{"id":"a8737624-32d3-4cb5-a3ab-1438f658e95b","arxiv_id":"2507.18985","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Monolayer hexagonal boron nitride films were grown as a single crystal on stepped Ge(110) by tuning hydrogen pressure, enabling clean layer-by-layer assembly.","lead":"This paper reports a way to grow one-atom-thick hexagonal boron nitride crystals that all point in the same direction on a germanium wafer, making a large single-crystal film. The advance could enable cleaner, larger dielectric layers for two-dimensional electronics and stacked ferroelectric devices.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Wafer-scale single-crystallinity rests on 1 mm² LEED and a 1 µm SHG map; no orientation-sensitive measurement spans the 2-inch wafer, making the headline extrapolative.","rationale":"I read the paper as a strong experimental contribution: the local LEED/SEM/SHG data are internally consistent with unidirectional hBN growth on vicinal Ge(110), the PH2 dependence is interesting, and the exfoliation and layer-by-layer assembly results are valuable. My concern is not about internal consistency or fabrication; it is about the scale of the headline claim. The reader's weakest assumption (the DFT selective de-passivation mechanism in §2.4) is real, and the 1×1×1 k-point mesh and imported edge energy do make the sub-0.01 eV energy differences fragile. However, that mechanism is supporting rather than constitutive: even if the DFT picture were incomplete, the empirical phenomenon of optimized-PH2 unidirectional growth would still stand. What would not stand is the 'wafer-scale single-crystalline' statement if the orientation is verified only over ~1 mm² and ~1 µm. The current evidence cannot distinguish a truly single-crystal wafer from a mosaic of large uni-directional domains, especially given the ~3% anti-parallel grains seen in partial-coverage statistics. This gap is directly addressable with a wafer-scale orientation map, and it does not require rejecting the local results. Because the reader's CONDITIONAL verdict already captures the need for additional evidence, I do not move the verdict; I would keep it CONDITIONAL with the explicit condition being wafer-scale crystallographic characterization rather than further DFT refinement alone.","tokens_in":14190,"tokens_out":6719,"duration_ms":71141,"concrete_test":"Map polarization-resolved SHG at 10–20 positions spread across the full 2-inch transferred film, extracting the azimuthal angle of the cos²(3Φ) pattern at each position and checking for local SHG-intensity suppression at putative grain boundaries. If the extracted angle varies by more than the experimental uncertainty, or if the SHG intensity drops at any location, the wafer-scale single-crystal claim is not supported. A complementary check is azimuthal grazing-incidence X-ray diffraction at multiple wafer positions to map crystallographic orientation over centimeter scales.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—monolayer, wafer-scale, single-crystalline hBN with uni-directional crystallography (Abstract; §2.3; Table S1)—requires that the crystallographic orientation be uniform over the full 2-inch Ge(110) wafer. The orientation-sensitive evidence is local only: LEED over ~1 mm² (Methods: 'All LEED patterns were obtained over an area of ~1 mm2'; Figs. 1g,h and 2c,d) and one polarization-resolved SHG pattern plus a 1 µm-scale SHG map (Fig. 3d,e). The wafer-scale data in Fig. S9 are a photograph and SEM micrographs, which show morphology but cannot detect lattice orientation; the transmission spectra in Fig. 3c confirm monolayer thickness, not crystallinity. The growth-time series in Fig. 3b shows aligned triangular voids only in the imaged fields, so this too is local. Figure S4 itself reports 97% alignment on Sub(A), leaving ~3% anti-parallel grains in partial-coverage samples; the reported LEED/SHG measurements cannot rule out that such minority domains persist in the full film outside the probed areas. The transition from 'unidirectionally aligned grains in the sampled regions' to 'single-crystalline across the entire wafer' is therefore an extrapolation. Since the practical novelty claimed in Table S1 is wafer-scale single crystallinity with clean exfoliation, this gap is load-bearing.