Skip to content

Bin-Cao/awesome-xrd2crystal

BranchesTags

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

35 Commits
tops
tops
 
 
.DS_Store
.DS_Store
 
 
CITATION.cff
CITATION.cff
 
 
CONTRIBUTING.md
CONTRIBUTING.md
 
 
LICENSE
LICENSE
 
 
README.md
README.md
 
 

Repository files navigation

Awesome XRD → Crystal Awesome PRs Welcome License: MIT

Eng

A curated, chronologically organized collection of papers, code, and datasets on (powder) X-ray diffraction (XRD/PXRD) → crystal structure — identification, symmetry classification, ab-initio structure determination, generative modeling, multi-phase decomposition, refinement, and related machine-learning topics.

The PXRD → structure problem is an ill-posed inverse problem: many candidate structures can produce similar 1D patterns, and experimental profiles contain instrumental broadening, finite-size effects, texture/preferred orientation, background, impurity phases, and peak overlap. Work in this space spans (1) phase identification and search-match, (2) symmetry / lattice / space-group prediction, (3) multi-phase decomposition and Rietveld automation, (4) autonomous/in-situ analysis, and (5) end-to-end generative ab-initio structure solution.

Within every section, entries are listed chronologically (oldest → newest) in a standardized format.

📢 Contributions, corrections, and translations are warmly welcomed — 欢迎补充、纠正与翻译！ If a paper is missing, a link is broken, or an author/affiliation is wrong, please open an issue or send a PR. See CONTRIBUTING.md. All contributors are credited in the Contributors section at the bottom.

Maintainers

Bin Cao
Bin Cao
@Bin-Cao 		Rui Jiao
Rui Jiao
@jiaor17 		Zhixun Lee
Zhixun Lee
@ZhixunLEE 		Yang Liu
Yang Liu
@yliukj
Yunyue Su
Yunyue Su
@OPilgrim 		Shuchen Pu
Shuchen Pu
@komusama0930 		Hanyu Gao
Hanyu Gao
@Licht0812 		Ruiming Zhu
Ruiming Zhu
@RaymondZhurm
Liming Wu
Liming Wu
@ManlioWu 		Tong Xie
Tong Xie
@0xTong 		Wei Nong
Wei Nong
@Tosykie

For paper additions, please open an issue or PR; for scope or organizational questions, feel free to tag any maintainer directly.

Classification vs. Generation — quick guide

Two complementary problem formulations recur throughout this list. Pick the section that matches your task:

Task Input Output Sections
Phase / structure identification (classification, retrieval) PXRD pattern Index of a known phase, crystal system, or space group Phase Identification · Crystal System / Space Group / Lattice
Multi-phase decomposition Mixed PXRD pattern Per-phase weight fractions and component patterns Multi-phase Decomposition & Rietveld
Ab-initio structure generation PXRD pattern (± composition / cell) Full 3D crystal (atom species + fractional coordinates + lattice / CIF) End-to-End PXRD → Crystal Structure (Generative)
Autonomous / in-situ analysis Live PXRD stream from beamline / lab Phase calls, next-measurement policy, synthesis decisions Autonomous / In-situ XRD

Classification models are typically lighter, require no symmetry equivariance, and degrade gracefully on out-of-distribution patterns; generative models target atomic-resolution structure but rely on stronger physical priors and are still constrained by training-set coverage.

Contents

Standard entry format

### YYYY · ShortName — One-sentence tagline
- **Title:** <full paper title>
- **Authors:** <first author> (<affiliation>), ..., <senior author> (<affiliation>)
- **Venue:** <journal vol, pp (year)>; arXiv:<id>
- **Paper:** <https://...>
- **Code:** <https://...>     (omit if none)
- **Data:** <https://...>     (omit if none)
- **TL;DR:** <1–3 sentences>

End-to-End PXRD → Crystal Structure (Generative)

Methods that take a PXRD pattern (optionally + composition / lattice) and output a full 3D crystal structure (atomic positions + lattice, typically as CIF or coordinates).

2024 · CrystalNet — Coordinate-based VAE for electron density from PXRD

2024 · XtalNet — Contrastive PXRD-crystal pretraining + conditional equivariant diffusion

2024 · Crystalyze — XRD encoder + CDVAE diffusion

  • Title: Crystal Structure Determination from Powder Diffraction Patterns with Generative Machine Learning
  • Authors: Eric A. Riesel (MIT), Tsach Mackey (Yale), et al.; seniors: Danna E. Freedman (MIT), Jure Leskovec (Stanford)
  • Venue: J. Am. Chem. Soc. 146, 30340–30348 (2024)
  • Paper: https://pubs.acs.org/doi/10.1021/jacs.4c10244
  • Code: https://github.com/ML-PXRD/Crystalyze
  • TL;DR: Pretrained XRD encoder feeding CDVAE; ~42% match rate on experimental and ~67% on simulated patterns; solved previously-unsolved Powder Diffraction File entries.

