{"id":"eb560317-52d9-45c7-b18b-57d957c0b70f","arxiv_id":"2412.08663","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Reverse sputtering at 75 W for 10 s converts the MoS2 under metal contacts into a more conductive, defective phase and cuts the extracted contact resistance to less than half the untreated value.","lead":"The paper shows that reverse sputtering of the contact areas of molybdenum disulfide transistors can lower contact resistance from about 1126 to 413 kΩ·µm in the best case. This is a standard semiconductor process that could improve 2D-material transistors, though the measured spread between devices is large.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Table 1's pristine Rc error bar (±1053 kΩ·µm) already spans the claimed 50% reduction; without a significance test or per-device data, the quantitative headline is not established.","rationale":"The reader's weakest_assumption is exactly the load-bearing issue. The quantitative claim is 'Rc < 50%' from Table 1; the error bars and lack of significance test mean this could be noise. Independent support: raw Iout increase at 75 W is large (~150x) and BOT-FET control losing gate modulation supports metallization/defect creation. These are real evidence but they do not establish the specific Rc ratio. A secondary issue is the XPS fitting: the 1T component is constrained 0.9 eV higher than 2H, while the common literature assignment for 1T-MoS2 is ~0.9 eV lower than 2H; this weakens the phase-transition interpretation but is not the main quantitative claim. Therefore, I agree with the conditional verdict and recommend keeping it conditional pending the significance test and raw data.","tokens_in":14595,"tokens_out":3581,"duration_ms":39446,"concrete_test":"Request the raw per-device Rtot vs Lch data for the pristine and 75 W groups. For each device, extract Rc from the linear TLM fit Rtot = 2Rc + Rsh·Lch; then compare the two groups with a Welch two-sample t-test and a bootstrap 95% confidence interval for the difference in means. If the CI includes 0 (or p > 0.05), the 'less than 50%' claim is unsupported by the reported data. To rule out sample-transfer confounding, a follow-up experiment using multiple independently processed chips per condition, or splitting one transferred film into treated and untreated device areas, would settle whether the effect is causal.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim—that reverse sputtering at 75 W for 10 s reduces Rc to less than 50% of untreated—rests entirely on the TLM means in Table 1: 1126 ± 1053 kΩ·µm (pristine) vs 413 ± 167 kΩ·µm (75 W). The pristine group's spread is almost as large as its mean and covers the entire claimed reduction. If the ± values are standard deviations across n=7 devices, a Welch two-sample t-test gives p ≈ 0.13; if they are standard errors, p is much larger. No significance test is reported. Because each condition was measured on a different transferred sample, chip-to-chip transfer or processing variability is a confound that cannot be separated from the sputter treatment. The 150x increase in output current at 75 W is encouraging and the BOT-FET loss of gate modulation supports a metallic-phase change, but those observations do not by themselves validate the specific 'less than 50% Rc' number. The TLM extraction itself is also fragile with only 7 channels (Lch 4–10 µm) and a noisy Rtot(Lch) intercept. The claim can be repaired by reporting per-device Rc values with error propagation and a significance test; until then the abstract overstates what the data establish.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a process to reduce contact resistance (Rc) in multilayer MOCVD-grown MoS2 field-effect transistors by applying reverse (Ar-ion) sputtering to the MoS2 regions underneath the source/drain contacts, while protecting the channel with photoresist. The authors characterize the treated MoS2 by PL quenching, Raman J-peak emergence, and XPS, interpreting these as signatures of sulfur-vacancy formation and a 2H-to-1T phase transition. Electrical measurements on top-contacted FETs (TOP-FETs) show that a 75 W, 10 s sputter treatment increases output current by about 150 times relative to pristine devices, and Transfer Length Method (TLM) extraction yields Rc = 413 kOhm·um for the 75 W sample versus 1126 kOhm·um for pristine, corresponding to a reduction to less than 50% of the untreated value. Control bottom-contact FETs (BOT-FETs), in which the entire MoS2 was sputtered, lose gate modulation, supporting a metallic phase transformation. The paper concludes that