{"id":"394a19f6-46bd-4043-a0fd-79d238c96b51","arxiv_id":"2608.12508","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"The single-particle gap in disordered superconductors survives and grows across the superconductor-insulator transition, and fills rather than closes with temperature.","lead":"Tunneling measurements on disordered indium oxide films show that the superconducting gap survives deep into the insulating phase and grows with disorder, confirming a long-predicted Cooper-pair insulator. Combined with quantum Monte Carlo simulations, the data link the pseudogap above Tc and the insulating gap to pairing without global phase coherence.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The identification of the measured insulating gap as a Cooper-pair gap is not tested against Coulomb-blockade or the validity of the H=9T normalization reference.","rationale":"The reader's weakest assumption was that the measured tunneling gap arises from local Cooper pairing rather than from Coulomb blockade or charging effects. My analysis converges on the same load-bearing concern, sharpened by the observation that the H=9T normalization reference may itself be gapped in the insulating state, which would make the normalized gap an artifact even before any fitting to a pairing model. This is the same concern because both routes (Coulomb blockade and an invalid reference) undermine the pairing-gap identification. The paper provides no control measurement on a non-superconducting disordered insulator and no full-spectrum comparison of the 0T and 9T densities of states, so the central claim is conditional rather than established. The reader's CONDITIONAL verdict is therefore appropriate; my stress test does not move the verdict, but it identifies a specific, checkable experiment that would resolve the interpretation. I find no internal logical inconsistency in the argument, and the DQMC simulations independently support the trend, so a REJECT or UNVERDICTED verdict is not warranted at this stage.","tokens_in":10981,"tokens_out":5764,"duration_ms":65502,"concrete_test":"Inspect the raw dI/dV(V) spectrum at H=9T for the most insulating sample (Fig. 5). If this 'normal-state' reference itself exhibits a zero-bias dip or any energy-dependent suppression, then the normalized conductance dI/dV(0T)/dI/dV(9T) will acquire spurious structure, and the reported gap cannot be attributed to an intrinsic zero-field Cooper-pair gap. If the H=9T spectrum is flat over the relevant energy range, the Coulomb-blockade and field-reference concerns are substantially weakened, and the pairing interpretation gains support.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the insulating-phase tunneling gap is direct evidence for localized Cooper pairs requires that the gap extracted from normalized conductance dI/dV(0T)/dI/dV(9T) be an intrinsic zero-field single-particle feature. Two assumptions are load-bearing and untested in the paper. First, the H=9T spectrum is used as a 'normal state' reference, but in the deep insulating regime the high-field state is still an insulator, not a metal; if this reference itself contains a gap-like suppression (from Coulomb blockade, magnetoresistance, or field-dependent localization), dividing by it will imprint structure in the normalized spectrum that is not a zero-field pairing gap. Appendix A and Fig. 6 only address zero-bias conductance versus field orientation, not the full field-dependent DOS or the shape of the 9T reference. Second, the gap is extracted by fitting to the Dentelski et al. fluctuating-superconductivity model, which already presupposes pairing; a single-electron charging gap could plausibly produce similar spectral features, and no control experiment or observable is presented to exclude that alternative. The paper's own stated discrepancy in the spectral-weight redistribution energy scale (two orders of magnitude between experiment and theory) further weakens the 'quantitative' confirmation, but the core unresolved issue is whether the measured gap is a pairing gap at all. Until a test distinguishes Cooper-pairing from Coulomb-blockade or validates the 9T reference, the headline conclusion remains conditional.