{"id":"9af62697-c30d-4213-af61-15a6aa4cd2a2","arxiv_id":"2507.16361","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Adding molybdenum interstitials between MoS2 layers during growth stabilizes a fault-free hexagonal stacking order in bilayer films.","lead":"Researchers grew bilayer molybdenum disulfide with extra molybdenum atoms trapped between the layers, which lock the layers into a single crystal stacking order. The result could make multilayer 2D semiconductors more uniform and reliable for future electronics, since stacked layers usually twist or slide.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The load-bearing causal claim that Mo interstitials select 2H stacking is not isolated from other Mo-rich-growth covariates, and the acknowledged DFT N=7 vs N=11 mismatch leaves the mechanism quantitatively unsupported; verdict stays conditional.","rationale":"Reader's conditional verdict is well placed. The empirical facts are credible: STEM contrast, XPS broadening, ARPES and transport are internally consistent enough for a moderate-confidence review. The specific obstruction to accepting the headline claim is not whether interstitials exist—they do under Mo-rich conditions—but whether they are the cause rather than a marker of the Mo-rich growth regime. Because only two growth conditions are compared, the correlation cannot exclude alternative Mo-rich-specific mechanisms. The DFT crossover mismatch is an independent sign that the proposed mechanism is not fully specified: the energy argument says the experimentally observed interstitial density is below the threshold needed to stabilize the AA' phase, and the paper's fix is an untested hypothesis. I would keep the verdict CONDITIONAL/UNCHANGED: the paper should be published only if the causal role is demonstrated by an independent-variation experiment or by post-growth interstitial insertion. I do not see a basis for rejection, because the growth result, if confirmed, is significant and the main missing piece is a specific, testable causal demonstration.","tokens_in":17495,"tokens_out":7814,"duration_ms":97224,"concrete_test":"Perform a four- or five-point dose-response series in the same MOCVD reactor in which only the Mo(CO)6 carrier flow is varied, keeping PS, H2 flow, total pressure, temperature, growth time, and substrate miscut fixed (compensating total Ar flow). For each sample, quantify both the interstitial areal density from aberration-corrected STEM and the bilayer 2H/3R phase fraction from DF-TEM. If the 2H fraction rises monotonically with measured interstitial density and samples with low interstitial density are always phase-mixed, the causal role survives; if any sample with high 2H fraction has low interstitial density, or if phase fraction changes at fixed interstitial density, the central claim is falsified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that Mo atoms inserted in the vdW gap are the operative cause of 2H stacking selection and stacking-fault-free growth. What is actually shown is a single two-condition correlation: Mo-rich growth (PMo/PS ≈ 4×10⁻²) gives ~99% 2H bilayers with bright interstitial-like atoms, while Mo-poor growth (PMo/PS ≈ 2×10⁻³) gives ~64% 2H with abundant solitons. The Mo-rich condition differs from the Mo-poor one in PMo/PS, absolute Mo and S flows, and therefore in growth chemistry that need not be mediated by interstitials: nucleation density on the first layer, edge termination, growth rate, and carbon incorporation from Mo(CO)6 all change together. No experiment varies interstitial areal density while holding these other variables fixed. The DFT part does not fully repair this gap: it shows that if a Mo interstitial is present, T-AA' becomes thermodynamically stable only for islands larger than N=11 unit cells, whereas the measured density (≈1 interstitial per 40 unit cells) corresponds to N≈7. The text acknowledges this and postulates that additional interstitials may enter after AA' formation, but provides no evidence for that step. The load-bearing causal mechanism therefore rests on an untested assumption, not on a demonstrated dose-response or an independent manipulation of interstitial concentration.