{"id":"9bb665de-0f23-4399-834d-8720ef66b88f","arxiv_id":"2507.19528","paper_version":1,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"The paper claims an asymptotic formula for the integral of Delta(x)^8 from 2 to X, with error O(X^{3-1/254+epsilon}) and constants defined by two divisor-function sums.","lead":"This paper studies the eighth power of the error term in the classical Dirichlet divisor problem and claims a new asymptotic formula with an error of order X^{3-1/254+epsilon}. If correct, it provides explicit constants for the average size of the eighth power of the divisor error term over long and short intervals.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 13, the key counting estimate behind S5/S6, is unproved: Lemma 11's proof uses an inverted inequality and Lemma 13's Fourier cutoff omits a δ^{-1}N_j^4 term, so the H^{3-1/254} error in Lemma 16 lacks support.","rationale":"The reader correctly identified Lemma 13 as the weak point, but focused on the N_j^{1/4} versus N_j^{1/2} mismatch. My reading finds a more basic problem: the proof of Lemma 13 depends on Lemma 11, and Lemma 11's proof contains a plainly inverted inequality. For k=2, nonzero fourth differences of √n can be of size n^{−5/2}, so the reciprocal is much larger than N^{−1/2}; the asserted bound on the reciprocal is false. Moreover, the Fourier splitting in Lemma 13 uses a cutoff bound that is too optimistic, omitting a δ^{−1}N^4 term. Since Lemma 16 relies exactly on Lemma 13 to bound S5 and S6, and the final error H^{3−1/254} is obtained from those bounds, Theorem 1 is not proved by the text as written. I do not claim the theorem is false; the constants and overall structure are plausible, and a repaired Lemma 13 might rescue the argument. Hence I would keep the reader's CONDITIONAL verdict, not moving to ACCEPT or REJECT, but the condition is stronger than the reader stated: a correct proof of Lemma 13 (and of the Lemma 11 bound it uses) is required before the paper can be accepted.","tokens_in":16521,"tokens_out":22331,"duration_ms":210517,"concrete_test":"Verify Lemma 11's asserted inequality by computing ∆ for k=2, n=10^6, h1=h2=h3=1; if 1/|∆|≫N^{5/2} as expected, the proof of Lemma 11 is invalid, and then redo Lemma 13's Fourier split with |φ(x)|≤min(δ,1/(π|x|)) to check whether the omitted δ^{−1}N_j^4 term can be absorbed into the final H^{3−1/254} bound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 13 is the sole estimate controlling the off-diagonal terms S5 and S6 in Lemma 16, but its proof is not valid as written. First, the proof of Lemma 11 asserts '1/|∆| ≪ N^{−1/k+ε}' in bounding the eighth-moment integral. For k=2, ∆ is the fourth difference of √n, and for h1=h2=h3≈1 one has |∆| ≈ n^{−5/2}; the correct upper bound on the reciprocal is ≫N^{5/2−ε}, not ≤N^{−1/2+ε}. The resulting N^{8−1/k} term in Lemma 11 therefore has no derivation. Second, in Lemma 13's Fourier argument, |φ(x)|≤δ is used on [0,p] with p=l/δ, but the actual bound is min(δ,1/(π|x|)); on [1/δ,p] the dyadic summation yields an additional δ^{−1}N_j^4 l term that the displayed estimate omits. Lemma 16 then applies Lemma 13 with N_j^{1/2} (not the stated N_j^{1/4}), so the claimed H^{3−1/254} error for S5/S6 depends on an unproved bound. Theorem 1 is therefore not established; the defect is repairable only if a correct version of Lemma 13 is supplied.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper claims an asymptotic formula for the eighth power moment of the Dirichlet divisor error term: ∫_2^X Δ(x)^8 dx = ((35C7 − 28C4)/(2048π^8)) ∫_2^X x^2 dx + O(X^{3−1/254+ε}), with C4 and C7 defined by explicit eight-variable Diophantine sums. The proof expands the eighth power of Voronoi's truncated formula, identifies diagonal contributions, and bounds off-diagonal oscillatory integrals through a sequence of counting lemmas; a dyadic decomposition yields the stated power saving. A second theorem states a short-interval analogue for H ≥ X^{7/32+δ}.","tokens_in":16763,"tokens_out":15495,"duration_ms":155356,"significance":"The main term is genuinely explicit and is not fitted to data; this is a strength. If the proof were complete, the paper would be the first asymptotic formula for the eighth moment of Δ(x), with a power-saving error, extending the line of work of Tsang, Zhai, and Ivić–Sargos. The claimed error exponent 3−1/254 is also falsifiable and would be a strong test of the conjectured bound Δ(x)≪x^{1/4+ε}. However, the argument currently rests on unverified counting estimates at its core, so the significance can only be assessed conditional on repairing those estimates.","major_comments":[{"comment":"The statement of Lemma 13 is internally