{"id":"62a50a34-5966-43ee-9ccb-55467ce9ec38","arxiv_id":"2508.01196","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every Gaussian-integer icube in dimension 4, and every one in dimension 3 whose norm is a sum of two squares, extends to an equal-length integral orthogonal basis; in dimension 4k+2 the norm must be a sum of two squares.","lead":"This paper asks when an integer vector, or an orthogonal family of integer vectors, can be completed to a full orthogonal basis of vectors of equal length. It proves the Gaussian-integer cases in dimensions 3 and 4, and it links the obstruction to a known difficulty in the sup-norm problem for the groups SU(n, n-1).","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.6's Case 2 uses an algebraically false identity, leaving the load-bearing 'enough divisors' step unproved; all main extension theorems depend on this lemma.","rationale":"The reader's weakest_assumption was Lemma 3.6, and my concern is also located in that lemma. However, the reader emphasized the reliance on the cited principal-right-ideal property of the Hurwitz order in large-norm cases, whereas I find a concrete algebraic error in the small-norm Case 2 construction: the identity defining u is false as printed. This strengthens rather than replaces the reader's conditionality: the central theorems likely remain true, but the proof as submitted does not establish Lemma 3.6. The proposed finite check can distinguish a genuine counterexample to the lemma from a merely typographical error in the construction, and therefore determines whether the condition should be a request for repair or a more serious objection. Because the reader already assigned CONDITIONAL, and my concern does not by itself force a lower verdict without the computational test, I leave the verdict unchanged.","tokens_in":22652,"tokens_out":22299,"duration_ms":251670,"concrete_test":"Enumerate the Lipschitz order A = Z[i]^4 (coordinate vectors in Z^4) up to norm N = 5000: for every t in A and every prime p <= N with p dividing |t|^2, check whether there exists u in A with |u|^2 = p and (conjugate(u) * t) / p in A, i.e. u is a left divisor of t of norm p. A single failure disproves Lemma 3.6; if all such instances pass, the lemma is almost certainly correct and the error in Case 2 is a repairable typo. The search can be restricted to Norm(t) < p^2, the exact regime where the faulty construction is invoked.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 3.6 is the hinge of the paper: through Corollary 3.5 and Proposition 1.9 it feeds every extension theorem (Theorems 3, 4, 7, 8 and Corollary 1.6). In Case 2 of its proof, for u' in S' \\ S and a chosen Hurwitz unit omega, the text defines u := u' * omega = ((u' + omega) / 2) * (2 * omega) - 1. This equality is false: for omega in S' \\ S with the stated parity properties, omega^2 is not 1, so ((u' + omega) / 2) * (2 * omega) - 1 equals u' * omega + (omega^2 - 1), not u' * omega. For example, omega = (1 + i + j + k)/2 has omega^2 = (-1 + i + j + k)/2. Thus the displayed construction does not produce the required left divisor u in A of norm p. This is exactly the step that converts a divisor in the Hurwitz order into a divisor in the Lipschitz order A, so the proof of 'enough divisors' is incomplete as printed. If Lemma 3.6 fails, Proposition 1.9(2) and the Gaussian extension theorems have no proven basis; if it holds, a corrected Case 2 is still required. The lemma itself can be tested finitely, which is the recommended next step.