{"id":"bd817e69-bf1d-4584-a45f-993745166aff","arxiv_id":"2506.09355","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"In multislice electron ptychography, increasing the atom-column integration radius changes Z-contrast from a monotonic Z^0.66 law to a shell-structure-dependent trend that can separate single-Z-apart elements such as Cu and Zn.","lead":"This simulation study shows that the atomic-number (Z) contrast of multislice electron ptychography depends strongly on the area used to integrate the reconstructed atom-column potential: small integration areas give a monotonic Z dependence, while larger areas give a non-monotonic, shell-structure-dependent contrast. The authors use this effect to show that elements one atomic number apart, such as copper and zinc, can be distinguished in simulated 20 nm thick samples.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Finite-source-size model for ptychography (random probe shift per scan position) is not equivalent to the Gaussian convolution used for HAADF/iDPC, so the quantitative Cu/Zn dose thresholds in Fig. 3 are not yet established.","rationale":"I read the paper as a simulation-based optimization study. The basic observation—that the apparent Z-dependence of ptychographic projected potentials changes from a Z^0.66 power law to shell-structure-modulated curves as the integration radius grows—is internally supported and is consistent with the transmission function because the transmission function is built from the same scattering factors. That part is not the weak point. The weak point is the headline comparative claim about Cu/Zn distinguishability under a finite source. A real finite source is a partial-coherence mixture; the random per-scan-position shift model is not that mixture, and it is not the same model as the convolution used for HAADF and iDPC. The direction of the bias is unclear a priori: if position refinement is off, ptychography is artificially degraded; if position refinement is on, some of the source blur may be removed as scan-position corrections, artificially helping ptychography. Hence the quantitative dose thresholds and the relative standing versus HAADF/iDPC are not yet established by the present simulations. This concern does not undermine the conceptual radius-dependence claim, which is well supported internally, so I would keep the reader's CONDITIONAL verdict; the condition should explicitly require the corrected source-size simulation before the Fig. 3 numbers are used.","tokens_in":15235,"tokens_out":12660,"duration_ms":164405,"concrete_test":"Repeat the beta-CuZn simulations at FWHM = 0.50 Å using a faithful partial-coherence forward model: for each scan position, average the simulated 4D intensities over 20–50 Gaussian-distributed probe shifts (the same S(delta) at every position), then reconstruct with identical PtychoShelves settings and recompute the CNR/dose map of Fig. 3. If the >95% regime and the HAADF/iDPC gaps change materially, the source-size comparison is model-dependent.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing quantitative claim—ptychography gives >95% Cu/Zn distinguishability above about 1.7e4 e-/Å^2 while HAADF/iDPC stay below 83%—depends on how finite source size is simulated. In Section II A, ptychographic data are generated by 'shifting the electron probe by a random amount at each real space pixel position in each thermal configuration,' whereas HAADF and iDPC source effects are included by convolving the final images with a 2D Gaussian. These are not equivalent partial-coherence models. A true finite source in focused-probe ptychography is an incoherent average over a distribution of probe shifts at every scan position; a single random shift per scan position injects per-position probe-position errors that are not the same physical effect. Depending on whether PtychoShelves position refinement is active, these random shifts will either act as an unmodeled noise source (over-penalizing ptychography) or be partially fitted out (treating source blur as scan distortion and removing it). Either way, the comparison between ptychography, iDPC, and HAADF in Fig. 3 and the derived dose threshold are not a clean test of source-size robustness. The 2.4% transmission-function agreement is also an internal consistency check because the transmission function derives from the same atomic potentials used to generate the data, so it does not independently validate the shell-structure prediction for a real alloy.