{"id":"a40f84f1-4c0f-44cd-8f25-dac10f203e26","arxiv_id":"2601.18071","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The eigenvalues of the connection and Dirac matrices of a finite simplicial complex are bounded above by the ordered connection and Dirac degrees, while the headline claim that the connection matrix dominates the Dirac and Green matrices in weak Loewner order remains an open conjecture.","lead":"This note proves eigenvalue bounds for the connection Laplacian and Dirac matrix of a finite simplicial complex, using the classical interlacing theorem to compare a complex with its subcomplexes. It also states an open conjecture that the connection spectrum dominates the Dirac and Green spectra, and reviews the discrete Lefschetz fixed-point theorem and its consequence, a Brouwer fixed-point theorem for simplicial maps.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1 proof relies on false degree inequality d_j(K) ≤ d_{j+1}(G); central λ_j≤d_j bound unproven as written.","rationale":"I read the paper's central claims as Theorems 1 and 3 plus Conjecture 1. The most load-bearing point is Theorem 1, whose proof contains a concrete false inequality, not the open-set interlacing issue the reader emphasized. For any subset U, both L(U) and D(U) are principal submatrices of the full matrices, so Cauchy interlacing applies; the reader's stated failure mode (entries among surviving elements changing under deletion) does not occur. The real gap is that after deleting a maximal simplex, the ordered connection degrees need not shift by one in the way the proof assumes. My explicit graph shows d_1(K) > d_2(G), breaking the induction. I do not claim the theorem itself is false, and I suspect a correct weak-majorization proof may exist; but as written the proof is invalid. The conjecture remains honestly labeled. Because the required fix is a proof repair, not a change of statement, the reader's CONDITIONAL verdict is appropriate: the manuscript should not be accepted until §2.5's induction is corrected or replaced. A small exhaustive search would determine whether the theorem statement itself survives.","tokens_in":16947,"tokens_out":34155,"duration_ms":322108,"concrete_test":"Brute-force enumerate all abstract simplicial complexes on ≤6 vertices (or all graphs on ≤7 vertices); for each, compute the connection matrix L, its ordered eigenvalues, and the ordered connection degrees, and test λ_j ≤ d_j for every j. Also evaluate the 13-simplex example above (vertices A,B,C,L,a1..a5 with edges AB,BC,CL,Aai) and check whether any eigenvalue exceeds the corresponding degree. If a violation exists, Theorem 1 is false; if not, the statement survives but §2.5 still needs a corrected proof.","verdict_should_be":"UNCHANGED","load_bearing_attack":"At §2.5, Theorem 1 is proved by induction: after deleting a maximal simplex x, the smaller complex K has eigenvalues μ_j interlacing with λ_j of G, and the proof uses μ_j ≤ d_j(K) ≤ d_{j+1}(G). The last inequality is asserted without proof and is false for connection degrees. Let G be the graph (1-dimensional complex) with vertices A, B, C, L, a1,...,a5 and edges AB, BC, CL, Aa_i (i=1..5). Connection degrees in G: AB has degree 9; each Aa_i has degree 8; so d_2(G)=8. Delete the maximal edge CL; in K=G\\{CL}, AB still has degree 9 because AB∩CL=∅, hence d_1(K)=9 > 8 = d_2(G). Thus the induction step λ_2 ≤ μ_1 ≤ d_1(K) ≤ d_2(G) collapses. The same unsupported comparison d_j(K) ≤ d_{j+1}(G) is used for every subsequent j. The theorem may still be true, but the manuscript's proof does not establish it. This is distinct from the reader's concern: interlacing via principal submatrices is fine; the broken link is the ordered-degree comparison after deletion.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the connection Laplacian L and the Dirac matrix D of a finite abstract simplicial complex. It claims general eigenvalue upper bounds λ_j ≤ d_j for both L (Theorem 1, §2.5) and D (Theorem 3, §3.3), with the d_j being ordered connection degrees or Dirac graph degrees respectively. It also states an interlacing theorem for subcomplexes and open sets (Theorem 2), conjectures that L weakly Loewner-dominates D and L^{-1} (Conjecture 1, §5.3), and introduces dynamical versions L_T, g_T, d_T for a simplicial map T, claiming that L_T remains unimodular with explicit inverse g_T^* and that the Dirac square is T-independent (§7). The final parts review the discrete Lefschetz fixed point theorem and contain wave-equation applications.","tokens_in":17235,"tokens_out":14669,"duration_ms":148453,"significance":"The paper is written in an expository style, includes reproducible