{"id":"207a8956-168d-475e-b1cb-c0ca9487d38c","arxiv_id":"2507.07886","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Local frustration centers in a single-hole Hubbard model bind one singlet each while sharing a delocalized hole, so embedding them on a graph tunes the net magnetization in steps of one.","lead":"The paper shows that adding frustrating diagonal bonds to a lattice with one mobile hole makes each such bond bind a pair of spins into a singlet while the hole stays spread out, lowering the total spin in predictable steps. This could let experimentalists sculpt magnetic ordering in cold-atom quantum simulators by choosing where to place the frustration.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Additivity of singlet binding is inferred from ≤20-site numerics; the key untested assumption is that each added frustration center continues to reduce S_total by 1 while the hole remains delocalized on larger graphs.","rationale":"The reader's weakest-assumption analysis already identifies the finite-size extrapolation and the possibility of hole localization or non-additive singlet binding as the central risk. My pass agrees: the paper is internally consistent, the numerics are reproducible (data and code posted), and the ladder DMRG up to l=32 and the 20-site random-graph ensemble are solid for what they directly show. The gap is that the stated 'general principle' is a claim about arbitrarily large graphs, and the evidence stops at 20 sites (random graphs) and 12 diagonals on a 4×5 grid. I could not find an internal inconsistency or a hidden assumption in the derivations; the concern is purely about whether the finite-size behavior is indicative of the thermodynamic limit. This is exactly why the reader's CONDITIONAL verdict is appropriate, and why I do not propose to change it. The proposed DMRG check on 6×6/8×8 grids and 30–40-site random graphs is the minimal experiment that would either corroborate the additivity/delocalization principle or expose its breakdown.","tokens_in":16813,"tokens_out":2229,"duration_ms":25763,"concrete_test":"Perform DMRG (as already used for ladders up to l=32) on 6×6 and 8×8 rectangular grids with N_diag = 0, 4, 8, 12, 16, 18 diagonals placed at most one per plaquette, and compute S_total and the hole inverse participation ratio (IPR). If ΔS_total per added diagonal drops below 0.5 before S_total reaches the saturation floor, or if the hole IPR grows beyond a value indicating localization (e.g., > 0.3 for a 64-site grid), the additivity/delocalization claim needs qualification. For random graphs, repeat the 20-site correlation analysis at 30 and 40 sites using DMRG in the low-S sector; if the negative S_total-versus-frustration-index correlation weakens substantially with size, the proposed general principle does not extend as stated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of the paper — that local frustration centers bind singlets while sharing a delocalized hole, so each center reduces S_total by about 1 — rests entirely on finite-size numerical evidence: a 4×5 grid with up to 12 diagonals (Fig. 4a) and random graphs with up to 20 sites (Fig. 5, SM). No scaling analysis, rigorous bound, or thermodynamic-limit argument is provided. The danger is not merely that finite-size numbers drift: the underlying mechanism itself could break down. If the hole becomes localized on larger graphs — e.g., because diagonals form clusters that act as potential wells, or because the hole cannot simultaneously 'share' many frustration centers — then the hole is no longer shared, singlet binding per center need not be additive, and the sharp S_total = N/2 − N_diag prediction fails. A related but distinct risk is saturation: S_total cannot go below 0 or 1/2 (for odd N), so if N_diag is large enough the reduction per diagonal must eventually drop below 1. The paper does not state the regime in which additivity is expected to hold, nor does it bound the crossover. Since the headline 'general principle' is exactly this additivity-plus-delocalization statement, the absence of any test at larger system size or larger N_diag is the most load-bearing weakness.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies how local kinetic frustration shapes magnetic order in the infinite-U Hubbard model with a single hole on a graph. The authors solve a single square with diagonal hopping exactly, identifying three regions: a Nagaoka ferromagnet, a vertical-singlet state, and a cross-dimer singlet, with an exact energy expression in the intermediate region. They then embed one or two frustrated plaquettes in ladders, using exact diagonalization and DMRG up to length 32, and show that the frustrated plaquette binds magnons with perturbatively controlled weights (P_out = 1/5 for the ladder, 1/9 for the grid). The central claim is a general principle: local frustration centers bind singlets while sharing a delocalized hole, so each added diagonal bond on a 4x5 grid reduces the ground-state total spin by roughly 1, and the total spin on 20-site random graphs is anticorrelated with loop counts and the frustration index. A half-filled Heisenberg comparison shows none of this sensitivity. An appendix