{"id":"be94e567-9689-4b68-a86c-b081ea6cadea","arxiv_id":"1908.02445","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For unitary Cayley graphs X_n, the author constructs n with arbitrarily many prime factors satisfying γ_t(X_n) ≤ g(n) - 16, and classifies all products of complete graphs with domination number t+2.","lead":"This paper studies domination in unitary Cayley graphs, showing that the total domination number can be much smaller than the Jacobsthal function: there exist integers n with arbitrarily many prime factors satisfying γ_t(X_n) ≤ g(n) - 16. It also completes the classification of direct products of complete graphs with domination number t+2.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.9's CRT condition h ≡ -z_i (mod r_i) is wrong; the constructed interval is not guaranteed to avoid integers coprime to n, so the amplification step needed for 'arbitrarily many prime factors' in Theorem 1.10 is unproved.","rationale":"The reader's weakest_assumption focused on Ziller's external computation H(41)=566, h(41)=550. That is a legitimate concern, but the more direct and load-bearing problem is internal: Theorem 1.9's proof contains a concrete CRT error that breaks the amplification argument. Without Theorem 1.9, Lemma 4.4 yields only a single integer Q with γ_t(X_Q) ≤ g(Q)−16, not the claimed family with arbitrarily many prime factors. The error appears to be a typo (the condition should be h ≡ x−z_i (mod r_i) rather than h ≡ −z_i (mod r_i)), and the theorem may be repairable, which is why the verdict remains CONDITIONAL rather than REJECT. The paper as submitted, however, has an unproved central claim, and the required correction should be stated explicitly in the revision. The Theorem 1.4 proof gap noted by the reader is also real but affects secondary lower-bound results, not the headline theorem directly. Therefore I recommend keeping the conditional verdict with an additional mandatory fix to Theorem 1.9.","tokens_in":14487,"tokens_out":16411,"duration_ms":157899,"concrete_test":"Run a small numerical check of the CRT construction from Theorem 1.9. Take s=6, k=1, g(6)=4, x=2, so S={2,...,10}, R={5,7}; choose r_1=11, r_2=13. Solve the paper's conditions h≡2 (mod 6), h≡−5≡6 (mod 11), h≡−7≡6 (mod 13), giving h=578. The interval 578..586 contains 581 = h+(5−2); gcd(581,6)=1, 581 mod 11 = 9, and 581 mod 13 = 9, so 581 is coprime to n=6·11·13. This directly contradicts the claim that the interval has no integer coprime to n. Recompute with the corrected condition h≡x−z_i (mod r_i), i.e. h≡8 (mod 11) and h≡8 (mod 13), and verify the interval avoids units, confirming that the proof needs the correction.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.10 promises n with arbitrarily many prime factors satisfying γ_t(X_n) ≤ g(n) − 16. This relies on Theorem 1.9, which amplifies a single seed s ∈ M_j to infinitely many prime factors. In the proof of Theorem 1.9 (Section 4), after choosing S = {x, ..., x+ks+g(s)−2} and R = {z_1, ..., z_ℓ} (the elements of S coprime to s), the author sets h ≡ x (mod s) and h ≡ −z_i (mod r_i) for each i, and claims the interval {h, ..., h+ks+g(s)−2} contains no integer coprime to n. This is false. For j = z_i − x, the integer h+j satisfies h+j ≡ z_i (mod s), so it is coprime to s. Modulo the associated prime r_i, h+j ≡ −z_i + (z_i − x) ≡ −x (mod r_i). Since 1 ≤ x ≤ s and r_i > ks+g(s) ≥ s for k ≥ 1, −x is not 0 modulo r_i, so h+j is not divisible by r_i. No argument shows it is divisible by any other r_l, and in general it is not. Thus the interval can contain integers coprime to n, so the claimed lower bound g(n) ≥ ks+g(s) is unsupported. The corrected condition would be h ≡ x − z_i (mod r_i), but as written the proof of the amplification theorem is invalid. Since the 'arbitrarily many prime factors' part of Theorem 1.10 depends entirely on Theorem 1.9, the central claim is not established even if Ziller's computational values H(41)=566 and h(41)=550 are correct.