{"id":"274f52c4-19bb-42f4-b7f0-2979b705e636","arxiv_id":"2501.14969","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Crystalline NbSe2/WSe2/NbSe2 Josephson junctions show exponential thickness dependence and a proximity-to-tunneling crossover near seven WSe2 layers, enabling a compact transmon whose measured frequency matches design.","lead":"Vertical stacks of superconducting NbSe2 and semiconducting WSe2 form Josephson junctions whose switching current and resistance change exponentially with WSe2 thickness, with a crossover from proximity-like to tunneling-like behavior near seven layers. The authors use this thickness dependence to design and build a compact superconducting qubit, a merged-element transmon, whose measured transition frequency matches the prediction.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Design-rule precision is undercut by ±1-layer AFM uncertainty: a one-layer error changes predicted f01 by roughly 60%, yet the 10% validation rests on a single device with no propagated error bars.","rationale":"The paper's central crossover observation is well supported by several independent signatures: the exponential thickness dependence of j_s0 and R_nA, the sudden change in I_s0/I_r0 near seven layers, the grouping of I_s R_n above and below the Ambegaokar-Baratoff limit, and the different temperature dependences of the retrapping current. These observations do not require the band-alignment doping model to be correct, so the reader's identified weakest assumption about the inferred 3-layer doping thickness is not the most load-bearing condition for the central claim. The design claim, however, requires the thickness-to-I_c relation to be quantitatively predictive. The paper's own stated ±1-layer AFM uncertainty, combined with the steep exponential slope of j_s0, produces a frequency uncertainty far larger than the claimed 10% agreement. The test of this concern is direct: re-measuring layer counts by an independent method and re-fitting the data would show whether the design rule survives. Until such a test is done, the appropriate verdict is CONDITIONAL, which matches the reader's verdict, so no verdict change is needed.","tokens_in":16006,"tokens_out":10609,"duration_ms":110971,"concrete_test":"Re-determine the WSe2 layer count for every junction in Fig. 1(c) and for the MET device using a technique independent of AFM step height, such as Raman spectroscopy or second-harmonic generation, which the authors themselves suggest in Appendix A. Re-fit the j_s0-vs-thickness exponential with the corrected layer numbers and recompute the predicted f01 of the 17-layer MET, propagating the ±1-layer uncertainty into the frequency prediction. If the corrected layer counts shift the predicted f01 by more than 10%, the quantitative design claim fails. A complementary test is to fabricate two additional METs with different WSe2 thicknesses and verify that each measured f01 falls inside the propagated prediction band.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing condition for the 'accurate design' claim is that the WSe2 layer count determines the junction critical current tightly enough to predict qubit frequency to the claimed accuracy. The data in Fig. 1(c) show j_s0 changing by roughly six orders of magnitude over 3-18 layers, i.e. about one order of magnitude per 2.5 layers. Appendix A states that AFM thickness determination carries a typical uncertainty of ±1 atomic layer. A one-layer error therefore changes j_s0 by a factor of about 2.5, and f01, which scales as the square root of E_J in the transmon regime, by about 1.58. That is a ~60% frequency uncertainty, not a 10% one. The reported agreement of the single 17-layer MET within 10% is one data point and is presented without propagated uncertainties from the j_s0 fit, junction area, capacitance, or layer count. The crossover observation itself is multiply supported and does not depend on this error budget, but the statement that DC transport characterization 'can be used to accurately design transmon qubits' is not established until the layer-count uncertainty is propagated and the design rule is tested on more than one device. The inferred 3-layer doping count in Section II is a plausible mechanism, but it is not the load-bearing part: even if that inference is wrong, the exponential thickness dependence used for design could still hold.