2025 · PXRDnet — SE(3)-equivariant score-based diffusion for nanocrystalline PXRD

  • Title: Ab initio structure solutions from nanocrystalline powder diffraction data via diffusion models
  • Authors: Gabe Guo (Columbia), Tristan Saidi (Columbia), Maxwell Terban (MPI Solid State Research), Michele Liu (Columbia), Hamed Salimian (Columbia), Simon J. L. Billinge (Columbia/BNL); senior: Hod Lipson (Columbia)
  • Venue: Nature Materials 24, 1726–1734 (2025); arXiv:2406.10796
  • Paper: https://www.nature.com/articles/s41563-025-02220-y · https://arxiv.org/abs/2406.10796
  • Code: https://github.com/gabeguo/cdvae_xrd
  • TL;DR: Score-based diffusion conditioned on finite-size-broadened PXRD + formula; works on simulated nanocrystals as small as 10 Å and on real experimental patterns spanning all seven crystal systems; median RMSD ≤ 0.11 Å on MP-20-PXRD.

2025 · PXRDGen — Contrastive XRD encoder + flow/diffusion + automated Rietveld

  • Title: Powder diffraction crystal structure determination using generative models
  • Authors: Qi Li (IoP, CAS), Rui Jiao (Tsinghua), Liming Wu, Tiannian Zhu, et al.; seniors: Wenbing Huang (Renmin Univ.), Shifeng Jin (IoP, CAS)
  • Venue: Nature Communications 16, 7428 (2025); arXiv:2409.04727
  • Paper: https://www.nature.com/articles/s41467-025-62708-8 · https://arxiv.org/abs/2409.04727
  • Code: https://codeocean.com/capsule/7727770/tree/v1 (DOI: 10.24433/CO.6347299.v1)
  • TL;DR: Contrastive-pretrained XRD encoder + DiffCSP-style diffusion/flow generator + integrated Rietveld refinement; 82% valid compounds from a single sample on MP-20, 96% within 20.

2025 · deCIFer — Autoregressive transformer over CIF tokens conditioned on PXRD

  • Title: deCIFer: Crystal Structure Prediction from Powder Diffraction Data using Autoregressive Language Models
  • Authors: Frederik Lizak Johansen, Ulrik Friis-Jensen, Erik Bjørnager Dam, Kirsten M. Ø. Jensen, Rocío Mercado, Raghavendra Selvan — University of Copenhagen (DIKU & Department of Chemistry)
  • Venue: TMLR / OpenReview 2025; arXiv:2502.02189
  • Paper: https://arxiv.org/abs/2502.02189 · https://openreview.net/forum?id=LftFQ35l47
  • Code: https://github.com/FrederikLizakJohansen/deCIFer
  • TL;DR: Decoder-only transformer that emits CIF text token-by-token conditioned on PXRD embeddings; ~94% structure match in the author setting on NOMA1/CHILI-100K-derived data.

2025 · DiffractGPT / AtomGPT — Generative pretrained transformer for XRD → structure

2025 · Uni-3DAR — Unified 3D autoregressive generation on compressed spatial tokens

2025 · CrystaLLM-π — Property- and PXRD-conditioned CIF language model

  • Title: CrystaLLM-π: Property- and PXRD-Conditioned Generation of Crystal Structures with a CIF Language Model
  • Authors: C-Bone-UCL group (University College London); builds on CrystaLLM (Antunes et al., Univ. of Reading)
  • Venue: preprint / open-source release (2025)
  • Code: https://github.com/C-Bone-UCL/CrystaLLM-2.0 · weights: CrystaLLM-pi_COD-XRD, CrystaLLM-pi_Mattergen-XRD on Hugging Face
  • Base model: https://arxiv.org/abs/2307.04340 · https://github.com/lantunes/CrystaLLM
  • TL;DR: Adds bandgap, density, photovoltaic efficiency and PXRD conditioning streams to the CrystaLLM CIF autoregressive language model. The base model is CrystaLLM (Antunes et al., Nat. Commun. 15, 10570, 2024) — a GPT-style decoder-only language model trained on millions of CIF files that emits crystal structures as CIF text token by token.

2026 · Xrd2Mof - Stable-Diffusion solver for MOF PXRD

  • Title: Title: Interpreting X-ray Diffraction Patterns of Metal–Organic Frameworks via Generative Artificial Intelligence
  • Authors: Bin Feng, Bingxu Wang, Linpeng Lv, Mingzheng Zhang, Zhefeng Chen, Feng Pan, Shunning Li(Peking University Shenzhen Graduate School)
  • Venue: J. Am. Chem. Soc. 148(1), 869–878 (2026)
  • Paper: https://pubs.acs.org/doi/10.1021/jacs.5c16416
  • Code: https://github.com/PKUsam2023/Xrd2Mof
  • TL;DR: Diffusion-based framework for MOF structure determination from PXRD using coarse-grained representations and chemical priors; >93% accuracy in recovering ground-truth MOF structures.

