The reviewed record of science sign in
Pith

arxiv: 2607.00756 · v1 · pith:5X3DCM5D · submitted 2026-07-01 · cond-mat.mtrl-sci · physics.atom-ph· physics.comp-ph

The BiP-PRISM algorithm for fast and scalable core-loss STEM-EELS simulations

Philipp Pelz This is my paper

Reviewed by Pith2026-07-02 10:14 UTCgrok-4.3pith:5X3DCM5Dopen to challenge →

classification cond-mat.mtrl-sci physics.atom-phphysics.comp-ph
keywords STEM-EELSPRISM algorithmcore-loss simulationbeam partitioningdynamical diffractionGPU accelerationatomic-resolution mapping
0
0 comments X

The pith

Beam partitioning in PRISM matrices lets core-loss STEM-EELS simulations skip per-scan propagation while keeping error local to each atom.

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

The paper develops the BiP-PRISM algorithm to make quantitative dynamical simulations of atomic-resolution core-loss electron energy loss spectroscopy practical on ordinary hardware. It splits the probe-forming and post-loss scattering matrices into sparse parent beams that are computed once and then interpolated locally around each ionized atom. A locality result shows that the overall error is controlled only by how well those local interpolations match the full matrices at the atom sites. This change removes the need to propagate the exit wave separately for every probe position and cuts memory use by a factor of five. The authors demonstrate the approach on a five-edge oxide interface and an FePt nanoparticle map while staying within the validity regime they map out.

Core claim

By calculating the S1 and S2 PRISM matrices only on a sparse set of parent beams and reconstructing the values at each ionized atom via natural-neighbor interpolation, the BiP-PRISM method removes per-scan exit-wave propagation; the total error is governed entirely by the on-atom reconstruction error, which the authors bound and validate on multimodal five-edge and Fe-L maps at 5x lower memory cost.

What carries the argument

Beam partitioning of the probe-forming (S1) and detector-propagating (S2) PRISM matrices with natural-neighbor interpolation on sparse parent beams at each ionized atom.

If this is right

  • Full-resolution elemental mapping, 4D data cubes, and momentum-resolved qEELS become feasible on consumer-grade GPUs.
  • Memory footprint drops by a factor of five while accuracy remains high enough for the demonstrated oxide-interface and nanoparticle cases.
  • The validity regime of the approximation is characterized so users can decide when the method applies.
  • Multimodal simulations involving multiple core-loss edges can be performed without prohibitive cost.

Where Pith is reading between the lines

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

  • The same locality principle could be tested on thicker specimens or different acceleration voltages to map the practical range further.
  • Integration with experimental workflows would let users decide scan density on the fly based on the reported error bound.
  • The approach might reduce the barrier to simulating momentum-resolved data for materials where only average spectra were previously affordable.

Load-bearing premise

The total error in the simulated EELS signal is governed entirely by the reconstruction error at the ionized atom sites.

What would settle it

Run both the full per-scan PRISM calculation and the BiP-PRISM version on the same atomic model for one probe position and check whether the difference in the core-loss spectrum exceeds the bound given by the on-atom interpolation error alone.

Figures

Figures reproduced from arXiv: 2607.00756 by Philipp Pelz.

Figure 1
Figure 1. Figure 1: Double-partitioned PRISM-EELS at a glance. Bi-partitioned S-Matrix pipeline for detector-integrated [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Measured STEM-EELS map scaling comparisons for conventional multislice, exact dual-matrix PRISM [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Accuracy, speed, and memory of the double-partitioned PRISM-EELS FePt Fe-L map vs. parent count, at a [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Validity regime of partitioned PRISM-EELS (single column, [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: LaAlO3/SrTiO3 interface: simultaneous atomic-resolution STEM-EELS maps (double-channeling, double￾partitioned PRISM; GPAW edges). (a) A-site Sr/La composite; (b)–(f) Sr-L, La-M, Ti-L, Al-K, and O-K maps. 3.1.5 Momentum-resolved qEELS The double-partitioned matrices yield momentum-resolved (qEELS) output by binning per-detector-beam intensities along one detector axis (Section 2.2.6); summing the resolved a… view at source ↗
Figure 6
Figure 6. Figure 6: Momentum-resolved (qEELS) output of double-partitioned PRISM-EELS on a row of oxygen columns (O-K). [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of growth-direction Ti-L EELS profiles simulated using abTEM and scatterem across the [PITH_FULL_IMAGE:figures/full_fig_p020_7.png] view at source ↗
read the original abstract

