{"id":"76aa4d1a-58dd-4dd1-b960-c54aa9e4b003","arxiv_id":"2502.07353","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A new combined ACPA-DMFT method is used to show that off-diagonal (hopping) disorder in the Anderson-Hubbard model shifts metal-insulator boundaries and produces a reentrant insulating-metal-insulating sequence as U increases.","lead":"This paper introduces a computational method, ACPA-DMFT, that extends a phonon-based disorder technique to electronic systems, allowing theorists to study the Anderson-Hubbard model when both on-site energies and hopping amplitudes are randomly disordered. The method reveals that hopping disorder shifts the metal-insulator transition and produces a reentrant transition, where a system turns metallic then insulating again as interactions grow.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No interacting benchmark supports the U>0 claim: the reentrant metallic pocket in Fig.8 rests entirely on the new ACPA-DMFT loop in the correlated regime, so an unquantified error in the disorder-treatment could remove it.","rationale":"The reader's named weakest assumption is the x-gauge freedom in Eq.2. I do not find that to be the load-bearing flaw: for the binary nearest-neighbor hopping matrix the decomposition tαβ=xαSxβ+λ has only an overall scale/sign freedom, and under x→cx, S→S/c², λ unchanged, the ACPA-DMFT equations scale coherently (P→P/c, g→cg, and Eq.16 remains invariant), so physical results are gauge-independent. The real soft spot is that the central reentrant claim is made in the interacting regime, where ACPA-DMFT is unvalidated by any independent calculation or known exact limit. The U=0 benchmarks and agreement with Ref.[32] are genuine supporting evidence, but they do not test the correlated off-diagonal-disorder physics that produces Fig.8. The CONDITIONAL verdict remains appropriate, but for a different reason than the one emphasized by the reader.","tokens_in":14844,"tokens_out":21887,"duration_ms":224918,"concrete_test":"Run determinant quantum Monte Carlo (or disorder-averaged exact diagonalization on 4x4x4/6x6x6 periodic clusters) for the same simple-cubic Anderson-Hubbard model with tAA=tBB=0.5, tAB=0.5 and 1.0, cA=cB=0.5, U=0,10,15,20, and ϵA=-5-U/2, ϵB=5-U/2, at a fixed inverse temperature such as β=20. Compute the local Green's function at iω0 and the real-frequency spectral function via stochastic analytic continuation. If the exact result does not show metallic behavior at U=10 for tAB=0.5 while showing insulating behavior at U=0 and U=15—or if the metallic pocket shifts by more than ΔU≈2.5—then the reentrant phenomenon in Fig.8 is not robust.","verdict_should_be":"UNCHANGED","load_bearing_attack":"All validation in the paper is at U=0 (Fig.4). The central claim—off-diagonal disorder shifts the Mott boundary and produces a reentrant insulator-metal-insulator sequence—is obtained solely from the ACPA-DMFT loop of Sec.II C for U>0, with no comparison against numerically exact or independent alternative calculations for the same lattice. In that loop a species-dependent local self-energy from CT-QMC is fed back into the auxiliary CPA medium via Eqs.26-31; the joint fixed point is an uncontrolled combination of two approximations. The only external consistency cited is BEB-CPA+DMFT on a Bethe lattice [32], i.e., a different lattice and a different method, not a validation of the simple-cubic phase diagram. Phase boundaries in Fig.8 are inferred from ImG(iω0) at the lowest Matsubara frequency without reporting the temperature, a threshold, or error bars, and the DOS in Fig.9 relies on maximum-entropy analytic continuation, which is unreliable for distinguishing small metallic quasiparticle peaks from spectral artifacts. If the ACPA medium distorts the hybridization Δ^Qi even moderately, the quasiparticle peak at U=10 for tAB=0.5 could be an artifact and the reentrant pocket could disappear.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript extends the auxiliary coherent potential approximation (ACPA) to electronic systems with off-diagonal disorder and combines it with dynamical mean-field theory (DMFT), calling the resulting scheme ACPA-DMFT. The method is applied to the Anderson-Hubbard model on a simple cubic lattice with both diagonal and off-diagonal disorder. For the noninteracting case, the ACPA density of states is benchmarked against BEB-CPA and supercell results and shows good agreement. For the interacting case, the paper reports that off-diagonal disorder substantially modifies the Mott metal-insulator transition, including a reentrant insulator-metal-insulator