{"id":"7206b02a-781c-4466-858c-2ff965b55e21","arxiv_id":"2608.10059","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Eigenstates of the band projector restricted to a single unit cell, called cell natural orbitals, provide a ranked, symmetry-labelled local basis for building minimal tight-binding models of topological and moiré materials.","lead":"This paper introduces 'cell natural orbitals', local building blocks extracted directly from the quantum states of a material's energy bands, and shows how to use them to construct tight-binding models of twisted bilayer WSe2. A generalist may want to read it because it offers a systematic, wavefunction-derived answer to the long-standing practical question of which few orbitals best represent the low-energy physics of a correlated material.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'optimal trial states' claim rests on an unbenchmarked spectral-weight metric; the sole supporting comparison is confounded by different disentanglement choices and multi-center geometry.","rationale":"The paper's mathematical core—the UC-1pRDM, its eigenstates as CNOs, the symmetry labels, the topology-enforced zeros—is standard and correctly presented, and the authors honestly concede several limitations (weaker dispersion test, qualitative second-band Berry curvature, deferred fractional-state claim, N_G must be converged). The most load-bearing weakness is the 'optimal' claim: it is the central assertion of the abstract and conclusion, yet the quality metric λ_A is defined internally and never validated against an external measure. The one comparison offered is confounded by differing disentanglement windows and by the multi-center geometry, so it cannot establish that higher λ_A causes better localization. The concrete test proposed would settle this by holding η_km and N_G fixed and sweeping λ_A, then checking monotonic correlation with spread, hopping range, and interaction locality. If the correlation fails, the central claim should be weakened to 'a systematic construction of local trial states' rather than 'optimal.' This reinforces the reader's CONDITIONAL verdict; no verdict change is needed.","tokens_in":22874,"tokens_out":22479,"duration_ms":201244,"concrete_test":"Fix θ=3.0°, fix one disentanglement envelope η_km (a single exponential decay with stated constant, identical for all trial sets), and state a converged N_G. Enumerate at least three CNO subsets of size 3 with distinct λ_A satisfying det S_k ≠ 0 (e.g., MM top-3, multi-centered, and a mixed set with intermediate λ_A). Wannierize each with the same η_km; compute (i) Wannier spread, (ii) longest hopping length, (iii) ratio of largest non-local interaction to on-site U. If ordering by λ_A does not monotonically order (i)-(iii), the 'optimal' claim is falsified; if it does, the metric gains independent support.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Sec. IV A ranks candidate sets by λ_A = Tr(P P_A)/N_k (Eq. 28), and the abstract asserts that a set maximizing λ_A (plus symmetry) provides 'optimal trial states for Wannierization.' This is underdetermined. λ_A is never benchmarked against an independent measure of model quality (e.g., interaction locality, many-body fidelity). The only supporting evidence is the θ=3° comparison in Sec. V B: the multi-centered set has λ_A=0.899 and smaller Wannier spreads than the MM-centered set (λ_A=0.755), but the two sets use different disentanglement functions η_km (Sec. IV B): the MM set required η_km 'slow enough to draw weight from the third and fourth bands,' with no guarantee the decay constants or windows match. The multi-centered set also places orbitals on three Wyckoff sites, which trivially reduces real-space spread. Moreover, the body's actual ranking uses λ_A together with 'the resulting Wannier spread' (Sec. IV A), so the abstract's single-metric 'optimal' overstates the criterion. Finally, the numerical support lacks stated convergence parameters: Sec. IV C says the plane-wave cutoff N_G 'must be converged' but specifies no value or convergence study; without it, the exponential decay of hoppings (Fig. 7e) and the Berry curvature agreement (Fig. 7d,f) are not reproducible. The paper's own concessions—dispersion is a 'weaker test' and the second-band Berry curvature 'deviates slightly'—further limit the empirical claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript introduces the unit-cell one-particle reduced density matrix L = ΠPΠ for a set of target Bloch bands and defines its eigenstates as cell natural orbitals (CNOs). It shows that CNO occupations λ_a quantify overlap with the target manifold, relate to unit-cell entanglement entropy via the singular value decomposition of U = ΠP, and carry Wyckoff-position and site-symmetry labels when the partition respects