{"id":"04582fe9-4006-461c-84cb-7f4f18c37c64","arxiv_id":"2507.18965","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Graphene films on Ge(110) can be transferred dry and intact with h-BN carriers because their interfacial adhesion energy is lower than typical van der Waals bonding.","lead":"This paper shows that graphene grown on Ge(110) can be peeled off intact using a boron nitride carrier and moved to another surface without any wet chemicals. The method works because graphene sticks only weakly to germanium, and the authors estimate that weak stickiness at roughly 23 meV per carbon atom.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 23 meV/C upper limit is not supported: Eq. (4) gives ~3 meV/C from stated values and ignores strain energy in the much thicker h-BN carrier.","rationale":"The paper reports a useful all-dry transfer method with strong characterization support, and the qualitative conclusion that graphene on Ge(110) is weakly adhered is consistent with the successful exfoliation. However, the central quantitative claim in the abstract—adhesion below typical vdW interactions—depends entirely on the 23 meV/C estimate, which has two unaddressed problems. First, the stated formula and input values do not yield 23 meV/C; straightforward substitution gives about 3 meV/C, a factor-of-8 discrepancy that the manuscript never explains. Second, the derivation neglects strain energy in the h-BN superlayer. Since h-BN is about 150 times thicker than graphene and comparably stiff, even modest h-BN strain dominates the energy balance. Consequently, the graphene-only strain energy cannot provide an upper bound on the interfacial adhesion energy; if anything, omitting h-BN energy would make the estimate a lower bound that could be far too low. The paper's own SI buckle calculation also appears inconsistent with its reported 7 meV/C value when the stated h, w, and t are inserted, indicating that the energy accounting needs revision. These issues are correctable in a revision but currently leave the headline comparison unverified. The reader's conditional verdict is appropriate; my analysis adds the numerical mismatch in Eq. (4) and the inconsistency in the SI buckle formula, so I agree only partially with the reader's stated weakest assumption.","tokens_in":15753,"tokens_out":10332,"duration_ms":110728,"concrete_test":"Measure the h-BN strain at the critical thickness on the same h-BN/graphene/Ge films used in Fig. 4b, using the h-BN E2g Raman shift (or XRD d-spacing) calibrated against a known strain response. Then recompute the delamination energy per unit area as the sum of graphene and h-BN biaxial strain energies, U = Σ E_i t_i ε_i²/(1−ν_i), and compare with the reported 23 meV/C. If ε_hBN is comparable to graphene's 0.67%, U exceeds the graphene-only estimate by roughly 150×, falsifying the claim that 23 meV/C is an upper limit. As an immediate arithmetic check, recompute Eq. (4) with the stated ε, E_Gr, t_Gr, and ν_Gr; if it gives about 3 meV/C instead of 23, the paper must explain the factor-of-8 discrepancy before the quantitative conclusion can stand.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline quantitative claim—that graphene/Ge(110) adhesion is below the typical vdW range—rests entirely on the 23 meV/C upper limit derived in Section 4 (Fig. 4c,d). Two independent problems undermine it. First, substituting the stated values (ε = 0.67%, E_Gr ≈ 1 TPa, t_Gr = 0.335 nm, ν_Gr ≈ 0.16) into Ee = ε²E_Gr t_Gr/(1−ν_Gr) yields only about 3 meV per carbon atom, not 23 meV. Reproducing 23 meV/C would require ε ≈ 1.9%, t_Gr ≈ 2.6 nm, or E_Gr ≈ 7.8 TPa; none of these is reported or defensible. So the central number cannot be reproduced from the given data. Second, even with corrected arithmetic, the derivation treats the h-BN superlayer as strain-free. h-BN at the critical thickness is 50–60 nm, roughly 150 times thicker than graphene, with a comparable Young's modulus (~865 GPa). If the h-BN carries any strain comparable to graphene's, its strain-energy contribution per unit area dominates by a factor of order 150. The delamination condition should be set by the total strain-energy release rate of the composite film, not by graphene alone. The text's justification that graphene is 'significantly softer' than h-BN is not correct: the two materials have comparable elastic moduli. The paper's own alternative estimate from buckle geometry (SI Eq. 3) gives a different value (reported as 7 meV/C), and recomputing that formula with the stated h = 440 nm, w = 3.6 μm, t = 60 nm does not reproduce 7 meV/C either, underscoring that the energy accounting is internally inconsistent. The qualitative transfer demonstration remains credible, but the specific claim that this is the first as-grown graphene film with adhesion below 40–50 meV/C is unverified.