{"id":"581d9e88-4d98-4839-8559-9f8472e8a0d8","arxiv_id":"2502.06353","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Quartic bicirculant nut graphs are exactly the B1, B2 and B3 parameter families described in Theorem 1.1 with the stated gcd, parity and congruence conditions; no B4 graph is a nut graph.","lead":"This paper gives a complete set of arithmetic conditions that decide which 4-regular bicirculant graphs are nut graphs, graphs whose zero eigenvalue has multiplicity one and a full eigenvector. It closes a natural classification case in algebraic graph theory and provides a reproducible computer check for the hardest family.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"B1 classification rests on finite cyclotomic check cited from [13] that is not reproduced or independently verified.","rationale":"The reader's weakest assumption is exactly where the paper is least self-contained. I found no error in the main structural reductions: Lemma 2.4 correctly passes from simple zero eigenvalue to full kernel vector for B1-B3; Corollary 2.6 correctly identifies simplicity with a unique k; the reductions to cyclotomic divisibility (Lemma 3.2, Lemma 4.2, Lemma 5.2) are sound. The B2 branch is supported by the Sage code in Appendix A, and the B3 branch is a short linear-congruence argument. For B1, however, the final step of the proof of Theorem 3.1 is an appeal to computational tables in [13] that are not included. Because Lemma 3.6 makes the B1 nut property equivalent to ten finite non-divisibility statements, a single missed residue class in [13] would break the 'if' direction. This is a verification concern, not evidence of incorrectness; the finite check is easy to settle. I also noticed the Lemma 4.5 typo flagged by the reader: in Case 1 the text says '4 ∤ c cannot be true' where the subsequent uniqueness argument needs '4 | c cannot be true' (c odd), and the case is in fact vacuous under condition 2(i); this is patchable and does not affect the final classification. Hence I do not move the verdict: CONDITIONAL remains appropriate, pending the B1 computation or an explicit reproduction.","tokens_in":20281,"tokens_out":27479,"duration_ms":237241,"concrete_test":"Run an independent Sage script that, for each f in {3,5,7,15,21}, enumerates even residues a,b modulo f with gcd(f,a,b)=1 (and, if 5|f, with one of a,b,a-b,a+b divisible by 5) and verifies Phi_f does not divide R_mod_{a,b}; for each f in {6,10,14,30,42}, enumerates even residues a,b modulo f with gcd(f/2,a,b)=1 (and the same condition 1(iv) when 5|f) and verifies Phi_f does not divide Q_mod_{a,b}. Any divisibility found would be a concrete counterexample to Theorem 3.1; if all checks pass, the B1 classification is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central reduction for class B1 is sound up to the last step, but the proof of Theorem 3.1 ends by invoking 'computational results given in [13, Sections A and B]' to assert that, under conditions 1(i)-(iv), no Phi_f divides R_mod for f in {3,5,7,15,21} and no Phi_f divides Q_mod for f in {6,10,14,30,42}. These finite checks are not reproduced here, and the cited [13] concerns cubic tricirculant nut graphs, so the reader cannot tell without going to the earlier paper whether the exact residue tables for these 8-term polynomials were computed and whether the conditions enforced match 1(i)-(iv). Since Lemma 3.6 reduces nut-ness for B1 precisely to these finite non-divisibilities, the 'if' direction of Theorem 3.1 is only as strong as that unstated computation. This is a verification gap, not an observed counterexample; the rest of the B1 argument (Lemmas 3.2-3.6) appears coherent. The B2 branch is better supported because Appendix A gives a runnable Sage enumeration.