Pith. sign in

REVIEW 3 major objections 6 minor 42 references

AES-Debye: an Accurate, Efficient, and Scalable Engine for Debye Scattering Calculations

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read AES-Debye preserves the numerical rigor of the Rose-X Debye scattering formulation while adding a parallel CPU/GPU engine that computes a 90-million-atom powder pattern in about 52 minutes.

desk verdict Useful parallel Debye engine with real benchmarks, but the printed bin-correction formula is dimensionally wrong and accuracy is validated only against the same group's old code. read the letter →

arxiv 2608.09916 v1 pith:RGEU56PY submitted 2026-08-10 cond-mat.mtrl-sci cond-mat.mes-hallphysics.app-ph

Navid Panchi , Sebastian Kuckuk , Markus Wittmann , Michael Engel , Alberto Leonardi This is my paper · ORCID
classification cond-mat.mtrl-scicond-mat.mes-hallphysics.app-ph MSC 65Y0565Z0578A45
keywords DebyescatteringequationpairdistributionfunctiontotalpowderdiffractionGPUaccelerationbincentercorrectionhigh-performancecomputingcelllistdomaindecomposition
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper presents AES-Debye, an engine for computing powder total scattering profiles from the Debye scattering equation without sacrificing accuracy for speed. The central claim is that by grouping pair distances into a pair distribution function with corrected bin centers and integer-based accumulation, the method reproduces the numerical accuracy of the earlier Rose-X code while running up to 20 times faster, scaling to about 90 million atoms in about 52 minutes. A sympathetic reader would care because accurate total scattering calculations for large disordered and compositionally complex systems have been practically out of reach, and this work reports making them routine on ordinary CPU clusters and GPUs.

What carries the argument

The carrying object is the pair distribution function with corrected bin centers, built through a two-stage formulation $I(Q)=\sum_{a,b} f_a f_b \sum_k N_k \frac{\sin(Q\nu_k)}{Q\nu_k}$. Each bin stores count $N_k$ and the accumulated squared pair-distance error $\psi_k$, from which the corrected center is obtained via a series expansion (Eq. 4). The paper also introduces a cell-list–based domain decomposition that sorts cell pairs by center-to-center distance; this is what turns random PDF updates into local ones and is the main reason disordered systems become tractable.

What would settle it

Compute the powder profile for a moderately sized disordered system (e.g., a 100,000-atom Pd nanoparticle with Gaussian disorder $\sigma$ ~ 0.5 Å) with AES-Debye and with a direct pairwise Debye summation using arbitrary-precision arithmetic (or long double with compensated summation) at low Q, say Q < 1 $Å^{-1}$, where cancellation is strongest; if the relative difference exceeds the claimed ~1.3e-7, the accuracy claim is refuted.

Watch

Extended reading notes

Core claim

The paper claims that accuracy and speed are not in conflict in Debye scattering calculations if the pair-distance histogram is constructed with two safeguards: bin centers are corrected using the accumulated squared-distance error (ESPD) so that each bin's representative distance tracks the true average distance, and all bin bookkeeping is done in 64-bit integers so that floating-point cancellation cannot corrupt the histogram. On top of this, a cell-list domain decomposition sorts cell pairs by center-to-center distance, making PDF updates land in nearby histogram regions and restoring cache locality that plain random access destroys, especially in disordered systems. The result is an implementation that matches Rose-X's intensity profiles to a maximum relative error of about 1.3e-7 while sustaining hundreds of millions of pair distances per second per core and reaching about 1.5 billion pair distances per second on an A40 GPU; the largest demonstration is a 90.1-million-atom polycrystalline copper sample computed on 64 CPU nodes in 3,159 seconds.

Load-bearing premise

The accuracy claim rests on agreement with Rose-X, a predecessor code from the same group, rather than on an independent brute-force reference; if Rose-X's own corrections contain shared systematic errors, the two codes could agree and both be wrong.