reverse sputtering is a scalable, CMOS-compatible contact-engineering route for MoS2 FETs.","tokens_in":1836,"tokens_out":2188,"duration_ms":46488,"significance":"If the main electrical claim holds, the work has practical significance because reverse sputtering is a standard, cleanroom-compatible process that could improve contact injection in 2D transistors without exotic metals or chemical treatments. The study has notable strengths: the electrical effect is supported by direct device measurements; the BOT-FET control experiment provides a clear qualitative contrast; the raw output-current enhancement is large and visually unambiguous; and the material characterization includes complementary PL, Raman, and XPS data. The claimed mechanism of sulfur vacancies promoting a 2H-to-1T phase transition is plausible and consistent with prior literature. However, the central quantitative claim of a 'less than 50%' Rc is not yet established statistically, and the XPS phase assignment is presented with insufficient fitting detail. The paper would be significantly strengthened by per-device data, a significance test, and a more cautious presentation of the XPS quantification.","major_comments":[{"comment":"The headline claim that optimized reverse sputtering reduces Rc to less than 50% of the untreated value rests on the TLM means in Table 1: pristine 1126 ± 1053 kOhm·um versus 75 W 413 ± 167 kOhm·um. The pristine standard error (or deviation, the caption is ambiguous) is almost as large as its mean and spans the entire claimed reduction. No significance test (e.g., Welch t-test or Mann-Whitney) is reported, and with n=7 per group the difference is not statistically robust if the quoted spreads are standard deviations. The authors should report per-device Rc values, specify whether error bars are standard errors or standard deviations, and provide a statistical test comparing the 75 W group with the pristine group.","section":"Table 1 and Conclusion"},{"comment":"The TLM extraction uses only seven channel lengths (4–10 um) with a single device per length per sample, and the authors do not show the individual Rtot(Lch) data points or the linear fits whose intercept yields Rc. Given the very large device-to-device spread visible in the transfer curves and the mobility histogram (Fig. 3c), the TLM intercept is fragile. The authors should present the raw Rtot(Lch) data for all 28 devices, the fitted lines, and confidence intervals for the extracted Rc and Rsh. Without this, the reader cannot assess whether the difference between 1126 and 413 kOhm·um reflects the sputter treatment or the choice of TLM fit.","section":"Fig. 4 and TLM extraction"},{"comment":"The XPS interpretation that reverse sputtering induces full 2H-to-1T conversion is based on a constrained fit in which the 1T Mo3d contribution is fixed at 0.9 eV higher binding energy than the 2H contribution, and even the pristine sample shows a minor fitted 1T component. The claim that after all the sputter treatments only a metallic 1T phase is obtained is surprisingly strong and could be a fitting artifact, for example from broadened peaks being absorbed into an unconstrained intensity. The authors should report the fitted peak widths, the 2H/1T area ratios with uncertainties, and a discussion of whether the reduced S/Mo ratio alone could explain the spectral broadening without invoking full phase conversion.","section":"XPS phase analysis (Fig. 2b and text)"},{"comment":"A confound in the experimental design is that each sputter-power condition was measured on a separately transferred MoS2 sample. As the authors note, wet transfer and patterning can introduce chip-to-chip variability in doping, layer number, or residue, which is not controlled for. The 25 W sample has a higher Rc than pristine (2632 kOhm·um), yet the channel mobility and Rsh values are similar across pristine, 25 W, and 75 W; this pattern could partly reflect sample-to-sample variation. The authors should either fabricate multiple samples per condition or provide evidence that the transfer and processing uniformity is high enough to exclude this confound, for example by measuring pristine reference devices on every transferred chip.","section":"Methods: Device fabrication"}],"minor_comments":[{"comment":"The abstract states Rc values reduced to less than 50% of their untreated counterparts, while the conclusion says a 50% reduction in Rc. These are mathematically different (reduced to 50% versus reduced by 50%); given the measured values (413 vs. 1126 kOhm·um, roughly 37%), the abstract wording is accurate but