Pith. sign in

REVIEW 2 major objections 4 minor 38 references

Using superpixels for interpretable feature reduction in large 2D diffraction datasets

T0 review · 2 major / 4 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read This paper argues that grouping detector pixels into variance-minimizing superpixels before non-negative matrix factorization turns a multi-hour 4D-STEM phase-mapping computation into a sub-minute one, with no visible change in the resultin

desk verdict Useful preprocessing recipe with a real speedup, but the 'without sacrificing accuracy' claim is asserted more than demonstrated — worth referee time. read the letter →

arxiv 2607.13979 v1 pith:DGYAFLCE submitted 2026-07-15 cond-mat.mtrl-sci

Andreas Werbrouck , Nikhila C. Paranamana , Andrew C. Meng , Xiaoqing He , Matthias J. Young This is my paper
classification cond-mat.mtrl-sci
keywords superpixelsfeaturereduction4D-STEMelectrondiffractionnon-negativematrixfactorizationphasemappingagglomerativeclusteringGaussianrandomprojection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes to compress each 2D diffraction pattern in a large dataset by grouping its detector pixels into superpixels—clusters of pixels that are averaged together, with small clusters where information density is high and large clusters where it is low. This shrinks the feature dimension of the data before machine-learning analysis, and the same superpixel layout can be reused across the whole dataset because the high- and low-information regions of the detector stay roughly fixed. The authors show that agglomerative clustering with a spatial-connectivity constraint produces superpixels that are easy to interpret (a beamstop becomes one pixel, Bragg peaks are densely sampled) and scale better than K-means when many superpixels are wanted. On a BaTiS3 4D-STEM dataset, this preprocessing drops the wall time for phase mapping by randomized non-negative matrix factorization from over 6000 seconds to roughly 40 seconds (or 26 seconds with an added Gaussian random projection), with no visible difference in the resulting phase maps and diffraction fingerprints. The larger point is that interpretable feature reduction can make a previously impractical analysis routine on commodity hardware.

What carries the argument

The load-bearing mechanism is variance-minimizing pixel clustering: for a diffraction dataset reshaped as an m×n matrix (m measurements, n detector pixels), the pixels are partitioned into k superpixels so that the sum of squared distances between each pixel and its superpixel average is minimized. Agglomerative clustering with a four-neighbor connectivity constraint solves this bottom-up, merging the two most similar adjacent clusters at each step using a variance-minimizing linkage; K-means solves the same objective top-down. Because both are distance-based, Gaussian random projection can reduce the dimensionality of the m-length feature vectors before clustering (approximately preserving

What would settle it

On a dataset with known phases (e.g., the BaTiS3 sample labeled by other TEM methods), compute phase maps with and without superpixel preprocessing and measure pixel-wise agreement with the labels. If a superpixel-based map misclassifies regions where a weak diffuse-scattering phase coexists with a strong crystalline peak—even though global reconstruction MSE stays low—the 'without sacrificing accuracy' claim is refuted.

Watch

Extended reading notes

Core claim

The central discovery is that diffraction datasets contain persistent spatial symmetries and redundant pixels that can be exploited by grouping pixels into superpixels of variable area, and that the most interpretable and scalable grouping is achieved by agglomerative clustering (bottom-up, merging the most similar adjacent clusters) rather than K-means (top-down), because the resulting superpixels are spatially connected and the runtime does not grow with the final number of superpixels. Applied as a preprocessing step before randomized non-negative matrix factorization, superpixels reduce the diffraction dimension from 65,536 pixels to 8,192 superpixels and cut the wall time of phase mappi

Load-bearing premise

The whole speedup rests on the assumption that minimizing pixel-averaging error (the mean squared error of reconstructing the data from superpixel averages) preserves exactly the information that phase-mapping later needs—so a phase that shows up only in low-intensity diffuse scattering or small peak shifts could be erased even while the reconstruction error stays small.