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports the MOCVD growth of single-crystalline, 2H-stacked bilayer MoS2 on miscut c-plane sapphire under Mo-rich conditions, and attributes the stacking selectivity and post-transfer structural robustness to Mo atoms intercalated at the van der Waals gap. The authors use HAADF-STEM to identify Mo interstitials at a single sublattice site, with areal density about one per ~40 unit cells in monolayers and ~2.4e13 cm^-2 in bilayers. DF-TEM shows >99% of bilayer domains are 2H under Mo-rich growth, versus ~64% under Mo-poor growth, and the Mo-rich bilayers survive transfer without forming strained solitons. Raman shows an E2g1 doublet, XPS shows a lower-binding-energy Mo shoulder with an area fraction (~11.5%) close to the STEM interstitial fraction (~11.3%), ARPES shows a single-crystal band structure with a pronounced G1st peak, and hBN-gated FETs reach on/off ratios near 1e9. DFT calculations identify two stable interstitial configurations (O-AB' and T-AA') and predict that T-AA' becomes thermodynamically favored for islands larger than N = 11 unit cells. The manuscript acknowledges that the measured average island size per interstitial (N ~ 7) is below this threshold and hypothesizes that additional interstitials may enter after AA'-host formation.","tokens_in":17799,"tokens_out":3947,"duration_ms":48895,"significance":"If the causal role of interstitials is firmly established, this work offers a new route to deterministic stacking control in vdW multilayers, addressing a long-standing challenge in 2D materials synthesis. The experimental characterization is broad and internally consistent: the XPS/STEM fraction match, the ARPES single-crystal signature, and the transport on/off ratio all support the quality and phase purity of the Mo-rich bilayers. The DFT calculations are an independent theoretical input, not fitted to the stacking fractions, so the central result is not circular. However, the load-bearing causal claim - that Mo interstitials are the operative cause of 2H stacking selection and of soliton-free robustness - is not isolated from other growth covariates, and the quantitative DFT threshold (N=11) does not match the observed interstitial density (N~7). These gaps make the mechanism currently a plausible hypothesis rather than a demonstrated result. The paper would be suitable for publication after these points are addressed experimentally or the claims are appropriately qualified.","major_comments":[{"comment":"The central causal claim that Mo interstitials select 2H stacking rests on a single comparison between Mo-rich and Mo-poor growth conditions. These conditions differ simultaneously in PMo/PS (4e-2 vs 2e-3), in absolute Mo and S precursor flows, and in carrier gas flow rates (Ar 500 vs 90 sccm, H2 2 vs 4 sccm). Consequently, the observed difference in stacking ratio (99% vs 64%) could in principle be caused by changes in growth rate, nucleation density, edge termination, or carbon incorporation from Mo(CO)6 decomposition, rather than by the interstitial atoms themselves. No experiment varies the interstitial areal density while holding these other parameters fixed. The paper should either provide a dose-response experiment in which only the Mo/S ratio is varied across several intermediate values with measurement of interstitial density and stacking fraction, or directly demonstrate that the 2H selection is mediated by interstitials, for example by post-growth interstitial insertion into a Mo-poor bilayer followed by annealing and stacking characterization. Without such evidence, the causal mechanism remains correlational.","section":"Results and Discussion, first paragraph; Figs. 2a-b and S7"},{"comment":"The DFT prediction that T-AA' becomes thermodynamically stable only for N > 11 is not consistent with the experimentally measured interstitial density (one per ~40 unit cells, corresponding to an average N ~ 7 per interstitial in a single-island picture). The manuscript acknowledges this mismatch and hypothesizes that 'additional Mo interstitials may be introduced following the formation of the AA'-stacked host,' but provides no evidence for such post-formation insertion. This is a load-bearing point because the proposed nucleation-seed mechanism (Fig. 3d) requires the interstitial to stabilize the AA' stacking from the initial stage; if the energy crossover only occurs at larger island sizes, the mechanism cannot explain the observed highly selective 2H formation. The authors should either present a revised DFT analysis that accounts for multiple interstitials per island at the measured density, or calculate the kinetic barrier for the O-AB' to T-AA' slide to show that it is accessible under growth conditions. Merely noting the discrepancy and invoking an untested hypothesis is insufficient for the central claim.","section":"Results and Discussion, Growth mechanism; Fig. 3c, p. 7"},{"comment":"The absence of strained solitons after transfer is presented as evidence that interstitial covalent bonds strengthen the interlayer coupling. However, the comparison is between Mo-rich (2H, with interstitials) and Mo-poor (mixed 2H/3R, without interstitials) samples, so the suppression of solitons could equally be a consequence of the uniform 2H stacking itself rather than a direct effect of the interstitials. To support the causal inference, the authors should compare a Mo-rich bilayer that has been mechanically strained (e.g., by