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports tunneling spectroscopy of amorphous indium oxide films driven deep into the insulating side of the disorder-driven superconductor–insulator transition (SIT), using planar junctions with resistances up to 100 MΩ and picoampere current sensitivity, combined with determinant quantum Monte Carlo (DQMC) simulations of the disordered attractive Hubbard model. The normalized conductance (dI/dV at 0 T divided by the spectrum at 9 T) shows a single-particle gap that survives across the transition and grows with disorder, reaching values more than twice those on the superconducting side; with increasing temperature the gap fills rather than closes, and spectral weight is redistributed to high energies. The DQMC results exhibit the same qualitative trends, and the paper interprets the combined observations as direct quantitative confirmation of a Cooper-pair insulating state—local pairing without global phase coherence—and as a unifying link to pseudogap physics in cuprates.","tokens_in":11265,"tokens_out":13138,"duration_ms":122220,"significance":"If the identification of the measured gap as a pairing gap holds, this is a significant advance: it is, to my knowledge, the first tunneling spectroscopic access well into the insulating phase of the SIT, a regime previously inaccessible because of the junction-resistance and current-sensitivity constraints that the authors have overcome. The theoretical side is genuinely independent: the DQMC parameters (U/t, V/t, μ, temperature) are inputs and are not fitted to the data, the method is sign-problem-free for the attractive Hubbard model, and the predicted gap growth with disorder agrees in trend with the experiment. The temperature-filling behavior and its similarity to cuprate pseudogap phenomenology are interesting and add breadth. These strengths are real. However, \"direct quantitative experimental confirmation\" is stronger than the evidence supports: the high-energy redistribution scale is admitted by the authors to disagree by nearly two orders of magnitude, and the pairing-gap identification rests on a normalization reference and a fitting model whose validity deep in the insulating regime is not established.","major_comments":[{"comment":"The normalized conductance is formed as dI/dV(0 T)/dI/dV(9 T), and the text refers to the 9 T curves as \"normal state\" spectra, but the paper nowhere demonstrates that at 9 T, deep in the insulating regime, the single-particle DOS is featureless. Appendix A and Fig. 6 track only the zero-bias conductance versus field orientation; the full field-dependent DOS and the bias-dependent shape of the 9 T reference are not shown. In the deep insulator the high-field state is itself an insulator, so a gap-like suppression in the 9 T reference (from charging effects or field-dependent localization) would directly distort the normalized spectrum and the extracted gap, and the distortion could evolve with disorder. Because the growing-gap trend in Fig. 1(b) is the central experimental result, the validity of the 9 T reference is load-bearing.","section":"The gap at high disorder; Appendix A and Figs. 5–6"},{"comment":"The gap values in Fig. 1(b) are extracted by fitting to the Dentelski et al. fluctuating-superconductivity model (ref. 40), whose functional form presupposes a locally paired state. The manuscript provides no test of whether a single-electron charging (Coulomb-blockade) gap or another non-pairing mechanism can account for the same spectra, no control experiment or distinguishing observable, and no fit-quality or residual analysis. As a result, the central interpretation—that the measured insulating gap is a Cooper-pair gap—is built into the analysis rather than independently established, and the abstract's wording \"direct quantitative experimental confirmation\" overstates what the measurement alone demonstrates. The authors should compare the fits against at least one non-pairing gap model and report goodness-of-fit for both.","section":"The gap at high disorder (Fig. 1b); Appendix A"},{"comment":"In the spectral-weight redistribution section the authors state that the energy scale \"differs significantly\" between theory and experiment—the calculated shift is of order Δ whereas the experimental redistribution \"extends nearly two orders of magnitude\"—and they attribute the difference to the finite numerical bandwidth. This self-stated limitation directly contradicts the abstract's claims of \"striking agreement\" and \"quantitative\" confirmation for this part of the data. Moreover, the dc bias was swept only over ±100 mV, so the \"two orders of magnitude\" scale may be constrained by the measurement window rather than by the underlying physics. The manuscript should either provide a quantitative metric for the agreement on the spectral-weight redistribution or restrict the quantitative-confirmation claim to the gap magnitude and its temperature filling.","section":"Redistribution of spectral weights (Fig. 3)"}],"minor_comments":[{"comment":"The main text states that the trend in Fig. 1(b) \"was found for nine measured films,\" while Appendix A states that \"Four samples were measured\"; the relationship between films and samples (e.g., number of junctions per film, annealing stages) should be clarified.","section":"The gap at high disorder; Appendix A"},{"comment":"The sentence \"fitting it to a model published by David et al. in 2018\" misnames the authors of ref. 40, which is the Dentelski et al. model; in addition, the main text contains a stray parenthesis in \"(see supplementary material) for details)\" and repeatedly references a supplementary material