{"id":"8825ddaa-ed30-4199-8ab0-360d4f98b798","arxiv_id":"2412.17190","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A partial Wannier construction with N-1 localized orbitals explains emergent Kondo lattice Mott states in fractionally filled topological bands, with exact-diagonalization evidence in BHZ and twisted MoTe2.","lead":"The paper shows that in fractionally filled topological bands, interactions can create an emergent Kondo lattice: electrons localize into N-1 tightly bound orbitals while one delocalized topological orbital remains, producing new Mott insulating states. It predicts such states in twisted MoTe2 and in the BHZ model, competing with fractional Chern insulators.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The partial Wannier basis construction lacks a proof of exponential localization for arbitrary supercell sizes; the numerical decay plots in Fig. 3 do not rule out algebraic tails, so the central mechanism rests on an unverified assumption.","rationale":"The reader's weakest_assumption precisely identifies the partial Wannier basis localization as the most load-bearing condition. The paper's central claim is that strong short-ranged interactions stabilize Mott insulators described by N-1 exponentially localized SCWOs plus one topological power-law orbital; if the SCWOs are not exponentially localized, the local-moment picture and the emergent Kondo lattice description collapse. The existence of a smooth frame for N-1 states is mathematically plausible via bundle splitting, but the specific projector construction requires non-vanishing projected trial states, and exponential decay requires analyticity in a strip around the rBZ. Neither is proven. The numerical decay plots for N up to 8 are suggestive but not conclusive: they do not extract decay lengths, do not distinguish exponential from superpolynomial or power-law behavior, and do not probe the asymptotic regime or larger supercells. The ED results themselves are finite-size computations and are acknowledged by the authors to be limited (e.g., the 6x3 torus bias toward striped AFM), but the localization assumption is more foundational because it underpins the interpretational framework. Therefore I agree with the reader's assessment, and the CONDITIONAL verdict remains appropriate. No change is needed; the concrete test above would provide the missing evidence.","tokens_in":16491,"tokens_out":12407,"duration_ms":122911,"concrete_test":"Compute SCWOs for the BHZ valence band on a large momentum mesh (e.g., 100x100) for supercell sizes N = 2, 4, 8, 16, 32 with the delta-function trial states of Sec. III.A. For each N, fit ln|W(r)| versus |r| for |r| beyond the supercell radius and compare a linear (exponential) fit with a log-log (power-law) fit; also extract the decay length versus N. If the tail is consistent with a power law or the decay length grows with N, the exponential-localization claim is falsified. Additionally, repeat with a different Chern-band model (e.g., Haldane model with nonuniform Berry curvature) to test generality.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central framework rests on the claim (Sec. II.A and Sec. II.B) that a U(N) rotation can concentrate all phase-vortex nonanalyticities into a single TPLO, leaving N-1 SCWOs exponentially localized. While a topological splitting of the band vector bundle guarantees a smooth global frame for N-1 states over the 2D torus, exponential localization additionally requires the projected trial states P(K)|psi_m(K)> to be nowhere vanishing and analytic for all K in the rBZ. The paper provides no general proof of non-vanishing; the delta-function trials could in principle develop zeros that the Gram-Schmidt procedure would turn into spurious vortices. The numerical evidence (Fig. 3) shows amplitudes decaying with distance for supercells up to N=8, but no fit distinguishes exponential from power-law tails, and no data for larger N is given. If the SCWOs were only algebraic or if the localization length grew with supercell size, the Mott-Hubbard treatment and the emergent-Kondo-lattice picture would lose quantitative support. This is the load-bearing condition for the paper's 'new theoretical tool' and for the claim that strong short-ranged interactions naturally favor the described states.