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies domination and total domination numbers of direct products of complete graphs, with applications to unitary Cayley graphs X_n of Z/nZ. Its main results are: (i) a lower bound for γ(∏_{i=1}^t K_{n_i}) in Theorem 1.4, leading to asymptotic lower bounds for γ(X_n) for squarefree n; (ii) a complete classification of products with γ(G)=t+2 in Theorem 1.5, answering a question of Defant and Iyer; and (iii) an amplification theorem, Theorem 1.9, stating that a single example with γ_t(X_s)≤g(s)−j yields examples with arbitrarily many prime factors satisfying the same gap, which combined with Ziller's computation H(41)=566, h(41)=550 gives n with γ_t(X_n)≤g(n)−16 in Theorem 1.10. The classification proof appears self-contained and convincing. The lower-bound proof and the amplification proof each contain a load-bearing gap, as detailed below.","tokens_in":14824,"tokens_out":16925,"duration_ms":179259,"significance":"If the proofs are repaired, the paper would make substantial progress: Theorem 1.5 would completely resolve the Defant–Iyer classification question, and Theorem 1.10 would answer Burcroff's question affirmatively with a much larger gap than the previously known value of 2. The monotonicity lemma and the reduction of the problem to Jacobsthal-function computations are elegant, and the paper is careful to identify the external computational input from Ziller rather than hiding it. The claimed asymptotic lower bounds for γ(X_n) would also be a genuine strengthening over the results of Mekiš and of Defant and Iyer. However, as written, the two central proof gaps mean that the main new claims are not yet established.","major_comments":[{"comment":"The CRT step that is supposed to produce a run of length ks+g(s)−1 with no integer coprime to n has the wrong congruence. The author sets h ≡ −z_i (mod r_i) for each z_i ∈ R. For j = z_i − x, the integer h+j satisfies h+j ≡ −x (mod r_i), which is nonzero because r_i > ks+g(s) ≥ s ≥ x; it is also coprime to s because h+j ≡ z_i (mod s). The proof gives no reason why h+j is divisible by any other r_l, and in general it is not. Thus the claimed lower bound g(n) ≥ ks+g(s) is unsupported. Replacing the congruence by h ≡ x−z_i (mod r_i) would make h+j divisible by r_i for every later-block element, so this appears to be a local fix; as written, however, Theorem 1.9 and consequently the 'arbitrarily many prime factors' conclusion of Theorem 1.10 are not established.","section":"Section 4, proof of Theorem 1.9"},{"comment":"The map f : S_{t−1}(k) → F_{v1}(1) is not well-defined. The vertex x_σ is formed by mixing coordinates of the selected vertices d_1,...,d_{t−1}, and while its first coordinate is v_1, nothing in the construction places x_σ in the dominating set D. Since F_{v1}(1) is defined as the set of vertices of D with first coordinate v_1, the cardinality bound |S_{t−1}(k)|/|F_{v1}(1)| > 4 does not imply the existence of four distinct σ with x_σ equal. This invalidates the subsequent use of Lemma 2.1 on the sets {d_{σ_j(q)}} and hence the lower bound in Theorem 1.4, together with Corollaries 2.2 and 2.4.","section":"Section 2, proof of Theorem 1.4"}],"minor_comments":[{"comment":"Near the end of the proof, 'we claim that n = s · ∏ r_i is in M16' should read 'is in M_j', since the argument is for a general j and the set M_16 is only one instance.","section":"Section 4, proof of Theorem 1.9"},{"comment":"The line 'there must be two distinct vertices x5, x6 ∈ K_t such that |F_{x5}(3)| = |F_{x5}(3)| = 2' should have |F_{x6}(3)| = 2 on the right-hand side.","section":"Section 3, proof of Theorem 3.4"},{"comment":"The notation 'log2e(t)' in the statement and proof of Theorem 1.4 is ambiguous and should be defined explicitly.","section":"Section 2, Theorem 1.4"},{"comment":"In the application of Lemma 2.1, the sets E_j are initially formed by subtracting previous F's so that they are disjoint, but the notation subsequently refers to E_j = F_{v_j}(j); the