{"id":"3df77db5-69e8-483a-b920-f538a8bf1c37","arxiv_id":"2504.17783","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"AFM-IR and sMIM can distinguish Bernal, rhombohedral, and intermediate stacking orders in multilayer graphene, including under a boron nitride coating.","lead":"This paper shows that two types of atomic force microscopy, one using infrared light and one using microwaves, can distinguish different layer arrangements in multilayer graphene. This gives researchers a practical way to check and build graphene devices with a desired layer stacking, which controls their electronic properties.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"AFM-IR 'absolute contrast' claim is undermined by the paper's own evidence that strain produces stacking-independent AFM-IR contrast; classification templates are not strain-calibrated.","rationale":"The reader's weakest assumption was that the ground-truth stacking labels, especially the inferred I4G assignment, rest on Raman spectroscopy in strained flakes. I agree that this is a real weakness, but I think the more load-bearing issue is the paper's own admission that AFM-IR contrast is not uniquely determined by stacking order: strain is present in the calibration flake and produces additional contrast and gradients that the authors explicitly cannot explain within the three-stacking-order picture. This directly threatens the 'absolute contrast' and 'unambiguous distinction' formulations, because it means a spectrum taken in an unknown region could be misinterpreted if strain mimics or shifts stacking-related contrast. The intermediate-order identification is important because it is the only support for the claim that intermediate stacking orders can be distinguished; if the I4G assignment is wrong, the technique merely shows three contrast levels without identifying them. The proposed test would settle both issues by checking whether the I4G cluster is reproducible and whether strain shifts are small compared to stacking contrast. Because the techniques still convincingly demonstrate domain-resolved, through-hBN imaging with high spatial resolution, the appropriate verdict remains conditional rather than rejection; the authors should soften the 'absolute/unambiguous' language and provide a strain-controlled calibration or an independent structural confirmation of the intermediate-order assignment.","tokens_in":14986,"tokens_out":7089,"duration_ms":75773,"concrete_test":"Re-analyze the existing hyperspectral data (Fig. 1 and Fig. S3): train a classifier on only the Raman-identified B4G and R4G pixels, then test whether the putative I4G strip forms a distinct, spatially contiguous cluster at the diagnostic wavenumbers (1170, 1310, 1500, 1576 cm^-1); if it does not, the intermediate-order identification is not supported. Then, on a fresh atomically flat tetralayer flake, confirm stacking domains by an independent structural probe (STM or LEEM) and acquire AFM-IR spectra to check that the B4G/R4G/I4G signatures reproduce. Finally, apply controlled uniaxial strain (e.g., by bending a flexible substrate) and measure the spectral shift at these wavenumbers. If strain-induced shifts are comparable to or larger than the inter-stacking contrast, the 'absolute' identification claim fails; if they are small, the claim survives.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that AFM-IR yields absolute, unambiguous stacking-order identification by IR spectral analysis. That claim requires the measured contrast at each diagnostic wavenumber to depend only on stacking order. The paper's own data undercut this premise. The calibration flake is explicitly strained (multiple wrinkles noted in Fig. S1 and the main text), and the authors attribute unexplained AFM-IR features to strain: Fig. 1i shows dark features they speculate are 'related to strain within the graphene layers,' and in the 'Imaging of domain wall defects' section they state that AFM-IR at strain-sensitive wavenumbers shows 'additional contrast and gradients' inconsistent with the three stacking orders, 'likely an indication of strain.' Thus strain can generate AFM-IR contrast that is not tied to stacking order. The spectral templates in Fig. 1d,e are averaged over regions of this same strained flake, so they are not strain-calibrated. The I4G assignment, which is the only evidence that intermediate stacking orders can be identified, is inferred from partial disagreement between Raman and AFM-IR rather than an independent structural measurement; strain can also shift the Raman 2D peak used for ground-truth labeling. The paper additionally concedes that domains related by translation or mirror symmetry are expected to show identical contrast, so the imaging cannot distinguish all crystallographically distinct stacking configurations. The 'absolute' and 'unambiguous' formulations are therefore not supported by the presented data.