2026 · XRDSol — Equivariant diffusion solver for inorganic PXRD

  • Title: Equivariant diffusion solution for inorganic crystal structure determination from powder X-ray diffraction data
  • Authors: Dongfang Yu, Zhewen Zhu (Zhejiang Univ. / Westlake Univ.), Fucheng Leng (Univ. College Cork); senior: Yizhou Zhu (Westlake Univ.)
  • Venue: Nature Communications 17, 3274 (2026)
  • Paper: https://www.nature.com/articles/s41467-026-70035-9
  • Code: https://github.com/ai4mat-zhu/XRDSol
  • TL;DR: EGNN-based denoising diffusion conditioned on stoichiometry + unit cell + PXRD; ~82.3% success on MP-20, 81.6% on ICDD-20 experimental benchmark; ~0.6 s per solution on one GPU.

2026 · RealPXRD-Solver — Flow-matching solver with experimental augmentation

  • Title: Experimental Powder X-ray Diffraction Crystal Structure Determination with RealPXRD-Solver
  • Authors: Qi Li et al. — Institute of Physics, CAS (with DP Technology, Tsinghua, Renmin Univ., SJTU, AISI)
  • Venue: arXiv:2603.00965 (Mar 2026)
  • Paper: https://arxiv.org/abs/2603.00965
  • Code: https://github.com/liqi-529/RealPXRD-Solver
  • TL;DR: Universal encoder of d-spacing–intensity fingerprints trained on 6,250,238 theoretical structures with experiment-mimicking augmentations; lattice-conditioned and lattice-free modes; 77.9%/91.9% Top-1/Top-20 on CNRS, 78.8%/92.9% on RRUFF; solved 39 previously-unreported PDF entries.

2026 · Hybrid Ab-initio PXRD Solver — Discrete search + continuous optimization

  • Title: Ab-initio Crystal Structure Determination from Powder X-Ray Diffraction
  • Authors: Kaixiang Su, Osman Goni Ridwan, Hongfei Xue; senior: Qiang Zhu — UNC Charlotte
  • Venue: arXiv:2605.24594 (May 2026)
  • Paper: https://arxiv.org/abs/2605.24594
  • Code: https://github.com/MaterSim/Ab-PXRD-Solver
  • Data: https://mmi.charlotte.edu/ab_pxrd_solver
  • TL;DR: Hierarchical pipeline: (1) discrete search over space group / unit cell / Wyckoff combinations, (2) continuous coordinate optimization with AI-based peak profile analysis, density estimation, and physics-informed energy constraints.

Phase Identification from XRD

Methods that, given a PXRD pattern, predict which known phase(s) it corresponds to (classification / multi-label / one-to-one ID).

2017 · Iwasaki et al. — Dissimilarity measures for clustering combinatorial XRD

  • Title: Comparison of dissimilarity measures for cluster analysis of X-ray diffraction data from combinatorial libraries
  • Authors: Yuma Iwasaki (NIMS), A. Gilad Kusne (NIST/UMD), Ichiro Takeuchi (UMD)
  • Venue: npj Computational Materials 3, 4 (2017)
  • Paper: https://www.nature.com/articles/s41524-017-0006-2
  • TL;DR: Systematic comparison of cosine, Pearson, JSD, NCDTW and other dissimilarity measures for grouping XRD patterns of Fe–Co–Ni composition spreads; foundation for downstream unsupervised phase mapping.

2019 · Oviedo et al. — Fast and interpretable XRD classification

  • Title: Fast and interpretable classification of small X-ray diffraction datasets using data augmentation and deep neural networks
  • Authors: Felipe Oviedo (MIT Photovoltaics Research Lab), et al.; senior: Tonio Buonassisi (MIT) — collab. NIST
  • Venue: npj Computational Materials 5, 60 (2019)
  • Paper: https://www.nature.com/articles/s41524-019-0196-x
  • Code: https://github.com/PV-Lab/autoXRD
  • TL;DR: Physics-informed augmentation lets 1D a-CNNs train from small experimental XRD sets; introduces CAM-based interpretability for crystal-system / dimensionality / space-group classification of thin films.

2020 · Lee et al. — Deep CNN for multiphase inorganic phase ID

  • Title: A deep-learning technique for phase identification in multiphase inorganic compounds using synthetic XRD powder patterns
  • Authors: Jin-Woong Lee, Woon Bae Park, Jin Hee Lee, Satendra Pal Singh; senior: Kee-Sun Sohn — Sejong University
  • Venue: Nature Communications 11, 86 (2020)
  • Paper: https://www.nature.com/articles/s41467-019-13749-3
  • TL;DR: CNN trained on ~1.78 M simulated PXRD patterns across the Sr-Li-Al-O quaternary; near-perfect identification on experimental multi-phase compounds.