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports CVD growth of monolayer hexagonal boron nitride (hBN) on vicinal Ge(110) substrates with a miscut toward [001], where Ge atomic steps guide unidirectional alignment of hBN domains. The authors show that the fraction of aligned triangular grains varies non-monotonically with hydrogen partial pressure, reaching near-unity at an optimized PH2, as characterized by SEM, AFM, and LEED. DFT calculations are used to propose that the optimum alignment arises from a balance between hydrogen passivation of the Ge terrace and direct Ge–N bonding at step edges. The grown films are mechanically exfoliated and assembled into bilayers with controlled rhombohedral or hexagonal stacking, and are demonstrated as dielectric interlayers in graphene and MoS2 field-effect transistors with reduced interfacial trap densities.","tokens_in":14445,"tokens_out":4547,"duration_ms":50450,"significance":"If the central claim holds, the work would provide a practical route to wafer-scale single-crystalline hBN that can be cleanly exfoliated, combining the advantages of metal-grown single-crystal hBN (large area, uniform thickness) with the clean-transfer capability of Ge-based growth. The demonstration of controlled stacking order through layer-by-layer assembly of such films is also valuable for applications in ferroelectric and nonlinear-optical devices. Important strengths of the manuscript include the systematic dependence of alignment on hydrogen partial pressure, the detailed local structural characterization (LEED, AFM, SEM, TEM), and the explicit DFT-based mechanistic model. However, as detailed in the major comments, the headline 'wafer-scale single-crystalline' claim is not directly evidenced, and the DFT energy differences underlying the proposed mechanism require convergence verification.","major_comments":[{"comment":"The claim that the film is 'single-crystalline' over a 2-inch wafer (Abstract; Table S1) is extrapolated from measurements that are local in extent. LEED patterns are obtained over an area of ~1 mm2 (Methods, 'All LEED patterns were obtained over an area of ~1 mm2'), the SHG polarization plot and map in Fig. 3d,e are from a single ~1 µm region, and the SEM images in Fig. 3b and Fig. S9 are local fields. No orientation-sensitive measurement spans the wafer: there is no SHG mapping or LEED at multiple positions across the 2-inch wafer, no XRD pole figure, and no dark-field TEM. The 97% alignment statistic in Fig. S4 also applies only to partial-coverage samples on one substrate, so the data do not exclude anti-parallel or misoriented domains elsewhere in the full film. The transition from 'unidirectionally aligned grains in the probed regions' to 'wafer-scale single-crystalline film' is therefore an extrapolation and should be either directly demonstrated or explicitly qualified.","section":"Section 2.3, Figure 3, and Methods"},{"comment":"The DFT conclusion that θ = 0° is a deep global minimum at intermediate PH2 rests on energy differences of order 0.01 eV (Fig. 4c,f) and on free-energy differences between H-H and Ge-N edge models (Fig. S10) of similar magnitude. All calculations use a 1×1×1 k-point mesh on a three-layer Ge slab (Computational Details), and no convergence tests with respect to k-point sampling, slab thickness, or the DFT-D3 dispersion correction are reported. Because the energy differences are comparable to the numerical uncertainty of such calculations, the predicted orientation preference and the relative stability of Ge-N versus H-passivated edges need to be verified with denser k-point meshes (e.g., 2×2×1 or 3×3×1) and, ideally, a second dispersion-correction scheme. Without these tests, the proposed hydrogen-mediated selection mechanism is not quantitatively supported.","section":"Section 2.4, Computational Details (Eqs. 1–4, Fig. 4, Fig. S10)"},{"comment":"The seamless stitching of aligned domains into a grain-boundary-free film is inferred from SEM morphology (uniform contrast and aligned triangular voids) rather than directly demonstrated. The Methods state that dark-field TEM imaging was performed, but no dark-field images are presented. Dark-field TEM over a statistically meaningful area, or an equivalent orientation-resolved technique, is needed to exclude the presence of small-angle grain boundaries or minority anti-parallel domains within the continuous film. This is load-bearing for the 'single-crystalline' claim and for the comparison in Table S1.","section":"Section 2.3, Figure 3b, and Methods"}],"minor_comments":[{"comment":"The notation '[11!0]', '[11!1]', and '[11!2!]' appears garbled; the overline notation for negative Miller indices should be typeset correctly to avoid confusion with exclamation