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a systematic dc transport study of vertical NbSe2/WSe2/NbSe2 Josephson junctions with WSe2 weak links from 3 to 18 atomic layers. The authors identify a crossover from proximity-type to tunneling-type behavior around 7 layers based on three correlated signatures: the switching-to-retrapping current ratio, the I_cR_n product relative to the Ambegaokar-Baratoff limit, and the temperature dependence of the retrapping current. They interpret the crossover via interfacial charge transfer and hole doping of the WSe2 near each NbSe2 electrode, with roughly 3 layers doped per electrode. Using the exponential thickness dependence of the switching current density, they design a merged-element transmon, fabricate a 17-layer device, and report a measured f01 within 10% of the predicted value, along with dispersive readout and anharmonicity consistent with a transmon.","tokens_in":16299,"tokens_out":5444,"duration_ms":49371,"significance":"If the design rule is quantitatively reliable, this work establishes a promising path toward compact, low-loss superconducting qubits with crystalline, atomically uniform semiconductor barriers. The crossover observation itself is well supported by multiple independent signatures and is an interesting physics result for superconductor-semiconductor heterostructures. The manuscript also contains an explicit control experiment with MoS2 weak links that supports the band-alignment interpretation, and the device demonstration includes two-tone spectroscopy and dispersive coupling. These strengths make the paper potentially influential for the 2D-materials quantum device community. The central empirical crossover observation does not depend on the fitting model, and the paper is transparent about the Thomas-Fermi model's inability to predict the crossover.","major_comments":[{"comment":"The claim that DC transport characterization can be used to accurately design transmon qubits is not yet established given the uncertainty budget. Appendix A states a typical AFM thickness uncertainty of ±1 atomic layer. From Fig. 1(c), j_s0 changes by roughly six orders of magnitude over 3–18 layers, about a factor of 2.5 per layer. Because f01 depends on sqrt(E_J) ∝ sqrt(j_s0) in the formula f01 = (sqrt(8E_J E_C) - E_C)/h, a one-layer uncertainty changes f01 by roughly 60%, not 10%. The reported validation is a single 17-layer device with no propagated errors from layer count, junction area, capacitance, or the switching-current proxy described in Appendix B. The authors should either propagate these uncertainties and test more devices, or explicitly soften the design-accuracy claim.","section":"Appendix A, Fig. 1(c), Fig. 4(a), Section II"},{"comment":"The Thomas-Fermi model is not an independent validation of the transport data. It contains fit parameters (ε = 5.5, G_M, and the Fermi-level position reported as 28.4 meV in Section II) that are adjusted to the same R_nA data shown in Fig. 1(d). The model also imposes zero-potential boundary conditions at the contacts, which effectively assumes hole doping rather than predicting it. The Conclusion correctly acknowledges that the model does not predict the crossover, but the earlier statement of 'quantitative agreement' should be framed as a fit, and the model's explanatory role should be restricted to the exponential resistance trend in the tunneling regime.","section":"Appendix D, Fig. 1(d), Section II, Conclusion"},{"comment":"The inference that '≈3 layers of WSe2 are doped by each NbSe2 electrode' is not directly measured but is inferred from the same dc datasets that show the crossover. This is a plausible and useful hypothesis, but it is not a direct determination. The wording should be softened to indicate that the 3-layer doping depth is an inferred value consistent with the crossover, not a measured quantity. This distinction matters because the designed 17-layer junction relies on the exponential j_s0 dependence, which does not require the 3-layer inference to be correct, whereas the mechanistic narrative does.","section":"Section II, paragraph beginning 'In the superconducting state...'"}],"minor_comments":[{"comment":"There is a typo in the abstract: 'demonstratedispersive' should be 'demonstrate dispersive'.","section":"Abstract"},{"comment":"In the sentence describing the 13L device, the stray word 'red' appears after the switching current value and should be removed.","section":"Section II, Fig. 2(discussion)"},{"comment":"The extracted capacitance C_J = 53 fF in Appendix B is presented as if it were a measured geometric capacitance, while Section II uses an 'expected geometric junction parallel-plate capacitance' for the calculation in Fig. 4(a). The relation between these two values should be clarified, and the model-dependence of the 53 fF extraction should be stated.","section":"Appendix B, Section II"},{"comment":"The term 'Thomas-Fermi' should use an en dash ('Thomas–Fermi') in the final version, and the hyphenated form should be made consistent.