sequence as the interaction strength U increases, plotted in a U-t_AB phase diagram. The paper also compares ACPA-DMFT with a constant-hopping CPA-DMFT to highlight the role of hopping disorder.","tokens_in":15114,"tokens_out":3587,"duration_ms":36137,"significance":"If the central claims hold, the ACPA-DMFT method would be a useful and computationally efficient tool for disordered strongly correlated systems, particularly because the U=0 benchmarks in Fig. 4 are convincing and the claimed O(N^3) per-k-point scaling relative to BEB-CPA is attractive. The paper also identifies an interesting physical scenario where off-diagonal disorder shifts and broadens the metallic region in the U-t_AB plane. However, the main new physical prediction, the reentrant metallic pocket for U>0, currently rests entirely on the ACPA-DMFT loop without an independent interacting benchmark, and the method contains an unspecified decomposition parameter that may affect the results. The work is therefore a promising methodological contribution whose central interacting-phase-diagram claim requires further validation before it can be accepted as established.","major_comments":[{"comment":"The decomposition t_{ij}^{Qi,Qj} = x_i^{Qi} S_{ij} x_j^{Qj} + λ_{ij} is not unique, and the manuscript never specifies how the amplitudes x_i^{Qi} are chosen or fitted. The text says 'x_{Qi}^{i} and x_{Qj}^{j} generically should be fitted from the chemical dependent hopping parameters,' but no fitting procedure, numerical values, or uniqueness criterion is provided. Since Eqs. (3)-(17) and Eq. (31) depend explicitly on x, different valid choices of x and S, λ could in principle lead to different ACPA self-consistency conditions and different phase boundaries. The authors should either specify a concrete, reproducible choice for x in the simple-cubic calculations or demonstrate numerically that the final Green's function and phase diagram are independent of this gauge freedom.","section":"Sec. II A, Eq. (2)"},{"comment":"The phase diagram in Fig. 8 is constructed from the value of ImG(iω0) at the lowest Matsubara frequency, but the paper does not report the temperature, the value of ω0, the threshold used to separate metallic from insulating regions, or any error bars or convergence tests. At finite temperature, a single Matsubara-frequency point is not a reliable order parameter for a Mott transition; metallic and insulating states can be distinguished only by extrapolating to T→0 or by using another criterion such as the quasiparticle weight Z^Q of Eq. (32). The paper should specify the temperature and the criterion, and ideally show the β-dependence, before the reentrant region in Fig. 8 is presented as a phase boundary.","section":"Sec. III B 3, Fig. 8"},{"comment":"All interacting results are generated solely by the ACPA-DMFT loop of Sec. II C, and there is no independent benchmark for U>0. The U=0 comparison in Fig. 4 validates the ACPA treatment of off-diagonal disorder in the noninteracting limit, but it does not validate the combination of a species-dependent CT-QMC self-energy with the ACPA medium for U>0. The comparison with CPA-DMFT in Fig. 5 uses the same DMFT framework with a different disorder treatment and is not an external check. The authors should provide at least one independent interacting benchmark, such as exact diagonalization of small disordered clusters, supercell CT-QMC, or a comparison with BEB-CPA+DMFT on the same simple-cubic lattice, to support the central claim that the reentrant metallic pocket in Fig. 8 is physical rather than an artifact of the combined approximations.","section":"Sec. III B, Figs. 5-9"},{"comment":"The spectral functions in Fig. 9 are obtained by maximum-entropy analytic continuation, which is known to sometimes produce spurious peaks. The claim that the system becomes metallic at U=10 for t_AB=0.5 and at U=15 for t_AB=1.0, and then insulating again, relies on identifying a quasiparticle peak in the continued DOS. This identification should be corroborated by direct Matsubara data, for example by showing ImG(iω_n) as a function of n or by computing Z^Q for the same parameter sets, rather than by visual inspection of MaxEnt spectra.","section":"Sec. III B 3, Fig. 9"}],"minor_comments":[{"comment":"The abstract contains minor language issues such as 'matsubara' and 'maximum entropy' capitalization, and the phrase 'the diagonal and off-diagonal disorders are treated in a unified and self-consistent framework' is a bit long; a careful proofread is recommended.","section":"Abstract"},{"comment":"The caption states 'cA = cB = 0.5' while