{"id":"2187995c-52a9-415b-8ece-0f3393315549","arxiv_id":"2505.04679","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"By fitting simulated muon stopping profiles to measured muonic X-ray depth profiles, the authors recover gold layer thicknesses of 3.5 to 11 micrometers on gilded copper alloys, matching scanning electron microscopy for most samples.","lead":"Muonic X-ray spectroscopy uses the X-rays emitted when negative muons are captured by atoms to identify elements deep inside objects. This paper shows that pairing the measurements with Monte Carlo simulations can estimate the thickness of a surface layer, such as a gold gilding, without cutting the object.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Validation is weakened by per-sample tuning of momentum spread, distance, and density; EM2's fitted 5 µm vs SEM 11 µm is omitted from Table 2, so the ±1 µm claim lacks systematic uncertainty.","rationale":"The reader's stated weakest assumption is the proportionality between stopped muons and X-ray intensity, which is a real simplification but is not the main load-bearing weakness here. The more directly falsifiable issue is that the analysis tunes several external inputs (momentum spread, source-to-sample distance, gold density) per sample and then quotes a ±1 µm uncertainty that ignores those tuning freedoms. This is visible in EM2, where the fitted 5±1 µm is inconsistent with the nominal SEM thickness of 11±1 µm unless a 20% density reduction is also introduced, and the SEM entry for EM2 is dropped from Table 2. A conditional acceptance remains appropriate because the method works for four of five samples and the authors are transparent about the EM2 difficulty, but the condition must be a proper sensitivity analysis of the tuned parameters and an independent check of the EM2 thickness/density. The reader's rationale already mentions the post-hoc adjustments and the Table 2 problem, so there is partial agreement, but the primary stated weakest assumption differs from the one identified here.","tokens_in":13061,"tokens_out":4779,"duration_ms":48968,"concrete_test":"Recompute the best-fit thickness for EP_A, EP_B, EP_C, and SM3 with the momentum spread fixed at the nominal 4% and the source-to-sample distance fixed at 10 cm, scanning only the gold thickness, and compare to the reported values. If any shift exceeds 1 µm, the quoted ±1 µm uncertainty is not robust to the tuned inputs. For EM2, repeat the fit with the gold density fixed at 19.32 g/cm³ and report whether any thickness gives an acceptable reduced chi-square.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim—that the simulated profiles determine gold thickness to ±1 µm and agree with SEM—depends on the fitted thickness being identifiable without per-sample tuning of other physical inputs. In this study, four inputs are adjusted: momentum spread is raised from the nominal 4% to 5% for all electroplated samples; source-to-sample distance is changed from 10 cm to 9.7 cm for EP_C; and for EM2 the gold density is reduced by 20% (to about 15.5 g/cm³) while the thickness is fitted to 5±1 µm against a SEM value of 11±1 µm. Table 2 lists EM2 as '5±1' without the SEM entry, obscuring that one of the five validation samples does not agree with the reference value. The reported ±1 µm uncertainty is derived solely from the χ² variation in thickness at fixed values of the tuned parameters, so it excludes the systematic shift these adjustments can induce; momentum spread and thickness, in particular, both influence profile width and position, making them partially degenerate. The paper itself acknowledges the spread was 'not extensively investigated' and that positioning is a 'source of error.' Thus the quantitative accuracy claim is less secure than the abstract implies, and the validation is inconclusive for EM2.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a data-analysis protocol for muonic atom X-ray emission spectroscopy (μ-XES) in which Monte Carlo simulations (GEANT4/ARBY and SRIM/TRIM) are matched to measured muon stopping depth profiles to determine the thickness of a superficial gold layer. The method is demonstrated on two sets of gilded copper-alloy standards: three electroplated foils (EPA, EPB, EPC) and two amalgam-gilded foils (SM3, EM2). The simulated depth profiles are compared with measured normalised peak areas via reduced chi-square, with the gold thickness as the fitted parameter. The authors report best-fit thicknesses of 3.5±1, 4.5±1, and 7.5±1 µm for EPA, EPB, and EPC (SEM: 3.3±0.2, 4.6±0.6, 7.3±0.8 µm), 11±1 µm for SM3 (SEM: 11±1 µm), and 5±1 µm for EM2, the last with no SEM value listed in Table 2. The central claim is that the two simulation tools, despite their different approaches, yield consistent thickness estimates that agree with the SEM characterisation to within about ±1 µm.","tokens_in":13313,"tokens_out":2721,"duration_ms":27566,"significance":"If the ±1 µm accuracy claim holds, the protocol would be a valuable non-destructive depth-profiling tool for heritage science, where gilding and alteration layers of tens of microns are otherwise difficult to assess. The paper's strengths include the use of two independent simulation codes, a straightforward reduced-chi-square comparison criterion, open data links for the ISIS beamtime, and a clear validation design against SEM on multiple standard samples. The approach is falsifiable in principle: each sample's fitted thickness can be checked against an independent reference. However, the significance is substantially reduced by the incomplete validation record for EM2 and by the post-hoc adjustment of several physical inputs, which means the reported ±1 µm uncertainty is not yet established as a general systematic error budget.","major_comments":[{"comment":"","section":"§2.2 and Table 2"},{"comment":"","section":"§2.1 and Discussion"},{"comment":"","section":"§1.1 and §1.4"}],"minor_comments":[{"comment":"","section":"Figure 4 and Figure 5"},{"comment":"","section":"§2.2"},{"comment":"","section":"Figure 4d"},{"comment":"","section":"§2.1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a useful proof-of-concept, but the validation is currently incomplete: the EM2 discrepancy is hidden in Table 2, and the ±1 µm claim does not yet include the systematic effects of the adjusted momentum spread, distance, and density. If the authors can either justify EM2 as a known inhomogeneous case with supporting evidence, or restrict the accuracy claim to the electroplated and homogeneous SM3 samples, the paper would be substantially stronger. The paper fits the journal's scope well; the issues are fixable within the manuscript's scope, hence major_revision rather than reject."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nThis is a solid proof-of-concept that muonic X-ray depth profiles, interpreted with MC simulations, can estimate gold layer thickness on metal artifacts. The genuinely new piece is the direct comparison of GEANT4/ARBY and SRIM/TRIM for this purpose, using real ISIS beam data and SEM thicknesses as ground truth. For four of the five standards, both codes fit the gold layer within the quoted ±1 µm, which is encouraging.\n\nThe paper does several things well: the protocol is clearly documented, the experimental data are openly available, and the authors are candid about the limits of the simulation codes (e.g., no full muon cascade in GEANT4, TRIM using a scaled hydrogen ion). That level of transparency is worth emulating.\n\nThe soft spots are mostly about the validation. EM2 is the clearest issue: the best fit gives 5 ± 1 µm of gold, while the SEM measurement is 11 ± 1 µm. The explanation—porosity in the amalgam layer, modeled by reducing the gold density by 20%—is plausible, but Table 2 lists only the simulated value for EM2, not the SEM value, so the mismatch is hidden. A reader who only looks at the table would think all five samples agree. Also, the ±1 µm error is derived only from the χ² variation in thickness at fixed tuning parameters; it does not incorporate the systematic shifts from changing the momentum spread (4% to 5%) or the sample distance (10 to 9.7 cm for EPC). Those parameters are partially degenerate with thickness, so the real uncertainty is larger than reported.\n\nNone of this sinks the central idea. It is a proof-of-concept, and the authors say so. But the quantitative claim of agreement with SEM, and the specific ±1 µm number, need to be softened or supported with a proper covariance analysis.\n\nThis paper is for heritage scientists using muon beams and for anyone building MC analysis pipelines for μ-XES. I would send it to peer review—referees can ask for a corrected Table 2 and a more rigorous uncertainty discussion. It is not a fundamental advance, but it is a useful, honest engineering contribution.\n\nI'd bring it to a reading group if the group works on muon techniques or depth profiling; otherwise it's a skimmer.","headline":"A useful proof-of-concept for muonic depth profiling of gold layers, but the EM2 discrepancy and tuned parameters mean the ±1 µm claim needs a stronger uncertainty treatment.","tokens_in":13886,"tokens_out":3263,"would_cite":false,"duration_ms":30185,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that coupling muonic X-ray depth profiles with Monte Carlo stopping simulations determines the thickness of a surface gold layer on copper alloys to about ±1 µm, in agreement with SEM measurements.","keywords":["muon spectroscopy","depth profiling","elemental analysis","archaeometry","muonic X-ray","Monte Carlo simulation","gilding thickness"],"falsifier":"Measure a multilayer standard of known composition (for example gold on nickel on copper) with a precisely machined gold thickness, run the full momentum scan, and fit it. If the best-fit simulated thickness disagrees with the known thickness by more than the claimed ±1 µm, or if switching to a substrate with very different muon capture properties shifts the fitted thickness while SEM confirms a constant gold thickness, the proportionality assumption is wrong. A more targeted check is to measure the actual beam momentum spread at the sample position; if it is 4% while simulations require 5%, the method is compensating for an unmodelled effect rather than measuring thickness cleanly.","tokens_in":12855,"feed_emoji":"🔬","tokens_out":8819,"duration_ms":81804,"temperature":0.7,"pith_summary":"Muonic atom X-ray emission spectroscopy (µ-XES) can see tens of microns to centimetres into a material, but turning its depth scans into a layer thickness has been mostly qualitative. The paper proposes a quantitative route: simulate where negative muons stop in a layered sample with two Monte Carlo codes, GEANT4/ARBY and SRIM/TRIM, and compare the simulated depth profile with the measured muonic X-ray profile by reduced chi-square. Applied to standard gilded copper-alloy foils, the method recovers the gold layer thickness to about ±1 µm, matching SEM measurements. If it works, restorers and curators could measure gildings, patinas, and corrosion crusts without cutting or sampling the object.","feed_headline":"Muon scans size hidden gold layers within 1 micron","feed_subtitle":"Matching muonic X-ray depth profiles to Monte Carlo simulations reads gilding thickness without cutting heritage metals.","key_machinery":"The load-bearing object is the momentum-scan depth profile: a series of muonic X-ray spectra taken at increasing beam momenta, in which each element's fitted peak area is plotted against momentum. The comparison engine is the reduced chi-square between measured and simulated profiles, computed over all momentum runs including zero-intensity runs. The simulation chain uses GEANT4/ARBY, a detailed-geometry Monte Carlo transport code, and SRIM/TRIM, a layered stopping-and-range code run with a hydrogen ion of one-ninth the muon mass, to output the number of stopped muons per layer; the paper assumes this number is proportional to the measured X-ray intensity. The muon itself is the physical enabler: at 207 times the electron mass, its cascade X-rays reach hundreds of keV to MeV, escaping from deep inside metals and giving the technique its non-destructive penetration.","core_discovery":"The paper's central claim is that a layer's thickness can be recovered by matching a simulated muon-stopping depth profile to a measured one. Because a negative muon stopped in a layer produces that layer's characteristic muonic X-rays, the authors assume the full-energy peak intensity from a layer is proportional to the number of muons stopped there. They scan the muon beam momentum so that muons stop at increasing depths, build momentum-dependent profiles for gold, nickel, copper, aluminium, and nitrogen, then vary the simulated gold thickness until the reduced chi-square between simulated and measured profiles is minimized. For electroplated samples, the best-fit gold thicknesses are 3.5±1, 4.5±1, and 7.5±1 µm against SEM values of 3.3±0.2, 4.6±0.6, and 7.3±0.8 µm; for the amalgam-gilded brass sample the best fit is 11±1 µm versus 11±1 µm by SEM, and for the bronze sample the best fit is 5±1 µm with a 20% reduced gold density to mimic surface unevenness and air pockets. Both simulation codes agree, with GEANT4/ARBY handling thin layers better than SRIM/TRIM.","pith_inferences":["If the stopped-muon-to-X-ray proportionality is verified layer by layer, the method could be extended from thickness fitting to quantitative per-layer composition, because the same machinery would then connect absolute peak intensities to elemental concentrations.","The reported ±1 µm uncertainty is set by the momentum spread and by the chi-square step size; a direct measurement of the beam's momentum distribution at the sample would likely tighten the error and remove the need to tune spread and source distance by hand.","The effective-density trick used for the uneven amalgam-gilded bronze suggests a testable extension: porosity of a gilding could be estimated from the density reduction needed to reach a best fit, giving conservators a quantitative measure of layer quality.","The same reduced-chi-square matching could be transferred to other depth-sensitive probes, but the muon's centimetre reach makes it uniquely suited to layered metal artefacts, where competing surface techniques lose signal."],"forward_implications":["Artworks with metal gildings, patinas, or corrosion crusts of ten to a few tens of microns can be measured non-destructively, recovering the thickness of each identifiable layer with about a micrometre of uncertainty.","The same profile-matching protocol applies to any layered material in which each layer has a distinct muonic X-ray line, including buried layers beneath thick outer shells that XRF and PIXE cannot see.","A quick open-source simulation route (SRIM/TRIM) gives usable thickness estimates, while a detailed-geometry code (GEANT4/ARBY) improves accuracy for thin or irregular layers.","Because the method reads the full layer sequence in one momentum scan, it yields not just the coating thickness but also information about the substrate and intermediate layers, such as nickel under gold.","For irregular or porous layers, the fit can absorb thickness variations as an effective density reduction, allowing the method to flag uneven gilding even when the nominal thickness is unknown."],"supporting_citations":[{"why":"Supplies the GEANT4 transport engine behind ARBY, simulating muon energy loss and stopping in the layered samples.","marker":"[18]"},{"why":"Provides the SRIM/TRIM stopping-power and ion-range calculations used as the second simulation route.","marker":"[22]"},{"why":"Earlier demonstration that negative-muon depth profiling can reveal elemental composition beneath a surface, the foundation this work builds on.","marker":"[10]"},{"why":"Documents that muon cascade X-ray generation in GEANT4 is not yet fully reliable, motivating the stopped-muon proxy rather than simulated X-ray spectra.","marker":"[25]"},{"why":"Describes the pulsed muon beams available at the source where the measurements were made.","marker":"[30]"},{"why":"Supplies the beamline parameters (momentum range, spread, pulse structure) that the simulations must reproduce.","marker":"[31]"},{"why":"Used to identify the muonic X-ray peaks (gold, nickel, copper, aluminium, nitrogen, oxygen) that make up the depth profiles.","marker":"[32]"},{"why":"Shows ARBY's prior validation in simulating germanium detector response, supporting its use for realistic experimental modelling.","marker":"[20]"}],"fun_headline_variants":["Muons reveal gold layer thickness in heritage metals","Muonic X-ray profiling measures gilding without cutting","Monte Carlo matches muon stops to size hidden layers","Negative muons gauge gold layers on ancient bronze","Non-destructive muon scans map gilding depth precisely"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole method assumes that the intensity of a layer's muonic X-ray peak is directly proportional to the number of muons stopped in that layer; if capture probabilities, cascade yields, or muon transfer between layers vary, the fitted thickness will be biased.","fun_headline_variants_meta":{"raw":{"variants":["Muons reveal gold layer thickness in heritage metals","Muonic X-ray profiling measures gilding without cutting","Monte Carlo matches muon stops to size hidden layers","Negative muons gauge gold layers on ancient bronze","Non-destructive muon scans map gilding depth precisely"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00059,"raw_usage":{"total_tokens":2842,"prompt_tokens":1091,"completion_tokens":1751,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":707,"completion_tokens_details":{"reasoning_tokens":1673}},"tokens_in":707,"tokens_out":1751,"duration_ms":12032,"temperature":1.0,"reasoning_tokens":1673,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:28:10.021786+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure a multilayer standard of known composition (for example gold on nickel on copper) with a precisely machined gold thickness, run the full momentum scan, and fit it. If the best-fit simulated thickness disagrees with the known thickness by more than the claimed ±1 µm, or if switching to a substrate with very different muon capture properties shifts the fitted thickness while SEM confirms a constant gold thickness, the proportionality assumption is wrong. A more targeted check is to measure the actual beam momentum spread at the sample position; if it is 4% while simulations require 5%, the method is compensating for an unmodelled effect rather than measuring thickness cleanly.","supporting_citations":[{"cited_title":"A Novel Non-Destructive Technique for Cultural Heritage: Depth Profiling and Elemental Analysis Underneath the Surface with Negative Muons,","cited_arxiv_id":null,"evidence_quote":"Earlier demonstration that negative-muon depth profiling can reveal elemental composition beneath a surface, the foundation this work builds on."},{"cited_title":"The Implementation of MuDirac in Geant4: A Preliminary Approach to the Improvement of the Simulation of the Muonic Atom Cascade Process,","cited_arxiv_id":null,"evidence_quote":"Documents that muon cascade X-ray generation in GEANT4 is not yet fully reliable, motivating the stopped-muon proxy rather than simulated X-ray spectra."},{"cited_title":"The RIKEN-RAL pulsed Muon Facility,","cited_arxiv_id":null,"evidence_quote":"Supplies the beamline parameters (momentum range, spread, pulse structure) that the simulations must reproduce."},{"cited_title":"Electronic catalogue of muonic X-rays,","cited_arxiv_id":null,"evidence_quote":"Used to identify the muonic X-ray peaks (gold, nickel, copper, aluminium, nitrogen, oxygen) that make up the depth profiles."}],"review_version":1}