Quantitative interpretation of atomic-resolution STEM-EELS requires dynamical simulation of the electron probe before and after core-loss transitions, which is computationally expensive. While the PRISM algorithm accelerates this by reusing scattering matrices, we introduce beam partitioning for both the probe-forming ($\mathcal{S}_1$) and detector-propagating ($\mathcal{S}_2$) PRISM matrices to further reduce computational and memory costs. Each matrix is calculated on a sparse set of parent beams and reconstructed via natural-neighbor interpolation locally at the ionized atom. A locality result demonstrates that the total error is governed entirely by this on-atom reconstruction error. The resulting BiP-PRISM algorithm removes per-scan exit wave propagation and significantly reduces memory requirements, enabling full-resolution elemental mapping, 4D cubes, and momentum-resolved qEELS on consumer-grade GPUs. We characterize the approximation's validity regime and demonstrate the simulation of a multimodal five-edge oxide-interface map and an FePt nanoparticle Fe-L map at 5x memory reduction, showing that the algorithm achieves high accuracy with significantly lower computational demands.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper introduces the BiP-PRISM algorithm as an extension of PRISM for core-loss STEM-EELS simulations. It partitions the probe-forming (S1) and detector-propagating (S2) matrices onto sparse parent beams, reconstructs them at ionized atoms via natural-neighbor interpolation, and invokes a locality result asserting that total simulation error is controlled solely by this local reconstruction error. This permits removal of per-scan exit-wave propagation, yielding ~5x memory reduction while enabling full-resolution elemental maps, 4D cubes, and qEELS on consumer GPUs. The validity regime is characterized and the method is demonstrated on a multimodal five-edge oxide interface and an FePt nanoparticle Fe-L edge map, reporting high accuracy relative to reference calculations.

Significance. If the locality result is rigorously established and the error remains localized under the stated conditions, the work would meaningfully advance quantitative EELS modeling by making previously intractable large-scale simulations practical. The reported 5x memory savings and GPU demonstrations directly address a recognized computational bottleneck in the field, with potential impact on routine analysis of interfaces and nanoparticles. The explicit partitioning and interpolation steps constitute a clear algorithmic contribution.

major comments (2)
  1. [Abstract and locality-result section] The locality result (abstract and method section) is load-bearing for the central claim that per-scan S1/S2 propagation can be eliminated while retaining controlled error. The manuscript must supply the full derivation, including all assumptions (probe delocalization bounds, neglect of long-range dynamical scattering, thickness limits) and a quantitative error bound showing that off-atom contributions are negligible; without this, the justification for skipping full propagation and the 5x memory claim cannot be verified.
  2. [Results section] Demonstration cases (oxide-interface and FePt maps): the reported accuracy must be accompanied by explicit error metrics (e.g., pixel-wise L2 or edge-integrated intensity differences) versus a full PRISM reference, together with the precise beam sparsity and interpolation parameters used, so that the claimed “high accuracy” can be assessed against the locality premise.
minor comments (2)
  1. Notation for the partitioned matrices S1 and S2 should be defined once with explicit dimensions and beam indices before the interpolation step is introduced.
  2. The validity-regime characterization would benefit from a single summary table or plot showing error versus sample thickness, probe convergence angle, and q-transfer range.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and constructive comments on our manuscript. We address each major comment point by point below.

read point-by-point responses

  1. Referee: [Abstract and locality-result section] The locality result (abstract and method section) is load-bearing for the central claim that per-scan S1/S2 propagation can be eliminated while retaining controlled error. The manuscript must supply the full derivation, including all assumptions (probe delocalization bounds, neglect of long-range dynamical scattering, thickness limits) and a quantitative error bound showing that off-atom contributions are negligible; without this, the justification for skipping full propagation and the 5x memory claim cannot be verified.

