{"id":"d01fab02-f595-43ab-b154-dab5c2fa1b68","arxiv_id":"2607.25963","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Stacking polarity in MoSe2/3R-MoS2 heterostructures tunes interlayer charge-transfer time (0.25 vs 0.37 ps) and interlayer exciton lifetime (40 vs 169 ps).","lead":"Using ultrafast pump–probe spectroscopy, this paper shows that flipping which face of a 3R-stacked MoS2 bilayer touches a MoSe2 monolayer changes the interlayer charge-transfer time from 0.25 ps to 0.37 ps and the interlayer exciton lifetime from 40 ps to 170 ps. The result suggests vertical stacking order could be a clean, chemistry-free knob for tuning carrier dynamics in 2D optoelectronics.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Directional claim rests on untested face assignment: 1L/M/X vs 1L/X/M labels are not independently established, so the 0.25 vs 0.37 ps ordering may be circular.","rationale":"The reader's weakest assumption—that the paper does not independently determine which face is M vs X, making the directional labeling potentially circular—is exactly the most load-bearing weakness in the central claim. The abstract and main text assert a specific, physically meaningful ordering: M-layer termination gives 0.25 ps CT, X-layer termination gives 0.37 ps, and this ordering is used to support the layer-polarization/wavefunction-overlap mechanism. Without independent face identification, this ordering could be an artifact of the labeling scheme; if the labels were swapped, the mechanism would predict the opposite ordering and the data would still look the same. The paper's own admission that the surfaces 'cannot be optically distinguished prior to stacking' flags this gap. I do not think the claim collapses entirely: the experiment still shows that opposite faces of a 3R bilayer yield different dynamics, and the 2H control rules out some trivial stacking-order effects. But the specific directional claim and the mechanism test are conditional on face identification. The reader's CONDITIONAL verdict is appropriate; my stress test does not move it. I considered the alternative concern that the 1.82 eV probe might be sensitive to M-layer population only after X→M relaxation, which could reinterpret the 0.37 ps as a relaxation time; this is real and mentioned in the reader's rationale, but it is secondary because the face-identification circularity would invalidate even the direction of the claim, while the relaxation ambiguity mostly affects the assignment of one number. Therefore the single most load-bearing concern is the unestablished face assignment, and a cross-sectional STEM or equivalent independent face characterization is the concrete test that would settle it.","tokens_in":10582,"tokens_out":7807,"duration_ms":74794,"concrete_test":"Independently determine the face identity of the same 3R MoS2 flake used in the experiment, without reference to the pump-probe dynamics. For example, prepare a cross-sectional lamella of the stacked sample with FIB and use high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) to image the atomic stacking at each interface, identifying whether the MoS2 layer adjacent to MoSe2 is the M layer (Mo atoms aligned over S of the adjacent layer) or the X layer. Then compare the independently assigned interface types with the reported τ_CT values: the region with the M-layer-adjacent interface should show 0.25±0.04 ps and the X-layer-adjacent interface 0.37±0.05 ps. If the independently identified M-face region corresponds instead to the 0.37 ps trace, or if the two regions cannot be distinguished by stacking at the interface, the directional stacking-polarity claim","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central directional claim—that 1L/M/X has τ_CT=0.25±0.04 ps and 1L/X/M has τ_CT=0.37±0.05 ps, with the M-layer-at-interface configuration being faster—requires knowing which face of the 3R MoS2 bilayer contacts MoSe2 in each of the two regions. The text explicitly states (paragraph after Fig. 1b) that 'the two surfaces of a 3R bilayer cannot be optically distinguished prior to stacking,' and no independent face identification is reported anywhere in the paper. The fabrication on a single MoS2 flake ensures the two interfaces use opposite faces, but it does not establish which face is the M layer and which is the X layer. Therefore the labels 1L/M/X and 1L/X/M are either arbitrary conventions or were inferred from the measured dynamics. If they were assigned after observing which region rose and decayed faster, then the reported ordering (M/X faster than X/M) is guaranteed by construction and cannot test the proposed wavefunction-overlap mechanism. This is a circularity in the load-bearing claim, not a cosmetic labeling issue: the specific quantitative values and the microscopic polarity assignment would both be unsupported. The fluence-dependence controls and the 2H inversion-symmetry check are good, but they do not resolve this face-identification gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports ultrafast