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"This is a solid systematic transport study with a promising device demonstration. The main limitation is the single-device validation of the design rule and the lack of propagated uncertainties, which I believe is fixable with additional analysis or a more cautious interpretation. The Thomas-Fermi model is fit-based, not predictive, and should be labeled accordingly. The crossover observation is robust and worth publishing. No concerns about novelty or fit with the journal's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real content here is the thickness series: vertical NbSe2/WSe2/NbSe2 junctions from 3 to 18 layers, with a crossover from proximity-type to tunneling-type behavior near 7 layers. That crossover is supported by three independent DC signatures — hysteresis ratio, IcRn relative to the Ambegaokar-Baratoff limit, and the temperature dependence of retrapping current. The all-vdW merged-element transmon with dispersive readout is a legitimate proof of concept, and the measured f01 close to the design value is encouraging. This is the first systematic study in this thickness range, and it will be useful to people building vdW Josephson devices.\n\nWhere the paper strains is the design-rule claim. The stress-test math is right: js0 changes by roughly an order of magnitude per 2.5 layers, and Appendix A admits ±1-layer AFM uncertainty. A one-layer error shifts f01 by about 60%, not 10%. The single 17-layer device's critical current itself was not directly measured but inferred from the dI/dV maximum because of premature switching. So the statement that DC transport 'can be used to accurately design transmon qubits' is not established by this data. It may be true eventually, but it needs more devices and propagated error bars.\n\nThe mechanistic story — about 3 hole-doped WSe2 layers per NbSe2 electrode and an undoped tunnel barrier beyond that — is plausible but inferred from the same transport data, not measured directly. The Thomas-Fermi model is fit with three parameters and does not predict the crossover; to the authors' credit, they say so plainly in the conclusion. That honesty matters. The crossover observation itself does not depend on the model, so I don't see circularity as a fatal flaw.\n\nBottom line: the empirical study deserves a serious referee. It is publishable after revision, but the revision must either soften the design claim or back it with multiple devices, direct thickness determination by a second method, and propagated uncertainties. I would bring it to a reading group and would cite the crossover data if I worked on vdW junctions.","headline":"A creditable DC transport study with a genuine thickness-driven crossover, but the 'accurate qubit design' claim rests on one device and unpropagated ±1-layer thickness uncertainty.","tokens_in":16889,"tokens_out":1244,"would_cite":true,"duration_ms":13420,"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":"This paper claims that the thickness of a semiconducting weak link in a vertical van der Waals Josephson junction controls whether the junction behaves as a proximity-type or tunneling-type element, with the crossover near seven atomic…","keywords":["Josephson junctions","van der Waals heterostructures","WSe2","NbSe2","transmon qubits","proximity effect","tunneling","merged-element transmon"],"falsifier":"Measure the actual carrier density profile across a WSe2 stack (e.g., by scanning capacitance or momentum-resolved photoemission) and determine the layer count by a non-AFM technique such as second-harmonic generation; if the doped-layer count is not about three per electrode, or if the crossover moves away from seven layers when thickness is reassigned, the design rule fails. A simpler check is to fabricate a second MET from an independently measured JJ and see whether the predicted $f_{01}$ still falls within 10%.","tokens_in":15784,"feed_emoji":"⚛️","tokens_out":5703,"duration_ms":45100,"temperature":0.7,"pith_summary":"This paper reports a systematic study of vertical Josephson junctions made by sandwiching semiconducting WSe2 between superconducting NbSe2 electrodes. As the WSe2 weak link is thickened from 3 to 18 atomic layers, the junction switches from proximity-type behavior, where the weak link is conductive, to tunneling-type behavior, where it acts as an insulating barrier, with the crossover near seven layers. The authors use the exponential dependence of critical current