2020 · Wang et al. — Interpretable CNN for one-to-one MOF XRD ID

  • Title: Rapid Identification of X-ray Diffraction Patterns Based on Very Limited Data by Interpretable Convolutional Neural Networks
  • Authors: Hong Wang, Yunchao Xie, Dawei Li, Heng Deng, Yunxin Zhao, Ming Xin; senior: Jian Lin — University of Missouri
  • Venue: J. Chem. Inf. Model. 60, 2004–2011 (2020); arXiv:1912.07750
  • Paper: https://pubs.acs.org/doi/10.1021/acs.jcim.0c00020 · https://arxiv.org/abs/1912.07750
  • TL;DR: CNN trained on theoretical MOF XRD patterns augmented with noise from limited experiments; 96.7% Top-5 identification accuracy with class-activation interpretation.

2021 · Lee et al. — Phase ID + quantitative phase-fraction prediction

  • Title: A data-driven XRD analysis protocol for phase identification and phase-fraction prediction of multiphase inorganic compounds
  • Authors: Jin-Woong Lee, Woon Bae Park, Minseuk Kim, Satendra Pal Singh, Myoungho Pyo; senior: Kee-Sun Sohn — Sejong University
  • Venue: Inorganic Chemistry Frontiers 8, 2492 (2021)
  • Paper: https://pubs.rsc.org/en/content/articlelanding/2021/qi/d0qi01513j
  • TL;DR: Extends Lee 2020 to the Li-La-Zr-O quaternary; 96.47% test phase-ID accuracy and R² ≈ 0.97 phase-fraction regression on simulated data (91.11% / 0.96 on experimental).

2023 · NeuralXRD — Synthetic + real ML for transition-metal phase ID

  • Title: Machine Learning-Assisted Close-Set X-ray Diffraction Phase Identification of Transition Metals
  • Authors: Maksim Eremeev et al.
  • Venue: ICLR 2023 ML4Materials Workshop
  • Paper: https://arxiv.org/abs/2305.15410
  • Code: https://github.com/maxnygma/NeuralXRD
  • TL;DR: Combines synthetic phase generation, peak integration, and post-hoc calibration to identify transition-metal phases from XRD; ships demo notebooks operating on COD-derived patterns.

2024 · CPICANN — Convolutional self-attention phase identifier

2024 · MLstructureMining — ML structure ID from atomic PDFs

2025 · XQueryer — 1.03 B-parameter transformer crystal identifier

  • Title: XQueryer: an intelligent crystal structure identifier for powder X-ray diffraction
  • Authors: Bin Cao, Zinan Zheng, Yang Liu, Longhan Zhang, Lawrence W. Y. Wong, Lu-Tao Weng, Jia Li, Haoxiang Li, Tong-Yi Zhang — HKUST(GZ) / HKUST
  • Venue: National Science Review 12(12), nwaf421 (2025); DOI 10.1093/nsr/nwaf421
  • Paper: https://doi.org/10.1093/nsr/nwaf421
  • Code: https://github.com/Bin-Cao/XQueryer · benchmark: https://github.com/WPEM/XqueryerBench
  • TL;DR: 1.03 B-parameter transformer trained on ~2 M high-fidelity simulated spectra; integrates with a powder diffractometer for real-time identification and ships RRUFF↔MP ID matching utilities and a PXRD simulator · website: https://xqueryer.caobin.asia/.

Crystal System / Space Group / Lattice Prediction from XRD

2017 · Park et al. — CNN for crystal system / extinction group / space group

  • Title: Classification of crystal structure using a convolutional neural network
  • Authors: Woon Bae Park, Jiyong Chung, Jaeyoung Jung, Keemin Sohn, Satendra Pal Singh, Myoungho Pyo, Namsoo Shin; senior: Kee-Sun Sohn — Sejong University
  • Venue: IUCrJ 4, 486–494 (2017)
  • Paper: https://journals.iucr.org/m/issues/2017/04/00/fc5018/index.html
  • TL;DR: Three CNNs trained on ~150 k synthetic PXRD patterns from ICSD; 94.99% crystal-system, 83.83% extinction-group, 81.14% space-group accuracy.

2019 · Aguiar et al. — CNN decoding crystallography from electron diffraction

  • Title: Decoding crystallography from high-resolution electron imaging and diffraction datasets with deep learning
  • Authors: Jeffery A. Aguiar (INL/Univ. of Utah), Matthew L. Gong, Raymond Unocic, Tolga Tasdizen, Brian Miller
  • Venue: Science Advances 5, eaaw1949 (2019)
  • Paper: https://www.science.org/doi/10.1126/sciadv.aaw1949
  • Code: https://github.com/SCIInstitute/DiffractionClassification
  • TL;DR: CNN trained on 571,340 diffraction patterns across 7 families / 32 genera / 230 space groups; reduces the candidate space groups to the top 2 with >70–95% confidence; cross-validated on alloys and 2D materials.