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports an all-dry, van der Waals transfer method for continuous monolayer graphene grown on Ge(110), using a CVD-grown h-BN carrier. The authors characterize transferred films by optical microscopy, Raman mapping, AFM, LEED, cross-sectional TEM-EDX, XPS, and electrical transport, reporting uniform monolayer films, low defect density, clean interfaces, and low charge-puddle density. They attribute the transferability to weak graphene/Ge(110) adhesion and estimate an upper limit of 23 meV/C atom for the interfacial adhesion energy from the compressive strain energy accumulated before spontaneous delamination; a second estimate based on buckle geometry gives 7 meV/C atom. They also demonstrate a 'contact and peel' extension to stack arbitrary overlayers, including twisted bilayer graphene.","tokens_in":16109,"tokens_out":7778,"duration_ms":74521,"significance":"The qualitative result—that continuous as-grown graphene can be transferred without wet chemistry while retaining good electronic quality—is valuable and appears well supported by the broad characterization. The transport data (Δn_w ≈ 3.6×10^11 cm^-2, uniform Dirac voltages) and the TEM/XPS cleanliness evidence are particular strengths, as is the demonstration of layer-by-layer assembly with twisted bilayer graphene. However, the paper's headline quantitative claim, that graphene/Ge(110) adhesion is below the van der Waals range, rests on the 23 meV/C atom estimate, and that estimate is not currently supportable as written. If the energy accounting is corrected or the claim is appropriately weakened, the qualitative finding may stand, but the strong 'first as-grown film below the vdW range' claim needs more work.","major_comments":[{"comment":"The headline value γ_Gr-Ge = 23 ± 7.5 meV/C atom is not reproducible from the stated inputs. Substituting ε = 0.67%, E_Gr ≈ 1 TPa, t_Gr = 0.335 nm, and ν_Gr ≈ 0.16 into E_e = ε² E_Gr t_Gr/(1−ν_Gr) gives about 1.8×10⁻² J/m², which converts to roughly 3 meV per carbon atom using a graphene areal density of 3.8×10¹⁹ atoms/m², not 23 meV/C atom. Recovering 23 meV/C atom would require ε ≈ 1.9% or t_Gr ≈ 2.6 nm, neither of which is reported. Please provide the actual arithmetic and the conversion to per-atom units; as written, the central number is not supported by the data.","section":"Section 4, E_e = ε² E_Gr t_Gr/(1−ν_Gr)"},{"comment":"The derivation assumes that 'most of the residual strain is accumulated at graphene, which is significantly softer than the much thicker h-BN superlayer.' This assumption is not valid: the Young's modulus of h-BN (~865 GPa, as cited in the Supporting Information) is comparable to that of graphene (~1 TPa), not significantly smaller, and the h-BN superlayer at the critical thickness is 50–60 nm, about 150 times thicker than monolayer graphene. If the h-BN carries even a modest fraction of the compressive strain measured in graphene, its strain-energy contribution per unit area dominates by a factor of order 10². The delamination condition should be written for the total strain energy of the composite h-BN/graphene film. Because energy stored in the h-BN is omitted, the value 23 meV/C atom cannot be justified as an upper limit on γ_Gr-Ge; at best it is the graphene-only strain energy, and the actual interfacial adhesion could be substantially larger.","section":"Section 4, strain-energy accounting for h-BN superlayer"},{"comment":"The reported estimate of 7 meV/C atom is not obtained from the stated geometry. With h = 440 nm, w = 3.6 μm, t = 60 nm, E = 865 GPa, and ν = 0.211, Eqs. (1)–(3) give σ_c ≈ 0.83 GPa, σ_d ≈ 34 GPa, and γ ≈ 40 J/m², which is roughly 6.5 eV per carbon atom and about three orders of magnitude larger than the stated 7 meV/C atom. Please show the full calculation or correct the input values; as it stands, the