Pith. sign in

REVIEW 2 major objections 5 minor 17 references

Quantification of Electron Energy-Loss Spectra

T0 review · 2 major / 5 minor · reviewed 2026-07-14 · grok-4.5

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

Pith's one-line read A nested fitting pipeline with smooth-background, fine-structure weights, and plural-scattering convolution makes automated EELS quantification reliable over multi-keV ranges.

desk verdict Solid, implementable EELS pipeline for wide-range spectrum-images; the algorithms are real and usable, the missing piece is quantitative residual benchmarks. read the letter →

arxiv 2607.10693 v1 pith:MM4PHMMY submitted 2026-07-12 physics.data-an

Pavel Potapov , Giulio Guzzinati This is my paper · ORCID
classification physics.data-an PACS 07.78.+s61.05.J82.80.Pv
keywords EELSquantificationsmoothbackgroundfine-structureweightingpluralscatteringnon-negativeleastsquarespower-lawexponentspectrumimagingatomiccross-sections
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 supplies the technical algorithms for a fully automated EELS quantification pipeline that works on modern, wide-energy spectrum images. It replaces the classic fixed-exponent power-law background with a cubic-spline “smooth background” whose slope can vary continuously while still matching the pre-edge exactly and remaining monotonic. Solid-state fine structure near edge onsets is handled by soft, noise-dependent Gaussian weighting envelopes derived from measured deviations, and plural scattering is removed by convolving theoretical atomic cross-sections with either experimental or approximate low-loss spectra. Relative and absolute atomic densities are then obtained by non-negative least-squares fitting of the corrected edges. The result is a practical software flow that can process large spectrum-images with little user intervention and still produce usable concentrations even when simple power-law extrapolation fails or white-line fine structure is strong.

What carries the argument

The “smooth background” cubic spline (defined by pre-edge power-law parameters y0, r0 and free end-point values y_fin, r_fin) together with the soft weighting function ω(ΔE) built from Gaussian envelopes of solid-state deviations.

What would settle it

Apply the published pipeline, with its fixed Table-2 envelopes and no manual exclusion windows, to a set of known-composition standards that exhibit strong fine structure (e.g., transition-metal L2,3 or rare-earth M4,5 edges) and check whether the recovered atomic fractions remain within a few percent of the certified values across multi-keV fitting windows.

Watch

Extended reading notes

Core claim

The authors show that a two-loop fitting procedure—outer non-linear optimization of two smooth-background end-point parameters under monotonicity constraints, inner non-negative least-squares or sequential edge fitting with empirical fine-structure and Poisson weights—recovers accurate elemental contributions over multi-keV ranges where classical power-law extrapolation and unweighted atomic cross-sections break down.

Load-bearing premise

The Gaussian envelopes measured on a few selected K, L and M edges are assumed to bound the solid-state deviations of every other edge well enough that the same soft weighting function can be used without per-edge retuning.

Editorial extensions

If this is right

  • Spectrum-images spanning several thousand eV can be quantified automatically without user-selected background windows.
  • Overlapping edges are fitted simultaneously inside merged patches while still using only the lowest pre-edge for background initialization.
  • Absolute areal densities (atoms nm⁻²) become available from a single vacuum zero-loss measurement, independent of probe current or detector efficiency.
  • When low-loss spectra are unavailable, a simple Drude model with one free thickness parameter still recovers usable edge shapes for typical TEM thicknesses.
  • Poisson weighting automatically down-weights white-line regions, giving a cheap partial correction for fine structure.

Reading between the lines

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

  • The same smooth-background spline and weighting machinery could be ported to soft X-ray absorption spectroscopy where power-law backgrounds and near-edge structure pose analogous problems.
  • If the Gaussian envelopes prove too tight or too loose for certain N edges, a single additional free amplitude per edge family would restore accuracy without abandoning automation.
  • Absolute quantification normalized to the full spectrum rather than a vacuum zero-loss peak may systematically overestimate thick-sample densities; a controlled thickness series on a pure elemental standard could decide the preferred normalization.
  • Because the outer loop is low-dimensional and convex, the pipeline is a natural candidate for GPU-accelerated batch processing of large spectrum-image stacks.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The manuscript is a technical reference for an automated EELS quantification pipeline to be implemented in CEOS Panta Rhei and TEMDM. It defines logarithmically scaled fitting ranges (pre-edge, margin, edge), a smooth-background model (cubic polynomial in log-space with boundary conditions y0, r0, y_fin, r_fin and monotonicity constraints derived in Appendix A), a nested outer non-linear / inner linear or NNLS fitting loop, empirical fine-structure and Poisson weighting functions, plural-scattering correction by convolution with measured or Drude-modelled low-loss spectra, and both relative and absolute quantification formulas. The central claim is that these constructive algorithms enable robust multi-keV quantification with minimal user interaction when simple power-law extrapolation or unweighted atomic cross-sections fail.