2019 · Vecsei et al. — Neural network classification of crystal symmetries

  • Title: Neural network based classification of crystal symmetries from x-ray diffraction patterns
  • Authors: Pascal Marc Vecsei, Kenny Choo; senior: Titus Neupert — University of Zürich
  • Venue: Phys. Rev. B 99, 245120 (2019)
  • Paper: https://journals.aps.org/prb/abstract/10.1103/PhysRevB.99.245120
  • TL;DR: Dense vs. CNN comparison on simulated and RRUFF experimental data — 54% (dense) vs. 42% (CNN) space-group accuracy on RRUFF.

2019 · Liu et al. — CNN space-group prediction from atomic PDF

  • Title: Using a machine learning approach to determine the space group of a structure from the atomic pair distribution function
  • Authors: Chia-Hao (Timothy) Liu, Yunzhe Tao, Daniel Hsu, Qiang Du; senior: Simon J. L. Billinge — Columbia University
  • Venue: Acta Crystallographica A 75, 633–643 (2019)
  • Paper: https://journals.iucr.org/paper?S2053273319005606 · https://arxiv.org/abs/1902.00594
  • TL;DR: PDF-based (not direct PXRD) CNN; 91.9% top-6 space-group accuracy on experimental PDF inputs.

2020 · Suzuki et al. — Interpretable XGBoost for symmetry prediction

2021 · Chitturi et al. — 1D-CNN lattice parameter prediction (DeepLPnet)

  • Title: Automated prediction of lattice parameters from X-ray powder diffraction patterns
  • Authors: Sathya R. Chitturi (Stanford/SLAC), Daniel Ratner, Richard C. Walroth, Vivek Thampy, Evan J. Reed, Mike Dunne, Christopher J. Tassone; senior: Kevin H. Stone — SLAC
  • Venue: J. Appl. Crystallogr. 54, 1799–1810 (2021)
  • Paper: https://journals.iucr.org/j/issues/2021/06/00/vb5020/
  • Code: https://github.com/sathya-chitturi/DeepLPnet
  • TL;DR: Per-crystal-system 1D-CNNs regress a/b/c lattice parameters from raw PXRD; ~10% MAPE corresponds to 100–1000× search-space reduction; trained on ICSD + CSD (~1 M structures).

2023 · Salgado et al. — Big-data symmetry CNN (SPGNet)

  • Title: Automated classification of big X-ray diffraction data using deep learning models
  • Authors: Jerardo E. Salgado, Samuel Lerman, Zhaotong Du, Chenliang Xu; senior: Niaz Abdolrahim, Jason Hattrick-Simpers — University of Toronto / Rochester
  • Venue: npj Computational Materials 9, 214 (2023)
  • Paper: https://www.nature.com/articles/s41524-023-01164-8
  • Code: https://github.com/AGI-init/XRDs
  • TL;DR: 7-way crystal-system + 230-way space-group CNN with augmented synthetic data approximating real experimental noise.

2023 · Schopmans et al. — Synthetic crystals → ICSD symmetry extraction

  • Title: Neural networks trained on synthetically generated crystals can extract structural information from ICSD powder X-ray diffractograms
  • Authors: Henrik Schopmans, Patrick Reiser; senior: Pascal Friederich — KIT (AiMat group)
  • Venue: Digital Discovery 2, 1414 (2023)
  • Paper: https://pubs.rsc.org/en/content/articlelanding/2023/dd/d3dd00071k
  • Code: https://github.com/aimat-lab/ML4pXRDs
  • TL;DR: Models trained on purely synthetic randomly-generated crystals surpass models trained on ICSD itself for space-group classification on ICSD test patterns.

2024 · CrySTINet — Structure-type assignment from PXRD

  • Title: Crystal Structure Assignment for Unknown Compounds from X-ray Diffraction Patterns with Deep Learning
  • Authors: Litao Chen, Bin Feng, Bingxu Wang, Linpeng Lv, Mingzheng Zhang, Zhefeng Chen, Feng Pan, Shunning Li(Peking University Shenzhen Graduate School)
  • Venue: J. Am. Chem. Soc. 146(12), 8098–8109 (2024)
  • Paper: https://pubs.acs.org/doi/10.1021/jacs.3c11852
  • Code: https://github.com/PKUsam2023/CrySTINet
  • TL;DR: RCNet-ensemble framework for assigning unknown inorganic PXRD patterns to 100 crystal structure types; combines neural confidence with global pattern similarity, achieving 80.0% accuracy.