buckle-geometry estimate does not provide independent support for the adhesion value.","section":"Supporting Information, 'Calculation of interfacial toughness', Eqs. (1)–(3)"}],"minor_comments":[{"comment":"The phrase 'has an substrate adhesion energy' should read 'has a substrate adhesion energy.'","section":"Abstract"},{"comment":"The symbol 'g' is used for adhesion energy; this should be the standard symbol γ to avoid confusion with a gravitational or coupling constant.","section":"Introduction, first paragraph"},{"comment":"The red dotted line marking the maximum strain and E_e would be easier to interpret if the number of samples and the criterion for selecting the maximum were stated, since the text reports a maximum value 'measured from multiple samples.'","section":"Figure 4c"},{"comment":"The deconvolution of strain versus hole doping relies on calibrations from Refs. 29 and 30; please state the uncertainty in the extracted strain and whether the calibration accounts for the 488 nm excitation wavelength used here.","section":"Section 4, Raman deconvolution"},{"comment":"The transfer procedure includes a PMMA spin-coating step before TRT attachment; the role of the PMMA layer in the claimed all-dry, clean transfer should be clarified, since PMMA is a polymer that is later removed by annealing.","section":"Methods, 'All-dry transfer process'"}],"recommendation":"major_revision","confidential_remarks":"The experimental demonstration of dry transfer is solid and likely publishable, but the adhesion-energy analysis is the least rigorous part of the manuscript and is central to the abstract's novelty claim. I recommend asking the authors to provide a corrected derivation with explicit arithmetic, or to substantially weaken the quantitative claim. The buckle-geometry estimate and the strain-energy estimate should be reconciled before the paper is accepted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper deserves a serious look for the method, not for the number. The all-dry transfer of centimeter-scale graphene from Ge(110) via an h-BN carrier is new, and the characterization is thorough: Raman, TEM-EDX, XPS, AFM, transport, and the tBLG demonstration with a clean van Hove singularity all hang together. The contact-and-peel approach is a nice generalization. If I worked on 2D assembly, I would keep this paper on the table.\n\nThe soft spot is exactly where the stress-test note lands. The central quantitative claim—23 meV/C as an upper limit for graphene/Ge(110) adhesion—does not survive contact with the stated formula. Plug ε = 0.67%, E = 1 TPa, t = 0.335 nm, ν = 0.16 into Ee = ε²Et/(1−ν) and you get about 3 meV/C, not 23. The paper never shows the intermediate steps, and the mismatch is large enough that the headline result is unverified. The justification for ignoring the h-BN strain energy is also wrong: h-BN is ~150× thicker than graphene and has a comparable modulus, so if it carries any strain, its energy contribution dominates. That flips the direction of the error: the 23 meV/C would be a lower bound on the total strain energy, not an upper limit on adhesion. The SI buckle-based estimate of 7 meV/C also does not reproduce from the given geometry, so the two independent estimates are mutually inconsistent. This is a load-bearing flaw because the paper's significance rests on being the first as-grown graphene film with adhesion below the 40–50 meV/C vdW range. That claim is currently unsupported.\n\nThat said, the qualitative inference—graphene/Ge(110) is weakly adhered and therefore transferable—is likely correct, given that the transfer works at scale. The paper is not a throwaway; it just needs the energy accounting fixed or the quantitative claim softened. I would send it to peer review, but with the expectation of major revision.","headline":"The dry-transfer method is genuinely useful, but the 23 meV/C adhesion claim is not reproducible from the paper's own equations and the strain-energy argument is backwards.","tokens_in":16709,"tokens_out":6402,"would_cite":true,"duration_ms":59540,"reading_group":"yes","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Continuous graphene grown on Ge(110) can be lifted dry with an h-BN carrier because its substrate adhesion, at most 23 meV per carbon atom, is weaker than typical van der Waals bonding.","keywords":["dry