Editorial extensions

If this is right

  • The superpixel set acts as a custom virtual detector for the sample: dense sampling at the beam center and at diffraction peak positions, coarse averaging elsewhere, and the same layout applies to every frame in the series.
  • The 100-to-200-fold speedup makes (R)NMF phase mapping on large 4D-STEM datasets feasible on a laptop with commodity processors and memory, not just on dedicated compute clusters.
  • Because superpixels preserve 2D information such as Bragg spots and beam deflections, the approach goes beyond azimuthal integration, which would collapse that information into 1D.
  • Adding Gaussian random projection before clustering further cuts the total pipeline to about 26 seconds, with only a small additional reconstruction error.

Reading between the lines

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

  • If this generalizes beyond the single demonstrated dataset, phase mapping could become a real-time or near-real-time analysis, letting experimenters adjust synchrotron or microscope conditions while the sample is still in the beam.
  • The spatial interpretability of agglomerative superpixels (a beamstop merged into one pixel, circular groupings that mirror rotational symmetry) could make the preprocessing step more defensible to domain scientists than an opaque binning or autoencoder.
  • The paper leaves open whether the same speedup and fidelity carry to other downstream tasks such as principal component analysis, dictionary learning, or supervised classification; that is a direct testable extension.
  • A subtle risk is that MSE minimization protects intense Bragg peaks more than weak diffuse scattering, so a phase whose signal lives in low-intensity regions might be averaged away even when global reconstruction error stays small.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper proposes a preprocessing strategy for large 2D diffraction datasets (4D-STEM and synchrotron) in which detector pixels are grouped into 'superpixels' by variance-minimizing clustering, reducing the feature dimension before non-negative matrix factorization (NMF) phase mapping. Two clustering approaches are compared: K-means (top-down) and agglomerative clustering with spatial connectivity (bottom-up). The authors show that agglomerative clustering yields more interpretable, spatially connected superpixels, and that Gaussian random projection speeds up superpixel construction. On a BaTiS3 4D-STEM dataset, the reported wall-time for RNMF phase mapping drops from 6361.1 s to 26.4 s (~241x speedup), with superpixel-preprocessed phase maps and diffraction fingerprints that the authors state are visually indistinguishable from the full-resolution results.

Significance. If the accuracy claim holds, this is a practical and timely contribution: it makes large 4D-STEM phase-mapping analyses tractable on commodity hardware and generalizes preprocessing beyond square binning and cropping. The paper has clear strengths: a well-posed objective (Eq. 1-3), standard open-source algorithms, internally consistent wall-time measurements, and a public benchmark dataset from Ren et al. (2025). The interpretability of connected superpixels and the use of random projection are sensible ideas that could benefit the microscopy community. However, the central clause 'without sacrificing accuracy' is not yet quantitatively established; the evidence is global MSE and visual inspection, which are insufficient for the load-bearing claim.