inconsistent. The displayed claim has N_j^{1/4}, while the proof concludes with N_j^{1/2}; Lemma 16 in the next section applies the N_j^{1/2} version. This matters because Lemma 13 is the sole estimate behind the off-diagonal terms S5 and S6, which are responsible for the O(H^{3−1/254+ε}) error. In addition, the Fourier-cutoff parameters in the proof cannot satisfy the hypotheses of Lemma 12: setting σ=7δ, the support conditions give b−δ=σ and b+δ=12σ/7, hence b=19σ/14 and δ=5σ/14, violating δ<b/4. The lemma must be repaired before the main theorem is supported.","section":"§2, Lemma 13"},{"comment":"The proof's final estimate uses 1/|∆|≪N^{−1/k+ε}, but for k=2 and h1=h2=h3≈1 the mixed difference ∆ is of size n^{−5/2}, so |∆|^{-1} is ≫N^{5/2−ε}; the asserted upper bound is false in that range. The displayed bound ∫_U^{2U}|S(x,N,k)|^8 dx ≪ (U N^4+N^{8−1/k})N^ε therefore lacks a valid derivation. Since Lemma 13 invokes Lemma 11 in its Hölder step, this invalidates the proof of Lemma 13 independently of the N_j^{1/4} vs N_j^{1/2} issue.","section":"§2, Lemma 11"},{"comment":"The counting lemmas for S5/S6 and S2/S3 are not proved in the text. Lemma 8's proof says the bound 'can be found by case analysis' and then gives only fragmentary cases; Lemma 9 says 'the proof is identical', although the equation in Lemma 9 has a different balance of square roots (six terms on one side, one on the other) and is not the same counting problem. Lemma 16 applies these lemmas, together with the unproved Lemma 13, to deduce the H^{3−1/254} bounds for S5 and S6. Thus the error term of Theorem 1 is not supported by the present exposition.","section":"§2, Lemmas 8–9 and Lemma 16"}],"minor_comments":[{"comment":"The proof's notation is inconsistent: after setting H1=N, H2=H^{1/2}, H3=H^{1/4}, it then sums over h1≤N, h2≤N^2, h3≤N^4; also ∂^3f/(∂t1∂t2∂t2) should presumably be ∂t1∂t2∂t3.","section":"§2, Lemma 11"},{"comment":"The function argument has 'S ∼ S' and the parameter list repeats L in A±(N,M,K,L,R,S,L,J); these typos obscure the ranges and should be corrected.","section":"§2, Lemma 8"},{"comment":"The notation 'δ ≍ δL^{1/2}' reuses δ for the threshold and for the small parameter; the two quantities should be renamed to avoid the appearance of a circular definition.","section":"§2, Lemma 16"},{"comment":"As stated ('Let A0>2 be fixed') the lemma claims a bound for all fixed A0; the paper only uses A0=267/27, and the cited result is known only in a restricted range. The statement should be restricted accordingly.","section":"§2, Lemma 15"},{"comment":"Theorem 1 and Theorem 2 are referred to as Theorem 1.1 and Theorem 1.2, and the abstract says 'the first author' where 'the authors' is meant; there are also name inconsistencies such as 'Dong Guangchang' versus 'Tong K.C.'.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is not in publishable form in its current state: the key counting lemma is stated in one form and used in another, and the proof of the underlying mean-value estimate appears to contain a false reciprocal bound. The overall strategy is plausible and the main constant is genuinely explicit, so I would not reject outright; the authors should supply a correct proof of Lemma 13 (or a correct replacement) and re-derive the error term from it."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Let me give you the short version. The paper claims an explicit asymptotic for the eighth power moment of the Dirichlet divisor error term, with constants C4 and C7 and error X^{3-1/254+epsilon}, plus a short-interval version. That would be progress, but the proof as written is not rigorous. The key counting estimates behind the off-diagonal terms are wrong or unproved, so Theorem 1 is not established.\n\nWhat is genuinely new: the explicit constants, the specific exponent 1/254, and the short-interval theorem. The expansion of Delta(x)^8 into S1–S7 and the identification of the diagonal contribution are competently done. The paper follows the Tsang–Zhai framework and is an honest attempt. What is not new is the existence of an eighth-moment asymptotic: Zhai's theorem (11) already covers k=8, so the \"for the first time\" claim is overstated.\n\nThe soft spots are serious. Lemma 11 claims 1/|Delta| << N^{-1/k+epsilon}; for k=2 and shifts of size N, Delta behaves like N^{-3/4}, so the reciprocal is >> N^{3/4}, not bounded above by N^{-1/2+epsilon}. The N^{8-1/k} term has no derivation. Lemma 13, which is the sole control on the off-diagonal S5/S6, is internally inconsistent: the statement has N_j^{1/4}, while the proof and Lemma 16 use N_j^{1/2}. The Fourier argument also omits a delta^{-1}N_j^4 term, so the H^{3-1/254} error in Lemma 16 is unsupported. Lemmas 8 and 9 are mostly delegated to \"case analysis\" and \"identical to Lemma 8\". And \"Kolesnik and Graham\" is cited without a matching reference.