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies when an integral vector or a set of vectors over Z or Z[i] can be completed to an orthogonal basis of equal-length vectors (an n-icube). It proves that a vector in Z^n with n=4k+2 that lies in an n-icube must have norm expressible as a sum of two squares (Theorem 1), and that the analogous necessary condition for odd n over Z[i] holds (Proposition 1.3). The main results are Theorems 7 and 8: any icube in Z[i]^3 of norm that is a sum of two rational squares extends to a 3-icube, and any icube in Z[i]^4 extends to a 4-icube; Theorem 4 (every vector in Z[i]^4 extends) is a consequence. The proofs go through a correspondence, developed in Section 3, between Q-orthobalanced bases and factorizations in quaternion orders, and a key lemma asserting that the relevant orders have 'enough divisors' (Lemma 3.6). The final section gives an application to the sup-norm problem for SU(n,n−1), claiming lower bounds #S_n ≫_n L^{2(n−1)} for n=2,3,4.","tokens_in":22793,"tokens_out":10264,"duration_ms":107166,"significance":"The paper introduces a quaternion-order framework (orthobalanced bases, 'enough divisors') that gives a unified proof of previously known integer results and new Gaussian analogues in dimensions 3 and 4. The extension theorems for Gaussian icubes (Theorems 7 and 8) are natural and nontrivial, and the necessary condition for n=4k+2 (Theorem 1) with the resulting Corollary 1.6 is a clean characterization for n=6. The paper also gives a concrete counting obstruction relevant to the sup-norm amplification method. Several components are carefully proved and machine-checkable in principle, including the Smith normal form symmetry (Lemma 1.10), the bijection in Proposition 3.3, and the cross-product identities in Lemmas 4.6 and 4.7. If the load-bearing issues identified below are repaired, this would be a solid contribution to the arithmetic of integral orthogonal sets.","major_comments":[{"comment":"The displayed identity in Case 2 is algebraically false. For ω ∈ S′\\S with ω^2 ≠ 1, one has ((u′+ω)/2)·(2ω) − 1 = u′ω + ω^2 − 1, not u′ω. For example, ω = (1+i+j+k)/2 satisfies ω^2 = (−1+i+j+k)/2. Consequently the constructed element u is not shown to lie in S, and the proof does not establish the required left divisor of t in A of norm p. Since Lemma 3.6 is the hinge of Corollary 3.5 and hence of Proposition 1.9 and Theorems 3, 4, 7, 8 and Corollary 1.6, a corrected argument (or an independent proof of the 'enough divisors' property for A) is required.","section":"Lemma 3.6, Case 2"},{"comment":"The assertion that 'for almost all choices of a1, the condition imposed in Proposition 1.13 below holds' is stated without proof. This condition — that d2(A0) is coprime to its conjugate, i.e. all prime divisors of d2(A0) have norm p ≡ 1 mod 4 and are pairwise non-conjugate — is exactly what allows Proposition 1.13 to be applied, and it is needed to pass from the ≫ L^6 choices of a1 to the lower bound #S4(ℓ1,ℓ2) ≫ L^6. The authors should either supply the missing density argument or downgrade the n=4 lower bound to a conditional statement.","section":"Section 1.4"}],"minor_comments":[{"comment":"The line 'νp(αj) ≤ νp(n + 1 − αj)' should read 'νp(αj) ≤ νp(α_{n+1−j})'.","section":"Proof of Theorem 1"},{"comment":"The definition of δ appears to have a missing conjugate: δ = d2(A0)α2/α2 as printed equals d2(A0), making the subsequent expression ωλ^2α2/α2 = ωλ^2 independent of α2; presumably δ = d2(A0)α2/\\bar{α2} is intended.","section":"Proof of Proposition 1.13"},{"comment":"The word 'irredubible' should be 'irreducible'.","section":"Section 1.4"},{"comment":"The word 'orhogonality' should be 'orthogonality'.","section":"Example 1.7"},{"comment":"The notation 'sum of (4,n) squares' should be clarified, e.g., as gcd(4,n).","section":"Corollary 1.2"}],"recommendation":"major_revision","confidential_remarks":"The false identity in Lemma 3.6 Case 2 may be a typographical slip, but as printed it invalidates the proof of the paper's central hinge; the authors should be asked to supply a correct argument or a finite verification. The Section 1.4 'almost all' assertion is a further gap that needs to be addressed. The paper would be strengthened by explicit statements of which results are unconditional."