that is not present in the submission, although Appendices A and B partially cover the same ground.","section":"Appendix A"},{"comment":"Several typos remain: \"insulting regime\" should be \"insulating regime\" in the gap-growth paragraph; \"This poses a question as to regarding the conservation of spectral density\" contains a duplicated preposition; \"Fig. 2 (b) depicts ... calculations. showing that\" has a lowercase word after a period; and \"InO/AlO/Alpla planar tunnel junctions\" appears to be missing a letter in the top electrode metal.","section":"Throughout"},{"comment":"The caption states that the gap is calculated \"at T=0\" using quantum Monte Carlo, but DQMC is a finite-temperature method and the simulations in Appendix C are presented at low but nonzero T/t=0.1; the caption should specify the temperature or the extrapolation procedure used.","section":"Fig. 1(d)"},{"comment":"The gap extraction points in Fig. 1(b) are shown without error bars or any measure of fit uncertainty; since the gap values are the core experimental quantity, the authors should report the fitting uncertainty and the number of spectra averaged per disorder level.","section":"Fig. 1(b)"}],"recommendation":"major_revision","confidential_remarks":"To the editor: the two decisive issues are (i) the unvalidated 9 T normalization reference and (ii) the fact that the gap-extraction model already presupposes pairing; both are addressable with data the authors appear to have in hand (field-dependent spectra, alternative model fits) or with a careful discussion, so I view major revision as the appropriate disposition rather than rejection. One citation-pattern note: the Dentelski et al. model (ref. 40) used for gap extraction includes a present coauthor (A. Frydman); the paper should state the model's provenance and its treatment of non-pairing gaps. The abstract's quantitative-confirmation claim should be softened regardless of the outcome of the added analysis."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: this is the first tunneling spectroscopy deep in the insulating phase of a disorder-driven SIT, and the headline trend — gap grows with disorder, fills with temperature — is credible. What is less clean is the stronger claim, that this is direct evidence for localized Cooper pairs; two load-bearing assumptions are not tested against the obvious alternative.\n\nWhat is genuinely new: nobody has run planar tunneling at ~100 MΩ junction resistance into films that far on the insulating side; previous STM and planar work stopped near the transition. The gap growth, across multiple films with annealing-controlled disorder, is a real experimental result. The DQMC work is real support: the sign-problem-free attractive Hubbard model with disorder reproduces both gap growth and gap filling. Credit also for honesty: the paper itself admits the spectral-weight redistribution energy scale differs from theory by two orders of magnitude. That is a large discrepancy, and it undercuts the abstract's 'striking agreement' and 'quantitative confirmation' language.\n\nSoft spots, in proportion. First, the normalization. The tunnel conductance at 0 T is divided by the spectrum at 9 T to define the normal state. Deep in the insulator, 9 T is still an insulator. If that reference has its own zero-bias suppression — charging, magnetoresistance, field-enhanced localization — the division imprints gap-like structure that is not a zero-field pairing feature. Appendix A and Fig. 6 address vortices and zero-bias conductance but not the shape of the full 9 T density of states. Second, the gap is extracted by fitting the Dentelski et al. fluctuating-superconductivity model, which presupposes pairing. A Coulomb-blockade gap also grows with disorder and fills with temperature; no control experiment separates the two. Third, the predicted gap growth comes from the same group's earlier work (Bouadim et al.), so the confirmation is in-house. That is not disqualifying — the measurement is independent — but it raises the bar for the fit and the normalization to be beyond reproach.\n\nMinor: no error bars on the fitted gaps, no posted data or code, and a confusing sample count (the text says nine films, Appendix A says four samples with annealing stages).