{"id":"82619567-0945-4558-9419-e80447c2a681","arxiv_id":"2509.00281","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In the half-filled 2D Hubbard model, the charge diffusion constant follows a robust 1/sqrt(T) scaling across the metal-insulator crossover, with a new Pseudogap Metal state where charge compressibility is insulator-like but transport remains metallic.","lead":"This paper computes the optical conductivity of the half-filled two-dimensional Hubbard model using a combination of diagrammatic Monte Carlo and self-consistent diagrammatic theory, and reports a universal 1/sqrt(T) growth of the charge diffusion constant across the metal-insulator crossover. The result suggests that strange-metal transport may share a common diffusive signature in a canonical model of interacting electrons, and identifies a distinct Pseudogap Metal regime w","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"NAC stretch-test error bars are a smoothness-based bound, not a uniqueness certificate; a second continuation method could shift σ(0) and change D(T).","rationale":"The reader identified the numerical analytic continuation as the weakest assumption, and I agree. The paper's central claims—the universal D ~ T^{-1/2} scaling and the pseudogap-metal classification—both rely on the DC limit of σ(ω), which cannot be uniquely determined from Matsubara data. The stretch test is a well-motivated and internally consistent way to assign error bars, but it only probes the family of smooth spectra. The SM explicitly states that the error bar is defined by where the smooth functional form changes, not by where all admissible spectra are excluded. The additional checks in the paper (SOM agreement at one point, Bold4 tail substitution) are helpful but do not systematically rule out a T-dependent continuation bias. This is a correctness risk, not a claim about the authors' methods being invalid. The right response is to demand a reproducibility check: either release the Matsubara data or run an independent continuation method. Because the paper is otherwise careful, with controlled DiagMC error bars and a plausible physical picture, the conditional verdict is appropriate; I would not upgrade the concern to a rejection without seeing the benchmark fail.","tokens_in":16554,"tokens_out":4906,"duration_ms":61297,"concrete_test":"Use the same Λ(iω_n) data and error bars as in Fig. 2 for all (U,T) points and perform an independent analytic continuation with a method that does not impose global smoothness—e.g. minimal-pole/Nevanlinna (Refs. 34–38) or sparse modeling (Refs. 30–31)—under the same high-frequency cutoff. Compare the inferred σ(0) and D(T). If any point moves by more than the quoted stretch-test error, or if the fitted D(T) exponent changes by more than ~0.1, the 1/√T claim is not robust. To make the test decisive, also run the same two methods on synthetic Λ(iω_n) generated from a Drude-plus-continuum σ(ω) with known σ(0) and noise matching the DiagMC error bars.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim D(T) ~ T^{-1/2} is inherited from σDC(T) via Eq. (2), and σDC is the output of a numerical analytic continuation of Λ(iω_n). The paper's own SM Sec. III defines the stretch-test error bar as the range of target σ(0) for which the MCC solution does not develop visible wiggles relative to the 'as-smooth-as-possible' spectrum. That is a smoothness-prior sensitivity test, not a statement about all spectra consistent with Λ(iω_n) within the DiagMC errors. The Fredholm kernel in Eq. (4) has small eigenvalues, so spectra with different σ(0) and different high-frequency structure can reproduce the same Matsubara data. If the true σ(ω) is less smooth than the MCC solution—for example a narrower Drude peak at low T—the smoothness constraint can systematically bias σ(0), and the bias can be T-dependent, faking or distorting a 1/√T law. The agreement with SOM for a single (T=0.2, U=3.2) point and the asymptotic matching to Bold4 for ω_n>15 do not close this gap: Bold4's own systematic error is admitted in SM Sec. III, and the ad hoc O(1/ω_n^3) uncertainty on the tail enters the inversion on the same footing as physical information. Since the universality claim rests on the scaling exponent of D(T) over a decade of T, an unquantified continuation systematic is load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies charge transport across the metal-insulator crossover in the half-filled two-dimensional Hubbard model. The authors compute the current-current