Mathematica code, and explicitly labels its main open problem as a conjecture. If Theorems 1 and 3 were correctly proved, they would give simple universal spectral bounds for two natural matrices attached to any finite simplicial complex. The dynamical extension in §7 would also be a neat observation if it were valid. However, the proof of Theorem 1 relies on a degree comparison that is false, Theorem 3 inherits the same flaw, and Theorem 7 is false as stated for general simplicial maps. These are load-bearing issues: the advertised results are not established as written.","major_comments":[{"comment":"The induction step uses the inequality d_j(K) ≤ d_{j+1}(G) without proof, via lines 'λ_2 ≤ μ_1 ≤ d_1(K) ≤ d_2(G)' and so on. This inequality is false for connection degrees. Example: let G be the 1-complex with vertices A,B,C,L,a1,...,a5 and edges AB, BC, CL, Aa_i (i=1..5). Then d(AB)=9 and each d(Aa_i)=8, so d_1(G)=9, d_2(G)=8. Delete the maximal edge CL; in K=G\\{CL}, d(AB) remains 9 because AB∩CL=∅, so d_1(K)=9>8=d_2(G). The interlace step therefore collapses. The same unsupported comparison is used for every subsequent j, so the proof as written does not establish Theorem 1.","section":"§2.5, proof of Theorem 1"},{"comment":"The proof is said to be 'identical to the connection case', so it inherits the same invalid degree comparison. In the Dirac graph of the example above, the vertex A has degree 6, while many simplices have degree 2, so d_1(G)=6 and d_2(G)=2. After deleting the maximal edge CL, A's Dirac degree is unchanged, so d_1(K)=6 while d_2(G)=2. The induction step μ_1 ≤ d_1(K) ≤ d_2(G) thus fails. A different argument would be needed even if the statement happens to be true.","section":"§3.3, proof of Theorem 3"},{"comment":"Theorem 7 states that L_T(x,y)^{-1} = g_T(x,y)^* for a simplicial map T. The proof reduces to Lemma 8 by asserting L_T = L P_T and g_T = g P_T with P_T a permutation matrix. This reduction is invalid unless T is a bijection on the n simplices and preserves dimension. For a non-injective simplicial map, no permutation matrix represents T. Concrete counterexample: let G={{1},{2}} and let T send both vertices to {1}. Then L_T is the 2×2 matrix [[1,1],[0,0]], which is singular, contradicting the claimed inverse. The theorem is false as stated; at most an automorphism version can be considered.","section":"§7, Theorem 7"}],"minor_comments":[{"comment":"The proof of antisymmetry of the weak Loewner order contains a typo: 'S_2(B)≤S_1(A)' should read 'S_2(B)≤S_2(A)'.","section":"§5.1"},{"comment":"The proof alternates between writing d_j(K) ≤ d_j(G) and later using d_j(K) ≤ d_{j+1}(G); the latter is the inequality actually needed, and it is false. The notation should be made consistent.","section":"§2.5"},{"comment":"There is a formatting/typo error in 'The interlacing story has a parallel situation for the Dirac matrix ELet D_K denote...'.","section":"§3.1"},{"comment":"'Leschetz' should be 'Lefschetz'.","section":"§6.9"},{"comment":"Lemma 9's proof that 'The transformation T just reshuffles the sum' is valid only when T is a permutation of the basis; for general simplicial maps additional justification is needed.","section":"§7.3"},{"comment":"The code testing L_T^{-1}=g_T^* uses FindAut, so it only exercises automorphisms and cannot detect the failure of Theorem 7 for general simplicial maps.","section":"§8.3"}],"recommendation":"reject","confidential_remarks":"The central new claims are not established: the eigenvalue-bound proofs fail on a specific false inequality, and the dynamical unimodularity theorem is contradicted by a simple constant-map example. The manuscript reads as working notes; substantial restructuring would be required to either repair these results or restrict the claims to automorphisms. The paper also relies heavily on the author's prior framework for the Green-Star formula and unimodularity, which is acceptable but should be stated more carefully."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take on Knill's remarks paper. The genuinely new content is small: the ordered-eigenvalue bounds for connection and Dirac matrices (Theorems 1 and 3) and the weak-Loewner conjecture. The conjecture is honestly labeled as open, with one figure as evidence. Theorems 7–9 in the dynamical section are true but trivial; the author himself says \"the proof is easy,\" and they all follow from applying a permutation matrix to the right of L or D. The Lefschetz half is a review of his own theorem.