\n\nWho it is for: the SIT and Cooper-pair-insulator community, and anyone interested in whether pseudogap phenomenology can arise from pairing alone. I would send it to referees. The experiment clears a real barrier, the trend is credible, and the open questions — the 9 T reference, the charging alternative, the energy-scale mismatch — are exactly what referees should push on.","headline":"First tunneling spectroscopy deep in the insulating side of the SIT; the gap-growth trend is credible, but the Cooper-pair reading rests on an untested 9 T normalization and a pairing-presupposing fit.","tokens_in":11767,"tokens_out":6256,"would_cite":true,"duration_ms":55149,"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":"The single-particle gap survives across the superconductor–insulator transition and grows with disorder to more than twice its superconducting-side value, while filling rather than closing as temperature rises.","keywords":["pseudogap","superconductor-insulator transition","Cooper-pair insulator","localized Cooper pairs","tunneling spectroscopy","amorphous indium oxide","attractive Hubbard model","quantum Monte Carlo"],"falsifier":"Measure the extracted gap as a function of the film's charging energy, or attempt to destroy pairing by applying a magnetic field while keeping disorder fixed: if the gap persists when pairing is suppressed, or tracks the grain charging energy $e^2/2C$, the Cooper-pair interpretation fails; if it disappears or follows the predicted localization scaling, the interpretation holds.","tokens_in":1537,"feed_emoji":"🔬","tokens_out":1657,"duration_ms":55948,"temperature":0.7,"pith_summary":"This paper asks whether a gap in the single-particle spectrum can arise purely from pairing without global phase coherence. It answers yes: tunneling measurements on amorphous indium oxide films, taken deep in the insulating regime, show a single-particle gap that survives the loss of superconductivity and grows with disorder, reaching more than twice the superconducting-side value. Quantum Monte Carlo simulations of the disordered attractive Hubbard model reproduce the same growth. The paper interprets this as spectroscopic evidence for a Cooper-pair insulator, where electrons remain paired in the valleys of the disorder potential but the phase of the order parameter is not coherent across the film. This unifies the pseudogap above $T_c$ and the insulating gap as two manifestations of pairing without global phase coherence.","feed_headline":"Gap grows as superconductivity dies in disordered films","feed_subtitle":"Tunneling and simulations show the gap persists and widens in the insulator, evidence for localized Cooper pairs.","key_machinery":"The load-bearing object is the Cooper-pair insulating state: electrons bind into local pairs in the minima of the random disorder potential, forming small superconducting islands whose phases are not coherent with one another. Experimentally, the enabling tool is ultra-high-resistance planar tunnel junctions, around 100 MΩ, with picoampere current sensitivity, which permit tunneling spectroscopy on films too resistive for scanning tunneling microscopy. The gap is extracted from the measured spectra using a model that accounts for spatial and temporal fluctuations of the superconducting order parameter. Theoretically, the argument is carried by determinant quantum Monte Carlo simulations of the disordered attractive Hubbard model, which include both quantum and thermal phase fluctuations and yield the density of states through maximum-entropy analytic continuation.","core_discovery":"The central discovery is that the single-particle excitation gap does not vanish when superconductivity is destroyed by disorder; instead it persists and grows monotonically with disorder, by a factor of more than two. The gap has a distinctive temperature signature: as temperature rises, states fill in below the gap and coherence peaks are suppressed, but the gap edge barely moves until a temperature $T^*$ where it disappears, which is not the BCS closing-gap behavior. The same behavior appears in determinant quantum Monte Carlo simulations of the attractive Hubbard model with random on-site disorder. The paper concludes that the underlying state is a Cooper-pair insulator with localized pairs trapped in potential minima and no global phase coherence, and proposes that the pseudogap in cuprates and the insulating gap in these films share the mechanism of pairing without phase coherence.","pith_inferences":["A direct test of the pairing interpretation would be to measure the same films in a magnetic field strong enough to suppress pairing: a pairing gap should evolve or close at a pair-breaking field, whereas a Coulomb-blockade gap would be insensitive to such a field; the paper does not report this check.","The localization-based expression $\\Delta \\propto U/2\\xi^2$ implies a quantitative relationship between the extracted gap and the disorder-driven localization length; measuring both on the same film would test whether the growth rate matches the mechanism.","The same ultra-high-resistance planar junction technique could be applied to other amorphous superconductors or thickness-tuned films to see whether the paired-insulator signature is a general feature of the disorder-driven superconductor–insulator transition."],"forward_implications":["In the paired-insulating regime, the single-particle spectrum remains gapped up to a disorder-dependent temperature $T^*$, so a film can be electrically insulating while still showing a spectroscopic pairing gap.","Gap filling with temperature, long considered