the conclusion wording is not. Please align them.","section":"Abstract and Conclusion"},{"comment":"The caption says standard errors, but the text in Fig. 3c says standard deviation was extracted from a Gaussian fit and the same ± notation is used. Please specify explicitly which quantity is reported and use consistent terminology throughout.","section":"Table 1 caption"},{"comment":"The sentence describing injection from Ni into the metallic 1T phase and then from the 1T into the 2T phase contains a typo: 2T phase should be 2H phase.","section":"Results, paragraph after Eq. (1)"},{"comment":"The histogram shows μ2p distributions, but the text says 28 TOP-FETs on each sample when it should be 7 per sample (28 across four samples). Please correct this wording to avoid ambiguity.","section":"Fig. 3c"},{"comment":"The mobility extraction using Eq. (1) is done from transfer curves at VDS = 1 V, but no justification is given that this VDS is in the linear regime for the two-probe measurement. Please state the linear-regime condition or note that the reported μ2p is an effective two-probe mobility with contact-resistance contributions.","section":"Methods: Electrical Measurements"}],"recommendation":"major_revision","confidential_remarks":"The paper reports a potentially useful process and contains a strong qualitative control experiment (BOT-FETs losing gate modulation). The main technical weakness is the absence of a statistical analysis for the headline Rc reduction; this is repairable with per-device data and a significance test. The XPS phase quantification also needs more cautious reporting. I would not reject the paper, but I would require the authors to address the statistical and fitting concerns before publication. There is no evident issue with novelty or scope for the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a useful, incremental process paper that maps a reverse-sputtering window for contact engineering on MOCVD multilayer MoS2. The central qualitative result is credible: 75 W for 10 s on the contact regions multiplies on-current by roughly 150, nearly doubles two-probe mobility, and the bottom-contact control – where the whole channel is exposed and gate modulation collapses – gives independent evidence that the sputtered material becomes metallic. That control is the right experiment and it carries a lot of the argument.\n\nThe material characterization is decent: PL quenching, the J Raman peak, and falling S/Mo ratio all point to sulfur vacancies and structural change. The 200 W case showing MoOx-related damage is consistent across XPS and the electrical data.\n\nThe soft spots are real, and they mostly concern the quantitative headline. Table 1 reports pristine Rc of 1126 ± 1053 kΩ·µm and 75 W of 413 ± 167 kΩ·µm. The caption calls them standard errors; if those are actually standard deviations, Welch's t-test gives p ≈ 0.13, and if they're standard errors the underlying scatter is even worse. Either way the claimed 'less than 50%' reduction isn't established. The groups come from different transferred chips, so transfer variability is a confound that can't be separated from the treatment. Also, only 7 TLM channels per group is thin. The raw on-current jump is large and encouraging, but the paper should show per-device Rc values and a significance test.\n\nThere's also a numerical inconsistency: the text says the 25 W sample's Rc increases by 'approximately 50%' compared to pristine, but Table 1 says 2632 vs 1126 kΩ·µm – a 134% increase. That's not fatal, but it hurts trust in the numbers.\n\nFinally, the XPS fitting constraint is suspect. The 1T component is constrained to be 0.9 eV higher binding energy than 2H. The standard literature (Eda et al. 2011, etc.) puts 1T about 1 eV lower than 2H. As written, that makes the 'full phase conversion' claim questionable. The BOT-FET control is the stronger evidence for metallization, so the mechanism probably survives, but the XPS section needs re-examination.