    Authors: We agree that the locality result is central and that its full derivation, assumptions, and quantitative bound must be provided for the claims to be verifiable. The current manuscript sketches the result and states the governing principle but does not contain the complete derivation. In the revised version we will expand the methods section to include the full mathematical derivation together with explicit statements of all assumptions (probe delocalization bounds, neglect of long-range dynamical scattering, and thickness limits) and a quantitative error bound demonstrating negligibility of off-atom contributions. revision: yes

  2. Referee: [Results section] Demonstration cases (oxide-interface and FePt maps): the reported accuracy must be accompanied by explicit error metrics (e.g., pixel-wise L2 or edge-integrated intensity differences) versus a full PRISM reference, together with the precise beam sparsity and interpolation parameters used, so that the claimed “high accuracy” can be assessed against the locality premise.

    Authors: We agree that explicit quantitative error metrics and parameter values are necessary to allow readers to assess the reported accuracy against the locality premise. The revised manuscript will add pixel-wise L2 differences and edge-integrated intensity differences relative to full PRISM references for both demonstration cases, together with the precise beam-sparsity factors and natural-neighbor interpolation parameters used in each simulation. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The BiP-PRISM algorithm is defined via explicit beam partitioning of S1/S2 matrices, natural-neighbor interpolation at ionized atoms, and a stated locality result that total error is governed by on-atom reconstruction error. This locality claim is presented as an internal demonstration justifying removal of per-scan propagation, with no reduction of the central result to a fitted parameter, self-citation chain, or definitional equivalence. The abstract and described steps remain independent of the target accuracy claims, consistent with a score of 0.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the validity of the stated locality result for error control and on the accuracy of natural-neighbor interpolation for the PRISM matrices; no free parameters or invented entities are mentioned in the abstract.

axioms (1)
  • domain assumption The total error is governed entirely by the on-atom reconstruction error
    Presented as a demonstrated locality result that justifies the sparse-beam approach.

pith-pipeline@v0.9.1-grok · 5712 in / 1227 out tokens · 48286 ms · 2026-07-02T10:14:00.308733+00:00 · methodology

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

15 extracted references · 15 canonical work pages

  1. [1]

    Allen, L. J., A. J. D'Alfonso, and S. D. Findlay. 2015. ``Modelling the Inelastic Scattering of Fast Electrons.'' Ultramicroscopy 151 (2015): 11--22. https://doi.org/10.1016/j.ultramic.2014.10.011

    work page doi:10.1016/j.ultramic.2014.10.011 2015

  2. [2]

    Brown, Hamish G., Jim Ciston, and Colin Ophus. 2019. ``Linear-Scaling Algorithm for Rapid Computation of Inelastic Transitions in the Presence of Multiple Electron Scattering.'' Physical Review Research 1 (3): 033186. https://doi.org/10.1103/PhysRevResearch.1.033186

    work page doi:10.1103/physrevresearch.1.033186 2019

  3. [3]

    Pelz, Colin Ophus, and Jim Ciston

    Brown, Hamish G., Philipp M. Pelz, Colin Ophus, and Jim Ciston. 2020. ``A Python Based Open-Source Multislice Simulation Package for Transmission Electron Microscopy.'' Microscopy and Microanalysis 26 (S2): 2390--91. https://doi.org/10.1017/S1431927620023326

    work page doi:10.1017/s1431927620023326 2020

  4. [4]

    Dwyer, C. 2005. ``Multislice Theory of Fast Electron Scattering Incorporating Atomic Inner-Shell Ionization.'' Ultramicroscopy 104 (2): 141--51. https://doi.org/10.1016/j.ultramic.2005.03.005

    work page doi:10.1016/j.ultramic.2005.03.005 2005

  5. [5]

    Dwyer, C., S. D. Findlay, and L. J. Allen. 2008. ``Multiple Elastic Scattering of Core-Loss Electrons in Atomic Resolution Imaging.'' Physical Review B 77 (18): 184107. https://doi.org/10.1103/PhysRevB.77.184107

    work page doi:10.1103/physrevb.77.184107 2008

  6. [6]

