{"id":"c7f59364-2a94-429d-9cf2-4b76404ad8f8","arxiv_id":"2511.12551","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"PCA applied to correlation matrices from ED data on 3x4 and 4x4 clusters identifies charge-dominated, spin-dominated, and pairing-dominated regimes in quarter-filled Hubbard models through variance condensation into leading components.","lead":"This paper applies principal component analysis to exact-diagonalization data from quarter-filled Hubbard models on small cylindrical clusters to identify how charge, spin, and pairing correlations compete as repulsion strength changes. A smart generalist might read it to understand a data-driven way to spot dominant behaviors in complex quantum systems without guessing order parameters first.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.3","headline":"PCA variance condensation on 3x4/4x4 cylinders may track finite-size boundary effects or correlation-function selection rather than bulk hierarchy","rationale":"The reader’s weakest assumption is precisely the finite-size and selection-effect vulnerability identified above; the abstract’s emphasis on 3x4/4x4 clusters makes this the single most load-bearing point. Full-text access does not remove the concern because the abstract already states the system sizes used, and no larger-system or scaling data are mentioned there.","tokens_in":1671,"tokens_out":386,"duration_ms":23282,"concrete_test":"Re-run the full PCA pipeline on the same U values but with an expanded correlation matrix that doubles the number of included distances and includes both x- and y-directed operators; if the identity or ordering of the top three components changes by more than one rank or the variance fraction in the leading component shifts by >15 %, the regime identification is sensitive to the operator selection and therefore not robust.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim requires that the leading principal components of the correlation matrix faithfully encode the physical competition among charge, spin, and pairing channels. The analysis is performed exclusively on 3x4 and 4x4 cylindrical clusters with ED. On these sizes, periodic or open boundaries truncate long-range correlations, impose artificial periodicity, and restrict the available momentum points, so that variance condensation into the first few components can arise from the restricted Hilbert space or from the particular subset of two-point functions (distances, directions, and operator types) that were chosen to populate the matrix. Without an explicit demonstration that the same leading components and regime boundaries survive under changes in cluster geometry, boundary conditions, or an enlarged operator set, the mapping from PCA output to “charge-dominated, spin-dominated, pairing-dominated regimes” remains vulnerable to these artifacts.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript applies principal component analysis (PCA) to correlation matrices constructed from exact-diagonalization (ED) data on 3×4 and 4×4 cylindrical clusters for the quarter-filled simple and extended Hubbard models. It reports that increasing on-site repulsion U induces localization and reorganizes the spectrum, and that PCA identifies charge-dominated, spin-dominated, and pairing-dominated regimes through variance condensation into leading components, without assuming predefined order parameters. The approach is benchmarked against conventional structure factors in the extended model at moderate (U=4) and strong (U=10) coupling and is presented as a model-agnostic framework bridging ED and machine-learning diagnostics for competing correlations in finite clusters.","tokens_in":1852,"tokens_out":624,"duration_ms":40730,"significance":"If the central claim is substantiated, the work demonstrates that unsupervised PCA applied directly to raw ED correlation data can recover known crossovers among charge, spin, and pairing channels in Hubbard models. This provides a transparent, parameter-free diagnostic that complements traditional order-parameter analyses and may be useful for exploring regimes where multiple instabilities compete. The use of exact ED input data on small clusters and the explicit comparison to structure-factor results constitute clear strengths.","major_comments":[{"comment":"The central claim that PCA captures the crossovers 'directly from the data' and identifies regimes via variance condensation requires that the leading components reflect the physical hierarchy rather than the restricted set of operators or finite-size truncation. The manuscript performs the analysis exclusively on 3×4 and 4×4 cylinders; without explicit checks (e.g., comparison of leading eigenvectors or explained-variance ratios when the operator set is enlarged or when periodic versus open boundaries are swapped), the mapping from PCA output to charge-/spin-/pairing-dominated regimes remains vulnerable to artifacts from the limited Hilbert space and momentum sampling.","section":"Results for the extended Hubbard model (U=4 and U=10)"},{"comment":"The abstract states that 'conventional structure factors reveal familiar crossovers' that PCA then captures, yet the manuscript does not report quantitative agreement metrics (e.g., overlap between PCA regime boundaries and structure-factor crossover points, or tables of leading-component weights for charge, spin, and pairing channels). This quantitative link is load-bearing for the assertion that PCA reproduces the known hierarchy without predefined order parameters.","section":"Comparison with conventional structure factors"}],"minor_comments":[{"comment":"The precise construction of the correlation matrix—i.e., the list of two-point