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes that for an isolated N-band manifold with net Chern number C, one can construct a partial Wannier basis consisting of N-1 exponentially localized supercell Wannier orbitals (SCWOs) and one power-law-localized topological orbital (TPLO) that retains the topological obstruction. It argues that strong short-ranged Coulomb interactions in fractionally filled topological bands naturally favor Mott insulating states described by this basis, yielding an emergent Kondo lattice of local moments in SCWOs coupled to an itinerant TPLO band. The authors support this with exact diagonalization (ED) of the interacting BHZ model at 1/2 and 3/2 valence filling and of twisted MoTe2 at -1/3 hole filling, reporting charge-ordered ferromagnetic or antiferromagnetic insulating phases whose low-energy many-body states have high overlap with the SCWO magnetic basis.","tokens_in":16706,"tokens_out":4345,"duration_ms":39802,"significance":"If the partial Wannier basis construction is rigorous, it would provide a genuinely new single-particle tool for correlated insulators in Chern bands and a new scenario for time-reversal-symmetric interacting topological bands, going beyond both Landau-level FQH paradigms and conventional Wannier-based Mott-Hubbard descriptions. The ED phase diagrams and the high overlaps f_n (Fig. 2i) and counts <n_SCWO> (Fig. 4e) are concrete and valuable evidence that the proposed physical picture captures real physics in small clusters. However, the central localization claim is not proven, and the predictive power of the construction is weakened by the post hoc selection of the supercell and the addition of a bias field to favor the symmetry-broken state. These issues are fixable within the scope of the manuscript, but they are load-bearing for the paper's central claim.","major_comments":[{"comment":"The claim that N-1 supercell Wannier orbitals are exponentially localized for arbitrary supercell sizes is the load-bearing assumption of the paper, but it is only argued heuristically. The assertion that a U(N) rotation can shift all phase-vortex nonanalyticities into a single state, leaving N-1 globally smooth states, is not proven, and the projective construction in Sec. II.B requires that the projected trial states P(K)|psi_m(K)> be nowhere vanishing and analytic over the rBZ; no general argument for non-vanishing is given. The numerical evidence in Fig. 3 shows decay of SCWO/TPLO amplitudes for supercells up to N=8, but the panels contain no fits distinguishing exponential from power-law decay, no localization lengths as a function of N, and no data for larger supercells. Since the Mott-Hubbard description of SCWOs and the emergent Kondo lattice picture depend on exponential localization, the paper should either provide a proof or a precise sufficient condition, or present numerical fits (e.g., log-linear decay and localization length vs N) that support exponential localization.","section":"Sec. II.A and II.B"},{"comment":"The supercell is chosen after seeing the ED charge order (\"The supercell is chosen based on the charge ordering found in the ED calculations\"), and a bias field V_b is added to select the translation-symmetry-broken partner encoded by the trial wavefunctions. This introduces a degree of circularity: the high overlap f_n and the near-unit occupancy <n_SCWO> are partly consequences of fitting the basis to the ED-identified order, rather than independent predictions. To support the claim that strong interactions 'naturally' stabilize these states, the authors should demonstrate that the supercell and SCWO can be predicted from filling fraction and lattice symmetries alone, or at least show that f_n and <n_SCWO> are robust to variations of the trial states and supercell orientation.","section":"Sec. III (before III.A), Fig. 2(i), Fig. 4(e-f)"},{"comment":"The artificially tuned valence band width Gamma is changed by adding longer-ranged hoppings while \"keeping the Bloch states and Chern number identical,\" but the SCWO localization depends on the quantum geometry, not only on the Chern number. The phase diagrams as functions of Gamma/U are therefore only meaningful if the SCWO spread is held fixed or its variation with Gamma is reported. Please provide the SCWO localization length as a function of Gamma and discuss whether the FM-to-AFM transition could be driven by a change in the Wannier localization rather than by the intended bandwidth effect.","section":"Sec. III.A, Fig. 2(d-e)"}],"minor_comments":[{"comment":"The statement that \"the final Nth projected trial state must vanish at discrete points in the rBZ\" is asserted without proof; it would be helpful to mark this as a conjecture or provide a proof sketch, since it is central to the topological splitting argument.","section":"Sec. II.B"},{"comment":"The caption says \"projection f_n for the first n = 1200 many-body eigenstates,\" while the text discusses the first 256 states, the next 256 states, and states beyond the 512th; please harmonize the caption with the text.","section":"Fig. 2(i) and Sec. III.A"},{"comment":"Please state whether the decay plots are on a linear, log-linear, or log-log scale, and include fits to exponential decay with the inferred localization lengths; otherwise