Significance. If the described procedures perform as illustrated, the work supplies a practical, software-ready reference that addresses real limitations of classical EELS quantification (wide energy ranges, solid-state fine structure, plural scattering, automation of spectrum-images). Strengths include the explicit monotonicity derivation in Appendix A, the use of publicly tabulated GOS, the noise-adaptive soft weighting, and the clear separation of relative versus absolute quantification. The algorithms are constructive rather than circular; free parameters (range widths, Gaussian envelopes, Drude Ep/Wp) are frozen after one-time empirical choice. The main limitation is the absence of systematic quantitative benchmarks, so the practical gain over existing model-based methods remains illustrated rather than measured. As a methods/reference paper for production software this is still useful to the EELS community.

major comments (2)
  1. [Section 6.1, Table 2, Eq. (4)] The Gaussian envelopes of Table 2 (magnitude, middle, σ for K/L/M) and the resulting soft weight ω(ΔE) of Eq. (4) are derived from a small set of manually tuned edges after interactive near-onset exclusion. The manuscript then applies these envelopes universally as limiting bounds. This is load-bearing for the claim that fine-structure effects can be handled automatically without per-edge tuning (Section 6.1). The paper should either enlarge the validation set, report residual errors when the envelopes are applied to edges outside the training set, or more explicitly restrict the claim to edges of similar character, and discuss failure modes when solid-state deviations exceed the shaded regions of Fig. 8.
  2. [Sections 3–8 overall; Figs. 4, 10, 11, 13, 16] Figures 4, 10, 11, 13 and 16 provide qualitative illustrations of improved residuals and thickness-dependent shape recovery, yet the manuscript contains no quantitative residual metrics, composition error bars, or systematic tests on samples of known stoichiometry. For a technical reference that positions the nested procedure as yielding accurate multi-keV quantification, even a modest benchmark table (e.g., recovered atomic fractions versus thickness or versus simple power-law) would substantially strengthen the central claim. Without it the accuracy assertion remains conditional on the illustrative cases shown.
minor comments (5)
  1. [Sections 3 and 4] Two independent equations are both labelled (1) (background log-linear fit and the double-differential cross-section). Renumber for clarity.
  2. [Throughout / References] Typographical issues: 'demostrated' (p. 6), 'fictive' (Fig. 9 caption), 'oscilation' (Eq. near (1)), 'pseudopothetials' (ref. [7]), 'cross-sectoions' (ref. [10]), 'Wilhelms-Universität' spelling, and occasional missing spaces around units.
  3. [Section 2 and 3.2] The fixed logarithmic widths (0.6 / 0.1 / 0.015 log(eV)) and the pre-edge reproducibility tolerance (±0.2) are stated without sensitivity analysis. A short remark on how results change when these defaults are varied would help users who need to override them.
  4. [Section 8.2, Fig. 16] Section 8.2 leaves open which absolute-normalization route (vacuum ZLP vs total spectrum counts) is preferred; Fig. 16 shows both deviate from linearity at high th. A one-sentence practical recommendation for software users would be useful.
  5. [Fig. 8 caption] Fig. 8 notes deconvolution artefacts (arrows) without sub-pixel precision; a brief statement that the envelopes remain conservative upper bounds despite these artefacts would avoid reader concern.

Circularity Check

1 steps flagged · score 1.0 of 10

No load-bearing circularity; the quantification flow is a constructive algorithmic definition with frozen empirical envelopes and fixed Drude parameters, plus one non-essential self-citation to a preliminary abstract.

  1. self citation load bearing [Section 6.1, paragraph on fine-structure envelopes]
    "In the preliminary release of this approach [16], we used an exponentially decaying function, but further testing and analysis on more data reveals that a Gaussian function is more appropriate for this purpose as illustrated in Fig. 8."