symmetry. The paper proves that Chern bands force zeros in CNO envelope functions and illustrates symmetry-enforced zeros in a monolayer TMD model, then proposes a construction of tight-binding models by selecting CNO subsets that span the target bands and maximize the spectral weight λ_A (Eq. 28), followed by single-shot Wannierization. The method is applied to continuum twisted bilayer WSe2 at θ = 3°, yielding two three-orbital local models whose hoppings decay exponentially and whose Berry curvature approximately reproduces the continuum result.","tokens_in":23046,"tokens_out":4710,"duration_ms":45992,"significance":"The construction is elegant and largely rigorous: L is Hermitian positive semidefinite, the SVD relations in Sec. II B are clean, and the symmetry arguments in Sec. III provide a genuine real-space diagnostic for Wannier obstructions. The explicit BHZ and monolayer TMD examples are convincing, and the tWSe2 demonstration addresses a timely problem. If the optimality claim can be properly benchmarked, the method would give a useful, parameter-light alternative to manual Wannier orbital selection. The main limitation is that the 'optimal' claim rests on a spectral-weight metric that is not validated against independent measures of model quality, and the numerical demonstration lacks stated convergence parameters.","major_comments":[{"comment":"The paper equates the 'best' local orbital set with a set maximizing the total spectral weight λ_A (Eq. 28) plus symmetry compatibility, but this quality metric is never benchmarked against an independent measure of model quality such as locality of interaction matrix elements, many-body spectra, or convergence of observables with respect to retained orbitals. As written, the abstract's 'optimal trial states' claim is therefore an assertion rather than a demonstrated result. I request either a relaxation of the wording ('well-suited' or 'systematically ranked') or an additional numerical test that compares the λ_A ranking with an independent criterion.","section":"Sec. IV A / Abstract"},{"comment":"The comparison between the MM-centered and multi-centered CNO sets does not isolate the effect of λ_A. The two constructions use different disentanglement weightings η_km outside P — the MM set needs η_km 'slow enough to draw weight from the third and fourth bands,' while the multi-centered set uses a different slow decay — and the multi-centered set places orbitals on three Wyckoff sites, which trivially reduces the real-space spread. I ask for a controlled comparison: either use the same disentanglement function for both sets, or report results for several η_km choices and show that the relative ranking and spreads are stable.","section":"Sec. V B / Fig. 7"},{"comment":"The numerical results are not reproducible without the plane-wave cutoff N_G. Section IV C states that N_G 'must be converged' but gives no value and no convergence study, while the exponential decay of hoppings (Fig. 7e) and the Berry curvature comparison (Fig. 7d,f) depend on the real-space grid resolution. Please state the N_G used and provide a convergence test showing that the reported spreads, hopping decay constants, and Berry curvature profiles are converged.","section":"Sec. IV C / Sec. V"},{"comment":"Because single-shot Wannierization with N_τ = N_P retains the target bands exactly, reproducing the continuum dispersion (Fig. 7c) is a consistency check and not evidence that the CNO-selected model is better than an arbitrary trial set. The paper acknowledges that dispersion is a weaker test, but the Berry curvature comparison shows a 'slight' deviation in the second band that is not quantified. I recommend reporting a quantitative deviation (e.g., relative L2 error of Ω(k) or of the quantum metric) for both models, so the reader can evaluate the fidelity claim.","section":"Sec. IV B / Eq. (31)"}],"minor_comments":[{"comment":"There is a typo in the Introduction: 'translational symetry' should be 'translational symmetry'.","section":"Sec. I"},{"comment":"Equation (9) uses the notation ⟨N_cell⟩ but this quantity is not defined before use; please define it explicitly (the text says it is the total electron number per unit cell in the target bands, but the notation is introduced only in the equation).","section":"Sec. II A"},{"comment":"In the text following Eq. (32), η_km is called 'the projector' in one place ('the projector ηkm must extend beyond P'); since η_km is a weighting factor, not a projection operator, the phrasing should be corrected.","section":"Sec. IV B"},{"comment":"The figure caption appears to contain duplicated or rearranged panel labels (c) and e) appear twice, and some labels are inconsistent between the top and bottom panels); please make the panel callouts precise.","section":"Fig. 7 