a fingerprint of cuprates, can appear in a weak-coupling s-wave superconductor once disorder suppresses phase coherence.","Spectral weight removed from the gap is redistributed to energies well above the gap scale, with the high-energy scale set by the disorder landscape.","Ultra-high-resistance planar junctions extend tunneling spectroscopy deep into the insulating phase, a regime that was previously inaccessible to STM.","The agreement between experiment and quantum Monte Carlo strengthens the case that local pairing alone, without competing electronic orders, can produce pseudogap phenomenology."],"supporting_citations":[{"why":"Theoretical prediction that the single-particle gap grows in the paired insulating state; this paper's experimental trend confirms that prediction.","marker":"[39]"},{"why":"The fluctuating-order-parameter model used to fit the measured tunneling spectra and extract the single-particle gap.","marker":"[40]"},{"why":"Tunneling experiments establishing localization of preformed Cooper pairs near the superconductor–insulator transition.","marker":"[20]"},{"why":"Earlier measurement of a superconducting energy gap in a homogeneously amorphous insulator, providing the first evidence of a gap on the insulating side.","marker":"[37]"},{"why":"Study of Coulomb interactions in the disorder-driven superconductor–insulator transition that frames the interpretation of the insulating gap.","marker":"[38]"},{"why":"Reports gap filling rather than closing with temperature in cuprates; used as the comparison for the non-BCS temperature behavior observed here.","marker":"[41]"},{"why":"Theory of inhomogeneous pairing and superconducting puddles in highly disordered s-wave superconductors, underlying the island picture in the disorder landscape.","marker":"[42]"}],"fun_headline_variants":["Disorder makes the gap grow, not vanish, in superconductors","Cooper pairs linger in insulating phase, gap doubles","Gap widens as superconductivity fades in disordered films","Insulating phase reveals gap from localized Cooper pairs","Pseudogap survives—and grows—without superconductivity"],"cache_read_input_tokens":13952,"weakest_assumption_plain":"The interpretation rests on the fitted tunneling gap's being a pairing gap from localized Cooper pairs; the paper does not explicitly rule out single-electron Coulomb-blockade or charging effects in the strongly insulating film.","fun_headline_variants_meta":{"raw":{"variants":["Disorder makes the gap grow, not vanish, in superconductors","Cooper pairs linger in insulating phase, gap doubles","Gap widens as superconductivity fades in disordered films","Insulating phase reveals gap from localized Cooper pairs","Pseudogap survives—and grows—without superconductivity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000185,"raw_usage":{"total_tokens":1304,"prompt_tokens":911,"completion_tokens":393,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":527,"completion_tokens_details":{"reasoning_tokens":313}},"tokens_in":527,"tokens_out":393,"duration_ms":3931,"temperature":1.0,"reasoning_tokens":313,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:06:19.927911+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the extracted gap as a function of the film's charging energy, or attempt to destroy pairing by applying a magnetic field while keeping disorder fixed: if the gap persists when pairing is suppressed, or tracks the grain charging energy $e^2/2C$, the Cooper-pair interpretation fails; if it disappears or follows the predicted localization scaling, the interpretation holds.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Theoretical prediction that the single-particle gap grows in the paired insulating state; this paper's experimental trend confirms that prediction."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The fluctuating-order-parameter model used to fit the measured tunneling spectra and extract the single-particle gap."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Tunneling experiments establishing localization of preformed Cooper pairs near the superconductor–insulator transition."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier measurement of a superconducting energy gap in a homogeneously amorphous insulator, providing the first evidence of a gap on the insulating side."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Study of Coulomb interactions in the disorder-driven superconductor–insulator transition that frames the interpretation of the insulating gap."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports gap filling rather than closing with temperature in cuprates; used as the comparison for the non-BCS temperature behavior observed here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Theory of inhomogeneous pairing and superconducting puddles in highly disordered s-wave superconductors, underlying the island picture in the disorder landscape."}],"review_version":1}