    Citation [16] is the authors’ own M&M2025 abstract. It is not load-bearing: the present paper replaces the earlier exponential with new Gaussian parameters derived from the same data set and supplies the full algorithmic specification; the reference merely records the change of functional form and does not underwrite uniqueness or the quantification results.

full rationale

The manuscript is a technical reference describing a nested fitting procedure (outer L-BFGS-B optimization of smooth-background y_fin/r_fin under Appendix A monotonicity bounds; inner sequential or NNLS edge fitting with ω(E) weights). All steps are constructive definitions of software algorithms rather than first-principles derivations or predictions that reduce to their own inputs. The Gaussian fine-structure envelopes of Table 2 are extracted once from a handful of manually-tuned edges and then frozen as upper-bound inputs to Eq. (4); they do not re-enter as free parameters that force the elemental fractions a_i or areal densities N_i/A. Likewise the Drude parameters E_p=20 eV, W_p=15 eV are fixed after a sensitivity check (Appendix C) and do not close a loop. The sole self-citation [16] merely notes a prior exponential form that was later replaced by Gaussians; it is not invoked to justify uniqueness, forbid alternatives, or underwrite the central claim. No self-definitional identity, fitted-input-as-prediction, uniqueness import, or ansatz smuggling appears in the derivation chain. Score 1 reflects only the minor, non-load-bearing self-reference; the procedure remains independently implementable from the stated equations and tables.

Assumptions & free parameters 5 free parameters · 4 assumptions · 3 invented entities

The quantification pipeline rests on standard EELS physics plus a modest set of empirical modeling choices and free parameters that are fixed once and then used for all spectra. No new physical entities are postulated; the ‘smooth background’ and ‘margin regions’ are algorithmic constructs.

free parameters (5)
  • edge/pre-edge/margin widths = 0.6 / 0.1 / 0.015 log(eV)
    Fixed at 0.6 / 0.1 / 0.015 log(eV); chosen by hand to balance noise and overlap (Section 2, Table 1).
  • fine-structure Gaussian envelopes = K: 0.8/10/25; L: 1.5/25/20; M: 4.0/10/12
    Magnitude, middle and σ for K/L/M edges (Table 2) extracted from selected elements after manual exclusion; used to build ω(ΔE).
  • Drude plasmon parameters = Ep=20 eV, Wp=15 eV
    Average Ep=20 eV, Wp=15 eV fixed for approximate plural-scattering correction when low-loss data are absent (Section 7.2).
  • cubic-spline bounds k, Δ = 3/2 < k < 3, Δ < 1/3
    3/2 < k < 3 and Δ < 1/3 derived in Appendix A to enforce monotonicity; still free within those bounds during optimization.
  • pre-edge reproducibility tolerance = ±0.2
    r = r0 ± 0.2 used for automatic ‘crab-like’ pre-edge selection (Section 3.2).
assumptions (4)
  • domain assumption Within a few hundred eV the EELS background is well approximated by a power law a E^{-r}; over wider ranges r varies smoothly enough to be captured by a cubic polynomial.
    Stated in Section 3 and used as the foundation of the smooth-background model; supported by Fig. 2 but not derived from first principles.
  • domain assumption Theoretical single-atom GOS (Segger et al.) integrated over the experimental collection angle give adequate edge shapes once solid-state fine structure is down-weighted.
    Section 4; standard Born-approximation practice, with the fine-structure correction added empirically.
  • domain assumption Plural scattering of core-loss edges is adequately described by convolution with a low-loss spectrum (or its Drude approximation).
    Section 7; classical multiple-scattering treatment (Egerton).
  • ad hoc to paper Monotonic decrease of background intensity with energy loss is a physical requirement that must be enforced on the cubic model.
    Appendix A; authors note no experimental counter-examples but acknowledge it is not a fundamental law.
invented entities (3)
  • smooth background model (cubic polynomial with boundary conditions y0,r0,y_fin,r_fin)
    purpose: Allow the power-law exponent to vary continuously while matching the pre-edge exactly and remaining monotonic.
    Defined in Section 3.1; algorithmic construct rather than a new physical object; independent evidence is the qualitative improvement shown in Fig. 4.
  • margin regions
    purpose: Buffer zones between pre-edge and edge-onset fitting windows to avoid transitional artifacts.
    Introduced in Section 2; purely algorithmic.
  • empirical limiting envelopes for fine-structure deviations
    purpose: Soft weighting that down-weights near-onset solid-state effects without hard exclusion windows.
    Section 6.1 and Table 2; derived from a small set of measured edges.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Quantification of Electron Energy-Loss Spectra." pith.science (2026). https://pith.science/paper/MM4PHMMY