major comments (2)
  1. [Application: preprocessing 4D-STEM data for phase mapping (Table 1, Figure 6)] The claim 'without sacrificing accuracy' is not quantitatively supported. The reported global reconstruction MSE values (1.173e-5 vs 1.187e-5) are dominated by intense features such as the direct beam and strong Bragg peaks. The phase-discriminating signal in this BaTiS3 dataset includes a sub-pixel direct-beam shift (ferroelectric deflection) and chiral-phase spot rotations—low-amplitude, low-variance features that can be averaged away without moving global MSE. The statement 'We leave the trade-off between computational cost and faithful data representation to the reader' explicitly defers the accuracy question. Please add quantitative agreement metrics between the full-resolution and superpixel-preprocessed phase maps/fingerprints (e.g., Pearson correlation, Dice overlap, or agreement with the original classification in Ren et al. 2025), and ideally test on a dataset with ground-truth
  2. [Superpixel construction (after Fig. 4)] The claim that 'this way of grouping pixels preserves more information than 2x2 square binning approaches' is asserted without a quantitative baseline comparison. Since square binning is the standard ad-hoc feature reduction that the paper aims to improve on, it needs to be tested at matched compression ratios. For example, compare k=4096 connected superpixels (n=16384 -> k=4096, a 4x reduction) against 2x2 binning on the same detector size in terms of reconstruction MSE and, more importantly, downstream phase-mapping accuracy. The word 'inherently' is not justified without this experiment; variable-size superpixels could, in principle, merge low-variance but phase-discriminating pixels with high-variance background.
minor comments (4)
  1. [Methods, Eq. (2)] The index notation in Eq. (2) is confusing (the first summation is over the m dimension, the second over n). Clarify that x_ij is the j-th diffraction measurement at detector pixel i, and mu_ij is the corresponding cluster center, or rewrite in matrix form.
  2. [Table 1] The table headings and first row seem misaligned (the 'No AC' row contains extra dashes and the column ordering is unclear). Please reformat so that the baseline row and the two superpixel rows have consistent entries for JL, AC, Post, RNMF, and wall time.
  3. [Figure 6] The figure caption appears to reverse the order of the superpixel and no-preprocessing panels relative to the description in the text. The text says 'The final phase maps and fingerprints are displayed in Figure 6 g-l' after describing the superpixel run, while the caption labels g-l as 'obtained without preprocessing.' Please verify and correct the panel labeling.
  4. [Increasing computational efficiency (Figure 5)] The Gaussian random projection target dimension is set to 32 with no sensitivity analysis. Since the JL-based acceleration is a claimed contribution, at least a small sweep of target dimensions (e.g., 8/16/32/64) should be reported, or a justification given for why 32 is sufficient.

Circularity Check

0 steps flagged · score 0.0 of 10

No load-bearing circularity; the 100-fold speedup is measured wall-clock time and the accuracy claim is supported by independent visual comparison of NMF outputs, not only by the minimized MSE objective.

full rationale

The central claim (Abstract: 'we show over 100-fold acceleration for phase mapping with Non-negative matrix factorization') is an empirical timing result: Table 1 reports wall times of 6361.1 s (no superpixel preprocessing) versus 26.4 s (JL projection + agglomerative clustering + post-processing), a measured speedup rather than a quantity derived from a fitted parameter. The superpixel objective in Eq. 1 and the reported MSE in Eq. 3 are algebraically the same criterion (Eq. 3 = Eq. 1 / mn), so the paper's low reconstruction errors are partly by construction; however, the paper does not rely solely on this MSE to assert accuracy — it additionally compares the resulting RNMF phase maps and diffraction fingerprints visually ('we find no difference between interpreting Figure 6 a-f and Figure 6 g-l') and explicitly leaves the cost/accuracy trade-off to the reader ('We leave the trade-off between computational cost and faithful data representation to the reader'), so the accuracy claim is an empirical, albeit qualitative, check rather than a definitional equivalence. The self-citation to the authors' prior RNMF work (Werbrouck et al., 2025) is used as a tool/baseline for NMF phase mapping and does not presuppose the conclusion; it is not load-bearing. No fitted parameter is renamed as a prediction, and no uniqueness argument is imported from the authors' prior work. Any concern that variance-minimizing superpixels could discard weak phase-discriminating signal is a correctness/validation risk, not circularity.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

No new physical entities, forces, or conserved quantities are introduced; the contribution is a data-processing procedure built from existing algorithms.