bending or indentation) with a Mo-poor bilayer subjected to the same strain, and show that soliton nucleation is suppressed or its onset is delayed in the interstitial-containing sample. Without such a test, the robustness claim is not specifically attributable to the interstitials.","section":"Results and Discussion, Robustness after transfer; Figs. 2e-f"}],"minor_comments":[{"comment":"Please clarify the Brillouin zone sampling: the text 'Γ, K-point was used' is ambiguous; it should state the k-point grid (e.g., a Γ-centered mesh) and the number of k-points.","section":"Methods, DFT calculations"},{"comment":"The caption labels the three curves as grey, orange, and red, but the y-axis label and units are not defined; please add explicit definitions of ΔEMoi, ΔEBulk, and ΔEF in the caption or in the main text.","section":"Fig. 3c caption"},{"comment":"The statement that 'AA'-MoiB is the inversion counterpart of AA'-MoiA and thus exhibits the same formation energy under free-standing conditions' should be reconciled with the earlier claim that only AA'-MoiA is kinetically accessible; please make this distinction explicit in the main text.","section":"Fig. S8 and text p. 7"},{"comment":"The green arrow points to an extra spectral feature that is attributed to AA stacking at solitons; please clarify that this assignment is based on reference 42 and provide the calculation level, as the energy position may depend on functional and vdW correction.","section":"Fig. 5c and related text"},{"comment":"The assignment of the lower-energy Raman peak to E2g1(-) and the interpretation of the 5.7 cm^-1 red-shift as ~1.6% tensile strain should be accompanied by a reference for the strain-phonon coupling coefficient in bilayer MoS2, or a direct strain calibration.","section":"Raman analysis, Fig. 4a"},{"comment":"The fitting of the Mo 3d shoulder peaks would benefit from a description of the fitting procedure, including peak shapes, constraints, and error estimates; please also discuss why alternative assignments (e.g., MoOx or Mo-C species from the MOCVD precursor) are ruled out.","section":"XPS analysis, Fig. 4b"},{"comment":"Please specify the number of independently grown samples and total analyzed area for the 'over 99% (159 out of 160)' statistic, and state the criteria for classifying a domain as 2H or 3R.","section":"DF-TEM statistics, Fig. S6"},{"comment":"The term 'hexagonal bilayer' is used throughout to mean 2H stacking; since 'hexagonal' also refers to the crystal system, consider defining 'hexagonal phase' at first use to avoid ambiguity with the lattice symmetry.","section":"Abstract and Conclusions"}],"recommendation":"major_revision","confidential_remarks":"This is a technically strong manuscript from a leading group, with a comprehensive set of characterization tools. The main weakness is the causal identification of the interstitial mechanism: the experimental correlations are compelling but do not exclude confounding variables in the Mo-rich growth condition, and the DFT threshold mismatch weakens the proposed mechanism. I see no reason to reject; the paper can be made publishable by either adding a targeted experiment (e.g., varying interstitial density at fixed Mo/S ratio, or post-growth interstitial insertion) or by carefully qualifying the mechanistic claims and reducing the emphasis on the unverified post-AA' interstitial insertion hypothesis. I would ask the editor to allow a major revision with the specific request to address the three major comments, especially the confound issue and the N=7 vs N=11 discrepancy."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about this paper. First, the headline result is real: Mo-rich MOCVD produces bilayer MoS2 that is overwhelmingly 2H-stacked, and the Mo interstitials are directly visible in STEM. It's the first experimental demonstration that a specific interstitial site can steer stacking in a semiconducting TMD—prior work was theoretical or in metallic systems. The XPS and STEM interstitial fractions agree to better than a percent, which is a solid internal check. Second, the causal story is still partly inferred. Mo-rich and Mo-poor growth differ in several ways at once—precursor ratio, absolute flows, growth rate, nucleation density, carbon incorporation—so the experiments don't isolate interstitial density as the operative variable. That's the main soft spot.\n\nThe DFT calculation is a genuinely independent input, not fit to the data, and it points in the right direction: interstitials stabilize the AA' 2H-like configuration. But the predicted threshold (N=11) is above the measured density (N≈7). The paper acknowledges this and floats a post-nucleation insertion hypothesis without supporting evidence. That's a real quantitative gap, though it doesn't sink the qualitative claim.