    Findlay, S. D., H. G. Brown, P. M. Pelz, C. Ophus, J. Ciston, and L. J. Allen. 2021. ``Scattering Matrix Determination in Crystalline Materials from 4D Scanning Transmission Electron Microscopy at a Single Defocus Value.'' Microscopy and Microanalysis 27 (4): 744--57. https://doi.org/10.1017/S1431927621000490

    work page doi:10.1017/s1431927621000490 2021

  7. [7]

    Hage, F. S., R. J. Nicholls, J. R. Yates, D. G. McCulloch, T. C. Lovejoy, N. Dellby, O. L. Krivanek, K. Refson, and Q. M. Ramasse. 2018. ``Nanoscale Momentum-Resolved Vibrational Spectroscopy.'' Science Advances 4 (6): eaar7495. https://doi.org/10.1126/sciadv.aar7495

    work page doi:10.1126/sciadv.aar7495 2018

  8. [8]

    Lovejoy, Niklas Dellby, Toshihiro Aoki, R

    Krivanek, Ondrej L., Tracy C. Lovejoy, Niklas Dellby, Toshihiro Aoki, R. W. Carpenter, Peter Rez, Emmanuel Soignard, et al. 2014. ``Vibrational Spectroscopy in the Electron Microscope.'' Nature 514 (7521): 209--12. https://doi.org/10.1038/nature13870

    work page doi:10.1038/nature13870 2014

  9. [9]

    Madsen, Jacob, and Toma Susi. 2021. ``The abTEM Code: Transmission Electron Microscopy from First Principles.'' Open Research Europe 1 (2021): 24. https://doi.org/10.12688/openreseurope.13015.2

    work page doi:10.12688/openreseurope.13015.2 2021

  10. [10]

    Mendis, B. G. 2023. ``A `Phase Scrambling' Algorithm for Parallel Multislice Simulation of Multiple Phonon and Plasmon Scattering Configurations.'' Microscopy and Microanalysis 29 (3): 1111--23. https://doi.org/10.1093/micmic/ozad052

    work page doi:10.1093/micmic/ozad052 2023

  11. [11]

    Ophus, Colin. 2017. ``A Fast Image Simulation Algorithm for Scanning Transmission Electron Microscopy.'' Advanced Structural and Chemical Imaging 3 (1): 13. https://doi.org/10.1186/s40679-017-0046-1

    work page doi:10.1186/s40679-017-0046-1 2017

  12. [12]

    Brown, Scott Stonemeyer, Scott D

    Pelz, Philipp M., Hamish G. Brown, Scott Stonemeyer, Scott D. Findlay, Alex Zettl, Peter Ercius, Yaqian Zhang, Jim Ciston, Mary C. Scott, and Colin Ophus. 2021. ``Phase-Contrast Imaging of Multiply-Scattering Extended Objects at Atomic Resolution by Reconstruction of the Scattering Matrix.'' Physical Review Research 3 (2): 023159. https://doi.org/10.1103/...

    work page doi:10.1103/physrevresearch.3.023159 2021

  13. [13]

    Savitzky, Mary C

    Pelz, Philipp M., Alexander Rakowski, Luis Rangel DaCosta, Benjamin H. Savitzky, Mary C. Scott, and Colin Ophus. 2021. ``A Fast Algorithm for Scanning Transmission Electron Microscopy Imaging and 4D - STEM Diffraction Simulations.'' Microscopy and Microanalysis 27 (4): 835--48. https://doi.org/10.1017/S1431927621012083

    work page doi:10.1017/s1431927621012083 2021

  14. [14]

    Sibson, Robin. 1981. ``A Brief Description of Natural Neighbour Interpolation.'' Interpreting Multivariate Data, 1981, 21--36

    work page 1981

  15. [15]

    Terzoudis-Lumsden, E. W. C., T. C. Petersen, H. G. Brown, P. M. Pelz, C. Ophus, and S. D. Findlay. 2023. ``Resolution of Virtual Depth Sectioning from Four-Dimensional Scanning Transmission Electron Microscopy.'' Microscopy and Microanalysis, ahead of print, 2023. https://doi.org/10.1093/micmic/ozad068. CSLReferences document

    work page doi:10.1093/micmic/ozad068 2023