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses multislice electron ptychography simulations, combined with HAADF and iDPC simulations, to show that the apparent atomic-number dependence of ptychographic projected potentials depends strongly on the radius over which each atom column is integrated. For a small integration radius (r0 = 0.20 Å) the per-atom integrated potential follows a monotonic power law (≈Z^0.66), while for larger radii (up to r0 = 0.90 Å) the Z-dependence becomes non-monotonic and tracks orbital shell structure. The authors then use the enhanced contrast at r0 = 0.90 Å to claim that Cu and Zn (Z = 29 and 30) columns in a 20 nm thick β-CuZn supercell can be distinguished with >95% accuracy for doses above about 1.7×10^4 e/Ų at 0.50 Å source size, whereas iDPC and HAADF remain below 83%. The paper also compares different quantification methods and maps the dose–source-size parameter space for four integration radii.","tokens_in":15483,"tokens_out":6836,"duration_ms":61355,"significance":"If the quantitative claims hold, the paper offers a practical, simulation-grounded insight: the choice of integration radius is a tunable parameter that can enhance or suppress Z-contrast in ptychography, and it can highlight elements that are only one atomic number apart even in moderately thick samples. The simulation parameters are specified in detail, the central trends are internally consistent across the four integration radii, and the comparison with iDPC and HAADF frames the result in a practically relevant way. The paper does not provide experimental validation, but the simulation study is sufficiently careful to be informative if the finite-source-size modeling is made symmetric across the compared techniques.","major_comments":[{"comment":"The finite-source-size model is not applied symmetrically across the compared methods. For ptychography, the source is emulated by a random probe shift at each scan position and thermal configuration, whereas for HAADF and iDPC it is included by convolving the final images with a two-dimensional Gaussian. These are not equivalent partial-coherence models: a single random shift per scan position injects per-position probe-position errors that the reconstruction algorithm may either suppress as noise or partly fit out as scan distortion if probe-position refinement is enabled, and neither behavior reproduces the incoherent Gaussian blur used for the other two methods. Because the dose threshold in Figure 3 (>95% ptychography distinguishability above about 1.7×10^4 e/Ų versus <83% for iDPC and HAADF) is the paper's central quantitative comparison, the authors should either generate the ptychographic data with an explicit incoherent average over a source distribution (e.g., multiple probe shifts per scan position) and state whether position refinement was used, or justify that the random-shift model is equivalent to the convolution model for focused-probe ptychography.","section":"Section II A and Figure 3"},{"comment":"The reported 2.4% average deviation between the reconstructed projected potential and the transmission function is an internal consistency check, not an independent validation, because the transmission function is computed from the same atomic potentials that feed the multislice forward simulation. The statement that this agreement 'indicates that the trends are a reflection of the electrostatic potential' should be rephrased to clarify that it demonstrates the reconstruction preserves the input potential; if the shell-structure prediction is intended as a general physical statement, the authors should add a comparison with an independent potential parameterization or with experimental data.","section":"Section III A and Figure 1c"},{"comment":"The paper reports a 95% true-positive identification threshold without specifying the number of atom columns per species used to construct the empirical cumulative distribution function, nor the uncertainty on the threshold. With a small number of columns (the simulated β-CuZn supercell is 7.36×7.36 nm), the threshold may carry substantial sampling error. The authors should state the number of Cu and Zn columns and, if possible, provide confidence intervals for the threshold doses.","section":"Section III C, Eq. (1) and Figure 3"}],"minor_comments":[{"comment":"The claim that 'ptychography recovered projected potentials are nearly independent of thickness' is difficult to reconcile with the reported 12.9% average decrease in per-atom projected potential with thickness (Supplementary Figure S1); a rewording to 'the Z-dependence trends remain qualitatively similar' would be more accurate.","section":"Section III A"},{"comment":"There is a typo in the text: 'the integration methods provde higher CNR' should read 'provide'.","section":"Section III D"},{"comment":"The text contains many '/gid...' artifacts and garbled references; the authors should ensure the production version is clean and all references are properly rendered.","section":"Throughout the provided manuscript"},{"comment":"The