pump–probe measurements of interlayer photocarrier dynamics in monolayer MoSe2/bilayer MoS2 heterostructures with 2H and 3R MoS2. For the 1L/2H system the charge-transfer (CT) rise is faster than the instrument response, while two regions of a single 3R MoS2 bilayer, labelled 1L/M/X and 1L/X/M, show CT times of 0.25±0.04 ps and 0.37±0.05 ps, respectively, and interlayer exciton lifetimes of ~40 ps and ~170 ps. The paper attributes these differences to stacking-induced layer polarization in 3R MoS2, which localizes the conduction-band minimum in the M layer and modulates interfacial wavefunction overlap. The study includes pump-fluence dependence and an inverted-2H control to support the interpretation.","tokens_in":11009,"tokens_out":11903,"duration_ms":116799,"significance":"If the face assignment is independently established, this would be a significant advance: vertical stacking polarity would provide a spatially uniform, chemically invariant control parameter for interlayer CT and recombination in TMD heterostructures. The same-flake fabrication for the two 3R configurations is a genuine strength, as is the fluence-series control showing linear, density-independent dynamics. The 2H inversion-symmetry check is a good internal control. However, the directional claim—M-layer-at-interface faster than X-layer-at-interface—currently rests on an unverified face identification, and the probe interpretation raises a further selectivity question. The underlying qualitative mechanism is grounded in prior literature on 3R MoS2 layer polarization, but the experimental evidence for the specific polarity assignment is incomplete.","major_comments":[{"comment":"The central directional claim (1L/M/X faster than 1L/X/M) requires an independent identification of which 3R MoS2 face contacts MoSe2 in each region. The paper states the two faces \"cannot be optically distinguished prior to stacking,\" yet no STEM, AFM, or other assignment is reported. If the labels were chosen after observing which region was faster, the 0.25 vs 0.37 ps ordering is circular. Provide an independent face assignment or recast the claim non-directionally.","section":"Fig. 1(b) and following paragraph; Fig. 3(c),(e)"},{"comment":"The 1.82 eV probe is degenerate for K-valley transitions of both M and X layers, as stated. The paper argues the signal is \"primarily associated with electrons in the M layer\" once quasi-equilibrium is established, but the CT time is extracted from the rising edge, which is precisely the pre-equilibrium regime. In 1L/X/M, electrons may first enter the interfacial X layer and subsequently relax to the lower M layer; the measured rise would then contain interlayer relaxation as well as CT. The comparison of 0.25 ps (direct transfer to M) with 0.37 ps (transfer to X plus relaxation) may conflate two processes. Please demonstrate that the X-layer contribution to the transient absorption is negligible on the 0.1–0.5 ps scale, or model the two-level population dynamics.","section":"Pump–probe description after Fig. 2; Fig. 3(c)-(f)"},{"comment":"The claim that \"MoS2 is not directly photoexcited at 1.59 eV\" is not self-evident for bilayer MoS2, whose indirect gap lies below 1.59 eV. Although the MoSe2 exciton absorption dominates, a small direct absorption in the MoS2 bilayer could contribute to the same 1.82 eV probe response and would be indistinguishable from CT. Since the 2H vs 3R comparison is central, please include a control measurement of the bare MoS2 bilayer (2H and 3R) under identical pump/probe conditions, or quantify the MoS2 absorption at 1.59 eV.","section":"Pump selectivity paragraph after Fig. 2"}],"minor_comments":[{"comment":"The fits in Fig. 3(c) and (e) report τCT = 0.22 ps and 0.38 ps, while Fig. 4 reports averages of 0.25±0.04 ps and 0.37±0.05 ps. Reconcile these values or state which dataset is shown in Fig. 3.","section":"Fig. 3 vs Fig. 4"},{"comment":"Please state how the 3R polytype and bilayer thickness were confirmed (e.g., Raman, SHG, PL line-shape), since the entire effect depends on 3R stacking.","section":"Sample characterization"},{"comment":"Report the number of independent measurements and a statistical test for the CT-time difference between 1L/M/X and 1L/X/M. The nominal separation of 0.12 ps is about 1.9σ of the reported errors, so a simple statement of overlapping errors would be more convincing.","section":"Statistics"},{"comment":"Ref. 34 is missing volume/page/DOI information, and Ref. 27 gives \"Bellus, B. Z.\" whereas Ref. 21 gives \"Bellus, M. Z.\"; please check the spelling.","section":"References"},{"comment":"The phrase \"global control parameter\" is strong for a demonstration on one 3R flake. Please acknowledge the single-flake nature or provide repeat measurements on additional flakes.","section":"Generality"}],"recommendation":"major_revision","confidential_remarks":"The same-flake design and the fluence controls are strengths, but the face-assignment circularity is serious and load-bearing. The probe-selectivity and direct-excitation concerns are additional correctness risks that should be addressed with controls or modeling. I do not think the current manuscript is publishable without either an independent face identification or a non-directional reframing, plus the requested control experiments."