marks.","section":"Throughout"},{"comment":"The group symmetry symbols 'C2V' and 'Cs or CV' should be written in standard form as C2v and Cs (or C1), with proper subscripts and italics.","section":"Section 2.1"},{"comment":"Equations (1) and (2) are typeset illegibly due to symbol encoding issues; please ensure all variables (E_B, E_f, E_total, E_Ge, E_hBN, A, L, ε_edge) render correctly and unambiguously.","section":"Computational Details"},{"comment":"The abstract states 'wafer-scale single crystalline hBN films' while the conclusion says 'large-scale hBN monolayer films with uni-directional crystallography'; the wording should be aligned with the actual evidence presented and with the revised claims made in response to Major Comment 1.","section":"Abstract and Conclusion"},{"comment":"The MoS2 FET mobility and subthreshold swing values (12 to 15.8 cm2/V·s and 2.3 to 1.2 V/dec) and the graphene mobility values would benefit from reporting the number of devices measured and an uncertainty estimate, since single numbers are presented without error bars.","section":"Section 5 (Device results)"},{"comment":"The XRD data in Fig. S11b are obtained on an assembled 10-layer hBN stack, while the main text and Figs. 5d,e describe bilayer assemblies; the relationship between these two sample types should be clarified so the reader understands the scope of the stacking-order demonstration.","section":"Figure S11 caption"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope and the local alignment evidence plus the device demonstrations are solid. The main risk is that the 'wafer-scale single-crystalline' headline claim is not backed by wafer-scale orientation-sensitive data; I would like the editor to impress on the authors that this gap is load-bearing, not cosmetic. The DFT convergence issue is also important because the proposed mechanism hinges on sub-0.01 eV energy differences. With these addressed, the paper could be suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a read. The paper's real result is new: unidirectionally aligned monolayer hBN grown on vicinal Ge(110), with orientation selectable by hydrogen partial pressure. The non-monotonic χalign(PH2) curve and the LEED switching between 6-fold and 3-fold are convincing, and the mechanical exfoliation plus controlled stacking of the assembled bilayers is a useful demonstration. For the 2D growth community, this is a meaningful step beyond prior Ge work, which only achieved aligned (not unidirectional) growth. The DFT story—H passivation on the terrace, de-passivation at the step edge allowing Ge–N bonds—is a plausible mechanism and consistent with the experiment, even if the numbers are rough.\n\nThe weak point is exactly where the stress-test note lands. The abstract and Table S1 claim wafer-scale single-crystalline hBN over a 2-inch wafer, but the orientation-sensitive data are local: LEED over ~1 mm² (stated in Methods) and one SHG map at 1 µm scale. The SEM and transmission spectra show morphology and thickness, not lattice orientation. The 97% alignment reported in Figure S4 for partial coverage means ~3% anti-parallel grains, and nothing in the current data rules out that such domains persist outside the probed regions of the full film. That is a load-bearing gap, because the paper's headline advantage over layer-by-layer assembly of polycrystalline film is precisely single crystallinity at scale.\n\nAlso worth flagging, but minor: the DFT uses a 1×1×1 k-point mesh and three-layer Ge slab, and the key energy differences on the flat terrace are below 0.01 eV—those specific numbers should not be leaned on. The formation-energy comparison for the step edge is more robust qualitatively, though Eq. 2 imports an edge energy from the authors' earlier work. Neither of these is fatal; they are secondary to the experimental demonstration.\n\nMy verdict: the core claim—unidirectional hBN on Ge with H2-controlled alignment—holds up as a new result. The wafer-scale single-crystal statement needs more evidence before it becomes part of the permanent record. A serious referee should see it; I would send it to peer review, but with a request for wafer-scale orientation mapping (SHG or other orientation-sensitive probe over cm scale), a clear statement of alignment statistics on full films, and a soft-pedaling of the DFT quantitative precision. I'd cite it for the growth mechanism, not yet for wafer-scale single crystallinity.","headline":"The unidirectional hBN-on-Ge growth and its hydrogen-pressure control are