller and Wouter Boomsma and Bradley Dice and Kostiantyn Lapchevskyi and Maurice Weiler and Michał Tyszkiewicz and Simon Batzner and Dylan Madisetti and Martin Uhrin and Jes Frellsen and Nuri Jung and Sophia Sanborn and Mingjian Wen and Josh Rackers and Marcel Rød and Michael Bailey , title =. doi...

    doi:10.5281/zenodo.6459381

  43. [43]

    arXiv preprint arXiv:2207.09453 , year=

    e3nn: Euclidean neural networks , author=. arXiv preprint arXiv:2207.09453 , year=

    Pith/arXiv arXiv

  44. [44]

    2022 , month = jul, journal =

    Local Inversion of the Chemical Environment Representations , author =. 2022 , month = jul, journal =

    2022

  45. [45]

    Bispectrum Inversion With Application to Multireference Alignment , year=

    Bendory, Tamir and Boumal, Nicolas and Ma, Chao and Zhao, Zhizhen and Singer, Amit , journal=. Bispectrum Inversion With Application to Multireference Alignment , year=

  46. [46]

    Kakarala, Ramakrishna , year =. The. Journal of Mathematical Imaging and Vision , volume =

  47. [47]

    , booktitle=

    Pinilla, Samuel and Mishra, Kumar Vijay and Sadler, Brian M. , booktitle=. Unique Bispectrum Inversion for Signals with Finite Spectral/Temporal Support , year=

  48. [48]

    Xie, Tian and Fu, Xiang and Ganea, Octavian-Eugen and Barzilay, Regina and Jaakkola, Tommi , year = 2021, month = oct, journal =. Crystal

    2021

  49. [49]

    Nature Materials , volume =

    Ab Initio Structure Solutions from Nanocrystalline Powder Diffraction Data via Diffusion Models , author =. Nature Materials , volume =. doi:10.1038/s41563-025-02220-y , urldate =

    doi:10.1038/s41563-025-02220-y

  50. [50]

    Nature Communications , volume =

    Powder Diffraction Crystal Structure Determination Using Generative Models , author =. Nature Communications , volume =. doi:10.1038/s41467-025-62708-8 , urldate =

    doi:10.1038/s41467-025-62708-8

  51. [51]

    Visser, J. W. , doi =. Journal of. 1969 , langid =

    1969

  52. [52]

    , address =

    Coelho, Alan A. , address =. An indexing algorithm independent of peak position extraction for X‐ray powder diffraction patterns , volume =. Journal of applied crystallography , keywords =

  53. [53]

    Indexing of powder diffraction patterns for low-symmetry lattices by the successive dichotomy method , volume =

    Boultif, A and Louër, D , address =. Indexing of powder diffraction patterns for low-symmetry lattices by the successive dichotomy method , volume =. Journal of applied crystallography. , lccn =

  54. [54]

    Acta Crystallographica , volume=

    On the determination of unit-cell dimensions from powder diffraction patterns , author=. Acta Crystallographica , volume=. 1957 , publisher=

    1957

  55. [55]

    Coelho, A. A. Indexing of powder diffraction patterns by iterative use of singular value decomposition. Journal of Applied Crystallography. 2003. doi:10.1107/S0021889802019878 , url =

    work page doi:10.1107/s0021889802019878 2003

  56. [56]

    Hauptman, H. A. , year = 1991, month = nov, journal =. The Phase Problem of. doi:10.1088/0034-4885/54/11/002 , urldate =

    work page doi:10.1088/0034-4885/54/11/002 1991

  57. [57]

    Powder Diffraction , author=

    Lattice metric singularities and their impact on the indexing of powder patterns , volume=. Powder Diffraction , author=. 2000 , pages=. doi:10.1017/S0885715600010873 , number=

    work page doi:10.1017/s0885715600010873 2000

  58. [58]

    and Dresselhaus, Gene and Jorio, Ado , title =

    Dresselhaus, Mildred S. and Dresselhaus, Gene and Jorio, Ado , title =. 2008 , doi =

    2008

  59. [59]

    npj Computational Materials , volume=

    A deep convolutional neural network for real-time full profile analysis of big powder diffraction data , author=. npj Computational Materials , volume=. 2021 , publisher=

    2021

  60. [60]

    The Journal of Physical Chemistry A , volume=

    Powder diffraction indexing as a pattern recognition problem: a new approach for unit cell determination based on an artificial neural network , author=. The Journal of Physical Chemistry A , volume=. 2004 , publisher=

    2004

  61. [61]

    Scientific reports , volume=

    Symmetry prediction and knowledge discovery from X-ray diffraction patterns using an interpretable machine learning approach , author=. Scientific reports , volume=. 2020 , publisher=

    2020

  62. [62]

    arXiv preprint arXiv:2512.04036 , year=

    The Loss Landscape of Powder X-Ray Diffraction-Based Structure Optimization Is Too Rough for Gradient Descent , author=. arXiv preprint arXiv:2512.04036 , year=

    arXiv

  63. [63]

    npj Computational Materials , volume=

    Automated classification of big X-ray diffraction data using deep learning models , author=. npj Computational Materials , volume=. 2023 , publisher=

    2023

  64. [64]

    Journal of Chemical Information and Modeling , volume =

    Shu, Ke and Gui, Dong-Yun and Yan, Wei-Xin and Wang, Chun-Hai , title =. Journal of Chemical Information and Modeling , volume =. 2025 , doi =

    2025

  65. [65]

    and Luo, Aileen and Yin, Xiangyu and Prince, Michael and Toby, Brian H

    Andrejevic, Nina and Du, Ming and Sharma, Hemant and Horwath, James P. and Luo, Aileen and Yin, Xiangyu and Prince, Michael and Toby, Brian H. and Cherukara, Mathew J. , year = 2026, month = mar, journal =

    2026

  66. [66]

    The Journal of Physical Chemistry Letters , volume =

    Choudhary, Kamal , title =. The Journal of Physical Chemistry Letters , volume =. 2025 , doi =

    2025

  67. [67]

    The Journal of Physical Chemistry A , volume =

    Convolutional. The Journal of Physical Chemistry A , volume =

  68. [68]

    Scripta Materialia , volume =

    Insights from the reciprocal space revealed by a convolutional neural network and transfer learning , author =. Scripta Materialia , volume =. 2025 , issn =. doi:10.1016/j.scriptamat.2025.116697 , url =

    arXiv 2025

  69. [69]

    and Bernstein, Herbert J

    Andrews, Lawrence C. and Bernstein, Herbert J. , year = 2014, month = feb, journal =. The Geometry of. doi:10.1107/S1600576713031002 , abstract =

    work page doi:10.1107/s1600576713031002 2014

  70. [70]

    Hollarek, Daniel and Schopmans, Henrik and. op. 2025 , langid =. doi:10.1002/aidi.202500044 , journal =

    work page doi:10.1002/aidi.202500044 2025