caption states 'The black dashed line shows the threshold for 95% distinguishability,' but the main text uses '95% true positive atom column identification' and '95% accuracy'; these terms should be defined consistently so that the reader knows exactly what quantity the threshold represents.","section":"Figure 3 caption"}],"recommendation":"major_revision","confidential_remarks":"The paper is simulation-only, which is appropriate for a methods-oriented manuscript, but the comparison with HAADF and iDPC would benefit from a more carefully controlled forward model for the finite source size. The provided manuscript version appears to contain serious text corruption ('/gid...' tokens); the editor may wish to verify that the submitted source manuscript is intact. The novelty is incremental but the practical guidance on integration-radius selection is likely to be useful to the ptychography community."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a serious simulation paper with a genuinely useful observation—the Z-dependence of multislice ptychography projections changes from a monotonic Z^0.66 power law to a shell-structure-driven non-monotonic curve as you increase the integration radius around an atom column. That is new and explains some conflicting trends in the literature. The Cu/Zn separability at r0=0.90 Å is a nice demonstration, and the CNR analysis across dose, source size, and quantification methods is thorough. The paper earns a real referee.\n\nThe core trend is well supported. The reconstructed potentials track the transmission function to 2.4%, though that is an internal consistency check because the transmission function comes from the same atomic potentials used to generate the data. The Z^0.66 exponent is consistent with Chen et al., and the shell-structure features line up with Kirkland's projected-potential radii, so the physics is plausible. I worry a bit that all elements share one Debye-Waller factor in the periodic-table supercells—fine for isolating Z but not directly transferable to real materials. They do use element-specific B-factors for CuZn, so that is partially addressed.\n\nThe real soft spot is the finite-source-size comparison in Figure 3. For ptychography they model a finite source by adding a random probe shift at each scan position; for HAADF/iDPC they convolve the images with a Gaussian. Those are not equivalent partial-coherence models. A true incoherent source is an average over a distribution of probe positions at every scan position. A single random shift per pixel either acts as scan-position error that PtychoShelves' position refinement can partially fit out, or as an unmodeled noise source if refinement is off. Either way, the ptychography dose thresholds and the >95% vs <83% comparison are not a clean test of source-size robustness. That does not break the central integration-radius story, but it does mean the quantitative thresholds in Fig. 3 and the conclusion should be softened unless the authors redo the source-size simulations with a proper incoherent average or mixed-state model.\n\nAlso, the single-Z demonstration rests on one simulated material pair (Cu/Zn in β-CuZn). That is a reasonable proof-of-concept, but it would be stronger with a second pair or an experimental dataset. No code or data are shipped, though the methods are detailed enough for an expert group to reimplement.\n\nVerdict: worth peer review. The core insight is solid and the paper is carefully executed. I would ask the authors to fix the source-size modeling and add at least one more material pair or experimental validation before publication. I would bring it to reading group, but I would not cite it in my own work until the quantitative claims are on firmer ground.","headline":"Simulation study shows integration radius controls Z-contrast in ptychography—solid core result, but the source-size comparison vs HAADF/iDPC is not as clean as the dose thresholds imply.","tokens_in":16060,"tokens_out":2785,"would_cite":false,"duration_ms":25366,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Atomic-number contrast in ptychography is set by the integration radius, not the technique alone.","keywords":["ptychography","atomic number contrast","Z-contrast","multislice electron ptychography","projected potential","atom column integration","CuZn","scanning transmission electron microscopy"],"falsifier":"An experimental 4D-STEM dataset of a β-CuZn lamella (approximately 20 nm thick) collected at 300 keV with a convergence semi-angle near 25 mrad and dose above 1.7e4 e/Å2: if a multislice ptychographic reconstruction integrated over a 0.90 Å radius does not separate the Cu and Zn column distributions at better than 95% accuracy, the central distinguishability claim