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents AFM-IR (photothermal infrared) and sMIM (scanning microwave impedance microscopy) as tools for nanoscale identification of stacking order in few-layer graphene. On exfoliated tetralayer flakes, the authors use Raman 2D-peak maps as a ground-truth label for Bernal (B4G) and rhombohedral (R4G) domains, and they assign a central strip to an intermediate stacking order (I4G, e.g. ABCB) based on partial agreement between Raman and AFM-IR contrast. They show wavenumber-dependent AFM-IR contrast, including a hyperspectral analysis where different wavenumbers highlight different stacking orders, and sMIM images with two/three contrast levels. Both techniques are applied to hBN-encapsulated trilayers and to imaging of domain walls, including a sub-20 nm-scale feature. The central claims are that AFM-IR provides 'absolute contrast' and unambiguous stacking-order identification via spectral analysis, that sMIM provides higher relative contrast and nanoscale resolution, and that both are suited for high-throughput characterization of van der Waals devices.","tokens_in":15245,"tokens_out":2549,"duration_ms":27147,"significance":"If the claims are established, this work offers a practical, high-throughput route to stacking-order mapping in multilayer graphene, including through an hBN capping layer, which would be valuable for fabricating rhombohedral-stacked devices. The experiments are carefully executed: Raman maps independently label the main domains, the AFM-IR and sMIM contrasts are internally consistent, and the subsurface imaging demonstrates a useful capability. The paper also shows a credible domain-wall analysis connecting the observed patterns to layer shifts and shear strain. However, the load-bearing claims of 'absolute' and 'unambiguous' identification go beyond what the evidence supports, because the I4G assignment is inferred rather than independently verified, and because the manuscript itself documents stacking-independent AFM-IR contrast attributed to strain. The strengths—independent Raman labels for the two extremal orders, consistent multispectral contrast, and sub-20 nm domain-wall resolution—deserve recognition; the overstatement needs correction rather than wholesale rejection.","major_comments":[{"comment":"The assignment of the central strip to the intermediate I4G stacking order is not independently verified: it is inferred from the observation that this region has a Raman signature similar to R4G but AFM-IR contrast similar to B4G. Because the flake is acknowledged to contain considerable strain (Fig. S1), and strain is later invoked to explain AFM-IR contrast that is inconsistent with stacking order (Fig. 4f,g), the inference to a specific stacking order is not unique. The statement in the text that 'we can unambiguously assign the three different stacking orders' overstates the evidence; an independent structural probe (STM, TEM, or a separately calibrated reference sample) or a quantitative exclusion of strain-induced contrast is required to support the I4G label.","section":"Photothermal infrared microscopy, Fig. 1c–1e"},{"comment":"The manuscript explicitly states that AFM-IR at some wavenumbers shows 'additional contrast and gradients' inconsistent with the three stacking orders, 'likely an indication of strain.' This is direct evidence that AFM-IR contrast is not a function of stacking order alone. The spectral templates in Fig. 1d,e were averaged over regions of this same strained flake and are not strain-calibrated, so the claim that AFM-IR 'provides absolute contrast between stacking orders' is not established. Please either add a strain-calibration measurement (e.g., comparing spectra under controlled strain) or revise the claim to specify that AFM-IR provides relative stacking contrast that is valid in the low-strain limit.","section":"Imaging of domain wall defects, Fig. 4f,g"},{"comment":"The paper concedes that 'domains related by translation symmetry or mirror symmetry about the center plane of the graphite are expected to show identical contrast.' This concession means that AFM-IR and sMIM cannot uniquely assign a stacking configuration among all crystallographically distinct possibilities, so the words 'absolute' and 'unambiguous' in the abstract and introduction are too strong. The claims should be restated as distinguishing the specific stacking orders probed in these samples (and, for sMIM, requiring an external calibration for absolute identification), rather than providing a complete and unique stacking-order determination.","section":"Imaging of domain wall defects, text after Fig. 4"}],"minor_comments":[{"comment":"There is a typo in the caption: 'few layer graphehne' should be 'few layer graphene.'","section":"Figure 2 caption"},{"comment":"In the sentence 'with the stacking order returning to the the original one over the course of 5-10 nm,' 'the the' is a typo and should be 'the original.'","section":"Imaging of domain wall defects"},{"comment":"The text refers to panel (c) as a cross-sectional profile, but the figure caption lists only (a), (b), and (d); the panel labeling should be made consistent.","section":"Supplementary Fig. S7"},{"comment":"The statement that the AFM-IR contrast appears 'almost all wavenumbers' would benefit from a quantitative statement of the signal-to-noise ratio and the contrast magnitude relative to the drift-correction uncertainty, since the normalization procedure in