free parameters (3)
  • Number of superpixels k = 64, 512, 4096 for demo; 8192 for BaTiS3 phase mapping
    User-selected trade-off between compression and reconstruction error; not derived from theory.
  • Gaussian random projection target dimension = 32
    Chosen empirically ('an empirical value of 32 for the target dimension was used'); affects clustering speed and accuracy.
  • Number of RNMF components = 3
    Chosen to match the expected phases (thickness component plus two chiral orientations) in the BaTiS3 dataset; NMF output depends on this choice.
assumptions (3)
  • domain assumption High- and low-information detector regions are persistent across the diffraction series, so one superpixel map serves all frames.
    Stated in 'Proposed strategy': 'the locations of high and low information density on the detector generally stay roughly the same during a series of measurements.' If patterns shift strongly during in situ experiments, fixed superpixels may lose efficiency.
  • domain assumption Within-cluster variance (Euclidean distance of pixel intensity vectors) is a valid proxy for information content relevant to downstream phase mapping.
    Central to Eqs. (1)-(3); downstream NMF accuracy is only checked qualitatively via visual inspection of maps and fingerprints.
  • standard math Johnson-Lindenstrauss random projection preserves distances well enough for clustering at target dimension 32.
    Invoked to justify Gaussian random projection; theoretical guarantee is approximate, and the paper uses an empirical target dimension.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Using superpixels for interpretable feature reduction in large 2D diffraction datasets." pith.science (2026). https://pith.science/paper/DGYAFLCE

@misc{pith2026260713979,
  author       = {Pith},
  title        = {Pith review of: Using superpixels for interpretable feature reduction in large 2D diffraction datasets},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DGYAFLCE}},
  note         = {Machine review of arXiv:2607.13979}
}
read the original abstract

Large 2D diffraction datasets, consisting of hundreds or thousands of measurements, are commonly acquired with 4D-STEM electron diffraction or at synchrotron X-ray beamlines. Machine learning and artificial intelligence offer great promise for analyzing these datasets. However, the sheer volume of data presents a significant data processing bottleneck. Cropping the detector and pixel binning are standard ways to reduce data size. Here we propose grouping and averaging pixels into superpixels of variable area. High-information regions are sampled densely, while low-information areas are collected into larger superpixels. In the process, symmetries in the data are captured and exploited, making this approach suitable for preprocessing 2D diffraction data. We compare two variance-minimizing methods: K-means clustering (top-down) and agglomerative clustering (bottom-up) and demonstrate superior scaling and interpretability for the bottom-up method. As these methods are distance-based, we demonstrate that the construction of superpixels can be accelerated using Gaussian random projection. Finally we show over 100-fold acceleration for phase mapping with Non-negative matrix factorization on a 4D-STEM dataset when superpixels are used as a preprocessing step.

Figures

Figures reproduced from arXiv: 2607.13979 by the authors.

Figure 1
Figure 1. a) In 4D-STEM or b) during a synchrotron measurement, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Vector quantization compared to agglomeration approaches for clustering pixels in an image with [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Visualization of pixel groups with 64, 512 and 4096 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: Reconstruction error and execution time of different [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: a-c) Phase maps and d-f) diffraction signatures [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

38 extracted references · 21 canonical work pages

  1. [1]

    Barry and Williams, David B

    Carter, C. Barry and Williams, David B. , title =

  2. [2]

    Microscopy and Microanalysis , year = 2019, month = may, volume = 25, number = 3, pages =

    Ophus, Colin , title =. Microscopy and Microanalysis , year = 2019, month = may, volume = 25, number = 3, pages =. doi:10.1017/s1431927619000497 , issn =

    doi:10.1017/s1431927619000497 2019

  3. [3]

    Sokolov, A. A. and Ternov, I. M. , title =

  4. [4]

    , title =

    He, Bob B. , title =

  5. [5]

    , title =

    Booth, C. , title =. Microscopy and Microanalysis , year = 2012, month = jul, volume = 18, number =. doi:10.1017/S1431927612002243 , issn =

    work page doi:10.1017/s1431927612002243 2012

  6. [6]

    and Purohit, Prafull and Chamberlain, Darol and Nguyen, Kayla X

    Tate, Mark W. and Purohit, Prafull and Chamberlain, Darol and Nguyen, Kayla X. and Hovden, Robert and Chang, Celesta S. and Deb, Pratiti and Turgut, Emrah and Heron, John T. and Schlom, Darrell G. and Ralph, Daniel C. and Fuchs, Gregory D. and Shanks, Katherine S. and Philipp, Hugh T. and Muller, David A. and Gruner, Sol M. , title =. Microscopy and Micro...