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a genuine advance. The enough-divisors idea and the Prop 3.3 bijection between orthoregular bases and factorizations in a quaternion order are new and they make the Z[i] extension theorems provable. Theorems 7/8 (any icube in Z[i]^3/4 extends), Theorem 1 (n=4k+2 necessary sum of two squares), Corollary 1.6 (full characterization in Z^6), and Example 1.7 (counterexample to two LG20 conjectures) are all real content. The application to the sup-norm problem is a nice bonus, and the counting bound #S_n >> L^{2(n-1)} for n=2,3,4 is a quantitative obstruction to amplification for SU(n,n-1). I verified the valuation arguments in Prop 3.4(3), Lemmas 4.6/4.7, Lemma 1.10, Lemma 3.8; they check out.\n\nBut the manuscript as printed has three problems. First, the final determinant step in Theorem 1 is wrong as written: from S11^T S11 + S21^T S21 ≡ 0 mod p you cannot conclude (det S11)^2 + (det S21)^2 ≡ 0. The standard fix is to hit the matrix congruence with x and use that S11 is invertible mod p, so this is repairable. Second, the 4x2 matrix (16) is not an icube over Z[i]; the sign pattern from the rational case needs conjugation to work over Gaussian integers. Again repairable, because the theorem is true, but the displayed construction is false. Third, and most serious: Lemma 3.6, Case 2 contains an algebraically false identity. The text sets u := u'ω = ((u'+ω)/2)(2ω)-1, but the last step expands to u'ω + ω^2 -1, and for a Hurwitz unit like (1+i+j+k)/2, ω^2 ≠ 1. So the construction does not produce a divisor u in A of norm p. Since Lemma 3.6 feeds every Z[i] extension theorem through Corollary 3.5 and Prop 1.9(2), the proof of the central results is incomplete at this hinge. The lemma itself is probably true and is finitely checkable; a corrected Case 2 is needed.\n\nAlso: Conjecture 1.4's computational evidence is undocumented, and the 'almost all a1' claim in Section 1.4 is asserted without proof. These are minor in comparison.\n\nBottom line: the architecture is coherent, the results are likely correct, and the errors look repairable. This deserves a serious referee, but not acceptance in the current form.","headline":"Solid new quaternion-order framework for Gaussian icube extension, but two printed proof steps are wrong and the load-bearing enough-divisors lemma has a false identity in its proof.","tokens_in":23582,"tokens_out":3304,"would_cite":true,"duration_ms":34955,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11E25","11R52","52C07"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every Gaussian-integer icube in dimension 4 extends to an equal-length orthogonal basis, and in dimension 3 exactly when its norm is a sum of two squares; over $\\mathbb{Z}$, vectors in $\\mathbb{Z}^{4k+2}$ need sum-of-two-squares norm…","keywords":["integral vectors","orthogonal bases","icubes","Gaussian integers","quaternion orders","factorization in quaternion algebras","sup-norm problem of automorphic forms","Smith normal form"],"falsifier":"A finite computer search in the order $A = \\{r + sj : r, s \\in \\mathbb{Z}[i]\\}$ for a rational prime $p$ and an element $t$ with $p \\mid |t|^2$ but no left divisor of $t$ of norm $p$ would settle the load-bearing Lemma 3.6, since the proof splits into cases according to whether $|t|^2 < p^2$ and claims every such configuration factors. For the sup-norm application, one can compute the density of primitive vectors $a_1$ with $|a_1|^2 = |\\ell_1\\ell_2|^2$, with $a_2$ chosen as in display (16), for which all prime divisors of $d_2(A_0)$ have norm $4k+1$ and are pairwise non-conjugate; if that density fails to approach 1, the lower bound $\\#S_4(\\ell_1,\\ell_2) \\gg L^6$ collapses.","tokens_in":22296,"feed_emoji":"🧊","tokens_out":29723,"duration_ms":272404,"temperature":0.7,"pith_summary":"This paper studies when a set of mutually orthogonal integral vectors of equal length, an 'icube', can be completed to a full orthogonal basis of integral vectors of that same length, over both the ordinary integers $\\mathbb{Z}$ and the Gaussian integers $\\mathbb{Z}[i]$. Its main results are complete answers in low dimensions: over $\\mathbb{Z}[i]$, every icube in dimension 3 whose common norm is a sum of two rational squares extends to a 3-icube (Theorem 7), and every icube in dimension 4 extends unconditionally (Theorem 8); in particular, every single vector in $\\mathbb{Z}[i]^4$ extends (Theorem 