Fig. S2 removes line-by-line background and could mask slowly varying spatial features.","section":"Microscopy section"},{"comment":"The phrase 'absolute certainty of identification' in the discussion of far-field IR imaging is not precisely defined; the authors should clarify what 'absolute' means operationally (e.g., calibration against a known reference) so that the contrast with the present sMIM result is clear.","section":"Introduction"}],"recommendation":"major_revision","confidential_remarks":"The experimental work is of good quality and the core correlation between AFM-IR/sMIM contrast and Raman-labeled domains is convincing. The main issue is a mismatch between the strength of the evidence and the strength of the claims: the I4G assignment is inferred, and the paper's own strain observations undermine the 'absolute' and 'unambiguous' phrasing. These are fixable with revised wording and, ideally, one additional control experiment or a clear statement of the low-strain validity regime. The paper is within scope for the journal and likely to be of interest to the community."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know before you read this one. First, the core experimental demonstration is real: AFM-IR and sMIM both produce stacking-dependent contrast in few-layer graphene that tracks Raman labels, survive hBN encapsulation, and resolve domain walls at the tens of nanometers scale. That is a genuinely useful toolkit for anyone screening flakes before device assembly. Second, the word 'unambiguous' in the abstract is doing more work than the data support, and the stress-test note on strain is not a straw man—it lands, partly.\n\nThe new part is the combination: AFM-IR spectral analysis across wavenumbers and sMIM with sub-20nm resolution, applied to tetralayer intermediate stacking (I4G) and to encapsulated flakes. Prior s-SNOM and KPFM work didn't do this. The paper is honest about some limitations—for example, it notes that sMIM gives relative, not absolute, contrast, and it explicitly concedes that translation/mirror-related stacking configurations are indistinguishable. That is a real caveat, and it's in the text, not hidden.\n\nThe soft spots are where the claims are a half-step ahead of the evidence. The 'absolute contrast via IR spectral analysis' claim requires that the chosen wavenumbers respond only to stacking order. But the authors themselves show wavenumbers where AFM-IR contrast is not explained by the three stacking orders and attribute it to strain (Fig. 4f,g), and they speculate that features in Fig. 1i are strain-related. The spectral templates in Fig. 1d,e are averaged over a flake with acknowledged wrinkles and strain. So a reader cannot yet know whether a region with unexpected AFM-IR contrast is an intermediate stacking order or a strained region of a known one. The I4G assignment compounds this: it rests on partial agreement—same as R4G in Raman, same as B4G in AFM-IR at one wavenumber—rather than an independent structural probe. The main text calls it unambiguous; the actual evidence is suggestive. Also, the O2-plasma pretreatment is mentioned in the methods but not in the main text; that's an important condition for reproducing the effect.\n\nThat said, the paper is not a house of cards. The Raman labeling is independent, the domain-wall picture in Fig. 4 is internally consistent, and the hBN encapsulation result is a useful step for device characterization. The strain concern is real but it is a calibration issue, not a demonstration that the methods fail. A serious referee should ask for either strain-calibrated templates or a more guarded formulation of 'absolute.'\n\nWho should read it? Anyone who builds multilayer graphene devices and needs a faster way to find rhombohedral or intermediate stacking domains. The paper deserves to go to peer review. I'd accept it as a methods paper with revision, and I'd want the authors to qualify the 'unambiguous' language and address the strain calibration gap directly.","headline":"Solid AFM-IR/sMIM toolkit for stacking screening, but 'unambiguous' overstates what the data show; strain and the inferred I4G assignment need to be addressed.","tokens_in":15828,"tokens_out":2227,"would_cite":true,"duration_ms":22410,"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":"Two atomic force microscopy techniques identify graphene stacking orders at the nanoscale, including through a hexagonal boron nitride layer.","keywords":["graphene stacking order","rhombohedral graphite","Bernal graphite","AFM-IR","scanning microwave impedance microscopy","van der Waals heterostructures","hexagonal boron nitride encapsulation","domain walls"],"falsifier":"Take a tetralayer flake with the same mixed stacking domains and determine the stacking of the central strip region directly by cross-sectional atomic-resolution imaging, for example annular dark-field scanning transmission electron microscopy. If the strip is not ABCB stacking, then the claimed identification of the intermediate order is wrong; more generally, any region where AFM-IR and sMIM contrast disagrees with the independent structural measurement would falsify the claim of unambiguous