2026 · AlphaDiffract — 1D ConvNeXt for lattice & symmetry determination

  • Title: AlphaDiffract: Automated Crystallographic Analysis of Powder X-ray Diffraction Data
  • Authors: Nina Andrejevic, Ming Du, Hemant Sharma, James P. Horwath, Aileen Luo, Xiangyu Yin, Michael Prince, Brian H. Toby, Mathew J. Cherukara — Argonne National Laboratory
  • Venue: arXiv:2603.23367 (Mar 2026)
  • Paper: https://arxiv.org/abs/2603.23367
  • TL;DR: 1D ConvNeXt backbone on 8192-bin PXRD vectors jointly predicts crystal system, space group, and lattice parameters; trained on GSAS-II-simulated patterns from NIST/ICSD structures; intended as a preprocessing stage for automated Rietveld.

Multi-phase Decomposition & Rietveld Refinement

2018 · Stanev et al. — Unsupervised NMF phase mapping

  • Title: Unsupervised phase mapping of X-ray diffraction data by nonnegative matrix factorization integrated with custom clustering
  • Authors: Valentin Stanev (UMD / NIST), Velimir V. Vesselinov (LANL), A. Gilad Kusne (NIST / UMD), Graham Antoszewski, Ichiro Takeuchi; senior: Boian S. Alexandrov (LANL)
  • Venue: npj Computational Materials 4, 43 (2018); arXiv:1802.07307
  • Paper: https://www.nature.com/articles/s41524-018-0099-2 · https://arxiv.org/abs/1802.07307
  • Multi-phase strategy: Joint matrix factorization across a library of patterns.
  • TL;DR: Stacks N PXRD patterns from a composition library into a non-negative matrix and decomposes it into K shared end-member basis patterns and per-sample mixture weights via NMF, with custom clustering and cross-correlation to stabilize K and resolve peak shifts. Unsupervised and requires no candidate phase list, but assumes the same K end-members are shared across the library.

2021 · Szymanski et al. — Probabilistic multi-phase decomposition (XRD-AutoAnalyzer)

  • Title: Probabilistic deep learning approach to automate the interpretation of multi-phase diffraction spectra
  • Authors: Nathan J. Szymanski, Christopher J. Bartel, Yan Zeng, Qingsong Tu; senior: Gerbrand Ceder — UC Berkeley / LBNL
  • Venue: Chemistry of Materials 33, 4204–4215 (2021); arXiv:2103.16664
  • Paper: https://pubs.acs.org/doi/10.1021/acs.chemmater.1c01071 · https://arxiv.org/abs/2103.16664
  • Code: https://github.com/njszym/XRD-AutoAnalyzer
  • Multi-phase strategy: Sequential identify-and-subtract with branching tree search.
  • TL;DR: Ensemble CNN trained on a fixed reference set scores the most likely single phase, subtracts its simulated contribution from the pattern, and recurses on the residual; branching keeps several hypothesis paths to mitigate error accumulation. Handles up to ~4 phases per pattern with quantitative weight fractions and per-call uncertainty, but is bound to the reference phase set used at training time.

2022 · BraggNN — Fast Bragg-peak localization with deep learning

  • Title: BraggNN: fast X-ray Bragg peak analysis using deep learning
  • Authors: Zhengchun Liu, Hemant Sharma, Jun-Sang Park, Peter Kenesei, Antonino Miceli, Jonathan Almer, Rajkumar Kettimuthu, Ian Foster — Argonne National Laboratory
  • Venue: IUCrJ 9, 104–113 (2022)
  • Paper: https://journals.iucr.org/m/issues/2022/01/00/fs5198/
  • Code: https://github.com/lzhengchun/BraggNN
  • Multi-phase strategy: Per-peak regression (building block, not full decomposition).
  • TL;DR: Regression DNN that replaces pseudo-Voigt peak fitting at the level of individual Bragg spots in high-energy diffraction microscopy (HEDM). Orders of magnitude faster than conventional per-peak fitting with comparable precision; used as a front-end localizer feeding downstream phase / grain analysis rather than a pattern-level multi-phase decomposer.

2025 · Spotlight — Automated global optimization for Rietveld

  • Title: Spotlight: efficient automated global optimization in Rietveld analysis of diffraction data
  • Authors: C. M. Biwer, Z. Feng, D. Finstad, M. McDonnell, M. Knezevic, M. McKerns, D. J. Savage, S. C. Vogel — LANL (X Computational Physics / Materials Science & Tech.), with UNH, Syracuse, ORNL and the UQ Foundation
  • Venue: Scientific Reports 15, 8358 (2025)
  • Paper: https://www.nature.com/articles/s41598-025-92452-4
  • Code: https://github.com/lanl/spotlight
  • Multi-phase strategy: Known candidate phase set; parallel global optimization across a parametric series.
  • TL;DR: Python wrapper over MAUD / GSAS / GSAS-II that drives Rietveld refinement of multi-phase samples through an ensemble of optimizers running in parallel on HPC, with an iteratively learned surrogate of the refinement response surface. Targets parametric series (temperature, composition, in-situ time) where simple sequential refinement or DB-matching diverges as phase transformations occur; demonstrated on U–Mo, Ti–6Al–4V, Al₂O₃ and PbSO₄.