on layer count to predict the transition frequency of a compact 'merged-element' transmon qubit, and a prototype built with a 17-layer weak link comes within 10% of the predicted frequency. This suggests that crystalline van der Waals junctions can serve as reproducible, compact superconducting qubit elements.","feed_headline":"7 WSe2 layers flip a junction from proximity to tunneling","feed_subtitle":"Vertical NbSe2/WSe2/NbSe2 junctions switch by layer count; a prototype transmon matches design within 10%.","key_machinery":"The central object is the vertical van der Waals Josephson junction NbSe2/WSe2/NbSe2, in which the WSe2 layer count acts as a thickness knob. The mechanism carrying the argument is band-alignment-driven charge transfer: ab initio estimates place the NbSe2 Fermi level 1.07 eV below the WSe2 valence band edge, so each electrode hole-dopes about three WSe2 layers, making thin weak links fully conductive (proximity-type) while thicker weak links leave an undoped central tunnel barrier (tunneling-type). The paper also uses the Ambegaokar–Baratoff limit and the RCSJ washboard model to interpret the crossover, and the standard transmon formula $f_{01} = (\\sqrt{8E_JE_C}-E_C)/h$ to turn the measured $j_{s0}$ into a qubit frequency.","core_discovery":"The central claim is that in NbSe2/WSe2/NbSe2 vertical Josephson junctions, the number of WSe2 atomic layers controls the junction regime: below about seven layers the junction is proximity-type with a conductive, hole-doped weak link, and above seven layers it is tunneling-type with an undoped semiconducting barrier. The crossover is evidenced by switching-to-retrapping current ratios near unity for thin weak links, exponentially growing hysteresis for thicker ones, and the temperature dependence of the critical-current–resistance product crossing the Ambegaokar–Baratoff limit near the same thickness. The underlying mechanism is proposed to be band-alignment-driven charge transfer, with each NbSe2 electrode hole-doping roughly three WSe2 layers; the remaining undoped layers form the tunnel barrier. Because the critical current density varies exponentially with layer count over six orders of magnitude, the DC transport data can be used to design qubit frequencies, and a prototype all-crystalline merged-element transmon with a 17-layer WSe2 weak link shows a 0→1 transition at about 5.30 GHz, within 10% of the calculated value.","pith_inferences":["A direct measurement of the carrier density profile across the WSe2 stack (e.g., by capacitance or angle-resolved photoemission) would test the inferred ~3-layer doping per electrode; if the doped-layer count is actually larger or smaller, the layer-number design rule for transmon frequencies would shift.","The crossover may be exploitable as a materials-design principle: choosing a semiconductor with a different valence-band offset relative to NbSe2 should move the proximity-to-tunneling threshold, as the MoS2 comparison suggests, offering a second knob beside thickness.","If the low-damping tunnel regime is confirmed in microwave spectroscopy (e.g., through qubit coherence times), crystalline vdW junctions could replace amorphous AlOx barriers in compact transmon architectures, reducing two-level-system losses from grain boundaries and pinholes.","Independent layer-count determination via second-harmonic generation or cross-sectional imaging would tighten the design rule, since AFM thickness uncertainty of about one layer is comparable to the crossover width."],"forward_implications":["If the crossover at ~7 layers is a general feature of superconducting/semiconducting vdW pairs, then the thickness of a semiconducting weak link alone can be used to select between a low-loss tunnel junction and a highly transmissive proximity junction.","The exponential dependence of $j_{s0}$ on layer count means qubit transition frequencies can be tuned across 1–15 GHz by choosing 15–20 layers of WSe2, with a footprint set by the junction's own capacitance.","The measured $f_{01}$ within 10% of prediction validates that DC transport characterization of vdW JJs can be used as a design tool for superconducting qubits.","Because $Q^*$ (the hysteresis-derived quality factor) grows exponentially with WSe2 thickness up to ~$10^4$, thicker barriers may offer lower microwave loss, though Joule heating limits the DC estimate.","Substituting MoS2 for WSe2 shifts the crossover to a smaller thickness, indicating that band alignment, not just barrier height, controls junction damping."],"supporting_citations":[{"why":"Supplies the workfunction–ionization-potential difference of 1.07 eV between