distinction.","tokens_in":14808,"feed_emoji":"🔬","tokens_out":8356,"duration_ms":78409,"temperature":0.7,"pith_summary":"The paper argues that two atomic-force-microscopy techniques can do what far-field optics cannot: tell apart the different layer-stacking orders of multilayer graphene at the nanoscale and identify the faults between them. Photothermal infrared microscopy (AFM-IR) reads out local infrared absorption and, by recording spectra at many wavenumbers, gives contrast that can be assigned to Bernal, rhombohedral, and intermediate stacking orders. Scanning microwave impedance microscopy (sMIM) reads out the local tip-sample impedance and shows stronger relative contrast, resolving stacking faults below 20 nanometers. Both techniques work on flakes already buried under a hexagonal boron nitride dielectric, which matters because stacking order controls the electronic band structure and therefore which correlated and topological phases the material can host. The practical goal is high-throughput screening of flakes and devices so that samples with a definite, uniform interlayer registry can be made reliably.","feed_headline":"Two AFM methods spot graphene stacking faults at sub-20 nm scale","feed_subtitle":"Infrared and microwave AFM probes distinguish Bernal, rhombohedral, and intermediate graphene domains, even under hBN.","key_machinery":"The load-bearing objects are two scanning-probe contrast mechanisms. AFM-IR (photothermal infrared microscopy) is a local version of infrared absorption spectroscopy: a pulsed infrared laser heats the sample and the AFM cantilever senses the resonant thermal expansion, so the signal tracks absorption in a volume near the 20 nm tip apex; sweeping the laser wavenumber yields a spectrum at each point. sMIM (scanning microwave impedance microscopy) sends a microwave signal down the cantilever and detects the reflected signal, whose amplitude depends on the local tip-sample impedance; in dC/dV mode an AC modulation of the tip voltage extracts the capacitance derivative and recovers contrast through a dielectric layer. The assignments are anchored by Raman 2D-peak maps, which supply the ground-truth stacking labels, and by the known band-structure differences between stacking orders that the paper cites as making infrared absorption stacking-dependent below about 0.8 eV. The domain-wall analysis then works by tracking which of the top one, two, or three layers shifts between A, B, and C registry positions, identifying transitions such as ABCA to ABAB to ACBC to ACBA as a shear-strain pattern.","core_discovery":"On the paper's own terms, the central discovery is that AFM-IR and sMIM provide a full nanoscale picture of stacking order in few-layer graphene that Raman and far-field infrared imaging cannot. In tetralayer flakes, AFM-IR spectra acquired across 800-1800 cm-1 show contrast between Bernal (ABAB), rhombohedral (ABCA), and the intermediate ABCB stacking order, with rhombohedral identified by a peak near 1580 cm-1 and other wavenumbers separating each order from the other two; the intermediate assignment is corroborated by a region that matches Bernal in AFM-IR but rhombohedral in Raman. sMIM, calibrated by Raman maps and by AFM-IR, distinguishes the same domains with higher contrast and resolves domain walls, including narrow double walls with internal structure at the 10 nm scale. Both techniques image trilayer graphene buried under 5-10 nm of hBN, with AFM-IR contrast unchanged and sMIM requiring an AC voltage modulation (dC/dV mode). The paper also shows that certain AFM-IR wavenumbers respond to strain, revealing that micrometer-scale strain gradients and incomplete relaxation coexist with the discrete stacking domains.","pith_inferences":["Because AFM-IR contrast appears at nearly every probed wavenumber, including energies below the lowest interband separation, the mechanism is probably not simple interband absorption; a natural extension is to record spectra to lower wavenumbers and compare with a Bernal bilayer reference to separate thermal-expansion contributions from electronic ones.","The strain-dependent contrast at certain wavenumbers suggests AFM-IR could be developed into a quantitative strain mapper if the response is calibrated against flakes with known uniaxial or shear strain patterns.","For flakes with more than four layers the number of possible stacking orders grows, and AFM-IR spectral libraries combined with an independent structural probe could classify those higher-order stackings, which the paper does not attempt.","For production screening, sMIM's speed and signal-to-noise make it the natural inline tool, but only when paired with AFM-IR or Raman calibration; a testable workflow would be to pre-screen entire wafers with sMIM and verify selected regions with AFM-IR."],"forward_implications":["Stacking order can be mapped at the nanoscale in mixed-domain flakes, so uniform rhombohedral regions can be located and used for device fabrication instead of relying on rare large-area domains.","Because both techniques work on hBN-encapsulated flakes, the stacking order of finished van der Waals devices can be verified without electrical contact or disassembly.","sMIM's