2026 · RAPID — Rietveld Analysis Pipeline with Intelligent Deep Learning

  • Title: Automation of Rietveld refinement through machine learning
  • Authors: Suk Jin Mun (IBS-CINAP / Sungkyunkwan University), Yoonsoo Nam (Rudolf Peierls Centre for Theoretical Physics, University of Oxford); senior: Sungkyun Choi (IBS-CINAP / SKKU / IBS-CVQS)
  • Venue: Journal of Applied Crystallography 59, 564–577 (2026); accepted Feb 2026
  • Paper: https://journals.iucr.org/j/issues/2026/02/00/yt5169/ · open-access mirror: https://pmc.ncbi.nlm.nih.gov/articles/PMC13060465/
  • Code: https://github.com/DataForgeSci/RAPID
  • Multi-phase strategy: Single-phase only; included as Rietveld-automation baseline.
  • TL;DR: Assumes the crystalline phase is already known and trains one CNN per compound on FullProf-simulated patterns to predict structural and profile parameters (lattice, B_iso, scale, U/V/W, zero shift) in a single forward pass. Validated on CeO₂, Tb₂BaCoO₅, PbSO₄ and monoclinic BaLu₆(Ge₂O₇)₂(Ge₃O₁₀) with R-factors matching manual refinement; the authors note that extension to multi-phase mixtures remains future work.

2026 · XDecomposer — Prior-free set decomposition for multiphase XRD

  • Title: XDecomposer: Learning Prior-Free Set Decomposition for Multiphase X-ray Diffraction
  • Authors: Hanyu Gao, Bin Cao — co-first; Yunyue Su; Tong-Yi Zhang, Qiang Liu (CASIA, HKUST(GZ))
  • Venue: arXiv:2605.05866 (May 2026)
  • Paper: https://arxiv.org/abs/2605.05866
  • Code: https://github.com/Licht0812/XDecomposer
  • Multi-phase strategy: Permutation-invariant set prediction (one-shot blind source separation).
  • TL;DR: Reformulates multiphase PXRD analysis as one-dimensional blind source separation (BSS): a pretrained global bottleneck encodes peak structure, then a fixed-size set of K_max output slots jointly predicts component patterns + mixture proportions in a single forward pass, trained with permutation-invariant separation loss, slot-activity supervision and mixture-consistency regularization. Avoids the error accumulation of sequential identify-and-subtract and does not require a candidate phase list, structural templates, or the number of phases.

Autonomous / In-situ XRD Analysis

2021 · XCA — Crystallography Companion Agent

  • Title: Crystallography companion agent for high-throughput materials discovery
  • Authors: Phillip M. Maffettone, Lars Banko, Peng Cui, Yury Lysogorskiy, Marc A. Little, Daniel Olds, Alfred Ludwig, Andrew I. Cooper — Brookhaven National Laboratory (NSLS-II), Ruhr-Universität Bochum, Univ. of Liverpool
  • Venue: Nature Computational Science 1, 290–297 (2021)
  • Paper: https://www.nature.com/articles/s43588-021-00059-2
  • Code: https://github.com/maffettone/xca · examples: https://github.com/bnl/pub-Maffettone_2020_08
  • TL;DR: Pseudo-unsupervised CNN ensemble trained on fully synthetic, physics-informed diffraction; provides real-time, probabilistic phase identification during high-throughput beam-time experiments.

2023 · Szymanski et al. — Adaptively driven XRD (Adaptive-XRD)

  • Title: Adaptively driven X-ray diffraction guided by machine learning for autonomous phase identification
  • Authors: Nathan J. Szymanski, Christopher J. Bartel, Yan Zeng, et al.; senior: Gerbrand Ceder — UC Berkeley / LBNL
  • Venue: npj Computational Materials 9, 31 (2023)
  • Paper: https://www.nature.com/articles/s41524-023-00984-y
  • Code: https://github.com/njszym/Adaptive-XRD
  • TL;DR: Ensemble CNN predicts phase mixtures with uncertainty; an adaptive policy decides the next 2θ ranges to measure, reducing total measurement time on transient or dilute samples.

2023 · Banko et al. — Self-supervised PXRD classification of phase transitions

  • Title: Application of self-supervised approaches to the classification of X-ray diffraction spectra during phase transitions
  • Authors: Lars Banko, Phillip M. Maffettone, Dennis Naujoks, Daniel Olds, Alfred Ludwig — Ruhr-Universität Bochum, BNL, DESY
  • Venue: Scientific Reports 13, 9173 (2023)
  • Paper: https://www.nature.com/articles/s41598-023-36456-y
  • TL;DR: Self-supervised relational reasoning + contrastive learning to classify 1D PXRD spectra of in-situ phase transitions recorded at PETRA III (P02.2).

