{"id":"6b611546-9517-441c-9952-84602314bf67","arxiv_id":"2607.21490","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A weighted average of static Hartree-Fock solutions of the local impurity problem reproduces DMFT's frequency-dependent self-energy and local-moment persistence for NiO and FeO.","lead":"A new approximation, the on-site dephased ensemble, treats a Mott-insulating atom as a thermally weighted average of its static Hartree-Fock states, skipping Monte Carlo solvers. Tests on NiO and FeO match dynamical mean-field theory's self-energies and local-moment correlations closely, offering a cheap route to strongly correlated materials and a formal link between correlation and disorder.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Stationary-phase (one-loop) corrections to DE weights are uncontrolled; quantitative agreement with DMFT is post hoc, not derived.","rationale":"Good-faith reading: the paper proposes a clean, parameter-light approximation (DE) and provides direct numerical comparison with DMFT for NiO and FeO. The agreement in ImΣ(iω) and in spin-spin covariances (Table I) is strong quantitative evidence for the central claim within the stated regime. The derivation in Appendix A is a standard stationary-phase approximation, but it is uncontrolled in the sense that the size of the omitted fluctuations is not estimated independently of the DMFT agreement. The authors honestly list limitations (small gaps, metals, near-transition), but for the two benchmark systems the burden is on showing that the saddle-point sum is accurate. The single most load-bearing concern is therefore the stationary-phase (one-loop) correction to the weights; if that correction is large, the DE result is coincidental. A concrete one-loop computation would settle this. The reader's verdict CONDITIONAL is appropriate: the manuscript is promising but the SI and code are not available, and the technical check is needed. Our stress-test does not change the verdict; it sharpens the required test.","tokens_in":13713,"tokens_out":13097,"duration_ms":118845,"concrete_test":"Compute the one-loop Gaussian fluctuation correction to the saddle-point free energy for all dominant stationary states of the NiO and FeO impurity models. For each saddle φ_s, evaluate det(δ²A/δφ²) and include it as a prefactor in p_s. Recompute Σ_DE(iω_n) and the spin-spin covariances. If the corrected self-energy differs from the zero-loop result by more than the residual DE–DMFT discrepancy (roughly 6% in the covariances), the stationary-phase approximation is not controlled and the central claim needs qualification. As a complementary check, compare the DE partition function sum with the exact impurity partition function (CTSEG or exact diagonalization for a reduced model) to test the adequacy of the saddle-point weights.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central approximation is Eq. (7): the impurity partition function is replaced by a discrete sum over static saddle points of the Hubbard-Stratonovich action, with the Gaussian determinant around each saddle and inter-basin tunneling dropped. The Boltzmann weights p_s are then exp(-βΩ[φ_s]) with Ω from Eq. (11). The validity of this 'dephased' limit for NiO/FeO is justified only by the strong-U heuristic (tunneling barriers scale with U) and by the post hoc agreement with DMFT; no estimate of the one-loop free-energy correction is given. Since the weights enter Eq. (3) exponentially, even a small correction to Ω from the fluctuation determinant could shift the ensemble. In particular, the degenerate families of low-energy states (two for NiO, six for FeO) could have different curvatures, altering their relative weights. Without computing the determinant of δ²A/δφ² at each saddle, one cannot tell whether the saddle-point sum is a controlled asymptotic expansion. The completeness of the stationary-state enumeration (Ponet protocol) is a related but separate fragility: if additional low-energy states are missed, the DE sum is incomplete. Both issues are unverified in the v1 artifact.