    work page doi:10.1017/s1431927615015664 2016

  7. [7]

    and Eikenberry, E

    Broennimann, Ch. and Eikenberry, E. F. and Henrich, B. and Horisberger, R. and Huelsen, G. and Pohl, E. and Schmitt, B. and Schulze-Briese, C. and Suzuki, M. and Tomizaki, T. and Toyokawa, H. and Wagner, A. , title =. Journal of Synchrotron Radiation , year = 2006, month = mar, volume = 13, number = 2, pages =. doi:10.1107/S0909049505038665 , issn =

    work page doi:10.1107/s0909049505038665 2006

  8. [8]

    Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment , series =

    Dinapoli, Roberto and Bergamaschi, Anna and Henrich, Beat and Horisberger, Roland and Johnson, Ian and Mozzanica, Aldo and Schmid, Elmar and Schmitt, Bernd and Schreiber, Akos and Shi, Xintian and Theidel, Gerd , title =. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment , series...

    work page doi:10.1016/j.nima.2010.12.005 2010

Show all 38 references
  1. [9]

    Small , year = 2018, volume = 14, number = 46, pages =

    Wang, Chunpeng and Steiner, Ullrich and Sepe, Alessandro , title =. Small , year = 2018, volume = 14, number = 46, pages =. doi:10.1002/smll.201802291 , issn =

    2018 doi

  2. [10]

    Nature Reviews Physics , year = 2021, month = dec, volume = 3, number = 12, pages =

    Li, Jizhou and Huang, Xiaobiao and Pianetta, Piero and Liu, Yijin , title =. Nature Reviews Physics , year = 2021, month = dec, volume = 3, number = 12, pages =. doi:10.1038/s42254-021-00397-0 , issn =

    2021 doi

  3. [11]

    , title =

    Krizhevsky, Alex and Sutskever, Ilya and Hinton, Geoffrey E. , title =. Advances in Neural Information Processing Systems , year = 2012, volume = 25, organization =

    2012

  4. [12]

    and Huttenlocher, Daniel P

    Felzenszwalb, Pedro F. and Huttenlocher, Daniel P. , title =. International Journal of Computer Vision , year = 2004, month = sep, volume = 59, number = 2, pages =. doi:10.1023/B:VISI.0000022288.19776.77 , issn =

    2004

  5. [13]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year = 2012, month = nov, volume = 34, number = 11, pages =

    Achanta, Radhakrishna and Shaji, Appu and Smith, Kevin and Lucchi, Aurelien and Fua, Pascal and S. IEEE Transactions on Pattern Analysis and Machine Intelligence , year = 2012, month = nov, volume = 34, number = 11, pages =. doi:10.1109/TPAMI.2012.120 , issn =

    2012 doi

  6. [14]

    and Werbrouck, Andreas and Datta, Amit K

    Paranamana, Nikhila C. and Werbrouck, Andreas and Datta, Amit K. and He, Xiaoqing and Young, Matthias J. , title =. Advanced Energy Materials , year = 2024, pages =. doi:10.1002/aenm.202403904 , issn =

    2024 doi

  7. [15]

    Journal of Machine Learning Research , year = 2011, volume = 12, number = 85, pages =

    Pedregosa, Fabian and Varoquaux, Ga. Journal of Machine Learning Research , year = 2011, volume = 12, number = 85, pages =

    2011

  8. [16]

    and Lindenstrauss, Joram , title =

    Johnson, William B. and Lindenstrauss, Joram , title =. Contemporary Mathematics , publisher =