4). Over $\\mathbb{Z}$ it adds a necessary condition, a vector in $\\mathbb{Z}^{4k+2}$ that lies in an $n$-icube must have norm a sum of two squares (Theorem 1), which combines with the Gaussian results to give an if-and-only-if characterization in dimension 6 (Corollary 1.6). The method is a dictionary between such extension problems and factorizations in orders of the Hamiltonian quaternions, and the same dictionary yields a counting lower bound demonstrating that the standard amplification method cannot succeed for the unitary groups $\\mathrm{SU}_{2,1}$, $\\mathrm{SU}_{3,2}$, and $\\mathrm{SU}_{4,3}$.","feed_headline":"Every 4D Gaussian vector fits into an equal-length orthogonal basis","feed_subtitle":"No condition needed in Gaussian 4-space; in 6-dim integer vectors complete exactly when norm is a sum of two squares.","key_machinery":"The engine is a bijection (Proposition 3.3) between $Q$-orthoregular bases of a binary hermitian form $Q$ and factorizations in the quaternion order $S = \\{r + s\\sqrt{\\varepsilon}\\, j \\mid r, s \\in R\\} \\subset \\mathbb{H}$, where $K$ is $\\mathbb{Q}$ or $\\mathbb{Q}(i)$, $R$ is its ring of integers, and $\\mathbb{H}$ is the algebra of Hamiltonian quaternions. For a form with Gram matrix entries $\\alpha, \\beta, \\gamma$ and discriminant $\\mu = \\Delta\\varepsilon$, a $Q$-orthoregular basis of norm $\\lambda = \\nu\\Delta$ corresponds exactly to a factorization $uv = (\\lambda/\\Delta)(\\beta + \\delta\\sqrt{\\varepsilon}\\, j)$ with $|u|^2 = \\alpha\\lambda/\\Delta$, so the question 'can this icube be extended?' becomes 'does this quaternion element have a divisor of prescribed norm?'. The key arithmetic input is Lemma 3.6: for $R = \\mathbb{Z}$ and $R = \\mathbb{Z}[i]$ with $\\varepsilon = 1$, the orders $\\mathbb{Z}[j]$ and $A = \\{r + sj \\mid r, s \\in \\mathbb{Z}[i]\\}$ have 'enough divisors', meaning that whenever $|y|^2$ divides $|t|^2$, the element $t$ has a left divisor $u$ with $|u|^2 = |y|^2$. Around this sits the lattice calculus of Section 4: the module $\\Lambda$ of vectors orthogonal to a given $k$-icube has discriminant $\\mathrm{disc}(Q) = \\lambda^k/|d_k(A_0)|^2$ (Proposition 4.1), the cross-product identity of Corollary 4.3, and the Smith normal form pairing $\\alpha_j\\alpha_{n+1-j} = \\lambda$ of Lemma 1.10, which together reduce extension in higher dimensions to the binary hermitian case. The 'enough divisors' property is delicate: Example 3.7 shows it fails for other choices of $\\varepsilon$, which is why the theorems apply specifically to the two orders $\\mathbb{Z}[j]$ and $A$.","core_discovery":"The paper's central claim is that, over the Gaussian integers, extension of icubes is governed only by the norm in dimension 3 and is unconditional in dimension 4. Theorem 7 states that if $\\lambda$ is a sum of two rational squares, then every $k$-icube ($1 \\le k \\le 3$) in $\\mathbb{Z}[i]^3$ of norm $\\lambda$ extends to a 3-icube, and Theorem 8 states that every $k$-icube ($1 \\le k \\le 4$) in $\\mathbb{Z}[i]^4$ extends to a 4-icube; these are the $\\mathbb{Z}[i]$ analogues of the known extension theorems over $\\mathbb{Z}$ in dimensions 3 and 4 ([GKMS12], [KK12]). Over $\\mathbb{Z}$, Theorem 1 gives the necessary condition that a vector in $\\mathbb{Z}^{4k+2}$ contained in an $n$-icube has norm a sum of two squares, and Corollary 1.6 turns this into an if-and-only-if for $n = 6$ by embedding $\\mathbb{Z}[i]^3$ into $\\mathbb{Z}^6$. Finally, Section 1.4 converts the extension theorems into the counting statement $\\#S_n(\\ell_1,\\ell_2) \\gg_n L^{2(n-1)}$ for $n = 2, 3, 4$, which the paper presents as a quantitative explanation of why the amplification method cannot, as currently understood, solve the sup-norm problem for $\\mathrm{SU}_{n,n-1}$.","pith_inferences":["The enough-divisors property is the natural place to look for a higher-dimensional version: Conjecture 1.4 would follow from an 8-dimensional analogue of Lemma 3.6, and the proof pattern points toward searching for such a property inside the relevant quaternion or octonion orders rather