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces the on-site dephased ensemble (DE) approximation for the local impurity problem in DMFT and related embedding schemes. Instead of keeping a single static Hartree-Fock solution, DE sums over all stationary solutions of the embedding grand potential with Boltzmann weights p_s = exp(-βΩ_s)/Z, producing a CPA-like average Green's function G_DE = Σ_s p_s G_s and hence a frequency-dependent effective self-energy. The authors derive this approximation from a Hubbard-Stratonovich representation of the Anderson impurity model (Appendix A) and apply it to paramagnetic NiO and FeO in a five-orbital model with U=6 eV, J=1 eV. They report quantitative agreement with CTSEG DMFT for the Matsubara self-energy, the real-frequency spectral function, and the spin-spin covariance (within 4–6%). The central interpretive claim is that the strong frequency dependence of Mott-insulator self-energies can be viewed as the signature of a classical alloy of static symmetry-broken states.","tokens_in":13919,"tokens_out":7273,"duration_ms":78236,"significance":"If the DE approximation is reliable in the large-gap Mott regime, it provides a computationally inexpensive, sign-problem-free route to paramagnetic Mott insulators, bypassing QMC impurity solvers and large supercells. The derivation in Appendix A is clean, the CPA analogy is made explicit, and the numerical comparison is direct: the same impurity model is solved with DE and with CTSEG, with no fitting to the DMFT reference. The two-particle covariance prediction in Appendix B is a non-trivial additional validation. The main weakness is that the central saddle-point reduction is not quantitatively controlled; the authors themselves state that intra-basin fluctuations 'become relevant for smaller gaps' and that DE cannot describe correlated metals or the immediate vicinity of the Mott transition.","major_comments":[{"comment":"The replacement of the exact impurity partition function by a discrete sum over static saddle points omits the Gaussian fluctuation determinant around each saddle and inter-basin tunneling. Since the weights p_s enter Eq. (3) exponentially, even a modest difference in the curvature of the action among the degenerate low-lying states (two for NiO, six for FeO) could alter the ensemble. The text's statement that intra-basin fluctuations 'become relevant for smaller gaps' is not quantitative. Please provide an estimate of the one-loop correction—e.g., by evaluating det(δ²A/δφ²) at each saddle point or by benchmarking against an exact impurity solver for a simplified model—to justify the dephased limit for these materials.","section":"Appendix A, Eq. (7)"},{"comment":"The DE sum is over 'all accessible stationary points,' and the Ponet et al. protocol is asserted to find them, but no completeness certificate or convergence test is described. If a low-lying stationary state is missed, Z_loc and p_s in Eq. (3) are incomplete. Please report the number of stationary states found, the number of initializations used, and a test of protocol completeness (e.g., random restarts, or exhaustive enumeration for a reduced model).","section":"Eq. (2) and SI Sec. IC1"}],"minor_comments":[{"comment":"The symbol U is used both for the interaction tensor U^{ijkl} and for the Hubbard parameter U = 6 eV; please distinguish them (e.g., use a different font or an explicit superindex).","section":"Eq. (1) and throughout"},{"comment":"In panels (d,e), the labels 'spin-up (positive y-axis)' and 'spin-down (negative y-axis)' are ambiguous; please clarify the plotted quantity and axis labels.","section":"Fig. 1"},{"comment":"The text says 'ψ_iσ, ψ†_iσ represent fermionic Grassmann fields', but earlier the spin index was absorbed into a compound index; please make the convention consistent.","section":"Appendix A"},{"comment":"Reference [23] lists the DOI '10.1103/zjy7-4jqd', which appears to be a placeholder; please verify.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The SI is referenced for several load-bearing details (Secs. IC1, ID, II, III); it was not part of the reviewed artifact. Please ensure it is included in the next round so the completeness and continuous-symmetry claims can be checked."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does what it says: it defines the on-site dephased ensemble, a Boltzmann-weighted sum over static Hartree-Fock saddle points of the local embedding problem, and shows that for NiO and FeO this reproduces the DMFT self-energy and spin-spin covariance remarkably well. The Hubbard-Stratonovich derivation in Appendix A is clean, and the CPA analogy is made precise rather than hand-waved. No fitting to DMFT or experiment is involved, and the limitations section is refreshingly honest: the authors state clearly that DE cannot describe correlated metals or the vicinity of the Mott transition. That is real progress, not a dressed-up version of an earlier two-state ansatz. The novelty is the systematic ensemble over the full stationary landscape with thermodynamic weights, and the quantitative benchmark against CTSEG-DMFT on identical impurity models.