the panel legends specify cA = 0.1, 0.6, 0.6, and 0.5. This inconsistency should be corrected.","section":"Fig. 4 caption"},{"comment":"The sentence 'the the presence of quasiparticle peak' contains a duplicated article; there are several similar typos throughout the text (e.g., 'summarrize,' 'presnt,' 'theoretic') that should be corrected.","section":"Sec. III B 1, Fig. 5"},{"comment":"The computational-cost comparison between ACPA and BEB-CPA is stated only for the noninteracting Green's-function evaluation and does not include the CT-QMC cost in the DMFT loop; a brief clarification that the O(N^3) scaling applies to the ACPA part only would avoid overstatement.","section":"Sec. III A"}],"recommendation":"major_revision","confidential_remarks":"The methodological U=0 validation is solid, but the central interacting phase diagram is not yet independently supported. The unspecified x decomposition is a genuine well-definedness concern, and the phase criterion in Fig. 8 lacks essential numerical details. I would like to see these addressed before publication; the paper is not at the reject stage because the issues are fixable within the manuscript's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the ACPA-DMFT combination for electrons, not the reentrant MIT itself, which was already on the Bethe lattice in Ref. 32. The U=0 benchmarks against BEB-CPA and supercell are solid and give me real confidence that the disorder part works in the noninteracting limit. The method is also clearly cheaper than BEB-CPA for multi-component problems, so there is a practical payoff if it holds up.\n\nWhat worries me is the interacting regime. The x_i amplitudes in Eq. 2 are never specified, and the auxiliary decomposition t = x S x + lambda has a gauge freedom that can affect the CPA self-consistency. The paper says these 'should be fitted' but never says how or which values were used. That is a missing piece, not a fatal flaw, but it makes the method hard to reproduce.\n\nMore seriously, the central U>0 claim, the reentrant pocket in Fig. 8, comes from the ACPA-DMFT loop with no independent check against ED, cluster QMC, or another lattice method. The stress-test worry that a moderately distorted hybridization could kill the quasiparticle peak is plausible. I would not call it a demonstrated flaw, but it needs to be answered. The phase boundary is inferred from ImG(iomega0) with no temperature, no threshold, and no error bars, and the DOS relies on MaxEnt analytic continuation, which is not reliable for small quasiparticle peaks. These are all fixable in a revision, but they are exactly the details that separate a convincing method paper from an exploratory one.\n\nThe paper reads as honest and the authors know the limitations, but they do not flag the decomposition ambiguity or the missing benchmark themselves. The citation pattern is appropriate, and the relation to previous work, including their own ACPA for phonons, is clear.\n\nBottom line: this deserves a serious referee. It is a plausible new tool with a working noninteracting limit and a potentially interesting phase diagram. I would send it to review, but I would ask pointed questions about the x decomposition, the temperature, and an independent interacting calculation before accepting the reentrant behavior as physical. If that benchmark comes out fine, I would cite it.","headline":"A credible but under-validated extension of ACPA to electronic systems; the U>0 results need a benchmark before I'd trust the reentrant phase diagram.","tokens_in":15641,"tokens_out":1886,"would_cite":false,"duration_ms":18347,"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":"This paper argues that off-diagonal hopping disorder cannot be treated as a small correction: combined with dynamical mean-field theory, it shifts the Mott metal-insulator boundary and produces a reentrant insulator-metal-insulator…","keywords":["Anderson-Hubbard model","off-diagonal disorder","coherent potential approximation","dynamical mean-field theory","metal-insulator transition","reentrant behavior","auxiliary coupling space","simple cubic lattice"],"falsifier":"Take the Fig. 8 model (c_A = c_B = 0.5, t_AA = t_BB = 0.5, crystal-field splitting 10) and run ACPA-DMFT with two different valid choices of $x_i^{{Q_i}}$ and lambda_ij that reproduce the same physical hoppings; if the insulator-metal-insulator boundaries in the U-t_AB plane shift beyond numerical accuracy, the central prediction is not well-defined. A complementary check is to compare the same phase diagram with exact quantum Monte Carlo on small