the claim of exponential localization cannot be assessed from the figures.","section":"Fig. 3"},{"comment":"The authors note that the 6x3 torus breaks rotational symmetry and may artificially favor striped AFM over competing Neeel order; please estimate the impact of this choice on the phase boundary and clarify whether the AFM phase remains stable on more symmetric clusters.","section":"Sec. III.B"},{"comment":"The embedding operator T(g) and the matrix O_{t,t'} are introduced compactly; a concrete example for the N=2 doubled supercell would improve the readability of the construction.","section":"Eq. (2.3)"},{"comment":"The discussion of direct Coulomb exchange versus superexchange would benefit from explicit definitions of these couplings in terms of the projected interaction matrix elements and the SCWO overlap integrals.","section":"Sec. III.A, paragraph on FM-to-AFM transition"}],"recommendation":"major_revision","confidential_remarks":"The paper presents a stimulating idea and solid ED evidence for the physical mechanism in small clusters, but the central 'new theoretical tool' - the exponential localization of N-1 supercell Wannier orbitals - is not proven and the numerical evidence is not yet convincing. The post hoc supercell selection and the use of a bias field further weaken the predictive claim. I would advise the editor that the manuscript is suitable for major revision rather than rejection, provided the authors can either prove the localization statement under generic conditions or substantially strengthen the numerical evidence, and clarify the predictive versus descriptive status of the construction. There is also some overlap with known results on Wannier obstructions and 'half Wannier' constructions; the novelty should be sharpened against that literature."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper is a genuine new idea: for an N-band manifold with total Chern number C, you can choose a gauge where N-1 bands become smooth and exponentially localizable, leaving one power-law orbital carrying the obstruction. The authors use that to construct supercell Wannier functions and propose that fractionally filled topological bands can spontaneously form a Kondo lattice of local moments on those orbitals coupled to the residual topological orbital. That is a real organizing principle, and it leads to concrete predictions: an AFM Mott state in tMoTe2 at -1/3 filling competing with the FCI, and a quantum spin Hall crystal at 3/2 filling in BHZ.\n\nThe ED results are the meat. They find FM and AFM Mott phases in the BHZ model and an AFM state in tMoTe2 at larger twist angles. The overlaps fn between low-energy eigenstates and the SCWO magnetic basis are high (about 95% over 256 states), which is good evidence that the partial Wannier basis describes the local moments. The tMoTe2 prediction is new and testable.\n\nSoft spots: the central localization statement is asserted heuristically, not proven. I don't think this is fatal – standard vector-bundle splitting suggests N-1 smooth sections exist – but the paper's specific construction using delta-function trials and Gram-Schmidt could have zeros, and Fig. 3 only shows decay for supercells up to N=8 without a fit distinguishing exponential from algebraic tails. The authors call it heuristic, which is honest, but the 'new theoretical tool' would be stronger with a proof. Second, the supercell and the small bias field are chosen after seeing the ED charge order. That weakens the predictive claim but not fatally; the overlap calculation is a postdiction diagnostic, not a fit. Third, finite-size: the 6x3 torus favors striped AFM, and the authors flag this. No code or data is shipped.\n\nOn balance the paper deserves a serious referee. The concept is important enough that the unproven localization step should be either proven or sharply tested in revision. I'd send it out.","headline":"New organizing principle for Mott states in Chern bands, with testable tMoTe2 predictions; the unproven localization step is real but not fatal.","tokens_in":17262,"tokens_out":3639,"would_cite":true,"duration_ms":39328,"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 argues that strong short-ranged repulsion in a fractionally filled topological band can reorganize the band into N-1 localized supercell orbitals plus one extended topological orbital, creating an emergent Kondo lattice and a…","keywords":["topological Mott insulator","emergent Kondo lattice","supercell Wannier functions","partial Wannier basis","fractional Chern insulator","twisted MoTe2","Bernevig-Hughes-Zhang model","antiferromagnetic order"],"falsifier":"A direct test is to compute the SCWO decay for a Chern band on a 16- or 32-unit-cell supercell, or for a band with strongly nonuniform Berry curvature; if