\n\nWhere I part company with the stress-test note is on the severity of the stationary-phase concern. The saddle-point sum is an approximation; of course the one-loop determinant and inter-basin tunneling are dropped. The real question is whether the comparison to DMFT supports the claim within the stated regime, and it does. The paper openly says smaller gaps will bring in fluctuations, and that is where DE will fail. So the uncontrolled corrections are a boundary, not a contradiction. What is genuinely missing is an a priori estimate of the one-loop weight corrections or a scan of the curvature at the low-lying saddles; without that, the agreement stays post hoc. I also would not call the completeness of the landscape search a load-bearing flaw: the Ponet protocol is a practical protocol, and the exponential suppression of the higher states in NiO makes the enumeration issue minor there. But the SI is not in the v1, so I cannot verify the six-state FeO degeneracy, the spin-quantization-axis treatment, or the computational parameters. That is a real reproducibility gap, as is the absence of a code/data repository and of error bars on the DMFT reference.\n\nNet: this is a solid, honest paper with a genuinely new twist on the polymorphous picture. It deserves serious peer review. I would want the SI and code available before publication, and an error-bar or convergence statement on the QMC data. The central claim holds up in the large-gap regime it targets.\n\nFor you: worth a reading-group slot, and I will cite it if a later version ships the missing artifact.","headline":"A genuinely new ensemble construction with honest limits; the NiO/FeO benchmarks are direct and convincing, but SI/code are missing and the saddle-point control is unquantified.","tokens_in":14479,"tokens_out":1477,"would_cite":true,"duration_ms":18358,"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":"A Boltzmann-weighted average over all static, symmetry-broken Hartree-Fock solutions reproduces the dynamical self-energy of paramagnetic Mott insulators, matching dynamical mean-field theory for NiO and FeO.","keywords":["Mott insulators","dynamical mean-field theory","self-energy","symmetry breaking","coherent-potential approximation","local moments","on-site dephased ensemble","transition-metal oxides"],"falsifier":"Compute the Gaussian fluctuation determinant around each stationary Hartree-Fock solution for NiO and reweight the ensemble; if the DE self-energy or spin-spin covariance shifts noticeably away from the dynamical mean-field benchmark, the saddle-point sum underlying the DE is missing physics.","tokens_in":13507,"feed_emoji":"🧲","tokens_out":7709,"duration_ms":63566,"temperature":0.7,"pith_summary":"Paramagnetic Mott insulators defeat standard density-functional theory, which predicts metals unless symmetry is artificially broken. This paper introduces the on-site dephased ensemble (DE): instead of picking one Hartree-Fock solution, it forms a Boltzmann average over all accessible static solutions of the local embedding problem. The average produces a frequency-dependent self-energy, and for NiO and FeO the DE self-energy, spectral function, and spin-spin covariance closely match dynamical mean-field theory at a fraction of the cost. The message is that the characteristic frequency structure of 'strong correlations' in gapped Mott systems can be read as the signature of a disordered alloy of symmetry-broken static states.","feed_headline":"Mixing static magnetic states reproduces Mott self-energies","feed_subtitle":"A Boltzmann ensemble of Hartree-Fock solutions matches costly quantum Monte Carlo for NiO and FeO.","key_machinery":"The central object is the on-site dephased ensemble (DE), defined by a Boltzmann-weighted sum over all stationary Hartree-Fock self-energies of the local embedding problem: G_DE = sum_s p_s G_s, with p_s proportional to exp(-beta Omega[Sigma_s]). This is mathematically the single-site coherent-potential approximation of an alloy whose 'elements' are symmetry-broken static solutions. The workhorse identity is the stationary-phase reduction of the impurity partition function, which replaces the functional integral over auxiliary fields by a discrete sum over saddle points, converting a dynamical problem