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The new thing here is real: opposite faces of a 3R MoS2 bilayer give measurably different charge-transfer times (0.25 vs 0.37 ps) and interlayer exciton lifetimes (40 vs 169 ps) in MoSe2/MoS2 heterostructures, while the 2H control stays faster than the instrument response. The fluence-dependence measurements are clean and the 2H stacking-inversion check is a good internal control. That part is solid.\n\nThe soft spot is exactly what the stress-test note flags. The paper never independently identifies which face is M and which is X. The text says the two faces cannot be optically distinguished prior to stacking, and the labels 1L/M/X and 1L/X/M appear to be assigned after measuring which device is faster and slower. If that is what happened, the ordering (M/X fast, X/M slow) is built into the labeling and cannot test the wavefunction-overlap mechanism. What is not circular is the empirical fact that the two faces differ; that stands regardless of which face is which.\n\nThere is a secondary, smaller concern: the 1.82 eV probe sees both layers' K-valley populations, so the extracted \"CT time\" could include a contribution from X-to-M relaxation. That would blur the physical meaning of the numbers, though it probably does not erase the qualitative difference between the two faces.\n\nBottom line: the paper deserves a serious referee, but it needs a substantial revision on the face-identification issue. The authors could either perform an independent face assignment (there are published methods, e.g., using layer-polarized exciton energies or the built-in dipole) or explicitly reframe the result as \"opposite faces differ\" without naming which is M and which is X. I would not cite the directional mechanism until that is resolved, but I would note the empirical observation as worth knowing.","headline":"Opposite faces of a 3R MoS2 bilayer really do show different CT and lifetimes, but the M-vs-X assignment is not independently established, so the directional claim is circular.","tokens_in":11472,"tokens_out":3665,"would_cite":false,"duration_ms":39107,"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":"Vertical stacking polarity sets the speed of interlayer charge transfer and the lifetime of interlayer excitons in MoSe2/MoS2 heterostructures.","keywords":["transition metal dichalcogenide","carrier dynamics","rhombohedral stacking","van der Waals heterostructure","transient absorption","charge transfer","interlayer exciton","stacking polarity"],"falsifier":"Independently identify the interfacial face in an assembled device—for example, by cross-sectional scanning transmission electron microscopy or by polarization-resolved second-harmonic generation of the 3R flake before stacking—and check whether the region labeled 1L/M/X really has the M layer at the interface. If the face that the paper calls X/M shows faster charge transfer, or if both faces give the same transfer time, the stacking-polarity mechanism is wrong.","tokens_in":10487,"feed_emoji":"⚡","tokens_out":4416,"duration_ms":37859,"temperature":0.7,"pith_summary":"This paper claims that which atomic face of a 3R-stacked MoS2 bilayer touches a MoSe2 monolayer is a global, chemical-free control knob for interlayer photocarrier dynamics. Using ultrafast pump–probe spectroscopy, the authors resolve charge-transfer times of 0.25 ps when the metal (M) face sits at the interface and 0.37 ps when the chalcogen (X) face does, while 2H-stacked MoS2 transfers charge faster than about 0.1 ps. The interlayer exciton lifetime also shifts, from roughly 40 ps (M face) to 170 ps (X face), versus about 130 ps in the 2H stack. The mechanism is stacking-induced layer polarization in 3R MoS2, which localizes the conduction-band electron in one layer and thereby modulates wavefunction overlap across the interface. If correct, this makes vertical stacking order a deterministic tuning parameter for optoelectronic devices without altering chemistry or introducing moiré disorder.","feed_headline":"Flipping a 3R MoS2 face slows charge transfer to 0.37 ps","feed_subtitle":"The same MoSe2/MoS2 pair gains a tuning knob: exciton lifetimes shift between 40 and 170 ps with stacking order.","key_machinery":"The central object is stacking polarity: in 3R MoS2 the two layers are not rotated, so sulfur atoms (the X layer) align with molybdenum atoms (the M layer), making the layers inequivalent and giving an intrinsic out-of-plane polarization. This lifts the K-valley conduction-band degeneracy, localizing the lower-energy branch in the M layer. Because the 1.82 eV probe senses electrons in MoS2 and the M layer holds the quasi-equilibrium electron population, the measured rise time gives the interfacial electron-transfer