supercells.","tokens_in":14644,"feed_emoji":"🔄","tokens_out":7019,"duration_ms":59779,"temperature":0.7,"pith_summary":"The paper extends the auxiliary coherent potential approximation (ACPA) to electronic systems and couples it with dynamical mean-field theory (DMFT), producing a scheme in which diagonal (on-site energy) and off-diagonal (hopping) disorder are treated by the same coherent-potential self-consistency. Applied to the Anderson-Hubbard model on a simple cubic lattice at half filling, the method predicts that off-diagonal disorder substantially changes the Mott metal-insulator transition: it raises the critical interaction needed to enter and leave the metallic phase and widens the metallic window. At equal concentrations of two atomic species, the system is reentrant, going from insulating to metallic and back to insulating as U increases. The authors intend this as a practical computational route for disordered strongly correlated materials where hopping integrals vary from site to site, a regime that conventional CPA-based DMFT handles poorly.","feed_headline":"Hopping disorder triggers insulator-metal-insulator reentrance","feed_subtitle":"New DMFT-based scheme treats diagonal and off-diagonal disorder on equal footing and widens the metallic window.","key_machinery":"The key object is the decomposition of each random hopping amplitude into a site-dependent amplitude and a fixed reference hopping, t_{ij}^{Q_i,Q_j} = $x_i^{{Q_i}}$ S_{ij} $x_j^{{Q_j}}$ + \\lambda_{ij}, which turns off-diagonal disorder into diagonal-like disorder in an enlarged 'coupling space' C = S \\otimes T (orbital space times neighbor translation vectors). Within this space, a CPA self-consistency loop (Eqs. 10-14) determines an effective medium for the auxiliary Green's function g; the physical Green's function is recovered as G = g $X^{{-1}}$. DMFT is then grafted on: the self-energy is local, the impurity hybridization \\$Delta_i^{{Q_i}}$ is computed from the ACPA physical Green's function, and the impurity solver (CT-QMC) returns a new self-energy that re-enters the ACPA loop. This machinery is what allows diagonal and off-diagonal disorder to be treated on the same footing.","core_discovery":"On the paper's own terms, the central discovery is that ACPA-DMFT correctly captures the combined effect of diagonal and off-diagonal disorder on the Mott transition, and that the off-diagonal part is not a minor perturbation. For a binary simple-cubic alloy with fixed hoppings t_AA = 0.8, t_BB = 0.5, t_AB = 0.63, the metal-insulator transition as a function of concentration or of interaction U occurs at different thresholds than with the averaged hopping alone; increasing t_AB or t_BB drives a first-order transition seen as a discontinuous jump of the quasiparticle weights Z^A and Z^B from zero to finite values. In the symmetric half-filled case, the U-t_AB phase diagram shows a reentrant sequence: a band insulator at U=0, a metal for an intermediate window, and a Mott insulator again at large U. The authors attribute this to off-diagonal disorder renormalizing the effective bandwidth and the hybridization bath of each species, so the ratio U/D controlling the Mott transition is displaced and the metallic region grows with t_AB.","pith_inferences":["An open question the paper leaves implicit is the gauge freedom in Eq. (2): the amplitudes x_i^{Q_i} are never specified or fitted, so a natural follow-up is to test whether different valid decompositions of the same physical hoppings yield the same phase boundary.","The coupling-space construction is not tied to single-band DMFT, so it could plausibly be combined with multi-orbital or nonequilibrium impurity solvers to study orbital-selective Mott transitions or disorder effects on superconductivity.","The reentrant window suggests a materials-level prediction: in real disordered compounds, alloy composition and strain (which change hopping integrals) should be able to tune the metallic window without changing U, which could be tested in transport experiments on doped transition-metal oxides.","A benchmark against exact small-cluster quantum Monte Carlo with the same hopping disorder would tell whether the reentrant boundaries survive beyond the single-site approximations used here."],"forward_implications":["The Mott transition in a disordered alloy is not a simple average of its pure-component behaviors, because alloying changes the hybridization bath that each species feels; off-diagonal disorder shifts the critical doping and interaction thresholds.","Hopping amplitudes act as a continuous control knob: for fixed U, changing t_AB or