outlines an optical-lattice scheme for realizing tunable diagonal hops.","tokens_in":17063,"tokens_out":19285,"duration_ms":223636,"significance":"Strengths: the single-square solution is exact and its three phases are characterized analytically with parity assignments; the perturbation-theory prediction for the magnon occupation outside the frustrated plaquette (P_out = 1/5) is verified against ED across the whole cross-dimer region; the ladder phase diagrams are stable to length 32 in DMRG; the Heisenberg contrast cleanly isolates the kinetic mechanism; and the data and source codes are archived on Zenodo (refs. 80-81). The most valuable outcome, if confirmed at larger size, is a falsifiable design rule: placing a diagonal bond in an otherwise bipartite ferromagnetic background locally binds a singlet and lowers S_total by one, enabling spatially resolved control of magnetization that exchange magnetism cannot provide. The variational ansatz for the two-plaquette collective state is presented with overlap data rather than as a derivation, and the fitted parameter alpha is not used as an input to the additivity claim, so no circularity is apparent.","major_comments":[{"comment":"The central additivity claim is not what Fig. 4(a) shows over its full range. For the 4x5 grid with 19 fermions, S_max = 9.5 and the hard lower bound is S = 1/2, so at most 9 of the 12 possible diagonals can reduce S by a full unit. The average total spin plotted in Fig. 4(a) falls from about 9.5 at N_diag = 0 to about 1.5 at N_diag = 12, i.e., an average reduction of roughly 8 units over 12 diagonals, or about 2/3 per diagonal, not 1. The paper does not state this saturation bound, does not identify the crossover where the slope-1 rule must fail, and does not report how the distribution over the 1000 graphs interacts with the S = 1/2 floor. In addition, the text claims that the Shastry-Sutherland-constrained placement reduces S by 1 per diagonal, but Fig. 4(a) shows the unconstrained case; the two should be shown separately, since the unconstrained data as plotted are consistent with a much weaker average slope near the upper end of N_diag. Because the headline general principle is exactly the additive slope-1 behavior, a quantitative regime statement, including the dependence on the diagonal-to-side hopping ratio (held fixed at t' = t in the grid data), is required.","section":"Grid with diagonal bonds (Fig. 4a)"},{"comment":"The extension to random graphs rests on exact diagonalization of 20-site graphs, with SM Fig. S5 extending the loop-count correlations down to 10 sites. This is encouraging but is not a scaling analysis: no statement is made about whether the about -0.5 correlation with the frustration index or the loop-count correlations converge with N, and the mechanism advanced for grids (each frustration center binds one singlet while the hole stays delocalized) is not directly testable on random graphs because, as the authors note, these do not have well-defined frustration centers. The abstract's claim that the principle extends to random graphs therefore currently bridges from a local mechanism to aggregate loop statistics by analogy. Either add a finite-size study showing a trend (e.g., correlation strength and S-distribution moments for N = 22-28 in fixed S_z sectors), or explicitly delimit the random-graph claim to what is computed.","section":"Random graphs (Fig. 5; SM Fig. S5)"},{"comment":"The stated general principle, that local frustration centers bind singlets while sharing a delocalized hole, is not the operative mechanism in all regimes presented in the paper. In region V of Fig. 3 the hole is localized in one plaquette while the other stays polarized (S is still reduced by 2), and in region III the hole is confined between the plaquettes with two domain walls; a shared, delocalized hole occurs only in regions II and IV. Furthermore, all grid data in Fig. 4 are taken at one parameter point, equal hopping on all bonds, which for a single square sits exactly at the II-III degeneracy in Fig. 1(b); for a weaker diagonal (u < 1/5 in Fig. 1) a single square remains a fully polarized Nagaoka ferromagnet and the reduction per diagonal is zero, so the additivity rule must depend on the frustration strength and not just on the number of centers. The general principle should be reformulated with its regime of validity (hopping ratio, distance between centers, and proximity to saturation), or the claims should be explicitly presented as regime-specific.","section":"Abstract and Figs. 3-4"}],"minor_comments":[{"comment":"Please clarify how the N_diag axis is sampled: are the plotted points independent random graphs with exactly N_diag diagonals, or averages over growth trajectories in which bonds are added one at a time? The two procedures give different distributions, and the caption's 'up to 1000 graphs' does not resolve this.","section":"Fig. 4 caption"},{"comment":"The maximum variational overlap in region IV, |<psi(alpha*)|GS'>|^2 ~ 0.6-0.8, is moderate; the main text should state that the two-plaquette collective wavefunction is an illustration and that the primary evidence for region IV is the ED/DMRG reduction of S by 4.","section":"Appendix