caption"},{"comment":"The conclusion states that the leading CNO 'defines an impurity space fixed by the wavefunctions' and that 'the occupation hierarchy orders the interaction scales,' but these are expectations rather than results shown here; I suggest marking them clearly as outlook items or citing the companion work [49] with a concrete summary.","section":"Sec. VI"}],"recommendation":"major_revision","confidential_remarks":"The core theoretical framework is sound and well within the journal's scope. My major concerns are fixable without changing the formulation: the optimality claim needs benchmarking or softening, the comparative Wannierization study needs a controlled disentanglement protocol, and the missing convergence parameters should be supplied. I would also encourage the authors to make the tWSe2 code and data available, since the main numerical application is otherwise difficult to reproduce."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this paper gives a genuinely new construction that I expect people will use. Diagonalizing the unit-cell-restricted band projector gives a ranked set of local orbitals tied to specific Wyckoff positions, with occupations that quantify how many orbitals a band manifold actually needs. The envelope-function zeros for Chern bands are a known theorem applied cleanly, and the monolayer WSe2 and BHZ examples make the idea concrete. The twisted WSe2 application is a nice demonstration that CNOs can produce short-ranged tight-binding models with Berry curvature that matches the continuum model. The math in Sections II–IV is straightforward and correct as far as it goes: L is well-defined, the SVD relations are clean, and det S_k is the right spanning test. The authors also make several honest concessions (dispersion as a 'weaker test', the algebraic-tail warning, the second-band Berry curvature deviation), which raises my confidence that they are not overselling wildly.\n\nThe soft spots are real but addressable. The abstract's 'optimal trial states' overstates what is actually shown: the ranking in Section IV A uses both spectral weight and Wannier spread, and spectral weight is never benchmarked against an independent measure of model quality (interaction locality, many-body fidelity, or anything else). The one explicit comparison between MM-centered and multi-centered sets is confounded by different disentanglement functions and by the trivial geometric advantage of spreading orbitals over three sites. That does not invalidate the construction, but it does mean the 'optimal' claim needs a better test. Second, the numerical section gives no plane-wave cutoff, no convergence study, no code or data. The exponential decay of hoppings and the Berry curvature agreement are not independently checkable from the paper alone. These are fixable in a revision.\n\nThe central construction itself is not circular: CNOs are eigenstates of a well-defined operator built from the input projector, with no fitted parameters. The circularity concern applies only to the validation via dispersion, which the authors themselves flag as a weaker test. I think the core idea stands.\n\nBottom line: this deserves a serious referee, likely with revision. The paper is for people building real-space models for moiré TMDs, topological bands, and embedding methods like DMFT and fRG. If the authors add a benchmark of spectral weight against an independent model-quality measure and provide numerical data, I would be comfortable citing it. As is, I would take it seriously but with the caveats above.","headline":"A genuinely new wavefunction-derived route to local-orbital selection, with an overreaching 'optimal' claim and some reproducibility gaps, but the core construction is sound and worth engaging.","tokens_in":23814,"tokens_out":1138,"would_cite":true,"duration_ms":13602,"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":"The paper introduces cell natural orbitals, eigenstates of a unit-cell density matrix, as a way to identify the few local orbitals that best capture chosen bands, yielding short-ranged Wannier models for twisted WSe2.","keywords":["cell natural orbitals","unit-cell one-particle reduced density matrix","Wannierization","quantum geometry","twisted bilayer WSe2","moiré materials","topological bands","unit-cell entanglement"],"falsifier":"Vary the disentanglement weighting $\\eta_{km}$ and the plane-wave cutoff $N_G$ in the twisted WSe2 construction and check whether the exponential decay length of the hoppings or the Berry curvature agreement changes materially; alternatively, solve the interacting problem in the CNO-derived model and see whether a lower-$\\lambda_A$ set with more local interactions reproduces the continuum many-body spectrum better than the maximal-$\\lambda_A$ set.","tokens_in":22491,"feed_emoji":"⚛️","tokens_out":15520,"duration_ms":128160,"temperature":0.7,"pith_summary":"This paper aims to make the choice of local orbitals for a band-structure model something the wavefunctions decide, rather than a guess. The authors define the unit-cell one-particle reduced density matrix, obtained by restricting the projector onto a chosen set of bands to a single unit cell, and show that its eigenstates—the cell natural orbitals—form a local, symmetry-labelled basis whose eigenvalues rank how much each orbital participates in those bands. They argue that the subset of these orbitals with maximal total spectral weight, subject to spanning the target bands at every momentum, provides optimal trial states for building tight-binding models, even when the bands are topologically nontrivial. They demonstrate the construction on twisted bilayer WSe2, producing local models with exponentially decaying hoppings whose Berry curvature tracks the continuum model across twist angles. If correct, the construction turns a routine object—the band projector—into a systematic diagnostic of multi-orbital character and a practical tool for correlated-materials modeling.","feed_headline":"Cell natural orbitals single out the orbitals that span twisted WSe2","feed_subtitle":"A unit-cell density matrix ranks the few local degrees of freedom that capture dispersion, geometry, and topology.","key_machinery":"The central object is the unit-cell one-particle reduced density matrix $L = \\Pi P \\Pi$, where $P$ projects onto the target Bloch bands and $\\Pi$ restricts to a single unit cell; its eigenstates are the cell natural orbitals and its eigenvalues are their occupations. Equivalently, through the singular value decomposition of $U = \\Pi P$, each occupation $\\lambda_a$ is the squared singular value connecting a CNO $|\\tau_a\\rangle$ to a band-subspace state $|s_a\\rangle$ whose rescaled Bloch coefficients $s_{a,n,k} = \\langle u^{\\mathrm{per}}_{kn}|\\tau_a\\rangle/\\sqrt{\\lambda_a}$ form the envelope functions that carry the zeros diagnosing topological or symmetry obstructions. The construction also supplies the completeness criterion $\\det S_k \\neq 0$, with $S_k$ the overlap Gram matrix between CNO Bloch states and target bands, and the ranking functional $\\lambda_A$ for choosing among candidate CNO sets. These ingredients together convert the band projector into trial states for the single-shot Wannierization procedure used to obtain explicit tight-binding models.","core_discovery":"The central claim is that the eigenstates $|\\tau_a\\rangle$ of $L = \\Pi P \\Pi$, called cell natural orbitals, are the local degrees of freedom intrinsically selected by a target band manifold: they are uniquely determined by the Bloch wavefunctions and the real-space partition, carry site-symmetry irreps and Wyckoff positions, and their eigenvalues $\\lambda_a$ measure the occupation of each orbital in the manifold. For Chern bands the projected envelope functions $s_{a,n,k}$ necessarily contain zeros, so no single compact orbital can span the manifold; these zeros are irremovable, while symmetry-enforced zeros can be lifted by moving the partition to the appropriate Wyckoff position. A CNO subset that spans the target bands (det $S_k \\neq 0$ at every $k$) and maximizes the total spectral weight $\\lambda_A = \\mathrm{Tr}(P P_A)/N_k$ is proposed as the optimal trial-state set for single-shot Wannierization. Applied to twisted bilayer WSe2 at 3°, both an MM-centered three-orbital set and a multi-centered set produce exponentially localized Wannier functions and Wannier models whose Berry curvature reproduces the continuum result, with the multi-centered set carrying the higher total spectral weight (0.899 versus 0.755).","pith_inferences":["Editorial inference: the paper's optimality metric is not benchmarked; a natural test is to compare CNO-selected models against many-body spectra or the locality of interaction matrix elements, not only band dispersion and Berry curvature.","Editorial inference: the final localization depends on the manually chosen disentanglement function $\\eta_{km}$ for bands outside the target manifold and on the plane-wave cutoff $N_G$; CNOs inform but do not fix these choices, so the practical 'single-shot' claim is weaker than the construction itself.","Editorial inference: because the construction takes only the projector as input, it could be added as a routine diagnostic in automated Wannierization pipelines, flagging band inversions through the momentum positions of envelope-function zeros.","Editorial inference: the paper's closing speculation that the dominant CNO favors charge-density-wave order while subdominant modes stabilize fractional Chern states is testable by solving the interacting problem in the CNO basis, which would