correlation function directly in the thermodynamic limit with connected determinant diagrammatic Monte Carlo (CDet) up to order 8-10, combine it with a self-consistent Bold4 high-frequency tail, and use numerical analytic continuation (SOCC/MCC with a 'stretch test') to obtain the optical conductivity. Combining the DC conductivity with the compressibility computed earlier by the same group, they extract the charge diffusion constant D = sigma_DC / kappa. The central claim is that, across a broad temperature range where the DC resistivity has anomalous scaling rho_DC ~ T^alpha with 0<alpha<1, the diffusion constant displays a robust ~1/sqrt(T) 'strange metal' behavior. They also identify a 'Pseudogap Metal' regime characterized by insulating charge compressibility coexisting with metallic transport, and analyze the diagrammatic origin of this behavior, finding that opposite-spin vertex corrections transfer Drude weight to a high-frequency continuum.","tokens_in":16976,"tokens_out":7144,"duration_ms":89606,"significance":"If the reported 1/sqrt(T) law holds, it would provide a striking universal diffusive signature of incoherent transport in the half-filled 2D Hubbard model, with a direct connection to cold-atom measurements in the doped system. The numerical work is careful: the CDet series are pushed to high order and extrapolated with Pade approximants, error bars are propagated into observables, and the analytic continuation is guarded by a stretch test and cross-checked against SOM at one representative point. The diagrammatic decomposition also benefits from an exact order-by-order selection rule for spin-resolved vertex contributions. These strengths make the paper a serious candidate for publication, provided the principal scaling claim is robust against the known ill-posedness of the analytic continuation.","major_comments":[{"comment":"The stretch test defines the sigma(0) error bar as the range over which the MCC spectrum does not develop visible wiggles; this is a smoothness-prior sensitivity test, not a uniqueness bound. Because the kernel K(i omega_n, omega) has small eigenvalues, non-smooth spectra with different sigma(0) can fit Lambda(i omega_n) within the DiagMC error bars. The main claim D(T) ~ T^{-1/2} is obtained from sigma_DC(T)/kappa(T) across a range of T, so a T-dependent continuation bias could fake or distort the power law. Please cross-validate with an independent continuation method (e.g., MaxEnt, SOM, sparse modeling, or Nevanlinna continuation) at enough (U,T) points to map D(T), and/or demonstrate with synthetic spectra that non-smooth solutions consistent with Lambda(i omega_n) change D(T) by less than the quoted error.","section":"SM Sec. III, Eq. (4)"},{"comment":"The high-frequency tail for omega_n > 15 is replaced by Bold4 values with an ad hoc O(1/omega_n^3) uncertainty. This tail enters as data in Eq. (4) and can influence the low-frequency inversion through overall normalization and constraints. The systematic difference between Bold4 and the exact CCF is acknowledged but is not propagated into the final D(T) error bars in a well-controlled way. Please test the sensitivity of sigma(0) to the substitution threshold and to the assumed tail-error model, and include the resulting uncertainty in the reported diffusion constant.","section":"SM Sec. III, Fig. S2(a)"},{"comment":"The central claim of a robust ~1/sqrt(T) scaling is supported visually by reference lines, but no fitting procedure, exponent uncertainty, or temperature-window definition is reported. Provide a power-law fit D(T) = A T^{-gamma} for each U with statistical errors, state the criterion for the fitting window (e.g., DM1 through PGM), and report gamma and its uncertainty. This is needed to substantiate the universality statement and to make the claimed T^{-1/2} exponent falsifiable.","section":"Fig. 2(c), Table I"}],"minor_comments":[{"comment":"The panel labels and caption are confusing: both panels appear to be labeled 'a' and 'b', and the inset showing sigma_DC * Gamma_tr and D * Gamma_tr is not fully described. Please clarify the figure layout and the definition of Gamma_tr in the caption.","section":"Fig. 3"},{"comment":"The temperature T_II is defined through the inflection point of a 5th-order polynomial fit to d kappa / d log T. Please comment on the stability of T_II with respect to the fit order and the fitting range.","section":"SM Sec. IV"},{"comment":"The quoted ranges of the resistivity exponent alpha (0 <~ alpha <~ 0.5 and 0.5 <~ alpha <~ 1) are not tied to explicit fits. These ranges should be obtained from and reported with the same fitting procedure used for the diffusion constant.","section":"Table I"},{"comment":"Reference [48] is an arXiv preprint; if a published version exists, please cite it.