  9. [17]

    Proceedings of the Second International Conference on Knowledge Discovery and Data Mining (KDD'96) , year = 1996, month = aug, pages =

    Ester, Martin and Kriegel, Hans-Peter and Sander, J. Proceedings of the Second International Conference on Knowledge Discovery and Data Mining (KDD'96) , year = 1996, month = aug, pages =

    1996

  10. [18]

    and He, Xiaoqing and Young, Matthias J

    Werbrouck, Andreas and Paranamana, Nikhila C. and He, Xiaoqing and Young, Matthias J. , title =. arXiv preprint arXiv:2507.17068 , year = 2025, month = jul, doi =. 2507.17068 , archiveprefix =

    2025 arXiv

  11. [19]

    and Hachtel, Jordan A

    Ren, Guodong and Jung, Gwan Yeong and Chen, Huandong and Wang, Chong and Zhao, Boyang and Vasudevan, Rama K. and Hachtel, Jordan A. and Lupini, Andrew R. and Chi, Miaofang and Xiao, Di and Ravichandran, Jayakanth and Mishra, Rohan , title =. Advanced Functional Materials , yea...

    2025 arXiv

  12. [20]

    Benjamin and Mendible, Ariana and Wihlborn, Sophie and Kutz, J

    Erichson, N. Benjamin and Mendible, Ariana and Wihlborn, Sophie and Kutz, J. Nathan , title =. Pattern Recognition Letters , year = 2018, volume = 104, pages =

    2018

  13. [21]

    and Pekin, Thomas C

    Allen, Frances I. and Pekin, Thomas C. and Persaud, Arun and Rozeveld, Steven J. and Meyers, Gregory F. and Ciston, Jim and Ophus, Colin and Minor, Andrew M. , title =. Microscopy and Microanalysis , year = 2021, volume = 27, number = 4, pages =

    2021

  14. [22]

    and Bedford, Nicholas M

    Young, Matthias J. and Bedford, Nicholas M. and Yanguas-Gil, Angel and Letourneau, Steven and Coile, Matthew and Mandia, David J. and Aoun, Bachir and Cavanagh, Andrew S. and George, Steven M. and Elam, Jeffrey W. , title =. ACS Applied Materials & Interfaces , year = 2020, mo...

    2020 doi

  15. [23]

    and Bedford, Nicholas M

    Young, Matthias J. and Bedford, Nicholas M. and Jiang, Naisheng and Lin, Deqing and Dai, Liming , title =. Journal of Synchrotron Radiation , year = 2017, month = may, volume = 24, number = 4, pages =. doi:10.1107/s1600577517006282 , issn =

    2017 doi

  16. [24]

    Khamidy, N. I. and Kovacova, V. and Bernasconi, A. and Le Rhun, G. and Vaxelaire, N. , title =. Materials Research Express , year = 2017, month = aug, volume = 4, number = 8, pages =. doi:10.1088/2053-1591/aa5e5a , issn =

    2017 doi

  17. [25]

    Macromolecules , year = 2006, month = feb, volume = 39, number = 6, pages =

    Sawayanagi, Tomoharu and Tanaka, Toshihisa and Iwata, Tadahisa and Abe, Hideki and Doi, Yoshiharu and Ito, Kazuki and Fujisawa, Tetsuro and Fujita, Masahiro , title =. Macromolecules , year = 2006, month = feb, volume = 39, number = 6, pages =. doi:10.1021/ma052425h , issn =

    2006 doi

  18. [26]

    Cell , year = 2014, month = nov, volume = 159, number = 5, pages =

    Shi, Yigong , title =. Cell , year = 2014, month = nov, volume = 159, number = 5, pages =. doi:10.1016/j.cell.2014.10.051 , issn =

    2014 doi

  19. [27]