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper classifies connected quartic bicirculant graphs that are nut graphs. Using the four classes B1-B4 of connected quartic bicirculants, it proves Theorem 1.1: B1(m;a,b) is a nut graph exactly under conditions 1(i)-(iv), B2(m;a,b,c) under conditions 2(i)-(iv), B3(m;a,b) under the stated parity, coprimality and 2-adic conditions, and no B4 graph is a nut graph. The proof reduces the nut-graph condition to a simple zero eigenvalue (Lemma 2.4), applies the known spectral formula for bicirculants (Theorem 2.5), and translates the zero-eigenvalue condition into divisibility of eight-term lacunary polynomials by cyclotomic polynomials, which is then analyzed with the Filaseta-Schinzel theorem and finite residue checks.","tokens_in":20479,"tokens_out":27440,"duration_ms":218082,"significance":"The classification is a natural completion of the programme that classified quartic circulant nut graphs and cubic tricirculant nut graphs; if the B1 branch is verified, it gives a complete parametric description of quartic bicirculant nut graphs. The B2 branch is backed by a reproducible SageMath script in Appendix A, and the B3 branch has a short elementary proof; the spectral reduction to cyclotomic divisibility is clean and yields explicit forbidden congruence conditions. The main weakness is that the B1 sufficiency depends on a finite computation cited from the authors' earlier paper [13] and not reproduced here, so the completeness of the B1 branch is currently a verification statement rather than a demonstrated computation.","major_comments":[{"comment":"The 'if' direction of Theorem 3.1 rests on the sentence 'As indicated by the computational results given in [13, Sections A and B]...' after Lemma 3.6 has reduced the problem to showing that Phi_f does not divide R_mod for f in {3,5,7,15,21} and Phi_f does not divide Q_mod for f in {6,10,14,30,42} under conditions (i)-(iv). These finite non-divisibility checks are load-bearing: without them Lemma 3.6 only gives a reduction, not a classification. Neither the tables nor the code for these checks appear in this manuscript, and [13] is a different classification (cubic tricirculant nut graphs), so a reader cannot verify that the residue computations were performed for exactly the polynomials (2) and (3) with the hypotheses 1(i)-(iv). I recommend either reproducing the finite checks (e.g., a script analogous to Appendix A together with the resulting tables) or stating precisely which lemmas of [13] imply these non-divisibilities.","section":"Section 3, proof of Theorem 3.1 (final paragraph)"},{"comment":"In the proof of Lemma 4.5, Case 1, the claim '4 does not divide c cannot be true, since otherwise condition (i) would not hold' is false: for m=16, a=b=4, c=1, conditions (i) and (ii) of Theorem 4.1 hold and 4 does not divide c. The intended conclusion that a+b has a unique remainder modulo 8 remains true in the odd-c subcase because a+b+2c and a+b are the only even terms and are distinct modulo 8, but the proof as written does not cover that subcase. This proof step is used in the reduction of the B2 classification, so it should be corrected.","section":"Section 4, Lemma 4.5, Case 1"}],"minor_comments":[{"comment":"The complex conjugate of zeta is rendered as 'zeta != zeta' in the proofs of Lemma 3.2 and Lemma 5.2; it should be zeta-bar, the complex conjugate.","section":"Lemma 3.2 and Lemma 5.2"},{"comment":"The sentence 'with even n >= 4' and the condition '1 <= a < b < n' use the symbol n where the parameter m is meant; these should be m to match the rest of the paper.","section":"Section 5, Theorem 5.1 and its proof"},{"comment":"In the script, condition (ii) is coded using v2(f) rather than v2(m). Since f divides m, this filters fewer triples than the theorem's condition, so the script actually verifies a stronger statement; the proof of Theorem 4.1 says 'provided conditions (i) and (ii) hold' without commenting on this difference, which may confuse readers trying to reproduce the check.","section":"Appendix A, SageMath script"},{"comment":"The enumeration counts Cn, Bn, Nn, Vn, Yn, Zn in Table 1 are presented without a description of the enumeration procedure; a short paragraph on how the counts were obtained and whether the classification theorem was used would make the table reproducible.","section":"Section 6, Table 1"}],"recommendation":"major_revision","confidential_remarks":"The B1 branch depends on [13], which is an earlier paper by the same authors; the dependence is not merely a citation of a standard theorem but a finite computation whose output is not displayed. For a journal that values self-contained proofs, I would ask the authors to supply the data or code before acceptance; if they do, the classification is likely sound. The B2 script and B3 proof give me confidence in the overall approach."