\n\nNone of this makes me think the main constant is fitted or the theorem false. The diagonal contribution is what it is, and the structure follows known methods. The gap is in technical counting, and a correct version of Lemma 13 might save the paper. But as written I would not bet on the proof.\n\nWho should read it: a specialist in exponential sums and the divisor problem who can judge whether the counting lemmas are repairable. It deserves peer review rather than desk rejection, because the claim is substantial and the flaws are technical, but any referee should be told to focus on Lemmas 11 and 13 first.","headline":"Plausible main term and useful explicit constants, but the key counting lemmas are not rigorous as written; refereeable but not acceptable yet.","tokens_in":17374,"tokens_out":4889,"would_cite":false,"duration_ms":52305,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11L07","11B83"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that $\\int_2^X \\Delta(x)^8\\,dx$ equals an explicit constant times $X^3$, up to $O(X^{3-1/254+\\varepsilon})$.","keywords":["divisor problem","divisor error term","eighth power moment","truncated Voronoi formula","exponential sums","square-root Diophantine equations","asymptotic formula","short intervals"],"falsifier":"Take $k=2$, $N_1=\\cdots=N_8=N$, and $\\delta=N^{-1}$, $N^{-3/2}$, or $N^{-2}$ in Lemma 13, and enumerate all $n_j\\le 2N$ with $0<|\\sqrt{n_1}+\\cdots+\\sqrt{n_4}-\\sqrt{n_5}-\\cdots-\\sqrt{n_8}|<\\delta$. If the fitted exponent of the count exceeds the lemma's product bound at accessible $N$, the counting lemma is false and the proof of Lemma 16 loses its error term; a count tracking the claimed power law across these $\\delta$ would support the main theorem's exponent.","tokens_in":16249,"feed_emoji":"📐","tokens_out":19074,"duration_ms":201408,"temperature":0.7,"pith_summary":"The divisor problem asks how the summatory function of divisors deviates from its main term, and $\\Delta(x)$ is that deviation. The paper claims that the integral of the eighth power of $\\Delta(x)$ from 2 to $X$ has a genuine asymptotic main term proportional to $X^3$, with explicit coefficient $(35C_7-28C_4)/(2048\\pi^8)$ and error $O(X^{3-1/254+\\varepsilon})$. That would be the first asymptotic formula for the eighth moment, extending known results for the first through fourth powers and fitting the general pattern $\\int \\Delta^k(x)\\,dx \\sim C_k X^{1+k/4}$. Such moment formulae turn the hard pointwise question of how large $\\Delta(x)$ is into averaged information, with explicit arithmetic constants encoding sums over square-root equalities weighted by divisor functions.","feed_headline":"Eighth moment of divisor error term has size $X^3$","feed_subtitle":"The paper proves an explicit constant times $X^3$ with a power-saving error for the eighth power mean.","key_machinery":"The load-bearing object is the truncated formula for $\\Delta(x)$ as $x^{1/4}$ times a finite sum of terms $d(n)n^{-3/4}\\cos(4\\pi\\sqrt{nx}-\\pi/4)$, plus an error term. Expanding the eighth power of this sum produces seven pieces, and the work is to show that six of them are negligible. The negligible pieces are oscillatory integrals controlled by two mechanisms: first-derivative estimates that need a lower bound on nonzero sums of square roots, and counting estimates for near-solutions of equalities between sums of four square roots on each side. Higher-dimensional Weyl differencing enters exactly there, bounding the exponential sums that the smoothed counting argument produces.","core_discovery":"On the paper's own terms, the discovery is Theorem 1: for every fixed integer $X\\ge 10$, $$\\int_2^X \\$\\Delta$^8(x)\\,dx = \\frac{35C_7-28C_4}{2048\\$pi^{8}$}\\int_2^X $x^{2}$\\,dx + O($X^{{3-1/254+\\varepsilon}}$),$$ where $C_4$ and $C_7$ are the explicit eight-variable sums in (14) and (15). These sums run over 8-tuples of natural numbers satisfying an equality of sums of square roots, with weights involving $d(n)$ and powers of the variables. The proof first obtains the analogous statement for the truncated sum $\\Sigma_Y(x)$ on a dyadic interval $[H,2H]$, controlling all oscillating contributions; only the two non-oscillating matchings survive to produce the constants. Summing dyadic intervals yields the stated asymptotic, and a short-interval version appears as Theorem 2.","pith_inferences":["Going beyond the paper, positivity of the integral forces the constant in Theorem 1 to be positive; checking $35C_7-28C_4>0$ from the two series is a self-contained verification that the main term has the correct sign.","A direct numerical test of Lemma 13 with $k=2$ and a range of $\\delta$ would show how much of the $1/254$ saving is genuinely available, because the proof of that lemma establishes a weaker bound than the main argument uses.","The same route looks extendable to higher even moments, with the square-root counting lemma as the bottleneck; each additional pair of factors adds many more near-equality cases before any new analysis is needed."],"forward_implications":["The eighth moment of the divisor error term is of size $X^3$ with a genuine power saving, so the $L^8$ average of $\\Delta(x)$ over $[2,X]$ is of order $X^{3/8}$.","The leading coefficient is given by the explicit sums in (14) and (15), so the constant in the asymptotic is a concrete arithmetic quantity rather than an unknown parameter.","Theorem 2 gives the same main term on short intervals $[X,X+H]$ for $H$ between $X^{7/32+\\delta}$ and $X$, with relative error $O(X^{-k})$.","The exponent 3 matches the pattern $1+k/4$ at $k=8$, so the result extends the moment pattern previously established for the first four powers."],"supporting_citations":[{"why":"supplies the truncated formula for Δ(x) that the eighth-power expansion starts from","marker":"[14]"},{"why":"provides the lower bounds for nonzero sums of square roots used to make the oscillatory integrals small","marker":"[5]"},{"why":"supplies the counting estimates for near-solutions of two-square-root equalities used inside Lemma 8","marker":"[7]"},{"why":"supplies the smooth cutoff with rapid Fourier decay used to count near-solutions in Lemmas 13 and 14","marker":"[13]"},{"why":"supplies the higher-dimensional Weyl differencing estimates in Lemmas 10 and 17 for the exponential sums after the cutoff","marker":"[1]"},{"why":"supplies the upper-bound moment estimate for |Δ(x)|^A used to control the truncated error term","marker":"[3]"}],"fun_headline_variants":["Eighth moment of Δ(x) pinned down to X^3","Explicit asymptotic for eighth power mean of Δ(x)","Δ(x) eighth moment: constant times X^3 with power saving","Eighth power mean of divisor error term equals X^3"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The power-saving error rests on Lemma 13's count of near-solutions of a sum-of-square-roots equation, and that lemma's proof delivers a weaker bound than the main argument needs; if the weaker bound fails, the advertised $O(X^{3-1/254+\\varepsilon})$ error collapses.","fun_headline_variants_meta":{"raw":{"variants":["Eighth moment of Δ(x) pinned down to X^3","Explicit asymptotic for eighth power mean of Δ(x)","Δ(x) eighth moment: constant times X^3 with power saving","Eighth power mean of divisor error term equals X^3"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000257,"raw_usage":{"total_tokens":1497,"prompt_tokens":780,"completion_tokens":717,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":396,"completion_tokens_details":{"reasoning_tokens":645}},"tokens_in":396,"tokens_out":717,"duration_ms":7445,"temperature":1.0,"reasoning_tokens":645,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:40:13.977187+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $k=2$, $N_1=\\cdots=N_8=N$, and $\\delta=N^{-1}$, $N^{-3/2}$, or $N^{-2}$ in Lemma 13, and enumerate all $n_j\\le 2N$ with $0<|\\sqrt{n_1}+\\cdots+\\sqrt{n_4}-\\sqrt{n_5}-\\cdots-\\sqrt{n_8}|<\\delta$. If the fitted exponent of the count exceeds the lemma's product bound at accessible $N$, the counting lemma is false and the proof of Lemma 16 loses its error term; a count tracking the claimed power law across these $\\delta$ would support the main theorem's exponent.","supporting_citations":[{"cited_title":"Ann.´Ecole","cited_arxiv_id":null,"evidence_quote":"supplies the truncated formula for Δ(x) that the eighth-power expansion starts from"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the lower bounds for nonzero sums of square roots used to make the oscillatory integrals small"},{"cited_title":"Ivi´ c and P","cited_arxiv_id":null,"evidence_quote":"supplies the counting estimates for near-solutions of two-square-root equalities used inside Lemma 8"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the smooth cutoff with rapid Fourier decay used to count near-solutions in Lemmas 13 and 14"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the higher-dimensional Weyl differencing estimates in Lemmas 10 and 17 for the exponential sums after the cutoff"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the upper-bound moment estimate for |Δ(x)|^A used to control the truncated error term"}],"review_version":1}