transfer","graphene","van der Waals interactions","hexagonal boron nitride","adhesion energy","Ge(110)","CVD growth","twisted bilayer graphene"],"falsifier":"Measure the strain in the h-BN layer directly (for example, by Raman phonon shifts or X-ray diffraction of the h-BN film at the critical thickness) and compute the strain-energy split between h-BN and graphene. If h-BN carries strain energy comparable to or larger than graphene, the derived 23 meV per carbon atom upper limit is not supported; alternatively, an independent adhesion measurement by pressurized blister or buckle geometry on the graphene/Ge interface would settle whether the true adhesion is below the van der Waals range.","tokens_in":15542,"feed_emoji":"🧲","tokens_out":6148,"duration_ms":60390,"temperature":0.7,"pith_summary":"This paper reports that continuous monolayer graphene grown on Ge(110) can be lifted off the growth substrate in a dry process, using a hexagonal boron nitride (h-BN) film as a carrier, and then placed onto another substrate. The reason the lift-off works is that graphene bonds to Ge(110) unusually weakly: from the strain energy released just before the film buckles off, the authors estimate an upper limit of 23 meV per carbon atom for the interfacial adhesion energy, below the 40–50 meV per carbon atom typical of van der Waals bonding between layered materials. If that estimate holds, graphene on Ge(110) is the first as-grown graphene film that can be mechanically exfoliated by ordinary van der Waals forces, giving a clean, chemistry-free route to large-scale graphene devices and heterostructures.","feed_headline":"Graphene peels off Ge with a dry h-BN carrier","feed_subtitle":"At most 23 meV per carbon atom binds the film, so it lifts cleanly—no wet etching, no polymer residue.","key_machinery":"The load-bearing mechanism is the weak van der Waals adhesion between graphene and Ge(110), measured through strain-energy balance at the point of spontaneous delamination. Growing an h-BN superlayer compresses the graphene; as the h-BN thickens past a critical value (about 50 nm), the stored strain energy exceeds the interfacial adhesion and the film buckles off. The authors extract the graphene strain from the shifts of the Raman G and 2D peaks (deconvoluted against hole-doping shifts), reaching about $0.67\\%$, and convert it to energy using the elastic formula $E_e = \\varepsilon^2 E_{\\mathrm{Gr}} t_{\\mathrm{Gr}}/(1-\\nu_{\\mathrm{Gr}})$, giving $23\\ \\mathrm{meV}$ per C atom as an upper limit. The h-BN/graphene stack itself acts as the carrier, and the \"contact and peel\" variant extends the mechanism to arbitrary overlayers X.","core_discovery":"The central claim is that graphene/Ge(110) is an unusually detachable interface: the interfacial adhesion energy is bounded above by $23 \\pm 7.5\\ \\mathrm{meV}$ per C atom, lower than the van der Waals adhesion between graphene and other layered materials ($40\\text{–}50\\ \\mathrm{meV}$ per C atom). The paper demonstrates the consequence by growing a graphene film on Ge(110), depositing an h-BN superlayer on top, and peeling the h-BN/graphene stack off with thermal-release tape in an inert atmosphere. The transferred films are continuous over centimeter areas, show monolayer-thickness uniformity, low defect densities, and charge inhomogeneity as low as $3.6\\times10^{11}\\ \\mathrm{cm^{-2}}$, and can be used to make field-effect transistors. The same \"contact and peel\" procedure is used to fabricate graphene interfaces with Cu, Ni, h-BN, and another graphene layer, including a twisted bilayer with a $15^\\circ$ stacking angle that shows the expected van Hove optical absorption, evidence that the transferred interfaces are clean.","pith_inferences":["A consequence the authors leave implicit: if the graphene/Ge interface is truly below the van der Waals range, the as-grown graphene is nearly freestanding on Ge, which may make Ge(110) a useful platform for studying intrinsic graphene physics without transferring it.","The strain-energy split between graphene and h-BN is the easiest place to pressure-test the 23 meV number; repeating the Raman analysis while monitoring h-BN phonons would show whether the upper-limit logic holds.","The same peeling criterion could be tested on other Ge-grown layers, such as the aligned h-BN monolayer