\n\nBottom line: this is a process-engineering paper targeted at people working on MoS2 contacts. A serious referee should see it, mostly to push on the statistics and the XPS sign. If those are repaired, it would be a solid contribution to the toolset. I'd send it out, but not fast-track it.","headline":"A credible process-window study with a weak quantitative headline; the 75 W/10 s effect is real but not statistically established, and the XPS 1T-binding-energy constraint looks backwards.","tokens_in":15400,"tokens_out":5829,"would_cite":false,"duration_ms":51876,"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":"A 10-second reverse sputtering step at 75 W cuts contact resistance in MOCVD-grown MoS2 transistors to less than half of its untreated value.","keywords":["MoS2 field-effect transistors","contact resistance","reverse sputtering","2H to 1T phase transition","sulfur vacancies","MOCVD MoS2","transfer length method","Schottky barrier"],"falsifier":"A decisive check would be a larger, paired experiment in which the same MoS2 film is split across all four conditions and transfer length measurements are run on more devices per group; if a two-sample comparison of the 75 W and pristine groups no longer shows separation, the headline reduction fails. A second check targets the mechanism: since XPS already shows full 1T conversion at 25 W while Rc does not improve, measuring how Rc tracks the 1T fraction and S/Mo ratio on the same contacted dies would settle whether phase conversion or vacancy doping is the active ingredient.","tokens_in":14354,"feed_emoji":"⚡","tokens_out":8335,"duration_ms":82945,"temperature":0.7,"pith_summary":"Reverse sputtering, the same argon-ion plasma cleaning step already used in semiconductor fabrication, can be repurposed to engineer the contacts of MoS2 transistors. The paper reports that treating only the source/drain contact areas of MOCVD-grown multilayer MoS2 for 10 seconds at 75 W drops the transfer-length-method contact resistance from about 1126 kΩ·µm to 413 kΩ·µm, and raises ON-state currents. The proposed cause is that argon-ion bombardment creates sulfur vacancies that n-type dope the MoS2 and trigger a transition from semiconducting 2H-MoS2 to metallic 1T-MoS2, narrowing the Schottky barrier to nickel. This matters because the process is clean, controllable, and compatible with existing wafer-scale flow, unlike chemical phase engineering or semimetal contacts. The paper also finds that the window is narrow: 25 W does not help and 200 W degrades the devices through oxidation and structural damage.","feed_headline":"Reverse sputtering cuts MoS2 transistor contact resistance in half","feed_subtitle":"A 10-second argon plasma at 75 W turns contact MoS2 metallic, cutting Rc from 1126 to 413 kΩ·µm","key_machinery":"The load-bearing mechanism is the 2H-to-1T phase transition of MoS2, driven by sulfur vacancies created by argon-ion bombardment during reverse sputtering. In the stable 2H phase MoS2 is semiconducting; in the metastable 1T phase it is metallic. The paper uses photoluminescence quenching, the emergence of a Raman J peak at about 277 cm$^{-1}$, and XPS fitting with the 1T component held 0.9 eV higher in binding energy than the 2H component to argue that the transition occurs. Reverse sputtering also lowers the S/Mo ratio, which the authors read as vacancy doping that narrows the Schottky barrier to the nickel contact. The same mechanism predicts an optimum: too little sputtering leaves the contact unimproved, while too much oxidizes the damaged MoS2 to MoO$_x$ and destroys the electrical benefit.","core_discovery":"On its own terms, the paper's central claim is that reverse sputtering in the contact regions of MoS2 field-effect transistors is a working contact-engineering method. In an optimized split of 75 W for 10 s, the mean contact resistance extracted from transmission line measurements fell to 413 kΩ·µm from 1126 kΩ·µm for untreated devices, with the ON-state output current roughly 150 times higher at a 4 µm channel length and the two-probe mobility nearly doubled. Spectroscopic evidence, including photoluminescence quenching, the appearance of a Raman J peak, and XPS fits with a 1T Mo3d component, supports conversion toward metallic 1T-MoS2 and sulfur depletion. The authors attribute the improvement to vacancy-induced n-type doping that narrows the Schottky barrier together with formation of conductive 1T-MoS2 under the nickel contacts, while noting that too much sputtering creates MoOx and ruins the contacts and channel. Control devices in which the whole channel was sputtered lose gate modulation, confirming that the benefit is specific to treating the contact regions.","pith_inferences":["If the vacancy-driven 2H-to-1T mechanism is correct, the same reverse-sputtering approach should transfer to other sulfur-based transition metal dichalcogenides, though the optimum power and duration would need to be recalibrated for each material's bond strength.","A direct way to separate the treatment effect from sample-to-sample transfer variability would be a split-chip experiment where treated and untreated contacts share a single MoS2 film and are compared on the