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThis paper gives the first classification of quartic bicirculant nut graphs, splitting connected quartic bicirculants into classes B1–B4 and stating exactly which parameter sets yield nut graphs. That is a clean and useful completion of the program after quartic circulant and cubic tricirculant nut graphs. The main reduction to non-divisibility of two 8-term polynomials by cyclotomic polynomials is elegant, and the Filaseta–Schinzel step is applied carefully. The B2 branch is particularly good: the finite modular exceptions in Theorem 4.1 are fully supported by a short Sage script in the appendix, and the exceptions are stated explicitly. The B3 branch is short and elementary, and B4 is disposed of by bipartiteness. The overall architecture of the proof looks sound.\n\nThe soft spot is the B1 branch. The sufficiency proof of Theorem 3.1 ends with a claim that computational results in [13, Sections A and B] rule out the required cyclotomic divisibilities. Those tables are not reproduced, and [13] classifies cubic tricirculant nut graphs, a different family. The authors do state that divisibility of these exact polynomials was already investigated in [13, Section 6] and they cite specific lemmas from it, so the reference is not unprincipled. But the finite check is load-bearing and unstated. A referee should ask for the script or explicit tables for B1, as was done for B2. This is a verification gap, not an observed error; I found no mathematical mistake in Lemmas 3.2–3.6.\n\nMinor issue: Lemma 4.5, Case 1, contains what looks like a typo: '4 ∤ c cannot be true' should presumably read '4 | c cannot be true'. The surrounding argument makes the intended meaning clear, but the sentence should be fixed.\n\nOverall: the classification is new and mostly rigorous, with one external computational dependency that should be made self-contained or at least explicit. I would send this to a serious referee; with the B1 checks supplied, I would treat the classification as correct.","headline":"Solid classification with a real but repairable verification gap in the B1 branch.","tokens_in":21052,"tokens_out":5057,"would_cite":true,"duration_ms":42324,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C50","05C25","05C75"],"pacs":[],"model":"deepseek-v4-flash","headline":"A quartic bicirculant graph is a nut graph exactly when it is one of three explicitly described families, and no graph from the fourth family is a nut graph.","keywords":["nut graph","bicirculant graph","quartic graph","graph spectrum","cyclotomic polynomial","lacunary polynomial","Rose Window graph","eigenvalue zero"],"falsifier":"Run an exhaustive search over all connected quartic bicirculants of order $n$ from 52 up to, say, 200, computing the nullity of the adjacency matrix and checking whether a kernel vector has zero entries. Any graph outside the three families of the main theorem with nullity one and a zero-free kernel vector, or any listed graph whose parameters satisfy the conditions but whose nullity is not one, would refute the classification; the paper's own tables only certify orders up to 50.","tokens_in":20106,"feed_emoji":"🧮","tokens_out":11081,"duration_ms":94034,"temperature":0.7,"pith_summary":"Nut graphs are graphs whose zero eigenvalue is simple and whose corresponding eigenvector has no zero entries; they are of interest because the kernel then 'uses' every vertex. The paper asks which quartic bicirculants—connected 4-regular graphs admitting an automorphism with two vertex orbits of equal size—are nut graphs. It proves that the answer is completely finite: up to isomorphism, the nut graphs are exactly the graphs B1(m;a,b), B2(m;a,b,c), and B3(m;a,b) whose parameters satisfy explicit congruence and gcd conditions, and no graph from the fourth class B4 qualifies because those graphs are bipartite. If correct, the classification turns a spectral question into a small number of arithmetic checks and gives a complete structural description of all quartic bicirculant nut graphs.","feed_headline":"Three parameter families cover every