Editorial extensions

If this is right

  • Large disordered microstructures, such as polycrystalline metals with millions of atoms, can be checked against experimental powder patterns directly, without relying on Bragg-approximation models or parameterized peak shapes.
  • Total scattering profiles and high-resolution PDFs come from the same computation, so structural refinement workflows can exploit both without additional expensive runs.
  • Because the Debye scattering equation is length-scale independent, the same engine applies to small-angle scattering of colloidal assemblies, as the paper demonstrates with a 200,000-particle crystallization study.
  • The reported throughput advantage over a recent brute-force GPU code suggests that PDF-based methods with good cache locality, rather than direct pairwise summation, are the practical route for large-system total scattering.
  • The method's $O(N^2)$ pair enumeration remains the computational floor, so further gains must come from smarter pair pruning or hardware, not from the reciprocal-space evaluation step.
  • The bin-center correction removes unphysical negative intensities and improves the diffuse background, which is the region that carries information about disorder and is most sensitive to approximations.

Reading between the lines

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

  • If the accuracy claim holds at the reported 1.3e-7 level, then simulated diffuse backgrounds become effectively exact compared with experimental noise, allowing subtle disorder signatures to be interpreted with confidence.
  • A direct independent validation against arbitrary-precision direct summation on a moderately sized system would strengthen the claim beyond the Rose-X comparison, which the paper does not provide but the machinery makes straightforward.
  • The same bin-center correction could be extended to partial PDFs in multicomponent systems in a way that preserves each partial's accuracy, a capability the paper gestures at but does not fully exploit.
  • The huge gap in throughput versus DebyeCalculator likely reflects a change in algorithmic regime (PDF-based versus direct summation) rather than pure implementation quality; a controlled benchmark with matched arithmetic and thread counts would be informative.
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

3 major / 6 minor

Summary. The manuscript describes AES-Debye, an open-source implementation of the Debye scattering equation that constructs a pair-distance histogram with corrected bin centers and integer-based accumulation, and parallelizes the pair-distance computation through OpenMP, MPI, and CUDA. The authors claim that the method retains the numerical rigor of the Rose-X code while improving memory locality and scalability, with speedups up to 20x, and they demonstrate the implementation with benchmarks and a 90-million-atom polycrystalline copper calculation completed in about 52 minutes. Applications to nanocube superlattices, halloysite tubes, and colloidal systems are presented to show practical reach.

Significance. If the numerical accuracy claim survives scrutiny, AES-Debye would be a useful contribution: it addresses a real bottleneck in total scattering modeling for large disordered systems, is released as open source, and includes a very large demonstration that is beyond most existing DSE implementations. The strong-scaling measurements and the independent comparison against DebyeCalculator are concrete strengths. However, the accuracy case is weakened by an incorrect printed bin-correction formula and by validation performed against a same-group predecessor code rather than an independent reference; these points need to be resolved before the significance can be judged.