@misc{pith2026260710693,
  author       = {Pith},
  title        = {Pith review of: Quantification of Electron Energy-Loss Spectra},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MM4PHMMY}},
  note         = {Machine review of arXiv:2607.10693}
}
read the original abstract

This manuscript summarizes the recent developments in EELS quantification flow as will be implemented in the CEOS Panta Rhei and TEMDM software. This should serve as a technical reference for the algorithms used in the software.

Figures

Figures reproduced from arXiv: 2607.10693 by the authors.

Figure 1
Figure 1. Illustration of fitting ranges in logarithmic scale. Edges with higher onset [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. (a) Power law exponent r for Cu samples of varying thicknesses, calcu￾lated over 100 eV-wide energy intervals not consisting of any characteristic edges. Sample thickness ranges from 0.1 to 1.6 in terms of mean free path. (b) Power Law exponent for different materials measured in wider energy range. The drop in the 500–600 eV region of Cu samples is due to eventual residual O-K edge contributions. All measurements w… view at source ↗
Figure 3
Figure 3. Power law in standard (a) and logarithmic (b) coordinates. In log space, [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (20 more)
Figure 4
Figure 4. Figure 4: Comparison of smooth background and simple Power Law methods. The [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Computational flow of the ”smooth background” algorithm. The outer [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Typical loss function landscape for the ”smooth background” optimization [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Experimental (a) almost noise-free and (b) noisy Ti [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Normalized deviations from atomic cross-sections for (a) [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: (a) Normalized empirical deviations as in Fig. 8 and calculated weighting [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: Fit of La oxide spectrum before (a) and after (b) fine structure correction. [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: (a) Spectrum of a Si–O–C sample where large deviations around 300 eV [PITH_FULL_IMAGE:figures/full_fig_p016_11.png]
Figure 12
Figure 12. Figure 12: (a) Poisson weighting function exhibits dips at the places of white line [PITH_FULL_IMAGE:figures/full_fig_p017_12.png]
Figure 13
Figure 13. Figure 13: Experimental (a) low-loss and (b) background subtracted Cu [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]
Figure 14
Figure 14. Figure 14: Theoretical Cu L23 cross-section convolved with a zero-loss peak of differ￾ent energy resolutions. For easier comparison with the other edges, the energy-loss is counted from the Cu L23 onset (931eV). The unconvolved cross-section is dotted [PITH_FULL_IMAGE:figures/f…
Figure 15
Figure 15. Figure 15: (a) A simulated low-loss spectrum for Si for different values of relative [PITH_FULL_IMAGE:figures/full_fig_p021_15.png]
Figure 16
Figure 16. Figure 16: (a): Semiconductor sample with drastically changing thickness from left [PITH_FULL_IMAGE:figures/full_fig_p024_16.png]
Figure 17
Figure 17. Figure 17: Schematic plot of linearly descending function and cubic polynomials [PITH_FULL_IMAGE:figures/full_fig_p027_17.png]
Figure 18
Figure 18. Figure 18: Point x ∗ where the second derivative is zero as a function of k. The regions when intersection is between 0 and 1 are plotted in red 28 [PITH_FULL_IMAGE:figures/full_fig_p028_18.png]
Figure 19
Figure 19. Figure 19: Second derivative y ′′ in case of ∆ < 0 for different k. The region of discontinuity is marked in red. Taking into account these trade-offs, a reasonable compromise is to choose 3 2 < k < 3, which coincides with the condition for convexity at ∆ > 0 and the condition f…
Figure 20
Figure 20. Figure 20: (a) Simulation of Cu low-loss spectra with the Drude plasmon model and [PITH_FULL_IMAGE:figures/full_fig_p031_20.png]
Figure 21
Figure 21. Figure 21: The plasmon peak (blue) calculated in the Drude model with a plasmon [PITH_FULL_IMAGE:figures/full_fig_p032_21.png]
Figure 22
Figure 22. Figure 22: Theoretical O K cross-sections convolved with the simulated low-loss for varying sample thickness up to th=1.8. The tableau shows the results for different pseudo-plasmon energies Ep (different columns) and plasmon width Wp (different rows). For easier comparison with…
Figure 23
Figure 23. Figure 23: Theoretical Si K cross-sections convolved with the simulated low-loss for varying sample thickness up to th=1.8. The tableau shows the results for different pseudo-plasmon energies Ep (different columns) and plasmon width Wp (different rows). For easier comparison wit…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references