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is technically strong and the scientific claim is potentially important. The main obstacle is the analytic-continuation systematic: the stretch test is a smoothness test, not a uniqueness certificate, and the central 1/sqrt(T) scaling is read off from sigma_DC obtained by analytic continuation. If the authors can add independent continuation cross-checks and a quantitative exponent uncertainty, I would be supportive. The paper fits the journal's scope well."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper you'd want to know about: it reports a clean D ~ T^{-1/2} scaling of the charge diffusion constant in the half-filled 2D Hubbard model over a wide temperature window, obtained by combining connected determinant Monte Carlo (CDet) with a self-consistent Bold4 scheme for the high-frequency tail and then doing numerical analytic continuation (NAC). That scaling hasn't been seen at half-filling before, and the paper connects it to the cold-atom data on the doped side, suggesting a possible universal diffusive law for incoherent metals. The other genuinely new piece is the identification of a 'Pseudogap Metal' regime where the compressibility is insulating-like while the DC conductivity is still metallic.\n\nThe methodology is careful: the diagrammatic series converge to order 8–10 with Padé extrapolation and controlled errors; the high-frequency tail from Bold4 is substituted only where the CDet error bars make it beneficial; and the stretch test is used to define error bars on σ(0). They also cross-check one point with SOM. No fitted parameter is dressed as a prediction—D is read off the computed σ and κ. The spin-resolved decomposition of the current correlator, with an exact selection rule based on Furry's theorem, is a nice formal contribution.\n\nThe main soft spot is exactly the NAC step. The stretch test quantifies how far one can shift σ(0) before the spectrum develops wiggles under a smoothness constraint; it does not certify that the true spectrum is smooth. A less smooth spectrum—say a narrower Drude peak—could give a different σ(0), and if the bias varies with T it could distort the scaling exponent. The ad hoc O(1/ω³) error on the Bold4 tail is honest but doesn't close that gap. The absence of released data and code makes it harder to pressure-test this; a reader can't re-run the inversion or examine the stretch test protocol in detail.\n\nStill, the scaling is observed across a decade of temperature for three U values, and the compressibility is computed directly, not continued, so the D(T) behavior is not simply an artifact of the continuum inversion. The NAC concern is a real caveat, but it's a caveat, not a refutation. This paper deserves a serious referee—ideally someone deep in the NAC literature—and the referee should focus on the continuation error bars and ask for reproducibility materials.","headline":"D~T^-1/2 diffusion in the half-filled Hubbard model is the new result; the analytic continuation is the main caveat, but the paper deserves serious refereeing.","tokens_in":17426,"tokens_out":7264,"would_cite":true,"duration_ms":75594,"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 half-filled 2D Hubbard model's incoherent metal has a charge diffusion constant D ~ 1/√T across all anomalous-resistivity regimes down to the metal-insulator crossover, plus a Pseudogap Metal state where insulating compressibility coexi","keywords":["Hubbard model","strange metal","charge diffusion","Nernst-Einstein relation","diagrammatic Monte Carlo","analytic continuation","pseudogap metal","optical conductivity"],"falsifier":"A measurement of the optical conductivity in the half-filled 2D Hubbard model that does not rest on maximum-smoothness analytic continuation—for example a real-time dynamical quantum Monte Carlo calculation or a cold-atom density-response measurement at half-filling—should find σDC at U=4,T=0.4-4 within the stretch-test window; if σDC deviates significantly, then D=σ/κ will not follow the reported 1/√T law.","tokens_in":16494,"feed_emoji":"⚛️","tokens_out":5384,"duration_ms":60487,"temperature":0.7,"pith_summary":"The paper studies charge transport in the half-filled two-dimensional Hubbard