major comments (3)
  1. [Section 2.1, Eq. (4)] The mean-shift expansion in Eq. (4) is dimensionally inconsistent and mathematically incorrect. With ψ the accumulated squared-distance error (dimension L^2) and ν a distance (dimension L), the terms ψ^3/(4ν^2) and ψ^4/(8ν^3) have dimensions L^4 and L^5, whereas δ is a length. The correct expansion for δ = sqrt(ν^2+ψ)−ν is ψ/(2ν) − ψ^2/(8ν^3) + ψ^3/(16ν^5) + …, with powers and signs different from Eq. (4). Because the bin-center correction is the core accuracy mechanism, the submitted manuscript does not state a correct method; directing the reader to Ref. [25] for the derivation does not repair an equation that is presented as the method. Please correct Eq. (4) (including, if needed, the definition of ψ or the averaging procedure) and verify the corrected formula numerically.
  2. [Section 2.2, Eq. (5)] The overflow bound in Eq. (5) is not dimensionally consistent with the accumulated quantity ψ. Since each update adds δr·δr − ν^2, a pair at the upper bin boundary contributes (ν_ul^2 − ν^2) = (ν_ul−ν)(ν_ul+ν) to ψ, not (ν_ul−ν). For the final bin, where ν is large, the missing factor (ν_ul+ν) is substantial, so the printed n_max overestimates the maximum safe count and the headroom counter may not trigger before an actual overflow. Please replace Eq. (5) with n_max = INT64_MAX / (ν_ul^2 − ν^2) in the appropriate integer units and re-derive the reported 5.6×10^6 estimate.
  3. [Section 3, Fig. 3a] The only large-scale accuracy validation is against Rose-X, a predecessor code with overlapping authorship and the same binning-correction scheme. A 1.3×10^−7 relative difference between AES-Debye and Rose-X demonstrates that the two implementations agree with each other, but unless Rose-X is independently established as accurate, it does not demonstrate accuracy of the shared approximation. This is load-bearing because the headline claim is that AES-Debye retains the numerical rigor of Rose-X. Please add at least one independent validation: for example, compare against a brute-force DSE calculation for a small system (the 32,000-atom Pd nanoparticle of Section 4.4 is suitable) over the same Q grid and report maximum and mean relative deviations, and also test the corrected bin centers against exactly computed mean distances in representative bins.
minor comments (6)
  1. [Section 2.5.3] The text reads 'shared memory parallelsm'; this should be 'parallelism'.
  2. [Section 4.2] The opening sentence refers to 'GPU throughout'; this should be 'GPU throughput'.
  3. [Section 4.1] There are two typos: 'Strong-scaling results are sown' should be 'shown', and 'largly removes' should be 'largely removes'.
  4. [Section 5.1] The name 'AES-debye' should be capitalized consistently as 'AES-Debye'.
  5. [Figure 10 caption] The caption contains 'packing faction' and 'confinment'; these should be 'packing fraction' and 'confinement'.
  6. [Section 1 and Conclusions] The claim of speedups up to 20x over Rose-X is stated without a direct timing or speedup table; Fig. 5(a) shows a dashed comparison line but no quantitative speedup values. Please add a direct measurement or qualify the claim.

Circularity Check

1 steps flagged · score 2.0 of 10

No significant circularity; central performance claim is independently benchmarked, but the accuracy claim leans on same-group predecessor Rose-X as its reference.

  1. self citation load bearing [Section 3 (Accuracy and Numerical Validation); Section 2.1 Eq. (4)]
    "To assess the accuracy of the PDF-based implementation, we therefore compared the computed intensity profiles against those obtained with Rose-X [25] ... This confirms that AES-Debye reproduces the numerical accuracy of the Rose-X formulation while enabling the architectural and performance improvements described above."

    Accuracy is operationalized as agreement with Rose-X, a predecessor code by co-author A.L. ('A.L. conceived the method and contributed the earlier Rose-X software'). The same bin-correction scheme is adopted from that work ('Full derivations and error propagation details are given in Ref. [25]'). So the central accuracy claim reduces to consistency with a same-group implementation of the same correction: a 1.3e-7 match demonstrates code equivalence, not independent numerical accuracy, and shared systematic bin errors would be invisible. Separate from circularity, Eq. (4) as printed is dimensionally inconsistent (psi has L^2 but the displayed psi^3/nu^2 and psi^4/nu^3 terms scale as L^4 and L^5), which is a correctness risk in that same self-cited correction.

full rationale

The performance and scalability results are self-contained and externally benchmarked: Section 4.4 compares throughput against DebyeCalculator, an independent brute-force code, and the strong-scaling measurements use fixed workloads and hardware. The PDF binning, cell-list decomposition, and OpenMP/MPI/CUDA parallelization are algorithmic, with no fitted parameters renamed as predictions. The only self-citation concern is the accuracy validation against Rose-X (Ref. [25]), which shares a co-author and the same bin-correction approach; the 1.3e-7 agreement confirms implementation equivalence but does not independently validate the bin correction itself. This is a minor, non-forceful self-citation rather than a derivation that collapses into its inputs—AES-Debye's output intensities are not computed from Rose-X, and the paper explicitly targets 'reproduces the numerical accuracy of the Rose-X formulation.' I therefore score 2, noting additionally that Eq. (4)'s dimensional inconsistency is a numerical-rigor defect to fix, not a circularity.