sub-20 nm resolution makes domain walls between regions of the same stacking order visible, allowing device areas to be checked for these nanoscale faults.","AFM-IR spectral fingerprints give an absolute, calibration-free assignment of Bernal, rhombohedral, and intermediate stacking orders in few-layer flakes.","The strain-sensitive infrared wavenumbers could allow simultaneous monitoring of stacking order and strain during experiments that heat, strain, or apply electric fields to the flake."],"supporting_citations":[{"why":"Supplies the result that infrared absorption spectroscopy distinguishes graphene stacking orders via low-energy band-structure differences, the mechanism AFM-IR localizes.","marker":"[23]"},{"why":"Provides the Raman 2D-peak signature used as ground truth for stacking labels in the flakes.","marker":"[24]"},{"why":"Provides the complementary Raman 2D-peak assignment that anchors the stacking-order calibration.","marker":"[25]"},{"why":"Supplies the identification of intermediate (ABCB-type) stacking in tetralayer graphene that the paper uses for its I4G assignment.","marker":"[33]"},{"why":"Supports the near-field infrared signatures of different stacking orders that AFM-IR extends.","marker":"[34]"},{"why":"Documents the strain and domain-wall structure in graphene that the paper resolves with sMIM.","marker":"[27]"},{"why":"Introduces photothermal AFM-IR, the detection principle the paper applies to stacking imaging.","marker":"[37]"},{"why":"Establishes the sMIM measurement scheme for local impedance contrast.","marker":"[39]"},{"why":"Supplies the dC/dV modulation mode used to recover sMIM contrast through the hBN dielectric.","marker":"[43]"},{"why":"Provides the dry-transfer fabrication method for the hBN-encapsulated samples imaged subsurface.","marker":"[47]"}],"fun_headline_variants":["AFM-IR and sMIM map graphene stacking at nanoscale","Infrared and microwave AFM pinpoint graphene stacking faults","New AFM duo reveals graphene layer order under hBN","IR and microwave AFM expose hidden graphene stacking order"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim of unambiguous stacking identification rests on Raman 2D-peak maps as the ground truth that labels each domain; if Raman mislabels stacking in these strained, wrinkled flakes, the inferred AFM-IR and sMIM assignments would fail.","fun_headline_variants_meta":{"raw":{"variants":["AFM-IR and sMIM map graphene stacking at nanoscale","Infrared and microwave AFM pinpoint graphene stacking faults","New AFM duo reveals graphene layer order under hBN","IR and microwave AFM expose hidden graphene stacking order"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000639,"raw_usage":{"total_tokens":3016,"prompt_tokens":1089,"completion_tokens":1927,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":705,"completion_tokens_details":{"reasoning_tokens":1860}},"tokens_in":705,"tokens_out":1927,"duration_ms":12366,"temperature":1.0,"reasoning_tokens":1860,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:31:10.306410+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a tetralayer flake with the same mixed stacking domains and determine the stacking of the central strip region directly by cross-sectional atomic-resolution imaging, for example annular dark-field scanning transmission electron microscopy. If the strip is not ABCB stacking, then the claimed identification of the intermediate order is wrong; more generally, any region where AFM-IR and sMIM contrast disagrees with the independent structural measurement would falsify the claim of unambiguous distinction.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the result that infrared absorption spectroscopy distinguishes graphene stacking orders via low-energy band-structure differences, the mechanism AFM-IR localizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Raman 2D-peak signature used as ground truth for stacking labels in the flakes."},{"cited_title":"Experimental observation of ABCB stacked tetralayer graphene","cited_arxiv_id":"2203.07971","evidence_quote":"Supplies the identification of intermediate (ABCB-type) stacking in tetralayer graphene that the paper uses for its I4G assignment."},{"cited_title":"Beitner, S","cited_arxiv_id":null,"evidence_quote":"Supports the near-field infrared signatures of different stacking orders that AFM-IR extends."},{"cited_title":"Topological kink states at a tilt boundary in gated multi-layer graphene","cited_arxiv_id":"1301.1690","evidence_quote":"Documents the strain and domain-wall structure in graphene that the paper resolves with sMIM."},{"cited_title":"Dazzi, R","cited_arxiv_id":null,"evidence_quote":"Introduces photothermal AFM-IR, the detection principle the paper applies to stacking imaging."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the sMIM measurement scheme for local impedance contrast."},{"cited_title":"Drevniok, P","cited_arxiv_id":null,"evidence_quote":"Supplies the dC/dV modulation mode used to recover sMIM contrast through the hBN dielectric."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the dry-transfer fabrication method for the hBN-encapsulated samples imaged subsurface."}],"review_version":1}