\n\nThe problem is that the main proof in the paper, Theorem 1, has a false step. The induction removes a maximal simplex x to get a subcomplex K, and then uses d_j(K) ≤ d_{j+1}(G) to push the interlacing bounds down. That inequality is not true for connection degrees. The stress-test counterexample works: take the graph with vertices A,B,C,L,a1..a5, edges AB, BC, CL, Aa_i. The connection degree of AB is 9, each Aa_i is 8, so d_2(G)=8. Remove the edge CL (which is maximal and disjoint from AB); in K, AB still has degree 9, so d_1(K)=9 > 8 = d_2(G). Thus the step λ_2 ≤ μ_1 ≤ d_1(K) ≤ d_2(G) collapses. Theorem 3 uses the same induction, so it inherits the same gap. The theorems might be salvageable with a different argument—for instance, deleting a maximum-degree maximal simplex one at a time, and noting d_1(K) ≤ d_2(G) in that case—but the manuscript as written does not prove them.\n\nThere are other soft spots. The proof of Theorem 2's open-set case is one sentence; the reader says it works, but it deserves more detail. The paper repeatedly cites \"Schur's theorem\" without a reference, and the Barmak-McCord remark is missing a citation. The abstract presents the trivial dynamical facts as results, which overstates the contribution.\n\nOn balance: the paper deserves a serious referee because the statements are new and plausibly correct, and the proof gap is the kind a referee could catch. But as submitted, Theorem 1 and Theorem 3 are unproven. If the author fixes the induction, the paper becomes a modest but useful contribution; if not, it's a remarks note with an open conjecture and some examples. I'd conditionally recommend revision, with the main requirement being a repaired proof of the interlacing bounds.","headline":"The new interlacing bounds are plausible but unproven: the core induction in Theorem 1 rests on a false degree inequality, and Theorem 3 inherits the same gap.","tokens_in":17814,"tokens_out":8003,"would_cite":false,"duration_ms":78328,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05E45","05C50","15A18","55M20"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that for any finite abstract simplicial complex, the eigenvalues of the connection Laplacian and of the Dirac matrix are bounded above by the corresponding ordered degrees, and conjectures a weak Loewner dominance of the co","keywords":["simplicial complex","connection Laplacian","Dirac matrix","eigenvalue interlacing","weak Loewner order","Lefschetz fixed point theorem","dynamical matrices","unimodular matrix"],"falsifier":"Search all finite abstract simplicial complexes with, say, at most five vertices: compute the ordered spectra of L and D and the ordered degrees d_j, and check λ_j ≤ d_j for every complex; also compute S_k(L) and S_k(D) for every open set and check S_k(L) ≥ S_k(D). A single complex violating either inequality would refute Theorem 1, Theorem 3, or Conjecture 1.","tokens_in":16733,"feed_emoji":"📐","tokens_out":5212,"duration_ms":46777,"temperature":0.7,"pith_summary":"This paper tries to establish that two square matrices attached to any finite abstract simplicial complex — the connection Laplacian L and the Dirac matrix D — have eigenvalues controlled by simple degree counts. By removing maximal simplices one at a time, the smaller complex's matrix sits as a principal submatrix, so Cauchy interlacing applies; this yields the bounds λ_j ≤ d_j for both matrices, where d_j are ordered connection degrees for L and ordered Dirac-graph degrees for D. The same interlacing is claimed for open subsets in the Dirac case but not for L. The paper further conjectures that L dominates both D and its inverse g = L^{-1} in the weak Loewner order (spectral sums), and presents numerical evidence. The second half generalizes L and D to dynamical versions under a simplicial map T, proving the dynamical connection matrix is still unimodular with explicit Green inverse, and reviews a heat-flow proof of the Lefschetz fixed point theorem. A sympathetic reader cares because these are elementary, purely combinatorial spectral bounds that extend graph-theoretic interlacing to all dimensions of a simplicial complex.","feed_headline":"Simplicial spectra bounded by connection and Dirac degrees","feed_subtitle":"For every finite simplicial complex, connection and Dirac eigenvalues satisfy λ_j ≤ d_j; a stronger dominance conjecture remains open.","key_machinery":"The load-bearing tool is the Cauchy interlacing theorem applied to principal submatrices obtained by removing maximal simplices one at a time. The connection degree d(x) = |{y : C(x) ∩ C(y) ≠ ∅}| and the Dirac graph degree d_x = Σ_y |D_{xy}| provide the upper bounds in the row-sum estimate. The weak Loewner order, defined by spectral sums S_k(A) = λ_1 + ... + λ_k, is the comparison notion for the conjecture. A second device is the