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

No free parameters are fitted to experimental data. The PDF bin width, cell counts, and coordinate scaling factor are user or implementation choices that affect performance and resolution, not fitted scientific parameters. No new physical entities are postulated; the only imports from prior literature are the standard Debye equation, the bin center correction of Ref. [25], and common simulation data assumptions.

assumptions (4)
  • domain assumption The Debye scattering equation (Eq. 1) is the exact powder-averaged elastic scattering intensity for atoms with independent scattering factors.
    The engine's accuracy is defined relative to this standard equation. It ignores directional correlations or texture, which the author's note as a limitation.
  • domain assumption The bin center correction series expansion in Eq. (4), imported from Ref. [25], correctly estimates the mean distance shift within each bin.
    The paper does not derive this series; it references Leonardi and Bish. As printed, the second and third terms are dimensionally inconsistent, so the actual implementation may differ from the printed formula.
  • domain assumption Input coordinates are accurately represented by six significant digits, justifying the 10^9 int64 scaling.
    Section 2.2 states that LAMMPS and HOOMD-blue outputs contain no more than six significant digits, which may not hold for all input formats or simulation tools.
  • ad hoc to paper The worst-case overflow bound in Eq. (5) uses (nu_ul - nu) as the per-update increment to the accumulated squared-distance error psi.
    The formula is presented as a safe upper bound for psi accumulation, but psi stores squared distance differences, so the correct increment should be nu_ul^2 - nu^2. As written the bound is dimensionally wrong, though conservative.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AES-Debye: an Accurate, Efficient, and Scalable Engine for Debye Scattering Calculations." pith.science (2026). https://pith.science/paper/RGEU56PY

@misc{pith2026260809916,
  author       = {Pith},
  title        = {Pith review of: AES-Debye: an Accurate, Efficient, and Scalable Engine for Debye Scattering Calculations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RGEU56PY}},
  note         = {Machine review of arXiv:2608.09916}
}
abstract

Total scattering models are essential for characterizing the structure and disorder of nanoscale materials. The Debye scattering equation (DSE) provides a rigorous route to elastic total scattering, but its direct evaluation is computationally demanding because pairwise contributions must be accumulated at every scattering vector, whereas common acceleration strategies based on binned pair-distance distributions or gridded fast Fourier transforms can introduce discretization and aliasing artifacts that compromise diffuse-scattering accuracy. Here, we present AES-Debye, an accuracy-preserving DSE framework that aggregates pair distances into a pair distribution function (PDF) using corrected bin centers and numerically robust accumulation to suppress discretization and summation errors. A data-locality-aware parallel design enables efficient execution on CPUs and GPUs. We demonstrate strong scalability by computing a high-resolution total scattering profile for a system of 90 million atoms, $(0.1,\mu\mathrm{m})^{3}$, in minutes on a distributed-memory CPU platform. These capabilities extend accurate elastic total scattering calculations to large, complex systems while simultaneously providing high-resolution PDFs for downstream structural analysis.

Figures

Figures reproduced from arXiv: 2608.09916 by the authors.

Figure 1
Figure 1. Schematic of the cell-list-based domain decomposition in a 2D toy system. [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. CPU parallelization scheme using a hybrid MPI+OpenMP model. Each node runs a [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Validation of the PDF-based implementation and bin center correction. (a) Comparison of [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Diffraction profiles for a Pd nanocube (side length 10 nm) with increasing Gaussian [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Performance scaling across CPUs and GPUs for different disorder levels [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Strong-scaling performance as a function of node count for a 4 million atom system. Only [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: (a) Pd nanocubes (5 nm) assembled into a 5 [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: (a) Polycrystalline Cu (FCC) sample with 1000 grains; atoms colored by coordination [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]
Figure 9
Figure 9. Figure 9: (a) Atomic model of a halloysite nanotube generated from a kaolinite sheet. (b) Two [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: (a) Spherical nanoparticles (diameter: 200nm) crystallized in a spherical confinement, at [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 42 canonical work pages