presentation should make the disjoint versions explicit throughout.","section":"Section 2, proof of Theorem 1.4"}],"recommendation":"major_revision","confidential_remarks":"The two proof gaps are significant but different in nature: the CRT error in Section 4 is a local sign error that appears repairable, whereas the Section 2 gap in Theorem 1.4 may require a substantially different argument. Because the classification theorem appears correct and the amplification framework is promising, I would not reject the paper outright, but I would ask the authors to supply complete proofs for both claims before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper has one solid half and two broken halves. The classification of γ(G)=t+2 (Theorem 1.5) looks correct, and the proof of Theorem 3.4 is a genuine piece of casework. Lemma 4.4 also works: combining Ziller's computation with the monotonicity lemma gives a single n with γ_t(X_n) ≤ g(n) − 16, which answers Burcroff's question.\n\nBut the other main results have load-bearing gaps. In Theorem 1.4, the map f is defined from permutations into F_{v1}(1), but x_σ is never shown to be in D. It is just a vertex whose first coordinate is v1. The pigeonhole argument that needs |F_{v1}(1)| is therefore invalid, and the asymptotic lower bounds in Corollaries 2.2 and 2.4 fall with it.\n\nThe stress-test note is right about Theorem 1.9. The CRT condition h ≡ −z_i (mod r_i) makes h+j ≡ −x (mod r_i) when j = z_i − x, not 0. Since r_i > x, the constructed interval can contain integers coprime to n. The fix would be h ≡ x − z_i (mod r_i), but as written the amplification step is unproved. Theorem 1.10's claim of arbitrarily many prime factors therefore does not follow. This is not a minor typo; it is the core of the construction.\n\nTwo smaller points: Theorem 1.9's proof says \"n is in M16\" where it should say M_j, and the paper leans on Ziller's computational values without independent verification. That reliance is acceptable if stated clearly, but it does make the single-n example conditional on an external computation.\n\nNet assessment: the t+2 classification is a solid contribution and alone justifies referee attention. The single-n gap-16 example is also worth reporting. But the marquee result—arbitrarily many prime factors with gap 16—is not established, and the new lower bounds are not established. The flaws are likely repairable, but they are real. I would send this to peer review and tell the referee to focus on the two broken proofs. If those cannot be fixed, the paper should be cut down to the classification plus the single-n observation.","headline":"The classification and the single-n gap-16 example are real; the asymptotic claims and the amplification theorem are not currently proven.","tokens_in":15381,"tokens_out":3793,"would_cite":false,"duration_ms":41185,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C69","05C76"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that for infinitely many $n$ with arbitrarily many prime factors, the total domination number of the unitary Cayley graph $X_n$ is at least 16 less than Jacobsthal's function $g(n)$.","keywords":["domination number","total domination number","direct product of complete graphs","unitary Cayley graph","Jacobsthal's function","prime factors","asymptotic bounds","graph classification"],"falsifier":"Independently recompute $H(41)$ and $h(41)$: find whether some squarefree $k$ with 41 distinct prime factors satisfies $g(k)>566$, or whether $g(p_1\\cdots p_{41})\\ne 550$; either mismatch would remove the input to Lemma 4.4 and invalidate the proof of Theorem 1.10 as presented.","tokens_in":14252,"feed_emoji":"🔢","tokens_out":9479,"duration_ms":88028,"temperature":0.7,"pith_summary":"The paper studies the total domination number $\\gamma_t(X_n)$ of the unitary Cayley graph of $\\mathbb{Z}/n\\mathbb{Z}$ and compares it with Jacobsthal's function $g(n)$, the smallest $m$ such that every block of $m$ consecutive integers contains