model using numerically exact diagrammatic Monte Carlo combined with controlled analytic continuation. It claims that over a broad temperature range where the DC resistivity is anomalously temperature-dependent (~T^α with 0<α≲1), the charge diffusion constant D extracted from the Nernst-Einstein relation follows a robust ~1/√T law. It also identifies a Pseudogap Metal regime in which the charge compressibility is insulator-like while transport remains metallic. If correct, this establishes a simple, universal-looking diffusive signature of incoherent charge transport and sharpens the picture of how the system enters the insulating state.","feed_headline":"Half-filled Hubbard model shows 1/√T strange-metal diffusivity","feed_subtitle":"Resistivity stripped into compressibility and diffusion reveals a universal 1/√T charge-diffusion law.","key_machinery":"The central computational object is the imaginary-time current-current correlator Λ(iωn), continued via its spectral representation to the optical conductivity σ(ω) using the SOCC/MCC method with a multiscale high-frequency tail from Bold4. The central physical identity is the Nernst-Einstein relation D = σDC/κ, which particle-hole symmetry at half-filling makes exact because thermoelectric response vanishes. Its diagnostic power lies in decomposing anomalous resistivity into a universal diffusion law (1/√T) and a non-universal compressibility; the diagrammatic spin-resolved decomposition separates bubble, same-spin, and opposite-spin vertex contributions that explain the Drude-to-continuum","core_discovery":"For the half-filled 2D Hubbard model, the authors compute the current-current correlation function directly in the thermodynamic limit with diagrammatic Monte Carlo, splice the high-frequency tail with the self-consistent Bold4 diagrammatic theory, and continue to real frequencies with a stretch-test-controlled stochastic method. They find that the DC resistivity's anomalous scaling between high and low temperatures is the product of a near-universal diffusivity D∼1/√T multiplied by a compressibility that is non-universal; at lower temperatures the compressibility turns over and the system enters a Pseudogap Metal state with ∂κ/∂T>0 but metallic ρDC(T), until the metal-insulator crossover. D","pith_inferences":["One implication beyond the paper's claims is that D∼1/√T may be universal across doping, with all material-specific and model-specific variation residing in the compressibility; the cold-atom data at substantial doping already point in this direction.","A testable extension would be a real-time measurement of charge diffusion in an ultracold-atom realization at half-filling: if the Pseudogap Metal picture is right, density-response (compressibility) and cloud-expansion (transport) probes should show opposite temperature trends.","The spin-resolved vertex mechanism suggests an optical sum-rule-style diagnostic: the same-spin vertex adds spectral weight in a window set by temperature, while the opposite-spin vertex removes it from the Drude peak; this separation could be probed by spin-resolved or polarized light experiments in analog systems.","If the 1/√T law is the universal incoherent-metal signature, then future analytic-continuation studies should report D rather than only ρDC, since D is the quantity that exposes the regularity behind the non-universal resistivity exponents."],"forward_implications":["The 1/√T diffusivity extends through Diffusive Metal I, Diffusive Metal II, and the Pseudogap Metal, so in that window the anomalous resistivity exponent α is controlled mostly by the compressibility, not by the diffusion constant.","Because particle-hole symmetry makes the Nernst-Einstein extraction exact at half-filling, the reported D is a genuine charge diffusion constant rather than a thermoelectric mixture.","The Pseudogap Metal state is defined by dκ/dT>0 together with metallic transport, showing that insulating charge response and metallic conduction can coexist without an actual gap in the conductivity.","The opposite-spin vertex corrections deplete the low-frequency Drude peak and feed a high-frequency continuum, explaining why the central optical peak is suppressed before the system becomes insulating."],"supporting_citations":[{"why":"Supplies the cold-atom diffusion measurement in the doped Hubbard model whose D∼T^-0.6 behavior the paper's half-filling 1/√T result strongly resembles, motivating the universality conjecture.","marker":"[7]"},{"why":"Shows that thermoelectric