2023 · A-Lab — Autonomous materials laboratory closing the synthesis loop

  • Title: An autonomous laboratory for the accelerated synthesis of novel materials
  • Authors: Nathan J. Szymanski, Bernardus Rendy, Yuxing Fei, et al.; senior: Gerbrand Ceder — UC Berkeley / LBNL
  • Venue: Nature 624, 86–91 (2023)
  • Paper: https://www.nature.com/articles/s41586-023-06734-w
  • TL;DR: End-to-end autonomous lab in which ML-based PXRD analysis closes the loop between synthesis attempts and recipe revision; reports 41 successful syntheses in 17 days.

Datasets & Benchmarks

Listed chronologically. (Reference databases (RRUFF, COD, Materials Project, ICSD, JARVIS-DFT) are grouped at the end for reference.)

2015 · RRUFF — Empirical mineral XRD / Raman / IR database

  • Title: The power of databases: the RRUFF project
  • Authors: Barbara Lafuente, Robert T. Downs, Hexiong Yang, Nate Stone — University of Arizona
  • Venue: Highlights in Mineralogical Crystallography, De Gruyter (2015)
  • Paper: https://www.degruyter.com/document/doi/10.1515/9783110417104-003
  • Portal: https://rruff.info/
  • TL;DR: Curated experimental XRD / Raman / IR / chemistry database for minerals; the most widely used experimental benchmark for PXRD ML methods.

2023 · ML4pXRDs — Synthetic-crystal training pipeline

2024 · CHILI-3K / CHILI-100K — Nanomaterials graph + scattering dataset

  • Title: CHILI: Chemically-Informed Large-scale Inorganic Nanomaterials Dataset for Advancing Graph Machine Learning
  • Authors: Ulrik Friis-Jensen, Frederik L. Johansen, Andy S. Anker, Erik B. Dam, Kirsten M. Ø. Jensen, Raghavendra Selvan — University of Copenhagen
  • Venue: KDD 2024; arXiv:2402.13221
  • Paper: https://arxiv.org/abs/2402.13221
  • Code/Data: https://github.com/UlrikFriisJensen/CHILI
  • TL;DR: Nanomaterials graph dataset with 11 classification (atom, crystal system, space group) and 8 regression tasks including PXRD, ND, SAXS, SANS, xPDF, nPDF.

2025 · SimXRD-4M — Large simulated PXRD benchmark

2025 · SIMPOD — COD-based 1D and 2D PXRD benchmark

2025/2026 · opXRD — Open experimental PXRD database

  • Title: opXRD: Open Experimental Powder X-Ray Diffraction Database
  • Authors: Daniel Hollarek, Henrik Schopmans, Jona Östreicher, Jonas Teufel, et al.; senior: Pascal Friederich — KIT (AiMat group), with 18 contributing labs
  • Venue: Advanced Intelligent Discovery 2, e202500044 (2026); arXiv:2503.05577
  • Paper: https://arxiv.org/abs/2503.05577
  • Portal: https://xrd.aimat.science/ · https://zenodo.org/records/15298026
  • TL;DR: ~92 k labeled + unlabeled experimental PXRD patterns aggregated from multiple high-throughput labs, covering many materials classes.

2025 · CrystDB / HKUST-CrystDB — Curated crystal databases for XRD / CSP / CPP

  • Title: CrystDB / HKUST-CrystDB (Hugging Face datasets)
  • Maintainer: Bin Cao — HKUST(GZ); companion data to SimXRD-4M and XQueryer.
  • Data:
  • TL;DR: Two ASE-readable crystal databases targeted at XRD-based identification, crystal-property prediction (CPP), and crystal-structure prediction (CSP); each row exposes atomic species, fractional coordinates, lattice, and space group.

Reference structure databases (used by virtually every entry above)

Simulation, Refinement, and Utility Tools

Reviews & Surveys

Reviews directly focused on PXRD-driven analysis and structure determination:

Broader crystal-CSP / generative reviews (context for the methods above):

Related Curated Lists

Contributing

Additions, corrections, and translations are warmly welcomed — 欢迎补充、纠正与翻译！

There are three easy ways to contribute:

  • Open an issue — paste the paper title, arXiv id or GitHub link and (if possible) the section it belongs in.
  • Open a Pull Request — edit README.md following the format described in CONTRIBUTING.md.
  • Email the maintainer — for bulk additions (e.g. a whole research direction we missed).

See CONTRIBUTING.md for the full entry-format template, chronology rules, and link conventions.

Contributors

Thanks goes to these wonderful people who have contributed papers, corrections, and improvements to this list — 感谢以下贡献者对本仓库的补充、纠正与维护：

Avatars are rendered automatically by contrib.rocks from the GitHub contributor graph.

About

[OPEN Project] A collection of papers on XRD-based crystal structure identification and analysis.

Topics

machine-learning awesome-list crystal-structure materials-science generative-models rietveld-refinement pxrd phase-identification multi-phase-decomposition

Resources

Readme

License

MIT license

Contributing

Contributing
Activity

Stars

16 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors