{"id":"c84c3374-ff84-4d39-a3d9-06726d5f111e","arxiv_id":"2501.07605","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A tensor-based lock-in thermography method reconstructs all six independent thermal conductivity tensor components from multi-surface phase and amplitude maps, validated on quartz and sapphire.","lead":"This paper presents 3D SR-LIT, an optical method that measures the full three-dimensional thermal conductivity tensor, including all six independent components of an anisotropic material, from infrared phase and amplitude maps on one to three sample surfaces. The technique is claimed to measure in minutes rather than hours and to reduce uncertainty by over 50% for simple anisotropic materials relative to vector-based approaches.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The amplitude calibration coefficient β is calibrated once on fused silica and assumed to transfer to all samples; if Ti coating absorptivity varies, the amplitude-derived k_zz and all single-surface tensor fits are systematically biased.","rationale":"The reader's weakest_assumption identifies the power coefficient β and spot-size extraction as uncorrelated with the unknown thermal conductivity. I agree that this is the most load-bearing concern. The method's central quantitative claims for simple anisotropic tensors rely on absolute amplitude calibration, and the only calibration is a single β value from fused silica. The paper does provide experimental agreement for sapphire and x-cut quartz, which is encouraging, but it does not provide a controlled test of β transferability across different surface conditions or coating batches. The numerical validation in Section 3.5 is self-consistency checking: the same forward model generates and fits the data, so it cannot reveal systematic errors from calibration transfer or model mismatch. The uncertainty formalism could in principle propagate beta uncertainty, but the main text does not demonstrate that it does for each sample. My recommendation is unchanged from the reader's conditional acceptance: the method is promising and the reported results are internally consistent, but the authors should either provide direct evidence that β is sample-independent or, alternatively, restrict the single-surface amplitude-based k_zz claim to samples with the same surface preparation and coating batch as the calibration standard. The proposed test would settle whether the concern actually lands.","tokens_in":14546,"tokens_out":5078,"duration_ms":55019,"concrete_test":"Perform a controlled calibration-transfer experiment: prepare two fused-silica samples with the same nominal Ti coating but deliberately different surface roughness (e.g., as-polished vs. 200-nm roughened), or deposit Ti in two separate batches. Calibrate β on one sample, then use that same β to fit k_zz of the other sample. If the fitted k_zz deviates by more than the reported 2σ uncertainty, the assumption that β is sample-independent fails. As an additional check, run synthetic inversions where the true absorbed power is perturbed by ±10% while keeping the tensor fixed; if the fitted k_zz shifts by more than the claimed uncertainty, the reported uncertainties are underestimates. The same approach can test the spot-size assumption by perturbing w_x and w_y by ±10% and measuring the induced bias in k_zz.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 2.2.2 introduces a power coefficient β calibrated on fused silica to convert monitored reflected pump power into absorbed power, and states that 'the same coefficient β is applied to subsequent samples.' The amplitude map is then used to determine k_zz (Section 3.2, Fig. 3e–h), and in the single-surface protocol for simple anisotropic tensors the fitted k_zz is directly proportional to the assumed absorbed power. If the Ti transducer absorptivity or emissivity varies between samples—due to surface roughness, coating thickness differences, or deposition batch variations—the calibrated β no longer holds and k_zz is biased. The paper reports agreement with literature for sapphire and x-cut quartz, but those are only two favorable cases and do not establish sample-independence of β. Notably, for the 200-nm-roughness AT-cut surface, the amplitude channel is deliberately omitted, so the single-surface amplitude protocol is never validated on rough surfaces. The uncertainty analysis (Appendix 3) lists absorbed power as an input parameter, but the main text does not show that the covariance between β and k_zz is propagated for each new sample; treating β as a fixed constant makes the reported 2σ uncertainties for k_zz optimistic. A related assumption is the spot size extracted from the normalized amplitude map (Section 2.2.3), which shares the same amplitude data that carries k_zz information; a bias in spot size can therefore leak into k_zz. These assumptions are load-bearing because the central quantitative claims—'2σ uncertainties below 12%' and the x-cut quartz tensor—depend on an unvalidated transfer of a single calibration constant.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript introduces Three-Dimensional Spatially Resolved Lock-In Micro-Thermography (3D SR-LIT), an optical method aimed at determining the full 3D anisotropic thermal conductivity tensor, including all six independent components, from phase and amplitude maps acquired on one to three orthogonal surfaces of a sample. The method is demonstrated experimentally on fused silica, sapphire, x-cut quartz, and AT-cut quartz, and is further exercised on 1,000 numerically generated anisotropic tensors. The central claims are that the tensor-based analysis reduces uncertainty by 50–60% compared to a vector-based analysis, keeps 2σ uncertainties below about 12% for a six-component tensor such as AT-cut quartz, enables data acquisition in 40 seconds to 3 minutes per surface, and tolerates surface roughness of 200 nm in the demonstrated phase-based measurements.","tokens_in":14777,"tokens_out":4881,"duration_ms":49457,"significance":"If the central claims hold, the paper would be a valuable methodological advance: it replaces the lossy vector-based projection used by many existing thermoreflectance techniques with a direct tensor fit, and it exploits full-field detection to shorten acquisition times. The experimental results on fused silica, sapphire, and x-cut quartz agree with literature values, and the numerical study covers a broad range of anisotropy with mostly small errors. The uncertainty propagation follows a standard regression formalism, and the sensitivity-based discussion of which surfaces constrain which tensor components is instructive. However, the external validation is strongest for simple anisotropic tensors; the six-component experimental case is compared to a target computed from the same authors' x-cut quartz measurement rather than to an independent literature tensor, and the calibration-transfer and spot-size assumptions are not stress-tested. The method's practical value depends on these assumptions, so the load-bearing points need additional evidence.","major_comments":[{"comment":"The power coefficient β is calibrated once on fused silica and then applied to all subsequent samples, yet the amplitude map carries the primary sensitivity to k_zz (Section 3.2, Fig. 3e–h). Because β absorbs the absorptivity and emissivity of the Ti coating, any batch-to-batch or surface-roughness variation in the coating changes the effective absorbed power and directly biases k_zz; the agreement for sapphire and x-cut quartz does not establish that β transfers to arbitrary samples, and the 200 nm rough AT-cut surface is analyzed with phase only (Section 3.3), so the amplitude channel is never validated on a rough surface. The uncertainty formalism in Appendix 3 lists A_p as an input parameter but does not show how uncertainty in β itself enters the covariance matrix; treating β as a fixed constant makes the reported 2σ uncertainties conditional on the calibration holding.","section":"Section 2.2.2"},{"comment":"The pump spot size is determined from the same normalized amplitude map that carries the k_zz information, which creates a potential systematic correlation that is not discussed. If the spot-size extraction is biased by the unknown thermal conductivity or by the model used to interpret the normalized amplitude, the bias can leak directly into k_zz and indirectly into the phase-derived in-plane components. The manuscript refers to a validation in Supplementary S4, but the main text does not report how uncertainties in w_x and w_y are propagated, nor does it demonstrate that the extracted spot size is statistically independent of the tensor being fitted; a covariance or sensitivity analysis that explicitly includes the spot-size extraction would be needed to support the reported uncertainty levels.","section":"Section 2.2.3"},{"comment":"The six-component experimental result for AT-cut quartz is validated against a 'target tensor' computed from the same authors' x-cut quartz measurement (Section 3.3), not against an independent literature tensor. The agreement therefore demonstrates self-consistency of the fitting pipeline but does not independently confirm that the method recovers an arbitrary six-component tensor. The numerical study in Section 3.5 is also a self-consistency check, because the simulated data are generated with the same forward model and noise assumptions used in the regression; an independent test, either a material with a known six-component tensor or a synthetic-data study using a different solver and noise model, would substantially strengthen the central claim.","section":"Section 3.4"},{"comment":"The claimed 50–60% uncertainty reduction relative to vector-based methods rests on a vector-based benchmark whose implementation is only briefly described ('phase data from four directions' in Section 3.2 and 'twelve directional measurements' in Section 3.5), with no independent validation or reproducible code. Because the vector-based comparison is self-implemented by the same group, the factor of improvement should be treated as provisional until the benchmark is either fully specified or validated against an established vector-based technique.","section":"Sections 3.2 and 3.5"}],"minor_comments":[{"comment":"The sentence 'For determine simple anisotropic tensors with k_xz=k_yz=0...' contains a grammatical error and should be revised to 'For determining...'.","section":"Section 2.2.2"},{"comment":"The text contains the placeholder '(add citations)' after the sentence about traditional methods requiring manual reconfiguration of optics; this placeholder must be replaced with actual references before publication.","section":"Section 3.2"},{"comment":"The statement 'the procedure applies equivalently for non-zero k_yz is identical' is grammatically incomplete and logically unclear; the authors should state exactly what is being claimed for the k_yz nonzero case and how it is handled by the three-surface protocol.","section":"Section 3.3"},{"comment":"The claim that data acquisition is 'over 35 times faster than current methods' needs a defined baseline; it should be stated whether the comparison includes the full serial multi-directional vector-based measurement time or only the camera acquisition time per surface.","section":"Abstract and Section 3.1"},{"comment":"The statement that data are available 'upon reasonable request' is limiting for a methodology paper; depositing the raw phase/amplitude maps, the fitting code, and the uncertainty-propagation code would allow independent reproduction of the reported uncertainty reductions.","section":"Data Availability"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within scope for a thermal-characterization or applied-physics venue. The main risk is not the forward model but the transferability of the amplitude calibration and the lack of an independent validation for the six-component tensor case; these are fixable with additional experiments or a reanalysis that explicitly propagates calibration and spot-size uncertainty. I saw no indication of misrepresentation, but the self-implemented vector-based comparison and the 'target tensor' derived from the authors' own x-cut measurement should be framed more cautiously in the revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read the 3D SR-LIT paper. In short, it's a genuine advance: they extend their 2D SR-LIT tensor analysis to the full 3D thermal conductivity tensor, measuring all six independent components from phase maps on one to three orthogonal surfaces, with amplitude maps only where needed. Experimental results on fused silica, sapphire, x-cut quartz, and AT-cut quartz track the literature values, and the adaptive mapping idea—choosing surfaces based on tensor complexity—is clever and useful.\n\nWhat's new is the tensor-based framework and the multi-surface protocol, not the heat diffusion model, which is standard. The 1000-tensor numerical study tests the fitting procedure with the same forward model, so it's a self-consistency check, not a validation of the physics. The external experimental agreement carries that weight, and it's good to see.\n\nNow the soft spots. The power coefficient beta is calibrated once on fused silica and then applied to all samples. For the single-surface amplitude-based k_zz (sapphire, x-cut quartz), that transfer is load-bearing. If the Ti coating absorptivity varies between depositions, k_zz shifts. They don't propagate beta's uncertainty into the reported error bars, so the 2σ values for k_zz in those cases are optimistic. It doesn't affect the two- or three-surface phase-only measurements for AT-cut quartz, since beta is never used there. A quick fix is to report beta's standard deviation or validate it on a second isotropic reference.\n\nThe second soft spot is the comparison baseline. The 'reduces uncertainty by 50-60%' claim is against their own vector-based implementation, not a published state-of-the-art method. The speedup by 35x is plausible, but the baseline should be stated more carefully. The numerical 'validation' on 1000 tensors is fine as a robustness test, but don't oversell it.\n\nData and code are only available on request, which limits independent reimplementation, but the method description is detailed enough that a determined group could reproduce it.\n\nOverall, this is a solid paper with a real contribution. The issues are addressable. I'd send it to peer review, and ask for the beta uncertainty propagation and a more rigorous comparison baseline before accepting.","headline":"A real step forward for full-tensor thermal metrology, but the beta-calibration transfer and a self-implemented baseline need honest treatment before I'd fully trust the quoted uncertainties.","tokens_in":15367,"tokens_out":4528,"would_cite":true,"duration_ms":41195,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["44.10.+i","07.20.-n"],"model":"deepseek-v4-flash","headline":"Thermography reads all six thermal-conductivity tensor components","keywords":["thermal conductivity tensor","anisotropic thermal conductivity","lock-in thermography","spatially resolved thermal measurement","tensor-based analysis","phase and amplitude mapping","AT-cut quartz","high-throughput thermal characterization"],"falsifier":"Remeasure one of the demonstrated samples, such as the two-surface AT-cut quartz, while deliberately changing the pump laser power by about 20% between runs; if the fitted tensor components shift by more than the reported $2\\sigma$ uncertainties, the power coefficient $\\beta$ is not truly sample-independent and the amplitude channel is biasing the tensor fit.","tokens_in":14285,"feed_emoji":"🔥","tokens_out":13275,"duration_ms":179717,"temperature":0.7,"pith_summary":"Anisotropic heat flow in a crystal is governed by a symmetric $3\\times3$ tensor with up to six independent numbers, but almost every established technique can recover at most four of them, and usually only by scanning direction by direction. The paper introduces 3D SR-LIT (three-dimensional spatially resolved lock-in micro-thermography), an optical method that fits full-field phase and amplitude maps of the surface temperature directly to the tensor heat-diffusion equation, skipping the lossy projection onto vectors. For simple tensors such as $x$-cut quartz with $k_{xz}=k_{yz}=0$, a phase-plus-amplitude map from one surface suffices, and the reported uncertainties drop by more than half compared with vector-based analysis. For a fully asymmetric six-component tensor such as a rotated AT-cut quartz, phase maps from three mutually perpendicular surfaces keep every $2\\sigma$ uncertainty below 12%. Because an infrared focal-plane array replaces point-by-point scanning, the paper claims acquisition times of 40 seconds to 3 minutes, more than 35 times faster than current methods.","feed_headline":"Thermography reads all six thermal-conductivity tensor components","feed_subtitle":"One to three surface maps keep 2σ uncertainties under 12 percent and run in 40 s to 3 min.","key_machinery":"The load-bearing object is the full 3D heat-diffusion equation written with a symmetric tensor $\\mathbf{k}$ and its Fourier-domain Green's function solution for a multilayer film-substrate system, computed by thermal quadrupoles, which generates synthetic phase and amplitude maps at every pixel. A nonlinear least-squares regression then adjusts the six tensor components until the synthetic maps match the measured ones. The adaptive mapping protocol chooses which surfaces to measure: the phase map is most sensitive to in-plane components, the amplitude map adds the cross-plane $k_{zz}$, and measuring a perpendicular surface converts a cross-plane off-diagonal component into a strongly sensed in-plane one, so three orthogonal phase maps separate all six components. Two calibration inputs carry the experimental burden: a power coefficient $\\beta$, calibrated once on fused silica, converts monitored laser power into absorbed power for the amplitude channel, and the pump spot size is extracted from the normalized amplitude map rather than from a separate knife-edge measurement.","core_discovery":"The paper's central claim is that a full three-dimensional thermal conductivity tensor, including the cross-plane off-diagonal components that earlier methods could not see, can be recovered from two-dimensional phase and amplitude maps without any vector projection. The enabling observation is a rotation identity: a cross-plane component such as $k_{xz}$ in the sample's native frame becomes an in-plane component when the measurement is made on a perpendicular face, and the phase map is highly sensitive to in-plane components. Consequently, the experimental burden scales with tensor complexity: one surface for simple anisotropy, two surfaces for one non-zero cross-plane off-diagonal term, and three surfaces for the general six-component case. The paper validates the scheme on fused silica, sapphire, $x$-cut quartz, and two- and three-surface AT-cut quartz measurements, and reports that in a numerical sweep of 1000 random anisotropic tensors with diagonal entries from 1 to 1000 $\\mathrm{W m^{-1}K^{-1}}$, 98% of fitted components come out within 6% of their true values.","pith_inferences":["Beyond the paper: if the reported $2\\sigma<12\\%$ precision holds at the extremes of the tested range, the method could act as a pre-screening stage for materials discovery, flagging which candidates need slower, more precise measurements.","Beyond the paper: a direct test the paper leaves implicit is to measure the same sample at two different modulation frequencies; the forward model has no frequency-dependent free parameters, so a frequency drift in the fitted tensor would localize the error in the $\\beta$ or spot-size calibration.","Beyond the paper: the perpendicular-surface rotation trick suggests a general design rule for tensor-based thermometry: every cross-plane component can be made visible by choosing a measurement plane whose normal lies in the corresponding principal plane, so sensitivity limits are set by surface-normal alignment errors rather than by the physics of the probe.","Beyond the paper: the paper mentions machine learning only as a regression accelerator; a natural extension is to use the 1000-tensor sweep as training data for a surrogate that maps phase and amplitude maps to tensors directly, potentially eliminating the multi-hour regressions reported for extreme off-diagonal cases."],"forward_implications":["Routine full-tensor characterization becomes possible for crystals with unknown or low-symmetry anisotropy, filling a gap in thermal-property databases.","For simple anisotropic materials, one surface and one measurement suffice, cutting both experiment time and fitted uncertainties for diagonal and off-diagonal components.","The tolerance to 200 nm surface roughness, and phase-only tolerance to 5 µm, means samples that cannot be mirror-polished can still be measured.","Because the tensor-analysis core is decoupled from the specific detector, it can be grafted onto other thermometry schemes, so existing instruments could gain six-component capability without new hardware.","Numerical testing on 1000 random tensors with diagonal components spanning 1 to 1000 $\\mathrm{W m^{-1}K^{-1}}$ supports the claim that accuracy does not depend on a narrow conductivity window."],"supporting_citations":[{"why":"Establishes the 2D spatially resolved lock-in thermography tensor-analysis approach that this work extends to three dimensions.","marker":"[38]"},{"why":"Supplies the elliptical-beam thermoreflectance framework and multilayer Green's function solution on which the forward thermal model is built.","marker":"[30]"},{"why":"Provides the thermal-quadrupole transform used to solve the multilayer heat-diffusion equation for simulated phase and amplitude maps.","marker":"[58]"},{"why":"Gives the nonlinear-regression uncertainty formalism that produces the reported $2\\sigma$ error bars.","marker":"[42]"},{"why":"Extends the regression uncertainty analysis to account for input-parameter uncertainties in thermoreflectance-style measurements.","marker":"[43]"},{"why":"Supplies the normalized-amplitude spot-size determination and the earlier spatial-domain thermoreflectance results used for comparison.","marker":"[35]"},{"why":"Represents the vector-based beam-offset method limited to one off-diagonal component that the new tensor framework outperforms.","marker":"[32]"},{"why":"Provides the literature values for fused silica, quartz, and sapphire heat capacity and conductivity used as inputs and validation targets.","marker":"[44]"},{"why":"Documents a previous three-dimensional anisotropic tensor measurement whose off-diagonal anisotropy motivates the new method's broader range.","marker":"[39]"}],"fun_headline_variants":["Thermography reads full 3D thermal conductivity tensor","All six thermal tensor components from surface scans","3D thermal tensor in minutes: off-diagonals no longer hidden","35x faster thermal tensor mapping with full component capture","Surface thermography decodes complete anisotropic tensor"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The power coefficient $\\beta$ calibrated once on isotropic fused silica and the pump spot size extracted from the normalized amplitude map are assumed to stay valid, and uncorrelated with the unknown thermal conductivity, for every sample coated and measured afterward.","fun_headline_variants_meta":{"raw":{"variants":["Thermography reads full 3D thermal conductivity tensor","All six thermal tensor components from surface scans","3D thermal tensor in minutes: off-diagonals no longer hidden","35x faster thermal tensor mapping with full component capture","Surface thermography decodes complete anisotropic tensor"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000599,"raw_usage":{"total_tokens":2833,"prompt_tokens":1010,"completion_tokens":1823,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":626,"completion_tokens_details":{"reasoning_tokens":1746}},"tokens_in":626,"tokens_out":1823,"duration_ms":14586,"temperature":1.0,"reasoning_tokens":1746,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:57:22.193704+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Remeasure one of the demonstrated samples, such as the two-surface AT-cut quartz, while deliberately changing the pump laser power by about 20% between runs; if the fitted tensor components shift by more than the reported $2\\sigma$ uncertainties, the power coefficient $\\beta$ is not truly sample-independent and the amplitude channel is biasing the tensor fit.","supporting_citations":[],"review_version":1}