  1. [25]

    Leonardi, D

    A. Leonardi, D. L. Bish,Journal of Applied Crystallography2016,49, 1593–1608

    work page

  2. [1]

    J. A. Kaduk, S. J. L. Billinge, R. E. Dinnebier, N. Henderson, I. Madsen, R. ˇCern´ y, M. Leoni, L. Lutterotti, S. Thakral, D. Chateigner,Nature Reviews Methods Primers2021,1, 77

    work page

  3. [2]

    Zhang, L

    Y. Zhang, L. Wu, M. Ji, B. Wang, Y. Kong, J. Xu,Optical Materials Express2011,2, 92

    work page

  4. [3]

    Cernuto, N

    G. Cernuto, N. Masciocchi, A. Cervellino, G. M. Colonna, A. Guagliardi,Journal of the American Chemical Society2011,133, 3114–3119

    work page

  5. [4]

    Zhang, Z

    S. Zhang, Z. Huang, Z. Wen, L. Zhang, J. Jin, R. Shahbazian-Yassar, J. Yang,Nano Letters 2017,17, 3518–3526

    work page 2017

  6. [5]

    Rietveld,Acta Crystallographica1967,22, 151–152

    H. Rietveld,Acta Crystallographica1967,22, 151–152. 21

    work page

  7. [6]

    Scardi, M

    P. Scardi, M. Leoni inSpringer Series in Materials Science, Springer Berlin Heidelberg,2004, pp. 51–91

    work page 2004

  8. [7]

    Gelisio, P

    L. Gelisio, P. Scardi,Acta Crystallographica Section A Foundations and Advances2016,72, 608–620

    work page

Show all 42 references
  1. [8]

    Kirk C., Ross; Joseph A., M

    P. Kirk C., Ross; Joseph A., M. Andrew M.,Powder Diffraction2014,29, 337–345

  2. [9]

    K. J., W. J. P.,Powder Diffraction2013,28, S339–S350

  3. [10]

    Leonardi,IUCrJ2021,8, 257–269

    A. Leonardi,IUCrJ2021,8, 257–269

  4. [11]

    Debye,Mathematisch-Physikalische Klasse1915, 70–60

    P. Debye,Mathematisch-Physikalische Klasse1915, 70–60

  5. [12]

    Scardi, L

    P. Scardi, L. Gelisio,Scientific reports2016,6, 22221

  6. [13]

    B. E. Warren,X-Ray Diffraction, Dover Publications, Mineola, NY,1990

    1990

  7. [14]

    Cervellino, R

    A. Cervellino, R. Frison,Acta Crystallogr. A Found. Adv.2020,76, 302–317

    2020

  8. [15]

    Marcin, GitHub - wojdyr/debyer: Debye’s scattering equation & other analysis of atomistic models

    W. Marcin, GitHub - wojdyr/debyer: Debye’s scattering equation & other analysis of atomistic models. — github.com,https://github.com/wojdyr/debyer

  9. [16]

    M. C. Watson, J. E. Curtis,Journal of applied crystallography2013,46, 1171–1177

  10. [17]

    R. F. Grover, D. R. McKenzie,Acta Crystallographica Section A: Foundations of Crystallog- raphy2001,57, 739–740

  11. [18]

    N. W. Thomas,Acta Crystallographica Section A: Foundations of Crystallography2010,66, 64–77

  12. [19]

    Gelisio, C

    L. Gelisio, C. L. Azanza Ricardo, M. Leoni, P. Scardi,Journal of Applied Crystallography 2010,43, 647–653

    2010

  13. [20]

    F. L. Johansen, A. S. Anker, U. Friis-Jensen, E. B. Dam, K. M. Jensen, R. Selvan,Journal of Open Source Software2024,9, 6024

  14. [21]

    Neverov,SoftwareX2017,6, 63–68

    V. Neverov,SoftwareX2017,6, 63–68

  15. [22]

