{"id":"92df2bf8-d6ae-4e08-9e9e-517dbdd66c19","arxiv_id":"2607.03121","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.5,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Persistent UV photodoping in WSe2/hBN FETs is mediated by SiO2 defect states acting as charge reservoirs, not by hBN defects, as shown by oxide-removal controls and wavelength/polarity asymmetry.","lead":"UV light can permanently dope WSe2 transistors, but only when a SiO2 substrate is present; removing the oxide kills the effect. This reassigns the long-lived traps from hBN defects (the usual story) to oxide defect states, changing how 2D optoelectronic devices should be designed.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"The oxide-removal control is decisive only if wet etch leaves the WSe2/hBN stack and gate electrostatics otherwise equivalent; that equivalence is under-documented.","rationale":"The Reader correctly isolates the oxide-removal experiment as both the strongest evidence and the weakest assumption. The wavelength asymmetry, hBN-thickness independence (S4) and DFT excitation energies already make bulk-hBN defects unlikely; the SiO2-removal result is what elevates the substrate-trap claim from plausible to decisive. Because that result is under-controlled (single device, thickness mismatch, no post-etch metrology), the claim remains conditional on better documentation of etch equivalence. No deeper internal inconsistency or circularity is present, so the Reader’s CONDITIONAL verdict and high confidence are unchanged.","tokens_in":17043,"tokens_out":483,"duration_ms":5164,"concrete_test":"Fabricate matched WSe2/hBN devices on the same wafer, measure baseline mobility, hysteresis and photodoping, then wet-etch SiO2 on half the devices; re-measure mobility, contact resistance and transfer-curve shape under the reduced |VEx| used in S5. If mobility drops >20 % or hysteresis appears only after etch, the suppression cannot be attributed solely to oxide-trap removal.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim rests on Fig. 5 / S5: photodoping of both polarities is strongly suppressed once SiO2 is wet-etched away. That inference requires that the etched devices remain electrically comparable to the SiO2-supported controls except for the absence of oxide traps. After etch the gate dielectric is only the remaining hBN, so |VEx| is reduced from 50 V to 3 V (S5) to keep the field roughly constant; contact quality, residual adsorbates, interface cleanliness, and possible etch-induced damage to the hBN/Si interface are not independently verified. The paper reports only a single etched device pair with mismatched WSe2 thicknesses (16 nm vs 33 nm) and no post-etch AFM, Raman, or mobility statistics. If the etch itself degrades the channel or alters the gate-field distribution, the observed suppression could be an artifact rather than proof that SiO2 traps are the charge reservoirs.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript reports UV-induced persistent photodoping in multilayer WSe2/hBN FETs on SiO2/p-Si and attributes the effect to defect states in the SiO2 substrate rather than to hBN or WSe2 defects. Wavelength- and polarity-dependent transfer-curve shifts show that 405 nm n-type photodoping is orders of magnitude more efficient than 640 nm excitation or p-type doping, with sheet densities up to ~3.7e12 cm^-2. Control devices with different hBN thicknesses show no thickness scaling (Fig. S4); first-principles excitation energies of representative hBN defects remain accessible under 640 nm, inconsistent with the observed suppression; and wet-etch removal of SiO2 nearly eliminates both n- and p-type photodoping (Fig. 5/S5). Supporting KPFM work-function shifts and multi-hour retention are provided. The authors conclude that deep SiO2 traps act as long-lived charge reservoirs that enable charge transfer through hBN into the WSe2 channel.","tokens_in":17270,"tokens_out":1229,"duration_ms":9821,"significance":"If the SiO2-trap assignment holds, the work revises a widely cited picture in which photodoping of TMD/hBN devices is ascribed primarily to donor-like defects in hBN (Table S1). The combination of wavelength/polarity asymmetry, hBN-thickness independence, DFT exclusion of bulk hBN defects, and the oxide-removal control is a coherent experimental package that would reorient device design toward substrate engineering for rewritable optical doping, photogating, and neuromorphic optoelectronics. The stretched-exponential kinetics and KPFM confirmation of real carrier transfer are additional strengths. The result is therefore of clear interest to the 2D optoelectronics community, provided the oxide-removal control is shown to be free of process artifacts.","major_comments":[{"comment":"The decisive control is the near-total suppression of both polarities after wet-etch removal of SiO2 (Fig. 5 and Fig. S5). After etch the gate dielectric is only residual hBN, so |VEx| is reduced from 50 V to 3 V; the etched device also has a different WSe2 thickness (33 nm vs 16 nm). No post-etch AFM, Raman, mobility statistics, or contact I–V comparison is reported to establish that channel quality, residual adsorbates, and gate-field distribution remain equivalent. Without that equivalence the suppression could partly reflect etch-induced damage or altered electrostatics rather than the absence of oxide traps. Additional characterization of the etched stack (or a second independent control, e.g., devices fabricated on hBN/Si without ever seeing SiO2) is needed to make the central claim load-bearing.","section":"Fig. 5 / Fig. S5 and Methods"},{"comment":"The proposed microscopic pathway—optical activation of deep SiO2 traps followed by charge transfer through tens of nm of hBN into WSe2—is left largely schematic (Fig. 6B). The manuscript does not quantify tunneling or hopping rates, nor does it address how the gate field and the hBN barrier jointly enable the observed polarity asymmetry and the three-order-of-magnitude wavelength contrast. A short estimate of transfer probability or a comparison with known SiO2/hBN interface trap densities would strengthen the mechanistic claim that SiO2 defects, rather than interface states created by the etch or by residual polymer, are the actual reservoirs.","section":"§2 Discussion and Fig. 6"}],"minor_comments":[{"comment":"Main-text AFM thicknesses (47 nm hBN, 12 nm WSe2) disagree with Fig. S1 captions (44 nm / 9 nm) and with the device used for the SiO2-removal comparison (16 nm / 38 nm). Please reconcile all thickness values and state which device corresponds to which figure.","section":"Fig. 1 / Fig. S1 / Fig. S5"},{"comment":"Fig. 2B caption writes “IDS = 0.3 A”; the text and other figures use 0.3 µA. Correct the unit.","section":"Fig. 2B"},{"comment":"The capacitance used for mobility and sheet-density extraction (C = 9.75 nF cm^-2) is stated without an explicit series-capacitance formula for the 300 nm SiO2 + bottom-hBN stack; a one-line derivation would aid reproducibility.","section":"§2 Results"},{"comment":"Table S1 is a useful literature survey; a short sentence in the main text noting which prior works already mentioned BN/SiO2 interface traps (e.g., L2, L5, L7) would better position the novelty claim.","section":"Introduction / Table S1"},{"comment":"Several references appear with incomplete or slightly garbled formatting (e.g., “Appli. Surf. Sci.”, “Nat. Nanotechnol.” vs “Nat. Nanotechnology”). Standardize journal abbreviations.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The oxide-removal control is the single most important piece of evidence; if the authors can supply post-etch channel characterization or an independent no-SiO2 fabrication route, the paper becomes a strong candidate for acceptance. Without it the central claim remains plausible but not fully secured. Scope is appropriate for a solid-state / 2D-materials journal."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The new result is the near-total suppression of both n- and p-type photodoping once SiO2 is wet-etched away (Fig. 5/S5), combined with hBN-thickness independence (S4) and DFT that leaves hBN defects optically accessible under 640 nm where doping is negligible. That package cleanly reassigns the charge reservoir from the hBN bulk (the default story in Table S1) to oxide traps. It is not a new device concept, but it is a real mechanistic correction for a technique people already use for rewritable doping and photogating.\n\nWhat they do well: wavelength and polarity asymmetry (Fig. 3), stretched-exponential kinetics linear in photon flux (Fig. 4), KPFM work-function shifts confirming actual carrier transfer (S6), multi-hour persistence (S7), and a literature table that honestly maps prior interpretations. The SiO2 trap energies are taken from the external literature and used only for qualitative consistency; no free parameters are recycled as predictions. Circular burden is low.\n\nThe soft spot is exactly the one the stress-test flags: after etch the gate dielectric is only hBN, so they drop |VEx| from 50 V to 3 V, and they show only one etched pair with mismatched WSe2 thicknesses (16 vs 33 nm) and no post-etch AFM/Raman/mobility statistics. If the etch damages the stack or changes residual adsorbates, the suppression could be partly artifactual. That is a real documentation gap, not a load-bearing contradiction; the multi-control pattern still points the same way. I would not over-weight it into a rejection.