the amplitudes decay algebraically rather than exponentially, the partial Wannier basis fails and the Mott description would not survive. Experimentally, a measurement on twisted bilayer MoTe2 at twist angles around or above 4 degrees and at -1/3 filling that finds a fully spin-polarized fractional Chern insulator with no square-root-of-3 charge order and no local moments would rule out the predicted antiferromagnetic Kondo-lattice phase in that material.","tokens_in":16207,"feed_emoji":"🧲","tokens_out":7350,"duration_ms":61712,"temperature":0.7,"pith_summary":"This paper tries to show that strong, short-ranged Coulomb interactions in fractionally filled topological bands need not produce fractional quantum Hall physics: they can instead stabilize a Mott insulator with broken translation symmetry and magnetic order. The key move is a gauge choice that splits an N-band manifold with nonzero Chern number into N-1 exponentially localized supercell Wannier orbitals and one power-law-localized orbital that keeps the topological obstruction. In that basis, interactions naturally put one electron in each localized orbital, forming local moments coupled to the itinerant topological orbital, an emergent topological Kondo lattice. The claim is supported by exact diagonalization of an interacting Bernevig-Hughes-Zhang quantum spin Hall model at 1/2 and 3/2 filling and of twisted bilayer MoTe2 at -1/3 filling, where an antiferromagnetic Mott state competes with a fractional Chern insulator. If correct, the paper predicts a new class of time-reversal-symmetric topological Mott insulators in moiré materials.","feed_headline":"Fractionally filled Chern bands can become Mott insulators","feed_subtitle":"Strong repulsion creates local moments on supercell orbitals plus one topological band, an emergent Kondo lattice.","key_machinery":"The central object is the partial supercell Wannier basis, built from a supercell Hamiltonian and the projector P(K) onto the folded topological bands. Smooth trial wavefunctions are projected, orthonormalized by Gram-Schmidt, and Fourier transformed: the first N-1 produce SCWOs with exponential decay, while the last state retains the phase-vortex nonanalyticities and becomes a TPLO with power-law decay. This basis performs the work of translating a topological band with no local Wannier description into a Hubbard-like model of local moments plus an itinerant topological band, which is the emergent Kondo lattice.","core_discovery":"At the center of the paper is a constructive claim: for a manifold of N bands with net Chern number C, one can choose a smooth gauge on an enlarged supercell so that N-1 orbitals become exponentially localized supercell Wannier orbitals (SCWOs) while a single topological power-law orbital (TPLO) inherits all the vortices and carries the Chern number. Strong short-ranged interactions then naturally favor one electron per SCWO, leaving the TPLO band empty or dilutely filled, and the resulting low-energy Hilbert space is a Kondo lattice of local moments exchange-coupled to an itinerant topological band. The paper verifies this picture numerically: in the BHZ model the low-lying eigenstates of the interacting model have about 95% overlap with the magnetic Wannier basis, and in twisted MoTe2 the -1/3 filled system develops a square-root-of-3 by square-root-of-3 charge-ordered antiferromagnetic Mott phase at larger twist angles, competing with the Laughlin fractional Chern insulator.","pith_inferences":["One testable extension is that the same construction should produce spin-unpolarized charge-ordered insulators in other time-reversal-symmetric moiré systems, such as twisted bilayer graphene or pentalayer graphene, where the SCWO localization length can be checked directly.","The competition between FCI and Kondo-lattice phases should be tunable: as twist angle or screening changes, the SCWO localization length should control which phase wins, giving a prediction for the phase boundary that could be tested in transport and local-probe experiments.","If the construction extends to three-dimensional Z2 topological bands, the emergent antiferromagnetic Kondo lattice would become a magnetic topological Mott insulator with axion electrodynamics, a possibility the paper notes as future work."],"forward_implications":["Strong interactions in time-reversal-symmetric moiré bands can stabilize spin-unpolarized topological Mott insulators without breaking time-reversal symmetry, making them direct competitors to fractional Chern insulators.","At 3/2 filling of a spin Chern band the TPLO band is fully occupied, producing a quantum spin Hall crystal, an integer quantum spin Hall insulator with spontaneously broken translation symmetry.","Any rational filling p/q admits a q-fold enlarged supercell