solid, but the wafer-scale single-crystal claim outruns the evidence: only 1 mm² LEED and a 1 µm SHG map support it.","tokens_in":15024,"tokens_out":2181,"would_cite":true,"duration_ms":26787,"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 reports that tuning hydrogen partial pressure during CVD growth on a stepped Ge(110) surface yields wafer-scale, monolayer, single-crystalline hexagonal boron nitride films that can be mechanically exfoliated and stacked in…","keywords":["hexagonal boron nitride","epitaxy","hydrogen passivation","chemical vapor deposition","van der Waals assembly","single crystal","vicinal Ge(110)","stacking order"],"falsifier":"Grow hBN on the [001]-miscut Ge(110) substrate while sweeping PH2 around the claimed optimum and use high-resolution electron microscopy or X-ray photoelectron spectroscopy to look for direct Ge-N bonds specifically at step edges; if no Ge-N bonding appears at the pressure where alignment reaches nearly 100%, the mechanism collapses. Alternatively, recompute the θ=0° versus θ=60° formation-energy difference with a converged k-point mesh and thicker Ge slabs: if θ=0° is not the global minimum at the stated hydrogen chemical potential, the thermodynamic-selection claim is not supported.","tokens_in":13978,"feed_emoji":"🧊","tokens_out":6683,"duration_ms":62610,"temperature":0.7,"pith_summary":"The paper reports a way to grow hexagonal boron nitride (hBN) as a continuous monolayer film that is single-crystalline across a 2-inch wafer, on a germanium substrate cut at a small angle so that atomic steps guide every crystal domain to point the same way. The key lever is hydrogen partial pressure during growth: too little or too much hydrogen leaves domains pointing in two opposite directions, but at an intermediate pressure almost all domains align. Because the film sits on hydrogen-passivated germanium rather than on a metal, it can be mechanically exfoliated and stacked layer by layer with pristine interfaces. The authors demonstrate controlled rhombohedral and hexagonal bilayer stacking, and use the films as dielectrics that reduce trap states in graphene and MoS2 devices. The claim matters because wafer-scale single-crystal hBN with clean assembly would give electronics a scalable insulating buffer layer.","feed_headline":"Hydrogen tuning grows wafer-scale single-crystal hBN on germanium","feed_subtitle":"Optimized H2 pressure aligns every hBN domain on stepped Ge, enabling clean layer-by-layer assembly.","key_machinery":"The central mechanism is hydrogen-passivation-controlled Ge-N edge bonding at atomic steps. On the vicinal Ge(110) surface, the step edge provides Ge atoms with dangling bonds that can bond to N-terminated zigzag edges of hBN. The paper's DFT calculations show that when hydrogen pressure is high, van der Waals interaction gives nearly equal stability to two orientations (θ=0° and 60°); when hydrogen is scarce, the terrace and edges behave differently. At an intermediate chemical potential, the equilibrium favors direct Ge-N bonds at the step and H-passivated terraces, creating a deep energy minimum at θ=0° and a maximum at 60°, so every domain stitches into one orientation. This equilibrium binding-energy difference, rather than growth kinetics, is what the paper proposes as the orientation-selection mechanism.","core_discovery":"The central discovery is that step-directed epitaxy of hBN on vicinal Ge(110) can be made uni-directional by tuning the hydrogen chemical potential, and that the resulting monolayer film is single-crystalline, wafer-scale, and mechanically exfoliable. On a Ge(110) surface miscut toward [001], the substrate symmetry is reduced so that atomic steps run along one direction; triangular hBN grains nucleate with an N-terminated zigzag edge along the step. At the optimum H2 partial pressure, DFT-based free-energy calculations indicate that hydrogen passivates the Ge terraces but leaves the step-edge Ge atoms able to form direct Ge-N bonds with the hBN edge, making one crystallographic orientation the global energy minimum. Experimental LEED and SHG show 3-fold, uni-directional symmetry at this pressure, and the merged film has no measurable grain boundaries. The same films can be exfoliated and assembled into bilayers with chosen stacking order, verified by TEM diffraction and ARPES band splitting.","pith_inferences":["The DFT energy differences that select θ=0° over θ=60° are below 0.01 eV with a 1x1x1 k-point mesh; a converged calculation could shift the preferred orientation, so