    Rudolph, M

    M. Rudolph, M. Motylenko, D. Rafaja,IUCrJ2019,6, 116–127

  16. [23]

    Reuter, J

    K. Reuter, J. K¨ ofinger,Computer Physics Communications2019,236, 274–284

  17. [24]

    B. Hall, R. Monot,Computers in Physics1991,5, 414–417

  18. [26]

    A. P. Thompson, H. M. Aktulga, R. Berger, D. S. Bolintineanu, W. M. Brown, P. S. Crozier, P. J. in ’t Veld, A. Kohlmeyer, S. G. Moore, T. D. Nguyen, R. Shan, M. J. Stevens, J. Tranchida, C. Trott, S. J. Plimpton,Comput. Phys. Commun.2022,271, 108171

    2022

  19. [27]

    J. A. Anderson, J. Glaser, S. C. Glotzer,Computational Materials Science2020,173, 109363

  20. [28]

    Vaccari, P

    M. Vaccari, P. Fornasini,Journal of synchrotron radiation2006,13, 321–325. 22

  21. [29]

    Fritz - NHR@FAU HPC Documentation — doc.nhr.fau.de,https : / / doc . nhr . fau . de / clusters/fritz

  22. [30]

    Alex - NHR@FAU HPC Documentation — doc.nhr.fau.de,https : / / doc . nhr . fau . de / clusters/alex

  23. [31]

    Johansen, Andy S

    Frederik L. Johansen, Andy S. Anker, Ulrik Friis-Jensen, Erik B. Dam, Kirsten M. Ø. Jensen, Raghavendra Selvan,Journal of Open Source Software2024, DOI10.5281/zenodo.10659528

  24. [32]

    S. Toso, D. Baranov, C. Giannini, S. Marras, L. Manna,ACS Mater. Lett.2019,1, 272–276

    2019

  25. [33]

    Leonardi, M

    A. Leonardi, M. Leoni, P. Scardi,Computational Materials Science2013,67, 238–242

  26. [34]

    Leonardi, D

    A. Leonardi, D. L. Bish,Acta Materialia2017,133, 380–392

  27. [35]

    Rawtani, Y

    D. Rawtani, Y. Agrawal,Rev. Adv. Mater. Sci2012,30, 282–295

  28. [36]

    Fizir, P

    M. Fizir, P. Dramou, N. S. Dahiru, W. Ruya, T. Huang, H. He,Microchimica Acta2018, 185, 1–33

  29. [37]

    L. Yang, X. Yang, F. Xia, Y. Gong, F. Li, J. Yu, T. Gao, Y. Li,Chemistry–An Asian Journal 2023,18, e202300473

    2023

  30. [38]

    Heidari Pebdani,Computational Materials Science2023,218, 111948

    M. Heidari Pebdani,Computational Materials Science2023,218, 111948

  31. [39]

    D. A. Prishchenko, E. V. Zenkov, V. V. Mazurenko, R. F. Fakhrullin, Y. M. Lvov, V. G. Mazurenko,Physical Chemistry Chemical Physics2018,20, 5841–5849

  32. [40]

    Gianni, M

    E. Gianni, M. Pˇ seniˇ cka, K. Mackov´ a, E. Scholtzov´ a, L. Jankoviˇ c, M. Mareˇ s, D. Papoulis, M. Posp ´ ıˇ sil,Journal of Molecular Structure2023,1287, 135639

  33. [41]

    Jakob, J

    W. Jakob, J. Rhinelander, D. Moldovan, pybind11 — Seamless operability between C++11 and Python, https://github.com/pybind/pybind11,2016

    2016

  34. [42]

    Ask Hjorth Larsen, Jens Jørgen Mortensen, Jakob Blomqvist, Ivano E Castelli, Rune Chris- tensen, Marcin Du lak, Jesper Friis, Michael N Groves, Bjørk Hammer, Cory Hargus, Eric D Hermes, Paul C Jennings, Peter Bjerre Jensen, James Kermode, John R Kitchin, Esben Leon- hard Kolsb...

    2017

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.