operators, distances, and directions included for each channel—should be stated explicitly (ideally in a table or appendix) to ensure reproducibility.","section":"Methods"},{"comment":"Figure captions or text should clarify whether the PCA is performed on the full covariance matrix or on a normalized correlation matrix, and whether any centering or scaling is applied before decomposition.","section":"PCA implementation"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for their careful reading of the manuscript and for the constructive comments. We address each major comment in turn below, indicating where revisions will be made to improve clarity and strengthen the presentation.","responses":[{"response":"We agree that the restricted cluster sizes and operator basis inherent to ED calculations on small cylinders represent a limitation that must be discussed explicitly. The manuscript already compares results across the two available cluster geometries (3×4 and 4×4) and finds consistent identification of the charge-, spin-, and pairing-dominated regimes. In the revised version we will add a dedicated paragraph in the discussion section that quantifies the stability of the leading eigenvectors and explained-variance ratios when additional longer-range correlation operators are included in the matrix. We will also note that the cylindrical (open-boundary in one direction) geometry is the standard choice for these system sizes to minimize momentum discretization artifacts, and we will state the computational constraints that preclude immediate checks on larger or fully periodic clusters. These additions will make the robustness arguments more transparent without altering the central claims.","revision_made":"partial","referee_comment":"[Results for the extended Hubbard model (U=4 and U=10)] The central claim that PCA captures the crossovers 'directly from the data' and identifies regimes via variance condensation requires that the leading components reflect the physical hierarchy rather than the restricted set of operators or finite-size truncation. The manuscript performs the analysis exclusively on 3×4 and 4×4 cylinders; without explicit checks (e.g., comparison of leading eigenvectors or explained-variance ratios when the operator set is enlarged or when periodic versus open boundaries are swapped), the mapping from PCA output to charge-/spin-/pairing-dominated regimes remains vulnerable to artifacts from the limited Hilbert space and momentum sampling."},{"response":"The original text presents the alignment between PCA regimes and structure-factor crossovers through direct visual comparison in the figures for U = 4 and U = 10. We recognize that explicit quantitative measures would make this correspondence more rigorous. Accordingly, the revised manuscript will include a new table that reports the projection weights of the leading principal components onto the charge, spin, and local-pairing channels at the two coupling strengths. We will also define regime boundaries from the points where the leading-component variance exceeds a chosen threshold and tabulate the numerical agreement with the structure-factor crossover locations. These additions will furnish the quantitative link requested while preserving the model-agnostic character of the PCA approach.","revision_made":"yes","referee_comment":"[Comparison with conventional structure factors] The abstract states that 'conventional structure factors reveal familiar crossovers' that PCA then captures, yet the manuscript does not report quantitative agreement metrics (e.g., overlap between PCA regime boundaries and structure-factor crossover points, or tables of leading-component weights for charge, spin, and pairing channels). This quantitative link is load-bearing for the assertion that PCA reproduces the known hierarchy without predefined order parameters."}],"tokens_in":1467,"tokens_out":626,"duration_ms":43157,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper runs PCA on correlation matrices from exact diagonalization of quarter-filled simple and extended Hubbard models on 3x4 and 4x4 cylindrical clusters. It shows that the leading components separate regimes where charge, spin, or local-pairing correlations dominate as U increases, using the non-interacting limit as reference and aligning with conventional structure-factor crossovers at U=4 and U=10.","headline":"PCA on small-cluster ED data picks out charge-spin-pairing crossovers in quarter-filled Hubbard models without preset order parameters, though finite-size effects on 3x4 and 4x4 cylinders need direct checks.","tokens_in":2307,"tokens_out":169,"would_cite":false,"duration_ms":20757,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":{"model":"grok-4.3","evidence":[{"relation":"unclear","rs_module":"IndisputableMonolith/Cost/FunctionalEquation.lean","rs_theorem":"washburn_uniqueness_aczel","paper_passage":"PCA of the corresponding correlation matrices captures these crossovers directly from the data, without assuming predefined order parameters by identifying charge-dominated, spin-dominated and pairing-dominated regimes through variance condensation into leading components."