permutation-matrix lemma used for dynamical matrices: L_T = L P_T and g_T = g P_T, which immediately yields L_T^{-1} = g_T^* and independence of D_T^2 from T.","core_discovery":"The paper establishes Theorem 1 and Theorem 3: for any finite abstract simplicial complex G with n simplices, the ordered eigenvalues λ_j of the connection matrix L satisfy λ_j ≤ d_j, where d_j are the ordered connection degrees (the number of simplices whose cones intersect), and the ordered eigenvalues of the Dirac matrix D satisfy λ_j ≤ d_j with d_j the ordered Dirac graph degrees. The proof uses induction by deleting maximal simplices, which leaves the smaller matrix as a principal submatrix, and then applies Cauchy interlacing together with a Perron-Frobenius row-sum bound for the spectral radius. The same interlacing is claimed for open subsets in the Dirac case. The paper also proves","pith_inferences":["The open-set interlacing claimed for the Dirac matrix, if true, would yield a Morse-theoretic inequality relating the spectra of D on a complex to its open covers; a natural test is to check all open sets on small complexes.","The weak Loewner conjecture, though only stated for L vs D and L vs g, would imply a full majorization chain among L, D, and g that could be checked numerically by random search; a single counterexample would refute it.","The dynamical zeta function defined from iterates of T could be connected to the unimodular dynamical Green matrices, potentially making the rationality of the zeta function a corollary of the permutation-matrix identities.","The same interlacing argument may extend to higher-characteristic connection matrices w_k, though the paper notes those are not unimodular in general, suggesting the bounds might hold but the inverse story would differ."],"forward_implications":["For any finite simplicial complex, the spectral radius of the connection Laplacian is at most the maximal connection degree, and the spectral radius of the Dirac matrix is at most the maximal Dirac graph degree.","Adding a maximal simplex to a complex can only push the connection and Dirac eigenvalues upward in the interlacing sense, giving monotonicity bounds for subcomplexes.","The dynamical unimodularity result implies that the total dynamical energy Σ_{x,y} g_T(x,y) equals the Euler characteristic χ(G), independent of the simplicial map T.","If Conjecture 1 holds, the connection matrix would dominate both its inverse and the Dirac matrix in spectral-sum order, which would imply spectral-radius comparisons and strengthen the hydrogen-identity bounds in the 1-dimensional case.","The heat-flow proof of the Lefschetz fixed point theorem gives a purely combinatorial route to Brouwer's fixed point theorem for contractible complexes."],"fun_headline_variants":["Eigenvalues of connection and Dirac matrices bounded by degrees","Simplicial spectra bounded by degrees: new proof","Interlacing gives degree bounds for connection and Dirac","Connection and Dirac eigenvalue bounds proven via interlacing"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The interlacing bounds for L require that deleting a maximal simplex leaves the smaller connection matrix as a principal submatrix, which holds for subcomplexes but fails for open sets; the Dirac version extends interlacing to open sets on a one-sentence argument that deleting a maximal simplex changes only entries involving that simplex. If any deletion alters entries among the surviving simplices, both eigenvalue bounds collapse.","fun_headline_variants_meta":{"raw":{"variants":["Eigenvalues of connection and Dirac matrices bounded by degrees","Simplicial spectra bounded by degrees: new proof","Interlacing gives degree bounds for connection and Dirac","Connection and Dirac eigenvalue bounds proven via interlacing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000718,"raw_usage":{"total_tokens":3042,"prompt_tokens":706,"completion_tokens":2336,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":450,"completion_tokens_details":{"reasoning_tokens":2285}},"tokens_in":450,"tokens_out":2336,"duration_ms":18068,"temperature":1.0,"reasoning_tokens":2285,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T08:08:55.456452+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Search all finite abstract simplicial complexes with, say, at most five vertices: compute the ordered spectra of L and D and the ordered degrees d_j, and check λ_j ≤ d_j for every complex; also compute S_k(L) and S_k(D) for every open set and check S_k(L) ≥ S_k(D). A single complex violating either inequality would refute Theorem 1, Theorem 3, or Conjecture 1.","supporting_citations":[],"review_version":1}