fails.","tokens_in":14980,"feed_emoji":"🔬","tokens_out":6584,"duration_ms":55600,"temperature":0.7,"pith_summary":"The paper argues that the atomic-number (Z) contrast in multislice electron ptychography is not a fixed property of the technique but is controlled by the area over which the reconstructed projected potential is integrated at each atom column. Using simulated 4D STEM datasets of supercells spanning Z = 8 to 87, the authors show that a small integration radius (0.20 Å) yields a monotonic $Z^{0}$.66 dependence, while a larger radius (0.90 Å) produces a non-monotonic, shell-structure-dependent contrast that follows orbital filling. They then demonstrate that this shell-structure sensitivity lets ptychography distinguish Cu (Z = 29) from Zn (Z = 30) in a 20 nm thick $\\beta$-CuZn simulation with >95% accuracy at doses above about 1.7e4 e/Å2, where iDPC and HAADF stay below 83%. The practical claim is that choosing the integration radius lets the microscopist tune Z-contrast for a specific material system.","feed_headline":"Integration radius decides atomic-number contrast in ptychography","feed_subtitle":"With a 0.90 Å integration radius, ptychography tells Cu from Zn at >95% accuracy where HAADF and iDPC fail.","key_machinery":"The load-bearing object is the integration radius r0 used to measure the per-atom integrated projected potential from a ptychographic reconstruction: the reconstructed phase is converted to projected potential via the interaction parameter, summed over a circular region around each atom column, and divided by the number of atoms in the column. Varying r0 changes which part of the atomic potential contributes — the nuclear core at small r0 versus outer valence electrons at large r0 — and this radial selectivity is what converts a monotonic $Z^{0}$.66 law into shell-structure-dependent contrast. The distinguishability metric is the contrast-to-noise ratio between the Cu and Zn column distributions, with a 95% threshold set by the empirical cumulative distribution function.","core_discovery":"The central result is that the Z-dependence of ptychographic projected potentials is governed by which part of the atomic potential is sampled: integrating over a small disk around the column core sees mostly the nuclear potential and gives a near-linear power law ($Z^{0}$.66), whereas integrating over a wider disk (0.90 Å) includes valence-shell electrons, producing shell-structure oscillations with local minima at closed shells (Z = 10, 18) and dips across the 3d series. The reconstructed per-atom integrated projected potentials match the transmission function to within 2.4% on average, showing the trend reflects the true electrostatic potential. Because these shell-structure differences can be larger than the Z-step difference itself, elements a single atomic number apart — Cu and Zn — become separable with contrast-to-noise ratios high enough for >95% correct column identification at 0.90 Å integration radius, at doses above roughly 1.7e4 e/Å2 and with a 0.50 Å finite source size, while iDPC and HAADF remain below 83% at all doses considered.","pith_inferences":["Beyond the paper, the shell-structure mechanism should also operate in single-slice (non-multislice) phase reconstructions, so the optimal integration radius may need recalibration for thin samples where dynamical scattering is weaker.","Because the reconstructed potential tracks the transmission function to 2.4%, the Z-dependence curves could serve as a calibration library for converting measured integrated potentials into elemental occupancy maps for arbitrary element pairs, not just Cu/Zn.","Since larger integration radii capture bonding charge redistribution, the optimum radius could vary with local chemical environment, meaning a single global radius may not be optimal across heterogeneous interfaces.","The reported dose threshold of 1.7e4 e/Å2 is tied to the specific optics (25 mrad convergence, 50 mrad collection) and reconstruction settings; changing these would shift the threshold, though the qualitative radius dependence should persist."],"forward_implications":["For a given material, the integration radius can be chosen to maximize contrast between specific neighboring elements, turning Z-contrast into a tunable parameter rather than a fixed response.","Ptychography offers a quantitative edge over iDPC and HAADF for separating elements with ΔZ = 1 in relatively thick (>20 nm) samples, at least in simulation.","Peak-intensity measurements are systematically worse than integrated measurements for similar-Z distinguishability; at high dose all integration methods (fixed radius, Voronoi, Gaussian, Lorentzian) give CNR ≈ 6.4–7.6.","Lowering the integration radius from 0.90 Å to 0.64 Å drops the average CNR by 20% and requires higher dose or smaller