into a static statistical mixture. A second identity splits the two-particle covariance int","core_discovery":"The paper's central claim is that the local Green's function of a correlated site is well approximated by a Boltzmann-weighted sum over Green's functions built from each static Hartree-Fock self-energy that is a stationary point of the embedding grand potential. Even though each ingredient is static, the averaged inverse Green's function produces a strongly frequency-dependent effective self-energy. This 'on-site dephased ensemble' is shown to reproduce the dynamical mean-field self-energy and spectral function for paramagnetic NiO and FeO, and the ensemble variance in the two-particle covariance keeps local moments alive at long times, matching dynamical mean-field theory to within a few pe","pith_inferences":["If the DE reproduces DMFT generically, the distinction between 'dynamical' and 'static-disorder' origins of Mott gaps becomes a formal duality, and correlation strength in gapped systems could be recast in terms of the density of low-lying symmetry-broken states.","The DE's stationary-phase approximation omits Gaussian fluctuations around each saddle; a direct test would be to include those fluctuations and check whether the NiO/FeO weights shift. If they do, the current agreement may partly rely on error cancellation.","The temperature dependence of the Boltzmann weights makes DE's ensemble composition—not just occupations—temperature dependent; comparing DE and DMFT across temperatures could reveal how basin structure controls the crossover toward the metallic state.","Applying DE to doped Mott insulators or charge-transfer systems with smaller gaps would probe where the static-alloy picture breaks down, since those systems have stronger inter-basin coherences."],"forward_implications":["For large-gap paramagnetic Mott insulators, DE can replace quantum Monte Carlo impurity solvers, recovering the DMFT self-energy and spectral function at a small fraction of the computational cost.","The strong frequency dependence of the local self-energy is reinterpreted as the CPA-like averaging of a disordered alloy of symmetry-broken static states, providing a concrete bridge between the many-body and polymorphous supercell viewpoints.","The ensemble variance in the two-particle correlation formula gives a static mechanism for persistent local moments; a single static mean-field solution cannot produce this persistence.","Because DE has no fermion sign problem, it opens a route to first-principles treatment of f-electron Mott insulators.","DE can be used to precondition DMFT impurity solves and can be extended to cluster schemes for non-local spatial correlations."],"fun_headline_variants":["Boltzmann sum of static solutions reproduces Mott self-energies","On-site dephased ensemble matches DMFT without quantum Monte Carlo","Static ensemble yields dynamic self-energy for Mott insulators","Efficient ensemble approximates DMFT for paramagnetic NiO and FeO","Mott physics from a thermal mix of Hartree-Fock states"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The construction works only if the local problem's partition function can be represented as a discrete sum over static Hartree-Fock solutions, ignoring quantum fluctuations around each solution and tunneling between them.","fun_headline_variants_meta":{"raw":{"variants":["Boltzmann sum of static solutions reproduces Mott self-energies","On-site dephased ensemble matches DMFT without quantum Monte Carlo","Static ensemble yields dynamic self-energy for Mott insulators","Efficient ensemble approximates DMFT for paramagnetic NiO and FeO","Mott physics from a thermal mix of Hartree-Fock states"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000311,"raw_usage":{"total_tokens":1611,"prompt_tokens":752,"completion_tokens":859,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":496,"completion_tokens_details":{"reasoning_tokens":769}},"tokens_in":496,"tokens_out":859,"duration_ms":9080,"temperature":1.0,"reasoning_tokens":769,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T07:16:45.966364+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the Gaussian fluctuation determinant around each stationary Hartree-Fock solution for NiO and reweight the ensemble; if the DE self-energy or spin-spin covariance shifts noticeably away from the dynamical mean-field benchmark, the saddle-point sum underlying the DE is missing physics.","supporting_citations":[],"review_version":1}