    Bendersky, L. A. and Gayle, F. W. , title =. Journal of Research of the National Institute of Standards and Technology , year = 2001, month = nov, volume = 106, number = 6, pages =. doi:10.6028/jres.106.051 , issn =

    2001 doi

  20. [28]

    , title =

    Krawitz, Aaron D. , title =

  21. [29]

    Journal of Synchrotron Radiation , year = 2024, month = jun, volume = 31, number = 4, pages =

    Kawaguchi, Shogo and Kobayashi, Shintaro and Yamada, Hiroki and Ashitani, Hirotaka and Takemoto, Michitaka and Imai, Yasuhiko and Hatsui, Takaki and Sugimoto, Kunihisa and Sakata, Osami , title =. Journal of Synchrotron Radiation , year = 2024, month = jun, volume = 31, number...

    2024 doi

  22. [30]

    and Ivashko, Oleh and Igoa Salda

    Zimmermann, Martin v. and Ivashko, Oleh and Igoa Salda. Journal of Synchrotron Radiation , year = 2025, month = apr, volume = 32, number = 3, pages =. doi:10.1107/s1600577525002826 , issn =

    2025 doi

  23. [31]

    Journal of Synchrotron Radiation , year = 2021, month = jan, volume = 28, number = 1, pages =

    Sch. Journal of Synchrotron Radiation , year = 2021, month = jan, volume = 28, number = 1, pages =. doi:10.1107/s1600577520013223 , issn =

    2021 doi

  24. [32]

    and Franz, Hermann , title =

    Dippel, Ann-Christin and Liermann, Hanns-Peter and Delitz, Jan Torben and Walter, Peter and Schulte-Schrepping, Horst and Seeck, Oliver H. and Franz, Hermann , title =. Journal of Synchrotron Radiation , year = 2015, month = apr, volume = 22, number = 3, pages =. doi:10.1107/s...

    2015 doi

  25. [33]

    Allan, D. R. and Collins, S. P. and Evans, G. and Hall, D. and McAuley, K. and Owen, R. L. and Sorensen, T. and Tang, C. C. and von Delft, F. and Wagner, A. and Wilhelm, H. , title =. The European Physical Journal Plus , year = 2015, month = mar, volume = 130, number = 3, doi =

    2015

  26. [34]

    Journal of Synchrotron Radiation , year = 2023, month = jul, volume = 30, number = 5, pages =

    Fitch, Andrew and Dejoie, Catherine and Covacci, Ezio and Confalonieri, Giorgia and Grendal, Ola and Claustre, Laurent and Guillou, Perceval and Kieffer, J. Journal of Synchrotron Radiation , year = 2023, month = jul, volume = 30, number = 5, pages =. doi:10.1107/s160057752300...

    2023 doi

  27. [35]

    Powder Diffraction , year = 2014, month = dec, volume = 30, number =

    Bernasconi, Andrea and Wright, Jonathan and Harker, Nicholas , title =. Powder Diffraction , year = 2014, month = dec, volume = 30, number =. doi:10.1017/s0885715614001304 , issn =

    2014 doi

  28. [36]

    Liermann, H.-P. and Kon. Journal of Synchrotron Radiation , year = 2015, month = jun, volume = 22, number = 4, pages =. doi:10.1107/s1600577515005937 , issn =

    2015 doi

  29. [37]

    Journal of Synchrotron Radiation , year = 2022, month = jan, volume = 29, number = 1, pages =

    Desjardins, Kewin and Mocuta, Cristian and Dawiec, Arkadiusz and R. Journal of Synchrotron Radiation , year = 2022, month = jan, volume = 29, number = 1, pages =. doi:10.1107/s1600577521012492 , issn =

    2022 doi

  30. [38]

    Microscopy and Microanalysis , volume =

    Gulati, Sahil and Pakzad, Anahita , title =. Microscopy and Microanalysis , volume =. 2023 , month =. doi:10.1093/micmic/ozad067.536 , url =

    2023 doi

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.