directly tie the occupation hierarchy to interaction scales."],"forward_implications":["CNO occupations give a quantitative, projector-only measure of multi-orbital character for any set of bands, applicable to both lattice and continuum models.","For Chern bands, no single compact local orbital can span the manifold: the leading CNO envelope necessarily vanishes somewhere in the Brillouin zone, a real-space signature of the Wannier obstruction.","Symmetry-enforced zeros in CNO envelopes are removable by recentering the unit-cell partition at the appropriate Wyckoff position, which can reduce the number of orbitals needed to describe an obstructed atomic insulator.","In twisted bilayer WSe2 at 3°, the CNO trial states produce exponentially decaying Wannier functions and short-ranged tight-binding models whose Berry curvature matches the continuum result for the top band.","Because the only input is a projector, the same construction carries over to first-principles wavefunctions, so the occupation spectrum could become a routine companion to a band-structure calculation."],"supporting_citations":[{"why":"Supplies the natural-orbital concept that the paper extends to the periodic, band-projector setting.","marker":"[27]"},{"why":"Shows the reduced density matrix is obtained from a restricted correlation matrix, grounding the entanglement interpretation.","marker":"[47]"},{"why":"Extends that relation to entanglement entropy, used for the unit-cell entropy formula.","marker":"[48]"},{"why":"Defines Wannier functions and localization theory, the target the CNO trial states feed into.","marker":"[19]"},{"why":"Provides the single-shot/disentangled Wannierization procedure used to convert CNOs into explicit tight-binding models.","marker":"[20]"},{"why":"Establishes the topological quantum chemistry formalism linking local orbitals to band representations.","marker":"[24]"},{"why":"Defines elementary band representations, from which CNOs inherit Wyckoff positions and site-symmetry irreps.","marker":"[25]"},{"why":"Supplies the band-representation formalism underlying the symmetry labels of CNOs.","marker":"[26]"},{"why":"Provides the continuum model of twisted transition metal dichalcogenide homobilayers used as the starting point for twisted WSe2.","marker":"[14]"},{"why":"Supplies the parameter set $(V, w, \\psi, m^*)$ used for the twisted WSe2 calculations across twist angles.","marker":"[93]"}],"fun_headline_variants":["Cell natural orbitals pick optimal local basis for twisted WSe2","CNOs: local orbitals that span twisted WSe2 bands","Unit-cell density matrix reveals key orbitals in quantum materials","Cell natural orbitals identify local degrees of freedom in twisted WSe2"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim that a CNO set is optimal rests on equating 'best' with maximal total spectral weight plus symmetry compatibility, a proxy the paper never compares against an independent measure of whether a local model is good for correlated calculations.","fun_headline_variants_meta":{"raw":{"variants":["Cell natural orbitals pick optimal local basis for twisted WSe2","CNOs: local orbitals that span twisted WSe2 bands","Unit-cell density matrix reveals key orbitals in quantum materials","Cell natural orbitals identify local degrees of freedom in twisted WSe2"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000439,"raw_usage":{"total_tokens":2269,"prompt_tokens":1025,"completion_tokens":1244,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":641,"completion_tokens_details":{"reasoning_tokens":1174}},"tokens_in":641,"tokens_out":1244,"duration_ms":10408,"temperature":1.0,"reasoning_tokens":1174,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:16:54.653563+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Vary the disentanglement weighting $\\eta_{km}$ and the plane-wave cutoff $N_G$ in the twisted WSe2 construction and check whether the exponential decay length of the hoppings or the Berry curvature agreement changes materially; alternatively, solve the interacting problem in the CNO-derived model and see whether a lower-$\\lambda_A$ set with more local interactions reproduces the continuum many-body spectrum better than the maximal-$\\lambda_A$ set.","supporting_citations":[{"cited_title":"L¨ owdin, Quantum theory of many-particle systems","cited_arxiv_id":null,"evidence_quote":"Supplies the natural-orbital concept that the paper extends to the periodic, band-projector setting."},{"cited_title":"Peschel and V","cited_arxiv_id":null,"evidence_quote":"Extends that relation to entanglement entropy, used for the unit-cell entropy formula."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines elementary band representations, from which CNOs inherit Wyckoff positions and site-symmetry irreps."}],"review_version":1}