same substrate.","The 25 W case is a useful discrepancy: XPS already shows full 1T conversion at that power, yet Rc does not improve, so later work that explains this gap will clarify whether phase fraction or vacancy density is the actual controlling variable.","Because reverse sputtering can be combined in situ with metal deposition, the process could in principle be extended to other CMOS-compatible metals besides nickel, provided their adhesion and barrier properties are tested."],"forward_implications":["At 75 W for 10 s, contact resistance drops from 1126 kΩ·µm to 413 kΩ·µm, with ON-state output current roughly 150 times higher for a 4 µm channel while on/off ratios remain near $10^6$.","Because reverse sputtering is a standard, clean, CMOS-compatible process, the treatment can be inserted into wafer-scale fabrication without adding rare materials or chemical processing.","The process window is narrow: 25 W is ineffective and 200 W degrades performance, so production would need to hold power and duration near the optimum.","Treating the entire channel destroys gate modulation, so the method must be restricted to the contact regions rather than applied globally.","The structural markers, including PL quenching, the Raman J peak, the 1T XPS component, and the reduced S/Mo ratio, provide inline metrology for the intended phase change."],"supporting_citations":[{"why":"Establishes the prior chemical phase-engineering route to low-resistance contacts on MoS2 that the reverse-sputtering method is intended to replace.","marker":"[4]"},{"why":"Shows argon plasma induces the 2H-to-1T phase transition in monolayer MoS2, the physical precedent for the structural modification claimed here.","marker":"[7]"},{"why":"Shows sulfur-vacancy engineering moves MoS2 contacts toward ohmic behavior, supplying the doping mechanism invoked to explain the Rc reduction.","marker":"[8]"},{"why":"Links sulfur vacancies to the 2H/1T phase transition in multilayer MoS2, supporting the vacancy-before-phase sequence in the argument.","marker":"[5]"},{"why":"Provides the atomic displacement mechanism by which the semiconducting-to-metallic transition occurs, used to describe how bombardment restructures the lattice.","marker":"[19]"},{"why":"Supplies the XPS fitting constraint that places the 1T Mo3d contribution 0.9 eV above the 2H contribution, used to quantify the phase conversion.","marker":"[38]"},{"why":"Source of the MOCVD-grown multilayer MoS2 on sapphire used for all material and device experiments.","marker":"[43]"}],"fun_headline_variants":["Reverse sputtering slashes MoS2 contact resistance by more than half","Sputtering converts MoS2 contacts to 1T phase, cutting resistance","Plasma treatment drops MoS2 FET contact resistance below half","Metallic 1T phase from sputtering cuts MoS2 contact resistance"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative conclusion assumes that the four device groups differ because of the sputtering power and not because of run-to-run variation in MoS2 transfer or device yield, and the untreated group's own standard deviation of 1053 kΩ·µm around a mean of 1126 kΩ·µm is large enough to overlap the claimed 75 W result.","fun_headline_variants_meta":{"raw":{"variants":["Reverse sputtering slashes MoS2 contact resistance by more than half","Sputtering converts MoS2 contacts to 1T phase, cutting resistance","Plasma treatment drops MoS2 FET contact resistance below half","Metallic 1T phase from sputtering cuts MoS2 contact resistance"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000244,"raw_usage":{"total_tokens":1562,"prompt_tokens":1008,"completion_tokens":554,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":624,"completion_tokens_details":{"reasoning_tokens":474}},"tokens_in":624,"tokens_out":554,"duration_ms":6132,"temperature":1.0,"reasoning_tokens":474,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T19:36:41.345450+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive check would be a larger, paired experiment in which the same MoS2 film is split across all four conditions and transfer length measurements are run on more devices per group; if a two-sample comparison of the 75 W and pristine groups no longer shows separation, the headline reduction fails. A second check targets the mechanism: since XPS already shows full 1T conversion at 25 W while Rc does not improve, measuring how Rc tracks the 1T fraction and S/Mo ratio on the same contacted dies would settle whether phase conversion or vacancy doping is the active ingredient.","supporting_citations":[],"review_version":1}