\n\nWhat the paper does well: triangulation. DF-TEM phase statistics, Raman peak splitting that survives transfer, XPS oxidation-state signatures, ARPES interlayer band features, and a DFT mechanism all line up. The transport data is thinner—one device per condition—so I'd read the mobility and on/off claims as supportive, not decisive. And 'stacking-fault-free' is a bit strong for a survey of 160 grains and transferred regions; it's 'stacking-controlled across the sampled area,' which is already good.\n\nBottom line: solid, honest materials science with a real advance. The caveats are about whether the mechanism is the interstitials per se or something else about Mo-rich growth; that warrants referee scrutiny, not rejection. I'd send it to review.","headline":"First clear experimental demonstration that Mo interstitials can steer 2H stacking in bilayer MoS2, though the mechanism is correlated rather than fully isolated from other Mo-rich-growth factors.","tokens_in":18372,"tokens_out":3537,"would_cite":true,"duration_ms":34160,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Growing bilayer MoS2 with molybdenum atoms bridging the layers locks the film into a single hexagonal stacking order that survives transfer.","keywords":["van der Waals materials","MoS2 bilayer","interstitial atoms","stacking order","layer-by-layer epitaxy","stacking faults","MOCVD","2H phase"],"falsifier":"Grow bilayer MoS2 under Mo-rich conditions but remove or prevent the interstitials - for instance by annealing after growth to drive them out, or by using a Mo precursor that does not leave interstitial atoms - and check whether the 2H phase fraction and soliton-free transfer are preserved. If the film remains more than 99% 2H and soliton-free without interstitials, the anchor mechanism is not the active agent.","tokens_in":79,"feed_emoji":"🔗","tokens_out":9578,"duration_ms":192735,"temperature":0.7,"pith_summary":"This paper reports a growth route for bilayer molybdenum disulfide (MoS2) that produces a single stacking phase across the whole film, avoiding the twisting, sliding, and buckling that normally plague van der Waals crystals, whose layers are held together by weak forces. The key move is to grow the second layer under Mo-rich conditions so that individual molybdenum atoms become trapped between the two layers. These interstitial Mo atoms form covalent bonds with sulfur atoms on both layers, acting as atomic anchors that pin the layers into the hexagonal (2H) stacking order. The paper shows that more than 99% of bilayer domains adopt this order and that the structure survives transfer without forming stacking faults or strained solitons. If correct, the work would turn stacking control from a thermodynamic gamble into a chemically directed choice for a technologically relevant semiconductor.","feed_headline":"Mo anchors between layers make bilayer MoS2 stacking-fault-free","feed_subtitle":"Metal-rich growth places molybdenum bridges in the van der Waals gap, pinning the hexagonal phase even after transfer.","key_machinery":"The carrying object is the interstitial Mo atom in the van der Waals gap, sitting on the Mo sublattice of the bottom layer and bonded to sulfur atoms from both layers. In the final structure it occupies a tetrahedral site between AA'-stacked layers (T-AA'), after passing through an octahedral configuration (O-AB') during nucleation. The interstitial acts as a nucleation seed for the second layer, lowering the barrier to forming the hexagonal phase; as the grain grows, the top layer translates into the AA' registry. Because sliding or twisting the layers would require breaking the interstitial's covalent bonds, the anchor raises the energy cost of stacking defects to roughly 1.7 eV, far above the ~1 meV per unit cell differences between polytypes without interstitials, and blocks the formation of strained solitons.","core_discovery":"The central claim is that Mo interstitials incorporated during layer-by-layer growth are the active agents that deterministically select and stabilize the hexagonal (2H) stacking of bilayer MoS2. Under Mo-rich growth, atomic-resolution STEM shows single Mo atoms lodged at one sublattice site, vertically aligned with Mo atoms in the bottom layer, at a density of about one interstitial per 40 unit cells. Dark-field TEM finds 159 of 160 bilayer grains in the 2H phase, with no strained solitons even after transfer; under Mo-poor growth the same measurement yields a roughly 64/36 mixture of 2H and 3R phases with abundant solitons. DFT calculations attribute this to the interstitial's large stacking-selection energy, about 1.7 eV per atom, which dwarfs the ~1 meV per unit cell energy difference between polytypes without interstitials. The paper further links the interstitials to a split, red-shifted Raman E2g1 mode, an XPS shoulder from less-oxidized Mo, a