reported on Ge, by checking whether a metal or polymer overlayer of comparable thickness exfoliates it with similar ease."],"forward_implications":["Centimeter-scale monolayer graphene can be moved to arbitrary substrates without wet chemistry, so polymer or metal etching and their residue are avoided at the graphene interface.","Because the graphene/Ge bond is weaker than the h-BN/graphene bond, the h-BN carrier can remain on the graphene as a protective layer and gate dielectric, simplifying device fabrication.","The \"contact and peel\" route makes clean graphene interfaces with evaporated metals (Cu, Ni), h-BN, and even another graphene layer, enabling layer-by-layer assembly and twisted bilayer graphene with controlled angles.","Transferred channels inherit low disorder: the measured charge-puddle density is comparable to the cleanest exfoliated graphene on SiO2, so dry transfer does not limit electronic quality.","Other 2D materials grown on Ge (h-BN monolayers, graphene nanoribbons) are expected to be similarly detachable, since their adhesion to Ge appears to be below 60 meV per carbon atom."],"supporting_citations":[{"why":"Supplies the growth method and substrate that produce high-quality monolayer graphene on Ge(110), the starting point of the dry transfer.","marker":"15,16"},{"why":"Provides comparison adhesion energies for graphene on SiC and Cu, showing why typical growth substrates resist mechanical exfoliation.","marker":"10,11"},{"why":"Defines the 40–50 meV per carbon atom van der Waals reference range that the measured graphene/Ge adhesion is claimed to fall below.","marker":"12,13"},{"why":"Supplies the fracture-mechanics relations used to convert the critical compressive strain energy into an interfacial adhesion energy.","marker":"28,31"},{"why":"Demonstrates h-BN as a protecting and peeling carrier for graphene, motivating the decision to keep the h-BN superlayer in the final device.","marker":"6"},{"why":"Shows wafer-scale mechanical transfer by van der Waals forces for TMD films, the approach this paper extends to graphene.","marker":"7"},{"why":"Reports the weak electronic and mechanical coupling of the graphene/Ge interface, the premise that makes the dry peel feasible.","marker":"17"},{"why":"Provides reference optical-conductivity spectra for twisted bilayer graphene, used to show that the transferred graphene-graphene interface is clean and strongly coupled.","marker":"32"}],"fun_headline_variants":["Dry peel: Graphene lifts off Ge with weak 23 meV grip","23 meV per C atom: graphene peels cleanly from Ge","Graphene on Ge: weakest stick yet, dry lift-off via h-BN","No wet etching: graphene film lifts off Ge with h-BN carrier"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The estimate assumes that essentially all of the compressive strain energy measured before delamination is stored in the graphene layer, because the authors treat graphene as much softer than the h-BN superlayer; if the h-BN stores a sizable share of that energy, the 23 meV per carbon atom figure would not be a valid upper limit.","fun_headline_variants_meta":{"raw":{"variants":["Dry peel: Graphene lifts off Ge with weak 23 meV grip","23 meV per C atom: graphene peels cleanly from Ge","Graphene on Ge: weakest stick yet, dry lift-off via h-BN","No wet etching: graphene film lifts off Ge with h-BN carrier"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000802,"raw_usage":{"total_tokens":3515,"prompt_tokens":922,"completion_tokens":2593,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":2510}},"tokens_in":538,"tokens_out":2593,"duration_ms":17091,"temperature":1.0,"reasoning_tokens":2510,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:05:00.801109+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the strain in the h-BN layer directly (for example, by Raman phonon shifts or X-ray diffraction of the h-BN film at the critical thickness) and compute the strain-energy split between h-BN and graphene. If h-BN carries strain energy comparable to or larger than graphene, the derived 23 meV per carbon atom upper limit is not supported; alternatively, an independent adhesion measurement by pressurized blister or buckle geometry on the graphene/Ge interface would settle whether the true adhesion is below the van der Waals range.","supporting_citations":[],"review_version":2}