a number coprime to $n$. It establishes that the gap between the two can be as large as 16: there exist integers $n$ with arbitrarily many prime factors for which $\\gamma_t(X_n) \\le g(n)-16$, answering the question of [4] affirmatively. The proof rests on a monotonicity comparison that turns the domination problem into a comparison of two values of Jacobsthal's function, plus an amplification step that makes the number of prime factors arbitrarily large. The paper also proves new lower bounds on domination numbers of direct products of complete graphs and completely classifies the graphs $G=\\prod_{i=1}^t K_{n_i}$ with $\\gamma(G)=t+2$.","feed_headline":"Total domination beats Jacobsthal's function by 16","feed_subtitle":"For n with arbitrarily many prime factors, unitary Cayley graph domination trails Jacobsthal's function by 16.","key_machinery":"The central machinery is the comparison lemma (Lemma 4.3): if $n_i \\le m_i$ for every $i$, then $\\gamma_t(\\prod K_{n_i}) \\ge \\gamma_t(\\prod K_{m_i})$. This lets the paper replace the graph parameter by arithmetic data, because $X_n$ is, by the Chinese remainder theorem, a direct product of complete graphs indexed by the prime-power factors of $n$. A second load-bearing device is the amplification theorem (Theorem 1.9), which takes any $s$ witnessing a gap of size $j$ and constructs a new squarefree $n = s \\prod r_i$, with the $r_i$ chosen larger than $ks+g(s)$, whose total dominating set has size at most $ks + \\gamma_t(X_s)$ while $g(n) \\ge ks + g(s)$; this preserves the gap and adds arbitrarily many prime factors. The numerical input is the pair $H(41)=566$, $h(41)=550$, where $H(k)$ is the maximum of $g$ over integers with $k$ distinct prime factors and $h(k)$ is $g$ evaluated on the product of the first $k$ primes.","core_discovery":"The central discovery is that the inequality $\\gamma_t(X_n) \\le g(n)$ is not just sometimes strict but can be strict by 16, and that strictness is an amplification phenomenon: a single witness $s$ with $\\gamma_t(X_s) \\le g(s)-j$ generates witnesses $n$ with arbitrarily many prime factors and the same gap. The reduction is provided by the isomorphism $X_n \\cong \\prod K_{q_i^{\\alpha_i}}$ and by the monotonicity lemma that coordinatewise larger complete graphs have no larger total domination number. Feeding in the computed values $H(41)=566$ and $h(41)=550$ gives a concrete squarefree $Q$ with $\\gamma_t(X_Q) \\le g(Q)-16$; the amplification theorem then spreads this gap to arbitrarily many prime factors. Separately, the paper resolves the classification question for $\\gamma(G)=t+2$ by showing it occurs exactly in three listed configurations.","pith_inferences":["If $H(k)-h(k)$ is unbounded as $k$ grows, then the paper's Conjecture 1.11 follows immediately; thus the computational study of Jacobsthal's function is the natural test bed for the full conjecture.","The monotonicity lemma suggests a broader principle: for squarefree $n$, the total domination number of $X_n$ is controlled by the minimal Jacobsthal value among integers with the same number of prime factors, so any improvement in the comparison between $H$ and $h$ translates directly into stronger domination gaps.","The explicit dominating-set constructions in the classification theorem are concrete enough to be checked by computer search on small $t$ and $n_i$; an exception would indicate an error in the proof, while agreement would provide independent verification of that theorem."],"forward_implications":["For any gap value $j$ for which a single witness exists, the family of $n$ with $\\gamma_t(X_n) \\le g(n)-j$ contains integers with arbitrarily many prime factors; in particular this is now known for $j=16$.","For squarefree $n=q_1\\cdots q_t$, the domination number satisfies $\\gamma(X_n) \\ge (1-\\epsilon)t \\prod_{i=1}^t \\frac{q_i}{q_i-1}$ asymptotically, improving the earlier lower bound of [5].","For primorial $n=p_1\\cdots p_t$, the estimates become $C_1 t\\log t \\le \\gamma(X_n) \\le C_2 t\\log^2 t$ for absolute constants $C_1,C_2>0$.","The paper completely determines when $\\gamma(\\prod_{i=1}^t K_{n_i}) = t+2$, settling the classification problem raised in [5]."],"supporting_citations":[{"why":"Supplies the base lower bound $\\gamma(G)\\ge t+1$ for direct products of complete graphs that Theorem 1.4 extends.","marker":"[20]"},{"why":"Provides the key lemma on coordinate-agreeing subsets, the earlier domination bounds, and the classification problem that the paper resolves.","marker":"[5]"},{"why":"Poses the question answered affirmatively here and gives the previous best construction with gap 2.","marker":"[4]"},{"why":"Supplies the computed values $H(41)=566$ and $h(41)=550$ that seed Lemma 4.4.","marker":"[22]"},{"why":"Provides the earlier counterexample to Jacobsthal's conjecture, $H(24)>h(24)$, which the same method converts into a new gap-2 family.","marker":"[12]"},{"why":"Establishes the basic inequality $\\gamma_t(X_n)\\le g(n)$ that motivates the entire comparison.","marker":"[18]"}],"fun_headline_variants":["Total domination beats Jacobsthal by 16","Cayley graph gap of 16 found","Domination number improves by 16","New bound: gamma_t achieves g(n)-16","Dominating Cayley graphs by 16 less"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 16-gap construction rests entirely on the external computation, attributed to [22], that the largest value of Jacobsthal's function among integers with 41 distinct prime factors is 566 while its value on the product of the first 41 primes is 550; if that computation is wrong, the seed integer $Q$ need not exist.","fun_headline_variants_meta":{"raw":{"variants":["Total domination beats Jacobsthal by 16","Cayley graph gap of 16 found","Domination number improves by 16","New bound: gamma_t achieves g(n)-16","Dominating Cayley graphs by 16 less"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000147,"raw_usage":{"total_tokens":1191,"prompt_tokens":958,"completion_tokens":233,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":574,"completion_tokens_details":{"reasoning_tokens":163}},"tokens_in":574,"tokens_out":233,"duration_ms":3115,"temperature":1.0,"reasoning_tokens":163,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:44:52.024812+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Independently recompute $H(41)$ and $h(41)$: find whether some squarefree $k$ with 41 distinct prime factors satisfies $g(k)>566$, or whether $g(p_1\\cdots p_{41})\\ne 550$; either mismatch would remove the input to Lemma 4.4 and invalidate the proof of Theorem 1.10 as presented.","supporting_citations":[{"cited_title":"Meki ˘s, Lower bounds for the domination number and total domination number of direct product graphs, Discrete Mathematics 310 (2010), 3310-3317","cited_arxiv_id":null,"evidence_quote":"Supplies the base lower bound $\\gamma(G)\\ge t+1$ for direct products of complete graphs that Theorem 1.4 extends."},{"cited_title":"Defant and S","cited_arxiv_id":null,"evidence_quote":"Provides the key lemma on coordinate-agreeing subsets, the earlier domination bounds, and the classification problem that the paper resolves."},{"cited_title":"Domination Parameters of the Unitary Cayley Graph of $\\mathbb{Z}/n\\mathbb{Z}$","cited_arxiv_id":"1809.04769","evidence_quote":"Poses the question answered affirmatively here and gives the previous best construction with gap 2."},{"cited_title":"New computational results on a conjecture of Jacobsthal","cited_arxiv_id":"1903.11973","evidence_quote":"Supplies the computed values $H(41)=566$ and $h(41)=550$ that seed Lemma 4.4."},{"cited_title":"Hajdu and N","cited_arxiv_id":null,"evidence_quote":"Provides the earlier counterexample to Jacobsthal's conjecture, $H(24)>h(24)$, which the same method converts into a new gap-2 family."},{"cited_title":"Maheswari and M","cited_arxiv_id":null,"evidence_quote":"Establishes the basic inequality $\\gamma_t(X_n)\\le g(n)$ that motivates the entire comparison."}],"review_version":1}