NbSe2 and WSe2 that underlies the band-alignment-driven doping model.","marker":"[43]"},{"why":"Provides the relative band alignment of WSe2 and MoS2 to NbSe2 used to explain the crossover and the MoS2 comparison.","marker":"[44]"},{"why":"Gives the Ambegaokar–Baratoff limit for $I_cR_n$ that separates proximity-type (above) from tunneling-type (below) behavior.","marker":"[34]"},{"why":"Establishes that minimal hysteresis near $I_{s0}/I_{r0}\\approx1$ is typical of proximity-type Josephson junctions.","marker":"[31]"},{"why":"Introduces the RCSJ model and washboard potential used to interpret hysteresis in tunneling-type junctions.","marker":"[32]"},{"why":"Provides the quasiparticle tunneling model that explains the nonmonotonic temperature dependence of retrapping current in tunneling-type junctions.","marker":"[42]"},{"why":"Defines the transmon qubit formula $f_{01}=(\\sqrt{8E_JE_C}-E_C)/h$ and the transmon regime used to convert junction parameters into qubit frequencies.","marker":"[1]"},{"why":"Introduces the merged-element transmon concept that this paper realizes with a crystalline vdW junction.","marker":"[4]"},{"why":"Provides hBN tunneling resistance values used to compare the frequency tunability of WSe2 versus hBN weak links.","marker":"[28]"}],"fun_headline_variants":["Thickness of WSe2 decides junction type for compact qubits","Layer count flips NbSe2/WSe2 junction from proximity to tunneling","WSe2 layer number sets junction regime, enabling transmon qubit","Crossover in 2-12 nm WSe2 junctions for compact superconducting qubits","All-crystalline transmon with tunable junction from layer thickness"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The design rule assumes that each NbSe2 electrode hole-dopes roughly three WSe2 layers, a number inferred from the same DC transport data rather than measured directly, and that the AFM-determined layer counts are accurate to about one layer; the paper's own Thomas-Fermi model does not predict the crossover.","fun_headline_variants_meta":{"raw":{"variants":["Thickness of WSe2 decides junction type for compact qubits","Layer count flips NbSe2/WSe2 junction from proximity to tunneling","WSe2 layer number sets junction regime, enabling transmon qubit","Crossover in 2-12 nm WSe2 junctions for compact superconducting qubits","All-crystalline transmon with tunable junction from layer thickness"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000567,"raw_usage":{"total_tokens":2676,"prompt_tokens":925,"completion_tokens":1751,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":541,"completion_tokens_details":{"reasoning_tokens":1654}},"tokens_in":541,"tokens_out":1751,"duration_ms":12234,"temperature":1.0,"reasoning_tokens":1654,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T14:44:56.413663+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the actual carrier density profile across a WSe2 stack (e.g., by scanning capacitance or momentum-resolved photoemission) and determine the layer count by a non-AFM technique such as second-harmonic generation; if the doped-layer count is not about three per electrode, or if the crossover moves away from seven layers when thickness is reassigned, the design rule fails. A simpler check is to fabricate a second MET from an independently measured JJ and see whether the predicted $f_{01}$ still falls within 10%.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the workfunction–ionization-potential difference of 1.07 eV between NbSe2 and WSe2 that underlies the band-alignment-driven doping model."},{"cited_title":"Guo and J","cited_arxiv_id":null,"evidence_quote":"Provides the relative band alignment of WSe2 and MoS2 to NbSe2 used to explain the crossover and the MoS2 comparison."},{"cited_title":"Ambegaokar and A","cited_arxiv_id":null,"evidence_quote":"Gives the Ambegaokar–Baratoff limit for $I_cR_n$ that separates proximity-type (above) from tunneling-type (below) behavior."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes that minimal hysteresis near $I_{s0}/I_{r0}\\approx1$ is typical of proximity-type Josephson junctions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the quasiparticle tunneling model that explains the nonmonotonic temperature dependence of retrapping current in tunneling-type junctions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the merged-element transmon concept that this paper realizes with a crystalline vdW junction."},{"cited_title":"Britnell, R","cited_arxiv_id":null,"evidence_quote":"Provides hBN tunneling resistance values used to compare the frequency tunability of WSe2 versus hBN weak links."}],"review_version":1}