the quantitative predictive range is untested.","The same chemistry—hydrogen-passivated terraces plus chemically active step edges—might extend to other nitride monolayers on hydrogen-passivating semiconductor surfaces.","Controlling step bunching during growth, which the paper associates with wrinkle formation, could yield even flatter films.","The wafer-scale single-crystal claim is supported by SHG uniformity and LEED over about 1 mm2; full-wafer diffraction mapping would be a natural confirmation."],"forward_implications":["If correct, wafer-scale single-crystal monolayer hBN becomes available on a semiconductor substrate rather than only on metals.","Because the film sits on hydrogen-passivated germanium, it can be mechanically exfoliated and stacked layer by layer with pristine interfaces, enabling controlled stacking orders such as rhombohedral and hexagonal.","Controlled rhombohedral stacking would make large-area ferroelectric and nonlinear-optical hBN stacks practical.","The hBN dielectric suppresses interfacial trap density by roughly an order of magnitude in graphene FETs and improves MoS2 subthreshold swing, supporting use as a scalable insulator.","The non-monotonic dependence of alignment on PH2 gives a growth knob for switching between bi-directional and uni-directional crystal populations."],"supporting_citations":[{"why":"Supplies the mechanical exfoliation and layer-by-layer assembly method used to make the pristine hBN interfaces.","marker":"[18]"},{"why":"Prior demonstration of aligned hBN growth on germanium that did not achieve uni-directional alignment, setting the baseline this paper improves on.","marker":"[19]"},{"why":"Benchmark wafer-scale single-crystal hBN growth on copper, the metal-substrate approach that cannot be dry-transferred.","marker":"[9]"},{"why":"Another metal-substrate single-crystal hBN growth reference that the germanium approach is compared against.","marker":"[10]"},{"why":"Provides the epitaxy framework explaining why substrate symmetry must be broken for uni-directional 2D crystal growth.","marker":"[13]"},{"why":"Theoretical basis for step-directed growth of 2D monocrystals on vicinal surfaces.","marker":"[15]"},{"why":"Establishes the optical second-harmonic generation method used to verify the film's crystallographic symmetry and uniformity.","marker":"[21]"},{"why":"Identifies how TEM diffraction distinguishes rhombohedral and hexagonal stacking orders in hBN bilayers.","marker":"[23]"},{"why":"Provides the reference hBN band structure used to interpret the ARPES band splitting in assembled bilayers.","marker":"[24]"}],"fun_headline_variants":["Hydrogen pressure steers hBN into single crystal","Step-directed epitaxy yields uni-directional hBN","Hydrogen tuning aligns every hBN domain on Ge","Wafer-scale single-crystal hBN via stepped Ge"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that, at the right hydrogen pressure, the edges of the germanium steps shed their hydrogen and bond directly to nitrogen atoms along the hBN edges while the flat terraces stay hydrogen-covered, and that this energy difference—rather than growth speed—is what makes every crystal point the same way.","fun_headline_variants_meta":{"raw":{"variants":["Hydrogen pressure steers hBN into single crystal","Step-directed epitaxy yields uni-directional hBN","Hydrogen tuning aligns every hBN domain on Ge","Wafer-scale single-crystal hBN via stepped Ge"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000987,"raw_usage":{"total_tokens":4215,"prompt_tokens":1001,"completion_tokens":3214,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":617,"completion_tokens_details":{"reasoning_tokens":3150}},"tokens_in":617,"tokens_out":3214,"duration_ms":23797,"temperature":1.0,"reasoning_tokens":3150,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:03:14.508127+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Grow hBN on the [001]-miscut Ge(110) substrate while sweeping PH2 around the claimed optimum and use high-resolution electron microscopy or X-ray photoelectron spectroscopy to look for direct Ge-N bonds specifically at step edges; if no Ge-N bonding appears at the pressure where alignment reaches nearly 100%, the mechanism collapses. Alternatively, recompute the θ=0° versus θ=60° formation-energy difference with a converged k-point mesh and thicker Ge slabs: if θ=0° is not the global minimum at the stated hydrogen chemical potential, the thermodynamic-selection claim is not supported.","supporting_citations":[],"review_version":2}