  1. [1]

    R. F. Egerton.Electron Energy-Loss Spectroscopy in the Electron Microscope. Springer, 3rd edition, 2011

    2011

  2. [2]

    Verbeeck and S

    J. Verbeeck and S. Van Aert. Model-based quantification of EELS spectra. Ultramicroscopy, 101:207–224, 2004

    2004

  3. [3]

    Leapman, P

    R.D. Leapman, P. Rez, and D.F. Mayers. K, l and m generalized osciillator strengths and ionization cross sections for fast electron collisions.J. of Chemical Physics, 72:1232–1243, 1980

    1980

  4. [4]

    Cueva, R

    P. Cueva, R. Hovdenand, H. L. Xin, and D. A. Muller. Data processing for atomic resolution electron energy loss spectroscopy.Microsc. Microanal, 18:667–675, 2012

    2012

  5. [5]

    Van den Broeck, D

    W. Van den Broeck, D. Jannis, and J. Verbeeck. Convexity constraints on linear background models for electron energy-loss spectra.Ultramicroscopy, 254:113830, 2025. 34

    2025

  6. [6]

    L. Segger. Berechnung generalisierter oscillatorenst¨ arken f¨ ur die quantifizierung von EEL-spektren. Master’s thesis, Wilhelms-Universit¨ at M¨ unster, 2019

    2019

  7. [7]

    D. R. Hamann. Generalized norm-conserving pseudopothetials.Phys. Reb. B, 40:2980–2987, 1989

    1989

  8. [8]

    J. P. Perdew and A. Zunger. Self-interaction correction to density-functional approximations for many-electron systems.Phys. Rev. B, 23:5048–5049, 1989

    1989

Show all 17 references
  1. [9]

    Segger, G

    L. Segger, G. Guzzinati, and H. Kohl. Generalised oscillator strengths for the simulation of EELS spectra , with a broader coverage of high energy and minor edges.https://zenodo.org/records/7645765. Accessed: 2025-12-06

    2025

  2. [10]

    Zhang2025, I

    Z. Zhang2025, I. Lobato, H. Brown, D. Lamoen, D. Jannis, J. Verbeeck, S. van Aert, and P.D. Nellist. Relativistic EELS scattering cross-sectoions for micro- analysis based on dirac solutions.Ultramicroscopy, 269:1114083, 2025

    2025

  3. [11]

    H. Kohl. A simple procedure for evaluating effective scattering cross-sections in stem.Ultramicroscopy, 16(2):265–268, 1985

    1985

  4. [12]

    BFGS.https://docs.scipy.org/doc/scipy/reference/generated/ scipy.optimize.BFGS.html

    Scipy. BFGS.https://docs.scipy.org/doc/scipy/reference/generated/ scipy.optimize.BFGS.html. Accessed: 2025-12-06

    2025

  5. [13]

    EELS quantification.https://eels.info/book/export/html/44

    Gatan Inc. EELS quantification.https://eels.info/book/export/html/44. Accessed: 2025-12-06

    2025

  6. [14]

    Verbeeck, S

    J. Verbeeck, S. Van Aert, and G. Bertoni. Model-based quantification of EELS spectra: Including fine structure.Ultramicroscopy, 106:976–980, 2006

    2006

  7. [15]

    Jannis, W

    D. Jannis, W. Van den Broek, Z. Zhang, S. Van Aert, and J. Verbeeck. Improved precision and accuracy of electron energy-loss spectroscopy quantification via fine structure fitting with constraint optimization.Ultramicroscopy, 269:114084, 2025

    2025

  8. [16]

    Potapov and G

    P. Potapov and G. Guzzinati. A new approach for quantification of EELSpectra. InM&M2025 Proceedings, Microsc. Microanal. 31 (7), pages 195–197, 2025

    2025

  9. [17]

    Verbeeck and G

    J. Verbeeck and G. Bertoni. Deconvolution of core electron energy loss spectra. Ultramicroscopy, 109:1343–1352, 2009. 35

    2009

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.