rate, while the long decay gives the interlayer exciton lifetime. The 3R bilayer thus acts as a layer-polarized switch that controls how strongly MoS2 accepts and retains electron","core_discovery":"The central finding is that interlayer photocarrier dynamics in MoSe2/MoS2 heterostructures depend on the stacking polarity of the MoS2 bilayer. In the inversion-symmetric 2H bilayer, the electron wavefunction spreads over both layers, giving charge transfer faster than the roughly 0.3 ps instrument resolution and an interlayer exciton lifetime near 130 ps. In a 3R bilayer, broken inversion symmetry lifts the layer degeneracy: the conduction-band minimum is predominantly localized in the metal (M) layer, which lies at the interface in one configuration and away from it in the other. When the M layer contacts MoSe2, electron transfer takes 0.25 ± 0.04 ps and the interlayer exciton recombines","pith_inferences":["The directional conclusion that the M face transfers faster than the X face rests on assigning which face of one 3R flake touched MoSe2; the paper states the two surfaces cannot be optically distinguished before stacking and reports no independent face measurement, so the face labels should be treated as assigned post-hoc unless a structural probe confirms them.","If the layer-polarization mechanism is correct, the same face-dependent transfer and recombination should appear in other 3R TMD heterostructures, such as MoSe2/WSe2 or WS2-based pairs, where the M versus X interface can be tested with similar pump–probe experiments.","The slower, resolvable transfer in 3R stacks could let researchers measure charge-transfer efficiency under competing cooling and recombination channels, turning stacking polarity into a design trade-off between transfer speed and exciton lifetime in the same material pair.","A direct test of the face assignment would be to determine one flake's M and X faces before stacking (for example, by polarization-resolved second-harmonic generation or by cross-sectional atomic imaging of a reference device), then assemble MoSe2 on each face and check whether the 0.25 ps versus 0.37 ps split follows."],"forward_implications":["Stacking order alone can tune charge-transfer time several-fold, from under 0.1 ps in 2H to 0.37 ps in the 3R X/M configuration, without changing chemical composition or introducing twist-angle disorder.","The interlayer exciton lifetime can be adjusted from about 40 ps to 170 ps simply by choosing which 3R face is interfacial.","Because 3R-based heterostructures make charge transfer time-resolvable, they provide a platform for studying how strain, twist angle, temperature, and dielectric environment affect the transfer process.","The observed pump-fluence independence shows the tuning reflects structure rather than carrier-density effects.","The 1L/2H control with inverted stacking shows identical dynamics, confirming that the 2H behavior is intrinsic and not dominated by substrate effects."],"fun_headline_variants":["Stacking polarity controls photocarrier speed in MoSe2/MoS2","3R stacking slows charge transfer to 0.37 ps in MoSe2/MoS2","Interlayer exciton lifetime tuned 40–170 ps by stacking order","Stack order dictates photocarrier dynamics in MoSe2/MoS2 stacks","Time-resolved charge transfer: 0.25 to 0.37 ps via stacking"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The claim that charge transfer is faster when the M face touches MoSe2 assumes the two regions of one 3R flake were correctly labeled M versus X, but the paper reports no independent measurement of which face contacted MoSe2; if the labels were assigned after seeing which region was faster, the direction of the effect becomes circular.","fun_headline_variants_meta":{"raw":{"variants":["Stacking polarity controls photocarrier speed in MoSe2/MoS2","3R stacking slows charge transfer to 0.37 ps in MoSe2/MoS2","Interlayer exciton lifetime tuned 40–170 ps by stacking order","Stack order dictates photocarrier dynamics in MoSe2/MoS2 stacks","Time-resolved charge transfer: 0.25 to 0.37 ps via stacking"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000263,"raw_usage":{"total_tokens":1445,"prompt_tokens":758,"completion_tokens":687,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":502,"completion_tokens_details":{"reasoning_tokens":580}},"tokens_in":502,"tokens_out":687,"duration_ms":5867,"temperature":1.0,"reasoning_tokens":580,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T00:58:48.135530+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Independently identify the interfacial face in an assembled device—for example, by cross-sectional scanning transmission electron microscopy or by polarization-resolved second-harmonic generation of the 3R flake before stacking—and check whether the region labeled 1L/M/X really has the M layer at the interface. If the face that the paper calls X/M shows faster charge transfer, or if both faces give the same transfer time, the stacking-polarity mechanism is wrong.","supporting_citations":[],"review_version":1}