source size to maintain 95% distinguishability."],"supporting_citations":[{"why":"Reported the Z^0.67 relation for multislice ptychography that this paper confirms at small radii, and underpins the PtychoShelves reconstruction approach used here.","marker":"Chen et al. (2021)"},{"why":"Provides the multislice simulation method and the projected-potential radial trends used to interpret shell-structure contrast.","marker":"Kirkland (2010)"},{"why":"PtychoShelves is the software package used to perform the ptychographic reconstructions.","marker":"Wakonig et al. (2020)"},{"why":"Established that integrated signals are more robust than peak intensities for atomic-column quantification, a principle extended to ptychography here.","marker":"MacArthur et al. (2013)"},{"why":"Gives the method for measuring effective source size, which the simulations use to model spatial incoherence.","marker":"Dwyer et al. (2008)"},{"why":"Provides the center-of-mass imaging theory and shell-structure Z-dependence that the paper compares against iDPC.","marker":"Cao et al. (2018)"},{"why":"Defines the Z^1.6–1.8 HAADF scattering dependence that serves as the comparison baseline.","marker":"Treacy (2011)"},{"why":"Demonstrates deep sub-angstrom ptychography and shows how Poisson noise impacts reconstruction quality at low dose.","marker":"Jiang et al. (2018)"}],"fun_headline_variants":["Integration radius flips ptychography's Z-contrast behavior","Atom-column disk size decides element contrast in ptychography","Shell-structure Z-contrast via integration radius in ptychography","Ptychography: wider integration radius separates Cu from Zn","Tuning integration radius optimizes Z-sensitivity in ptychography"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The results assume the simulated 4D STEM datasets — with frozen-lattice thermal scattering, Gaussian source-size blur, and Poisson shot noise — faithfully represent a real electron microscope, and that a single Debye-Waller factor for all elements does not change the qualitative Z-trends.","fun_headline_variants_meta":{"raw":{"variants":["Integration radius flips ptychography's Z-contrast behavior","Atom-column disk size decides element contrast in ptychography","Shell-structure Z-contrast via integration radius in ptychography","Ptychography: wider integration radius separates Cu from Zn","Tuning integration radius optimizes Z-sensitivity in ptychography"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000249,"raw_usage":{"total_tokens":1575,"prompt_tokens":998,"completion_tokens":577,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":614,"completion_tokens_details":{"reasoning_tokens":489}},"tokens_in":614,"tokens_out":577,"duration_ms":5341,"temperature":1.0,"reasoning_tokens":489,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:50:48.171003+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An experimental 4D-STEM dataset of a β-CuZn lamella (approximately 20 nm thick) collected at 300 keV with a convergence semi-angle near 25 mrad and dose above 1.7e4 e/Å2: if a multislice ptychographic reconstruction integrated over a 0.90 Å radius does not separate the Cu and Zn column distributions at better than 95% accuracy, the central distinguishability claim fails.","supporting_citations":[{"cited_title":"& Muller, D.A","cited_arxiv_id":null,"evidence_quote":"Reported the Z^0.67 relation for multislice ptychography that this paper confirms at small radii, and underpins the PtychoShelves reconstruction approach used here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the multislice simulation method and the projected-potential radial trends used to interpret shell-structure contrast."},{"cited_title":"& Guizar-Sicairos, M","cited_arxiv_id":null,"evidence_quote":"PtychoShelves is the software package used to perform the ptychographic reconstructions."},{"cited_title":"& Nellist, P","cited_arxiv_id":null,"evidence_quote":"Established that integrated signals are more robust than peak intensities for atomic-column quantification, a principle extended to ptychography here."},{"cited_title":"& Etheridge, J","cited_arxiv_id":null,"evidence_quote":"Gives the method for measuring effective source size, which the simulations use to model spatial incoherence."},{"cited_title":"& Muller, D.A","cited_arxiv_id":null,"evidence_quote":"Provides the center-of-mass imaging theory and shell-structure Z-dependence that the paper compares against iDPC."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the Z^1.6–1.8 HAADF scattering dependence that serves as the comparison baseline."},{"cited_title":"& Muller, D.A","cited_arxiv_id":null,"evidence_quote":"Demonstrates deep sub-angstrom ptychography and shows how Poisson noise impacts reconstruction quality at low dose."}],"review_version":1}