single-crystalline ARPES band structure with a strong interlayer-hybridized valence band, and field-effect transistors with an on/off ratio near $10^{9}$.","pith_inferences":["If interstitials are the true cause, then removing them after growth, for example by annealing or electrochemical extraction, should restore stacking-fault formation and soliton mobility in the same film.","A cleaner test of the mechanism would be to vary interstitial density at a fixed Mo:S ratio, using temperature or a secondary flux, to separate the anchor effect from other consequences of Mo-rich growth.","The ~1.6% tensile strain implied by the Raman red-shift suggests the anchored bilayers are under built-in stress; this could shift band edges or exciton energies in ways that go beyond the paper's ARPES and transport measurements.","The same interstitial-anchor idea might be used to stabilize other stackings, such as rhombohedral, by choosing an interstitial species or site with a different bonding preference, which would give phase-selective growth in other TMD families."],"forward_implications":["Bilayer MoS2 can be grown with deterministic 2H stacking over large areas: 159 of 160 grains in the checked regions were hexagonal, compared to about 64% under Mo-poor conditions.","The resulting films remain single-phase and free of strained solitons after transfer, indicating that the interlayer registry is locked in place.","Transistors made from these films show stable operation with on/off ratios approaching 10^9, whereas devices from Mo-poor, mixed-stacking films are markedly worse.","The same interstitial-guided growth can set the stacking of thicker layers (a trilayer, where the third layer takes rhombohedral order) and extends to WS2 grown under W-rich conditions, suggesting the strategy is not limited to bilayers of MoS2."],"supporting_citations":[{"why":"Theoretical work predicting that interstitial Mo impurities can switch MoS2 bilayer stacking; the paper's growth mechanism builds on this proposal.","marker":"24"},{"why":"Report of covalently bonded self-intercalated TMDs grown under metal-rich conditions; motivates the metal-rich growth window used here.","marker":"12"},{"why":"Step-directed epitaxy of wafer-scale monolayer MoS2 on sapphire; provides the platform for the layer-by-layer growth.","marker":"26"},{"why":"Prior epitaxial bilayer MoS2 growth with uniform nucleation; serves as the baseline that this paper's stacking control improves on.","marker":"14"},{"why":"Calculated band structures of differently stacked bilayer TMDs; used to identify the interlayer-hybridized valence band in ARPES.","marker":"42"},{"why":"Reference ARPES band dispersion for bilayer MoS2, used for comparison with the measured single-crystalline bands.","marker":"39"},{"why":"Observation of Raman E2g1 splitting in Li-intercalated bilayer MoS2; supports assigning the split peaks to interlayer intercalants.","marker":"32"},{"why":"Demonstration of stacking-order-dependent dark-field contrast in bilayer van der Waals crystals; used to quantify 2H versus 3R domains.","marker":"21"}],"fun_headline_variants":["Mo interstitials anchor layers to yield fault-free MoS2","Interstitial Mo pins stacking of bilayer MoS2","Mo bridges between layers lock in the 2H phase","Fault-free bilayer MoS2 grown with Mo interstitials","Single Mo atoms between layers stop stacking faults"],"cache_read_input_tokens":20480,"weakest_assumption_plain":"The paper's argument depends on the claim that Mo interstitials, not some other feature of Mo-rich growth, are what determines the stacking order and its robustness.","fun_headline_variants_meta":{"raw":{"variants":["Mo interstitials anchor layers to yield fault-free MoS2","Interstitial Mo pins stacking of bilayer MoS2","Mo bridges between layers lock in the 2H phase","Fault-free bilayer MoS2 grown with Mo interstitials","Single Mo atoms between layers stop stacking faults"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000714,"raw_usage":{"total_tokens":3243,"prompt_tokens":1009,"completion_tokens":2234,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":625,"completion_tokens_details":{"reasoning_tokens":2153}},"tokens_in":625,"tokens_out":2234,"duration_ms":15216,"temperature":1.0,"reasoning_tokens":2153,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:11:21.233481+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Grow bilayer MoS2 under Mo-rich conditions but remove or prevent the interstitials - for instance by annealing after growth to drive them out, or by using a Mo precursor that does not leave interstitial atoms - and check whether the 2H phase fraction and soliton-free transfer are preserved. If the film remains more than 99% 2H and soliton-free without interstitials, the anchor mechanism is not the active agent.","supporting_citations":[],"review_version":1}