B / Fig. 8"},{"comment":"A Pearson correlation of about -0.5 is moderate, not 'strong' as stated in the text; please report the correlation type, the scatter, and confidence intervals (the SM already gives 95% intervals for Fig. S5), and soften the wording accordingly.","section":"Random graphs section"},{"comment":"Please specify how the five regions of the two-plaquette phase diagram are delimited (e.g., level crossings between symmetry sectors) and whether the transitions are first-order; the figure currently shows no gap or transition data for this case.","section":"Fig. 3b"},{"comment":"The parenthetical comment '(For t_i,j < 0, ferromagnetism is not frustrated [36-38])' is too compressed; one sentence explaining the role of the hopping sign relative to the loop orientation would help readers who are not specialists in kinetic frustration.","section":"Setup"},{"comment":"For the Heisenberg random-graph comparison, please report the fraction of nonseparable graphs in the ensembles that are non-bipartite and check the claim of 'negligible spread' of S around zero for r_min = 3, since an antiferromagnet on a graph containing an isolated odd cycle generically has a nonzero ground-state spin.","section":"SM Fig. S4"},{"comment":"Please state the role of the central mediator site in the U to infinity limit: whether it is empty in the initial state, whether its on-site interaction enters the virtual tunneling, and what condition keeps it from being occupied; as written, the second-order process is described only through the band-structure plots in Fig. 10.","section":"Appendix D"}],"recommendation":"major_revision","confidential_remarks":"The central issue is the gap between the advertised 'general principle' and the finite-size evidence. The requested revisions are feasible with the tools the authors already use (DMRG on wider ladders, Lanczos at fixed S_z for N ~ 22-28 random graphs) or by explicitly qualifying the claims. Please ask the authors to provide the underlying data behind Fig. 4(a) (median and quantiles vs. mean, and the Shastry-Sutherland-constrained curve), since the paper's slope-1 statement in the text and the plotted mean slope of about 2/3 are not obviously consistent. The exact single-square solution and the P_out = 1/5 perturbation check are solid and should be preserved. The archived code and data are a genuine asset for the journal's reproducibility standards."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things. First, this paper is honest, reproducible work: the single square is solved exactly, the perturbation theory for P_out matches ED, the DMRG and random-graph ensembles are documented, and the data and code are on Zenodo. Second, the advertised 'general principle' — that each local frustration center binds one singlet while the hole stays delocalized — is an extrapolation from 20-site random graphs, a 4x5 grid, and ladders up to length 32. The conditional verdict and the stress-test note are both on target about that gap, and I agree with them.\n\nWhat is actually new: the systematic observation that inserting diagonal bonds into a grid reduces total spin by roughly one per diagonal in the dilute limit, the two-plaquette collective regions III and IV where the hole mediates binding between centers, and the random-graph correlations with loop counts and girth. The region IV state, with distorted singlets on both plaquettes, is a genuinely collective effect that would not happen for independent plaquettes. The contrast with Heisenberg exchange in the SM is a nice control and strengthens the kinetic-magnetism interpretation.\n\nThe soft spot, in proportion: the paper does not state a precise conjecture or give a scaling analysis for when additivity holds. The sentence 'we do not expect the physics to be limited by system size' is hope, not evidence. The saturation visible in Fig. 4(a) — 12 diagonals reduce S from 9.5 to about 1.5, which is less than 1 per diagonal — already shows the rule has a crossover, and the authors do not specify where or why. If the hole localizes on larger graphs or the per-center reduction stops being additive, the principle fails. That is a real gap, not a nitpick. But the paper does not pretend to have a theorem; it presents a numerical discovery with a plausible mechanism, and the evidence across the studied sizes is broad and consistent.\n\nWho this is for: people working on kinetic magnetism, cold-atom quantum simulators, and possibly network science applications of correlated quantum systems. A serious referee can extract a clear task: weaken the 'general principle' language or provide a scaling analysis / bounded crossover. I would send it to peer review rather than desk-reject, and I'd cite it for the dilute-limit additivity on grids and the random-graph loop-count correlations.","headline":"A credible, reproducible numerical study of kinetic magnetism on frustrated graphs; the headline 'general principle' is extrapolated beyond the finite-size evidence, but the core observations are solid and deserve referee time.","tokens_in":17600,"tokens_out":4208,"would_cite":true,"duration_ms":43254,"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":"Local