and an associated partial Wannier basis, implying a hierarchy of translation-symmetry-broken Mott and Kondo-lattice states at nearby fillings.","The partial Wannier basis gives a concrete computational route to Hubbard parameters, spin exchange couplings, and RKKY interactions in topological bands, making quantitative predictions for magnetic order possible."],"supporting_citations":[{"why":"Defines the Bernevig-Hughes-Zhang model whose interacting fractional fillings are the first numerical demonstration of the emergent Kondo lattice.","marker":"[33]"},{"why":"Supplies the continuum model for twisted transition-metal dichalcogenide homobilayers whose topmost spin Chern band is used for the MoTe2 calculations.","marker":"[14]"},{"why":"Provides the model parameters (v, w, psi) for twisted MoTe2 used in the exact-diagonalization phase diagram.","marker":"[45]"},{"why":"Prior study of competing nonmagnetic states in twisted MoTe2 that the paper's AFM Mott phase extends and contrasts with the FCI.","marker":"[46]"},{"why":"Establishes the exponential-localization obstruction for bands with nonzero Chern number, the background the partial Wannier construction is designed to circumvent.","marker":"[27]"},{"why":"Supplies the projector and maximally-localized Wannier framework on which the supercell Wannier construction and gauge optimization are based.","marker":"[34]"},{"why":"Gives the smooth-gauge U(N) construction for composite bands with zero net Chern number that the paper generalizes to leave one obstructed orbital.","marker":"[36]"},{"why":"Experimental observation of fractional Chern insulators in twisted MoTe2 that defines the competing phase the paper's Mott state must be compared with.","marker":"[8]"}],"fun_headline_variants":["Chern bands can host emergent Kondo lattices","Supercell orbitals turn Chern bands into Mott insulators","Fractionally filled topological bands become Kondo lattices","Topological bands spawn emergent Kondo lattices","Mott states arise from supercell Wannier orbitals"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that for any topological band and any supercell size, a gauge rotation can push all the topological obstruction into a single extended orbital while the other N-1 orbitals stay exponentially localized; the paper shows this numerically for supercells up to eight unit cells but gives no general proof.","fun_headline_variants_meta":{"raw":{"variants":["Chern bands can host emergent Kondo lattices","Supercell orbitals turn Chern bands into Mott insulators","Fractionally filled topological bands become Kondo lattices","Topological bands spawn emergent Kondo lattices","Mott states arise from supercell Wannier orbitals"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000501,"raw_usage":{"total_tokens":2483,"prompt_tokens":1009,"completion_tokens":1474,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":625,"completion_tokens_details":{"reasoning_tokens":1398}},"tokens_in":625,"tokens_out":1474,"duration_ms":12251,"temperature":1.0,"reasoning_tokens":1398,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T05:43:51.619569+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct test is to compute the SCWO decay for a Chern band on a 16- or 32-unit-cell supercell, or for a band with strongly nonuniform Berry curvature; if the amplitudes decay algebraically rather than exponentially, the partial Wannier basis fails and the Mott description would not survive. Experimentally, a measurement on twisted bilayer MoTe2 at twist angles around or above 4 degrees and at -1/3 filling that finds a fully spin-polarized fractional Chern insulator with no square-root-of-3 charge order and no local moments would rule out the predicted antiferromagnetic Kondo-lattice phase in that material.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the continuum model for twisted transition-metal dichalcogenide homobilayers whose topmost spin Chern band is used for the MoTe2 calculations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Prior study of competing nonmagnetic states in twisted MoTe2 that the paper's AFM Mott phase extends and contrasts with the FCI."},{"cited_title":"Brouder, G","cited_arxiv_id":null,"evidence_quote":"Establishes the exponential-localization obstruction for bands with nonzero Chern number, the background the partial Wannier construction is designed to circumvent."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the smooth-gauge U(N) construction for composite bands with zero net Chern number that the paper generalizes to leave one obstructed orbital."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Experimental observation of fractional Chern insulators in twisted MoTe2 that defines the competing phase the paper's Mott state must be compared with."}],"review_version":1}