}],"headline":"PCA on finite-cluster Hubbard correlation matrices","alignment":"orthogonal","rationale":"The paper applies standard unsupervised PCA to site-resolved charge/spin/pairing correlation matrices obtained from ED on 3x4/4x4 cylinders. It identifies variance condensation into leading components that track crossovers among charge-, spin- and pairing-dominated regimes. No J-cost, cosh-cost, golden-ratio identities, φ-ladder spacings, 8-tick periodicity, or parameter-free constant derivations appear; the work is conventional condensed-matter numerics on the Hubbard Hamiltonian with tunable U,V.","tokens_in":55614,"confidence":"high","tokens_out":231,"duration_ms":10447,"cache_read_input_tokens":38528,"cache_creation_input_tokens":0},"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Principal component analysis of correlation matrices from exact diagonalization data identifies charge-dominated, spin-dominated and pairing-dominated regimes in quarter-filled Hubbard models without assuming order parameters in advance.","keywords":["Hubbard model","principal component analysis","correlation functions","exact diagonalization","competing orders","strongly correlated electrons","unsupervised learning","finite clusters"],"falsifier":"Repeating the identical PCA procedure on larger clusters or on clusters with different boundary conditions and checking whether the same sequence of charge-spin-pairing crossovers appears.","tokens_in":2580,"feed_emoji":"📊","tokens_out":741,"duration_ms":20243,"temperature":0.7,"pith_summary":"The paper applies principal component analysis to matrices built from various correlation functions computed on small 3x4 and 4x4 clusters for the quarter-filled simple and extended Hubbard models. As on-site repulsion increases, the variance in these matrices condenses into a few leading principal components that label the dominant correlation type at each coupling strength. Conventional structure factors already show crossovers among charge, spin and local-pairing channels; the PCA reproduces those crossovers directly from the raw data. A reader would care because the method requires no prior guess about which order parameter to measure and therefore supplies a model-agnostic diagnostic for competing correlations on finite clusters.","feed_headline":"PCA sorts charge, spin and pairing regimes in Hubbard clusters","feed_subtitle":"Unsupervised analysis of correlation matrices on small clusters detects crossovers without any assumed order parameter.","key_machinery":"Principal component analysis performed on matrices whose entries are different two-point correlation functions; the leading eigenvectors condense the total variance and thereby label which correlation channel dominates at given interaction strength.","core_discovery":"We present an unsupervised learning analysis of correlation hierarchies in the quarter-filled simple and extended Hubbard models by applying principal component analysis to exact-diagonalization data on 3x4 and 4x4 cylindrical clusters. While the non-interacting limit provides a finite-size reference, increasing on-site repulsion induces localization and reorganizes the low-energy spectrum. For the extended model at moderate and strong coupling, PCA of the corresponding correlation matrices captures the familiar crossovers directly from the data by identifying charge-dominated, spin-dominated and pairing-dominated regimes through variance condensation into leading components.","pith_inferences":["One could test whether including longer-range or higher-order correlation functions alters the ordering of the leading components or merely redistributes variance among sub-dominant ones.","The method may prove useful for mapping out crossovers in other fillings or geometries where conventional order parameters are harder to guess.","If the identified regimes remain stable under modest changes in cluster size, the technique could serve as a quick pre-screening step before more expensive calculations."],"forward_implications":["In the extended Hubbard model at U=4 and U=10 the leading components cleanly separate the three regimes already known from structure-factor analysis.","The approach supplies a transparent bridge between exact-diagonalization spectra and modern unsupervised diagnostics for strongly correlated systems.","No predefined order parameter is required; the data themselves determine which correlation channel carries the largest variance.","The same workflow can be applied to any set of correlation functions computed on finite clusters, making it portable across different Hubbard variants."],"fun_headline_variants":["PCA sorts Hubbard charge spin pairing regimes","PCA uncovers correlation crossovers in Hubbard clusters","PCA captures variance in Hubbard correlation matrices","Unsupervised PCA maps regimes in quarter-filled Hubbard"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The leading principal components extracted from correlation matrices on small finite clusters faithfully reflect the physical hierarchy of competing correlations rather than finite-size artifacts or the particular choice of functions included in the matrix.","fun_headline_variants_meta":{"raw":{"variants":["PCA sorts Hubbard charge spin pairing regimes","PCA uncovers correlation crossovers in Hubbard clusters","PCA captures variance in Hubbard correlation matrices","Unsupervised PCA maps regimes in quarter-filled Hubbard"]},"model":"grok-4.3","cost_usd":0.010038,"raw_usage":{"total_tokens":4369,"prompt_tokens":654,"num_sources_used":0,"completion_tokens":52,"cost_in_usd_ticks":100378000,"prompt_tokens_details":{"text_tokens":654,"audio_tokens":0,"image_tokens":0,"cached_tokens":64},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3663,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":654,"tokens_out":52,"duration_ms":28076,"temperature":1.0,"reasoning_tokens":3663,"cache_read_input_tokens":64,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-05-17T22:24:22.768445+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Repeating the identical PCA procedure on larger clusters or on clusters with different boundary conditions and checking whether the same sequence of charge-spin-pairing crossovers appears.","supporting_citations":[],"review_version":1}