frustration centers bind singlets while sharing a delocalized hole, so embedding diagonals lowers the ground-state spin by about one per bond.","keywords":["kinetic magnetism","Hubbard model","Nagaoka ferromagnetism","geometric frustration","singlet binding","random graphs","Shastry-Sutherland lattice","cold-atom quantum simulation"],"falsifier":"Compute the ground-state total spin of a much larger frustrated graph, such as a grid with dozens of random diagonal bonds or random graphs with hundreds of sites, by a scalable numerical method: if the average spin reduction per diagonal stops being roughly one, or if the hole's density localizes on the frustration centers, the general principle fails. A cold-atom experiment with two or three engineered diagonal bonds could also falsify it by measuring the total spin or spin correlations and finding a reduction different from one per bond.","tokens_in":16585,"feed_emoji":"🧲","tokens_out":9332,"duration_ms":102599,"temperature":0.7,"pith_summary":"This paper tries to establish a general principle for kinetic magnetism, the magnetism generated by a hole moving through strongly interacting fermions: a local frustration center, such as an odd loop or an added diagonal bond, binds one singlet while the same hole stays delocalized and the remaining spins stay aligned. If the principle is right, the magnetic order of a strongly correlated system can be shaped by editing the graph rather than by tuning exchange couplings. On a rectangular grid, each added diagonal bond lowers the ground-state total spin by roughly one. On random graphs, the total spin is anticorrelated with the number of odd loops and with the frustration index, and it oscillates with the smallest loop size. None of this happens for exchange magnetism at half filling.","feed_headline":"A diagonal bond cuts a magnet's total spin by one","feed_subtitle":"Local frustration centers bind singlets while the hole stays delocalized, giving stepwise control of magnetic order.","key_machinery":"The central object is the one-hole, $U\\to\\infty$ Hubbard model on a graph, whose low-energy physics is governed entirely by the hole's hopping while double occupancy is forbidden. The load-bearing mechanism is the frustration center, meaning a diagonal bond or odd loop that spoils the constructive interference behind the ferromagnetic ground state, binding exactly one singlet while the same hole is shared across the graph. The paper analyzes this with exact diagonalization of the combined hole-and-magnon Hilbert space, builds variational cross-dimer states for collective plaquettes, and defines frustration measures such as the number of odd loops, the frustration index, and the minimum loop size to organize the random-graph results.","core_discovery":"In the infinite-$U$ Hubbard model with a single hole, the paper claims that the ground state on a frustrated graph is organized by the rule that frustration centers bind singlets while sharing a delocalized hole. The hole's kinetic interference, which normally favors a fully aligned ferromagnet on bipartite lattices, is locally spoiled by an odd loop or a diagonal bond; that spoiled region traps a down spin into a singlet, while the hole keeps moving and the rest of the lattice stays spin-polarized. Consequently the ground-state total spin $S_{\\rm total}$ drops by approximately one per added diagonal on a rectangular grid, and two nearby frustrated plaquettes can share the hole and bind distorted singlets collectively. On random graphs the same principle shows up statistically: $S_{\\rm total}$ is anticorrelated with the frustration index and with odd-loop counts, positively correlated with even-loop counts, and its distribution swings as the minimum loop size is increased. The authors present this as a collective effect that exchange magnetism does not possess.","pith_inferences":["Beyond the paper, the additivity of the spin reduction suggests that any set of well-separated frustration centers on a large graph should give a predictable saturation value of $S_{\\rm total}$, turning the rule into a design recipe for synthetic quantum magnets.","Beyond the paper, if the principle survives the thermodynamic limit, standard network diagnostics such as odd-loop density, girth, and frustration index become direct predictors of a quantum ground-state property, connecting network science to strongly correlated electron physics.","Beyond the paper, one could test the hole-sharing claim directly by measuring the hole's inverse participation ratio on a large frustrated lattice: the principle predicts the hole remains delocalized even when all magnons are pinned to frustration centers.","Beyond the paper, multi-hole versions may inherit the same mechanism and produce interactions between singlet centers, since the shared hole mediates the collective binding seen in the two-plaquette calculation."],"forward_implications":["On a rectangular grid with one hole, adding a diagonal bond to a plaquette reduces the ground-state total spin by approximately one, so the net magnetization can be tuned in unit steps by embedding frustration centers.","On random graphs, ground-state total spin is strongly anticorrelated