\n\nThis is for people who build or model TMD/hBN FETs and care about where the traps actually sit. It deserves a serious referee who will demand better etch-control characterization and perhaps a second etched device. I would cite it when discussing photodoping origins and would bring it to reading group.","headline":"Solid experimental reassignment of photodoping from hBN defects to SiO2 traps; the oxide-removal control is decisive if you accept the etch equivalence, which is under-documented but not fatal.","tokens_in":17927,"tokens_out":500,"would_cite":true,"duration_ms":4738,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Persistent UV photodoping in WSe2/hBN FETs is enabled by defect states in the SiO2 substrate, not by defects in hBN.","keywords":["photodoping","van der Waals heterostructure","SiO2 defects","charge transfer","WSe2","hBN","field-effect transistor","photogating"],"falsifier":"Fabricate identical WSe2/hBN devices on a different dielectric (e.g., Al2O3 or high-quality thermal oxide free of E' centers) that still supports gating; if persistent UV photodoping of both polarities reappears at the same rates, the SiO2-defect claim is falsified.","tokens_in":17906,"feed_emoji":"⚡","tokens_out":736,"duration_ms":12010,"temperature":0.7,"pith_summary":"This paper shows that the long-lived n- and p-type carrier doping seen when WSe2/hBN field-effect transistors are illuminated by UV light comes from trap states in the underlying SiO2, not from the hexagonal boron nitride layer that most earlier work blamed. Wavelength and polarity dependence, the complete lack of thickness scaling with hBN, and first-principles defect energies all rule out hBN as the source. The decisive experiment is that both polarities of photodoping collapse once the SiO2 is etched away. The result matters because it reassigns the microscopic origin of a widely used rewritable doping method and tells device designers that the oxide substrate, not the 2D stack, is the charge reservoir that must be engineered.","feed_headline":"SiO2 traps, not hBN, drive UV photodoping in WSe2 FETs","feed_subtitle":"Etching away the oxide kills both n- and p-type doping, reassigning the charge reservoir in 2D devices.","key_machinery":"SiO2 defect states acting as long-lived charge reservoirs: deep electron traps (~2.8–3.0 eV below the conduction band) efficiently emptied by 405 nm photons and shallow/short-lived hole traps that limit p-type efficiency; carriers tunnel or hop across hBN into the WSe2 channel under the gate field.","core_discovery":"In multilayer WSe2/hBN heterostructure FETs on SiO2/p-Si, UV (405 nm) illumination under gate bias produces large, persistent n-type (and weaker p-type) photodoping that saturates at sheet densities of a few 10^12 cm^{-2}. The same process is orders of magnitude weaker under 640 nm light. Removing the SiO2 layer by wet etch suppresses both polarities almost completely, while photodoping rate is independent of hBN thickness and DFT shows hBN defect levels remain optically accessible under conditions where doping is experimentally absent. Therefore the dominant charge reservoirs are deep electron and hole traps in the amorphous SiO2 that exchange carriers with the WSe2 channel through the hBN","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["SiO2 traps not hBN drive UV photodoping in WSe2 FETs","Removing SiO2 suppresses photodoping in WSe2/hBN devices","SiO2 defect states act as charge reservoirs for WSe2 photodoping","Oxide traps enable persistent UV doping in WSe2 FETs","hBN defects ruled out as source of UV photodoping in WSe2"],"cache_read_input_tokens":128,"weakest_assumption_plain":"Wet-etch removal of the SiO2 leaves the WSe2/hBN interfaces, contact quality, residual adsorbates and gate-field distribution electrically equivalent to the control devices, so the disappearance of photodoping can be attributed solely to the missing oxide traps.","fun_headline_variants_meta":{"raw":{"variants":["SiO2 traps not hBN drive UV photodoping in WSe2 FETs","Removing SiO2 suppresses photodoping in WSe2/hBN devices","SiO2 defect states act as charge reservoirs for WSe2 photodoping","Oxide traps enable persistent UV doping in WSe2 FETs","hBN defects ruled out as source of UV photodoping in WSe2"]},"model":"grok-4.5","effort":"low","cost_usd":0.005702,"raw_usage":{"total_tokens":1573,"prompt_tokens":886,"num_sources_used":0,"completion_tokens":106,"cost_in_usd_ticks":57020000,"prompt_tokens_details":{"text_tokens":886,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":581,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":886,"tokens_out":106,"duration_ms":4703,"temperature":1.0,"reasoning_tokens":581,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-12T04:44:34.045715+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Fabricate identical WSe2/hBN devices on a different dielectric (e.g., Al2O3 or high-quality thermal oxide free of E' centers) that still supports gating; if persistent UV photodoping of both polarities reappears at the same rates, the SiO2-defect claim is falsified.","supporting_citations":[],"review_version":1}