quartic bicirculant nut graph","feed_subtitle":"A complete list of modular conditions decides when a 4-regular two-orbit graph has one zero eigenvalue with an all-nonzero eigenvector.","key_machinery":"The machinery is the eigenvalue formula for bicirculants as $\\mathbb{Z}_m$-covers: the eigenvalues are $\\frac{\\lambda^S_k+\\lambda^T_k \\pm \\sqrt{(\\lambda^S_k-\\lambda^T_k)^2+4|\\lambda^R_k|^2}}{2}$, where $\\lambda^S_k,\\lambda^T_k,\\lambda^R_k$ are sums of roots of unity coming from the parameter sets $S,T,R$. From this, zero is an eigenvalue with multiplicity one exactly when $|\\lambda^R_k|^2=\\lambda^S_k\\lambda^T_k$ for exactly one $k$. Substituting the parameters and clearing denominators turns this into a non-divisibility statement for explicit lacunary polynomials ($R_{a,b}$, $Q_{a,b}$, $P_{a,b,c}$, or $(x^{b-a}+1)(x^a+1)(x^b+1)$) by cyclotomic polynomials $\\Phi_f$. Two tools finish the job: a theorem on divisibility of lacunary polynomials by cyclotomic polynomials, which strips large prime factors from $f$, and a case analysis reducing the remaining divisors to short finite lists, so the whole classification ends in a handful of modular checks.","core_discovery":"On the paper's own terms, the central claim is a complete if-and-only-if description. A connected quartic bicirculant graph is a nut graph exactly when, up to isomorphism, it is $B1(m;a,b)$ with $m\\equiv_4 2$, $a$ and $b$ even, $\\gcd(m/2,a,b)=1$, and (if $5\\mid m$) one of $a,b,a-b,a+b$ divisible by $5$; or $B2(m;a,b,c)$ with the four conditions of its classification theorem: coprimality of four gcd expressions, a 2-adic valuation inequality, and forbidden residue triples modulo $12$ and $30$; or $B3(m;a,b)$ with $a,b$ odd, $\\gcd(m,a)=\\gcd(m,b)=1$, and $v_2(b-a)\\ge v_2(m)$. The fourth class, $B4$, contains no nut graphs because its members are bipartite cyclic Haar graphs.","pith_inferences":["The same spectrum-plus-cyclotomic reduction should apply to classify nut graphs among other two-orbit families of fixed degree; the hard part is only carrying the finite divisibility checks through.","The arithmetic conditions yield a fast recognition algorithm, so the enumeration table, which currently certifies only orders up to 50, could be extended to much larger orders, and the open question of graphs belonging to more than one of the classes could be tested computationally.","Because $B2$ gives nut graphs for almost all admissible parameter triples when $m$ is prime, the classification suggests a density analysis of quartic bicirculant nut graphs by order, a count not attempted in the paper."],"forward_implications":["If the theorem holds, the nut-graph question for quartic bicirculants is decidable by arithmetic: checking a graph means checking finitely many gcd and congruence conditions, never computing a spectrum.","Since every $B4$ graph is bipartite and bipartite graphs are never nut graphs, the family of quartic bicirculant nut graphs is exactly the union of the nut graphs in $B1$, $B2$, and $B3$.","The Rose Window graphs, obtained by setting $a=1$ in $B2$, receive a complete nut-graph classification as a special case.","For prime $m$, the classification implies that among connected quartic bicirculants of class $B2$ there is exactly one non-nut graph up to isomorphism, namely $B2(m;1,1,2)$.","The $B1$ remark ties the classification to $I$-graphs: $B1(m;a,b)$ is the product $I(m/2,a/2,b/2)\\square K_2$, so the $B1$ nut graphs correspond exactly to $I$-graphs having $1$ or $-1$ as a simple eigenvalue."],"supporting_citations":[{"why":"Supplies the closed-form eigenvalue formula for bicirculants that reduces the zero-eigenvalue condition to the equality $|\\lambda^R_k|^2=\\lambda^S_k\\lambda^T_k$.","marker":"[22]"},{"why":"Supplies the divisibility lemmas and the external computational tables for the B1 branch that cut the infinite cyclotomic check down to the finite sets of divisors.","marker":"[13]"},{"why":"Supplies the lacunary-polynomial divisibility theorem used repeatedly to strip large prime factors from $f$ and reduce to finite cyclotomic checks.","marker":"[19]"},{"why":"Establishes the four parameter classes B1–B4 of connected quartic bicirculants that the theorem classifies.","marker":"[26]"},{"why":"Identifies