with the frustration index and with the number of odd loops, and positively correlated with even loops; restricting the minimum loop size swings the magnetization distribution between nearly ferromagnetic and nearly singlet.","Two frustrated plaquettes placed close together share the hole and bind distorted singlets collectively, producing regions where the spin reduction exceeds what independent plaquettes would give.","A focused-laser protocol can create effective diagonal hops in selected plaquettes of a cold-atom optical lattice, making the key predictions testable in existing experiments.","Exchange magnetism at half filling shows no such sensitivity to diagonal bonds or loop structure, so this effect is diagnostic of kinetic magnetism."],"supporting_citations":[{"why":"It provides the baseline ferromagnetic ground state on bipartite lattices that the paper's frustrated graphs modify.","marker":"[10]"},{"why":"It establishes that odd loops kinetically frustrate ferromagnetism, the effect the paper generalizes to arbitrary graphs.","marker":"[5]"},{"why":"It gives the exact singlet ground state for one frustrated triangle, the smallest instance of a frustration center binding a singlet.","marker":"[11]"},{"why":"It reports the experimental observation of kinetic ferromagnetism on a square plaquette, the geometry the paper starts from.","marker":"[15]"},{"why":"It explains the effective attraction between hole and magnons from kinetic frustration, which the paper uses to account for singlet binding to plaquettes.","marker":"[39]"},{"why":"It supplies the Shastry-Sutherland construction of diagonal bonds on a square grid, the template for the grid calculations.","marker":"[41]"},{"why":"It is the earlier Shastry-Sutherland-lattice result whose singlet structure the grid calculations agree with.","marker":"[42]"},{"why":"It supplies the exact-ground-state uniqueness condition for nonseparable graphs used in the random-graph statistics.","marker":"[37]"},{"why":"It provides the exchange-magnetism comparison showing that total spin is insensitive to frustration on networks.","marker":"[33]"}],"fun_headline_variants":["Kinetic frustration binds singlets and tunes total spin","Frustration centers cut magnetization stepwise in Hubbard model","Local frustration turns ferromagnetism into singlet binding","Hole plus odd loop: a recipe for controlled spin drops"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The small-system numerical results, covering graphs with 20 sites, a 4-by-5 grid, and ladders up to 32 sites, are assumed to keep holding on arbitrarily large graphs, with every frustration center still binding exactly one singlet and the hole staying delocalized; the paper gives no proof or thermodynamic-limit scaling for that step.","fun_headline_variants_meta":{"raw":{"variants":["Kinetic frustration binds singlets and tunes total spin","Frustration centers cut magnetization stepwise in Hubbard model","Local frustration turns ferromagnetism into singlet binding","Hole plus odd loop: a recipe for controlled spin drops"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000181,"raw_usage":{"total_tokens":1282,"prompt_tokens":896,"completion_tokens":386,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":512,"completion_tokens_details":{"reasoning_tokens":320}},"tokens_in":512,"tokens_out":386,"duration_ms":4843,"temperature":1.0,"reasoning_tokens":320,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T18:30:10.388477+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the ground-state total spin of a much larger frustrated graph, such as a grid with dozens of random diagonal bonds or random graphs with hundreds of sites, by a scalable numerical method: if the average spin reduction per diagonal stops being roughly one, or if the hole's density localizes on the frustration centers, the general principle fails. A cold-atom experiment with two or three engineered diagonal bonds could also falsify it by measuring the total spin or spin correlations and finding a reduction different from one per bond.","supporting_citations":[{"cited_title":"Kim, Exact hole-induced resonating-valence-bond ground state in certainU=∞Hubbard models, Phys","cited_arxiv_id":null,"evidence_quote":"It gives the exact singlet ground state for one frustrated triangle, the smallest instance of a frustration center binding a singlet."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It explains the effective attraction between hole and magnons from kinetic frustration, which the paper uses to account for singlet binding to plaquettes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the Shastry-Sutherland construction of diagonal bonds on a square grid, the template for the grid calculations."},{"cited_title":"Tasaki, Extension of Nagaoka’s theorem on the large- UHubbard model, Phys","cited_arxiv_id":null,"evidence_quote":"It supplies the exact-ground-state uniqueness condition for nonseparable graphs used in the random-graph statistics."},{"cited_title":"Zheng, D","cited_arxiv_id":null,"evidence_quote":"It provides the exchange-magnetism comparison showing that total spin is insensitive to frustration on networks."}],"review_version":1}