t_BB across a critical value switches the system between insulator and metal through a first-order transition.","In the half-filled binary alloy, the system is reentrant in U, and raising t_AB pushes both transition points U1 and U2 to larger U while widening the metallic window |U2 - U1|.","Because the auxiliary medium in ACPA has a dimension independent of the number of alloy species, the method should remain computationally affordable for multi-component or multi-orbital disordered systems where BEB-CPA becomes expensive."],"supporting_citations":[{"why":"Introduces the auxiliary coherent potential approximation for disordered phonon systems, which the present work adapts to electronic systems.","marker":"[35]"},{"why":"The BEB-CPA+DMFT study on the Bethe lattice that reported the reentrant insulator-metal-insulator sequence and serves as the main comparison for the cubic-lattice result.","marker":"[32]"},{"why":"Earlier CPA+DMFT work showing reentrant behavior for diagonal disorder, the effect this paper extends to off-diagonal disorder.","marker":"[12]"},{"why":"Establishes the dynamical mean-field mapping from lattice to impurity model on which the combined scheme relies.","marker":"[13]"},{"why":"Provides the hybridization-expansion continuous-time quantum Monte Carlo impurity solver used to obtain the local self-energy.","marker":"[42]"},{"why":"Maximum entropy method used to analytically continue Matsubara Green's functions to real-frequency spectral functions.","marker":"[51]"},{"why":"Extends ACPA to cluster formulations, providing the coupling-space perspective adopted here.","marker":"[40]"}],"fun_headline_variants":["Off-diagonal disorder drives reentrant metal-insulator transitions","Reentrant metal phase from off-diagonal hopping disorder","Off-diagonal disorder shifts Mott transition to new metallic window","Hopping disorder widens metallic window in correlated system"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method's predictions are assumed not to depend on how each random hopping is split into $x_i^{{Q_i}}$ S_ij $x_j^{{Q_j}}$ + lambda_ij, but the amplitudes $x_i^{{Q_i}}$ are never specified or fitted, so different valid splits could in principle move the phase boundaries.","fun_headline_variants_meta":{"raw":{"variants":["Off-diagonal disorder drives reentrant metal-insulator transitions","Reentrant metal phase from off-diagonal hopping disorder","Off-diagonal disorder shifts Mott transition to new metallic window","Hopping disorder widens metallic window in correlated system"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0005,"raw_usage":{"total_tokens":2451,"prompt_tokens":952,"completion_tokens":1499,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":568,"completion_tokens_details":{"reasoning_tokens":1432}},"tokens_in":568,"tokens_out":1499,"duration_ms":11420,"temperature":1.0,"reasoning_tokens":1432,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T13:01:40.624571+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the Fig. 8 model (c_A = c_B = 0.5, t_AA = t_BB = 0.5, crystal-field splitting 10) and run ACPA-DMFT with two different valid choices of $x_i^{{Q_i}}$ and lambda_ij that reproduce the same physical hoppings; if the insulator-metal-insulator boundaries in the U-t_AB plane shift beyond numerical accuracy, the central prediction is not well-defined. A complementary check is to compare the same phase diagram with exact quantum Monte Carlo on small supercells.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the auxiliary coherent potential approximation for disordered phonon systems, which the present work adapts to electronic systems."},{"cited_title":"Milovanovi´ c, S","cited_arxiv_id":null,"evidence_quote":"The BEB-CPA+DMFT study on the Bethe lattice that reported the reentrant insulator-metal-insulator sequence and serves as the main comparison for the cubic-lattice result."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier CPA+DMFT work showing reentrant behavior for diagonal disorder, the effect this paper extends to off-diagonal disorder."},{"cited_title":"Chakraborty, T","cited_arxiv_id":null,"evidence_quote":"Establishes the dynamical mean-field mapping from lattice to impurity model on which the combined scheme relies."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the hybridization-expansion continuous-time quantum Monte Carlo impurity solver used to obtain the local self-energy."},{"cited_title":"Cheng, J","cited_arxiv_id":null,"evidence_quote":"Extends ACPA to cluster formulations, providing the coupling-space perspective adopted here."}],"review_version":1}