B4 as cyclic Haar graphs, from which bipartiteness and hence the absence of nut graphs in B4 follows.","marker":"[25]"},{"why":"Supplies the basic nut-graph facts used throughout, including that bipartite graphs are never nut graphs.","marker":"[30]"},{"why":"Provides the linear-congruence counting result that converts the B3 divisibility condition into the explicit gcd and 2-adic valuation conditions.","marker":"[31]"},{"why":"Provides the computer script in Appendix A whose finite search yields the forbidden residue triples in conditions 2(iii) and 2(iv) for B2.","marker":"[33]"}],"fun_headline_variants":["All quartic bicirculant nut graphs are in three families","Nut graphs among quartic bicirculants: exactly three families","Three families decide nut status for quartic bicirculants","Quartic bicirculant nut graphs: complete classification","All quartic bicirculant nut graphs come from B1, B2, B3"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the finite cyclotomic-divisibility tables used in the B1 branch—cited from the authors' earlier paper and not reproduced here—are correct and exhaustive, so that no missing divisor $f$ escapes the reduction.","fun_headline_variants_meta":{"raw":{"variants":["All quartic bicirculant nut graphs are in three families","Nut graphs among quartic bicirculants: exactly three families","Three families decide nut status for quartic bicirculants","Quartic bicirculant nut graphs: complete classification","All quartic bicirculant nut graphs come from B1, B2, B3"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001679,"raw_usage":{"total_tokens":6591,"prompt_tokens":813,"completion_tokens":5778,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":429,"completion_tokens_details":{"reasoning_tokens":5684}},"tokens_in":429,"tokens_out":5778,"duration_ms":35071,"temperature":1.0,"reasoning_tokens":5684,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T15:45:38.913428+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run an exhaustive search over all connected quartic bicirculants of order $n$ from 52 up to, say, 200, computing the nullity of the adjacency matrix and checking whether a kernel vector has zero entries. Any graph outside the three families of the main theorem with nullity one and a zero-free kernel vector, or any listed graph whose parameters satisfy the conditions but whose nullity is not one, would refute the classification; the paper's own tables only certify orders up to 50.","supporting_citations":[{"cited_title":"Gao and Y","cited_arxiv_id":null,"evidence_quote":"Supplies the closed-form eigenvalue formula for bicirculants that reduces the zero-eigenvalue condition to the equality $|\\lambda^R_k|^2=\\lambda^S_k\\lambda^T_k$."},{"cited_title":"Damnjanovi´ c, N","cited_arxiv_id":null,"evidence_quote":"Supplies the divisibility lemmas and the external computational tables for the B1 branch that cut the infinite cyclotomic check down to the finite sets of divisors."},{"cited_title":"Filaseta and A","cited_arxiv_id":null,"evidence_quote":"Supplies the lacunary-polynomial divisibility theorem used repeatedly to strip large prime factors from $f$ and reduce to finite cyclotomic checks."},{"cited_title":"Kov´ acs, B","cited_arxiv_id":null,"evidence_quote":"Establishes the four parameter classes B1–B4 of connected quartic bicirculants that the theorem classifies."},{"cited_title":"Hladnik, D","cited_arxiv_id":null,"evidence_quote":"Identifies B4 as cyclic Haar graphs, from which bipartiteness and hence the absence of nut graphs in B4 follows."},{"cited_title":"Sciriha and I","cited_arxiv_id":null,"evidence_quote":"Supplies the basic nut-graph facts used throughout, including that bipartite graphs are never nut graphs."},{"cited_title":"Tattersall, Elementary number theory in nine chapters , Cambridge University Press, New York, 1999","cited_arxiv_id":null,"evidence_quote":"Provides the linear-congruence counting result that converts the B3 divisibility condition into the explicit gcd and 2-adic valuation conditions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the computer script in Appendix A whose finite search yields the forbidden residue triples in conditions 2(iii) and 2(iv) for B2."}],"review_version":1}