power vanishes in half-filled bands, making the Nernst-Einstein relation exact and enabling D to be extracted from σDC and κ.","marker":"[12]"},{"why":"Supplies the numerically exact charge compressibility κ and spin/charge correlation data used in the Nernst-Einstein decomposition and in defining the temperature regime boundaries.","marker":"[13]"},{"why":"Documents the anisotropic single-particle self-energy and spectral pseudogap that the Pseudogap Metal regime is identified with.","marker":"[15]"},{"why":"Provides the multi-method cross-benchmark that supports the numerical exactness of the thermodynamic-limit data used as input for the transport calculation.","marker":"[17]"},{"why":"Introduces the connected determinant Monte Carlo algorithm used for high-order diagrammatic sampling of the current-current correlator.","marker":"[21]"},{"why":"Defines the stretch test and the consistent-constraints continuation method that control the systematic error on the inferred DC conductivity.","marker":"[25]"},{"why":"Introduces the Bold4 self-consistent diagrammatic theory used for the non-stochastic high-frequency tail of the current-current correlator.","marker":"[26]"}],"fun_headline_variants":["Universal 1/√T diffusion drives strange metal in 2D Hubbard","Strange metal diffusivity scales as 1/√T in half-filled Hubbard","Resistivity anomaly traced to 1/√T charge diffusion","Pseudogap metal: metallic transport, insulating compressibility","Half-filled Hubbard: diffusion law explains strange metal"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The numerical analytic continuation must recover the true optical conductivity near zero frequency; the smoothness-constrained spectrum is validated by a stretch test but is not uniquely determined by the Matsubara data.","fun_headline_variants_meta":{"raw":{"variants":["Universal 1/√T diffusion drives strange metal in 2D Hubbard","Strange metal diffusivity scales as 1/√T in half-filled Hubbard","Resistivity anomaly traced to 1/√T charge diffusion","Pseudogap metal: metallic transport, insulating compressibility","Half-filled Hubbard: diffusion law explains strange metal"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0002,"raw_usage":{"total_tokens":1218,"prompt_tokens":760,"completion_tokens":458,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":368}},"tokens_in":504,"tokens_out":458,"duration_ms":4889,"temperature":1.0,"reasoning_tokens":368,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T13:46:15.872385+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A measurement of the optical conductivity in the half-filled 2D Hubbard model that does not rest on maximum-smoothness analytic continuation—for example a real-time dynamical quantum Monte Carlo calculation or a cold-atom density-response measurement at half-filling—should find σDC at U=4,T=0.4-4 within the stretch-test window; if σDC deviates significantly, then D=σ/κ will not follow the reported 1/√T law.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the cold-atom diffusion measurement in the doped Hubbard model whose D∼T^-0.6 behavior the paper's half-filling 1/√T result strongly resembles, motivating the universality conjecture."},{"cited_title":"Beni and C","cited_arxiv_id":null,"evidence_quote":"Shows that thermoelectric power vanishes in half-filled bands, making the Nernst-Einstein relation exact and enabling D to be extracted from σDC and κ."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the numerically exact charge compressibility κ and spin/charge correlation data used in the Nernst-Einstein decomposition and in defining the temperature regime boundaries."},{"cited_title":"ˇSimkovic, J","cited_arxiv_id":null,"evidence_quote":"Documents the anisotropic single-particle self-energy and spectral pseudogap that the Pseudogap Metal regime is identified with."},{"cited_title":"Schaefer, N","cited_arxiv_id":null,"evidence_quote":"Provides the multi-method cross-benchmark that supports the numerical exactness of the thermodynamic-limit data used as input for the transport calculation."},{"cited_title":"Goulko, A","cited_arxiv_id":null,"evidence_quote":"Defines the stretch test and the consistent-constraints continuation method that control the systematic error on the inferred DC conductivity."},{"cited_title":"ˇSimkovic, Y","cited_arxiv_id":null,"evidence_quote":"Introduces the Bold4 self-consistent diagrammatic theory used for the non-stochastic high-frequency tail of the current-current correlator."}],"review_version":1}