than in the lattice geometry alone.","Corollary 1.6 is the only dimension in which the paper proves the converse of Theorem 1; testing $n = 10$, the next $4k+2$ dimension, would show whether 'norm is a sum of two squares' remains sufficient for single vectors there or whether new obstructions appear, because the proof of the necessity is uniform in $k$.","The Section 1.4 assertion that Proposition 1.13's condition $(d_2(A_0), d_2(A_0)) = 1$ holds for almost all choices of the first column $a_1$ is stated without proof; a density computation over the sphere $|a_1|^2 = |\\ell_1\\ell_2|^2$ would turn the $\\#S_4 \\gg L^6$ lower bound into a fully verified statement or reveal a hidden dependence on the choice of $\\ell_1, \\ell_2$.","The paper's non-extendable examples in $\\mathbb{Z}^{10}$, $\\mathbb{Z}^{18}$, and $\\mathbb{Z}^{36}$ (Example 1.7) show that beyond dimension 6 the norm condition is not sufficient for collections of vectors, so the complete answers in dimensions 3, 4, and 6 do not propagate naively to higher dimensions."],"forward_implications":["Every single Gaussian vector in $\\mathbb{Z}[i]^4$ extends to a 4-icube, and every icube in $\\mathbb{Z}[i]^4$ extends to a 4-icube (Theorems 4 and 8): in Gaussian 4-space the equal-length orthogonal basis completion problem has no exceptions.","In dimension 3 over $\\mathbb{Z}[i]$, the norm condition is exact: an icube of norm $\\lambda$ extends to a 3-icube if and only if $\\lambda$ is a sum of two rational squares (Theorem 7 together with Proposition 1.3).","For the ordinary integers, a vector $v \\in \\mathbb{Z}^6$ lies in a 6-icube if and only if $|v|^2$ is a sum of two squares (Corollary 1.6).","The counting lower bound $\\#S_n(\\ell_1,\\ell_2) \\gg_n L^{2(n-1)}$ for $n = 2, 3, 4$ shows that the amplification method as currently understood cannot yield a sup-norm exponent improvement for $\\mathrm{SU}_{2,1}$, $\\mathrm{SU}_{3,2}$, and $\\mathrm{SU}_{4,3}$ (Section 1.4).","If Conjecture 1.4 holds, that any icube in $\\mathbb{Z}^8$ extends to an 8-icube, then for every $n < 8$ any icube in $\\mathbb{Z}^n$ of norm a sum of $(4,n)$ squares extends to an $n$-icube (Proposition 1.5)."],"supporting_citations":[{"why":"Introduces the k-icube notion and supplies the Z-version results (Proposition 1.1, Theorem 2) that the paper extends to Z[i].","marker":"[GKMS12]"},{"why":"The original source of the square-norm theorem and the dimension-3 extension theorem over Z, which Theorem 1 and Theorems 5 and 7 generalize.","marker":"[Sár61]"},{"why":"Supplies Theorems 5 and 6, the Z^3 and Z^4 extension theorems whose Z[i] analogues are the paper's Theorems 7 and 8.","marker":"[KK12]"},{"why":"Provides the principal-right-ideal property of the Hurwitz order used in the critical large-norm case of Lemma 3.6.","marker":"[CS03, Section 5.1]"},{"why":"Gives the elementary approach to 3-cubes and proves the k = 1 primitive case of the discriminant identity (Proposition 4.1) on which the dimension-4 argument builds.","marker":"[Hor24]"},{"why":"Its Proposition 18 is generalized by the paper's Theorem 1, and the paper's non-extendable examples refute its Conjectures 3 and 4.","marker":"[LG20]"},{"why":"Sets the baseline sup-norm bound whose improvement is the target of the amplification method that Section 1.4 shows is blocked.","marker":"[Sar04]"},{"why":"Introduces the amplification method, and the matrix-counting step at the heart of that method is the very set whose size the paper's lower bound estimates.","marker":"[IS95]"}],"fun_headline_variants":["4D Gaussian orthonormal extension needs no condition","No condition needed to complete a Gaussian 4D icube","Gaussian 4D: every vector has an equal-length orthogonal completion","Unconditional extension for Gaussian 4D vectors"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The extension theorems all rest on Lemma 3.6, which asserts that the quaternion orders $\\mathbb{Z}[j]$ and $\\{r + sj : r, s \\in \\mathbb{Z}[i]\\}$ have 'enough divisors', and whose large-norm case depends on the cited principal-right-ideal property of the Hurwitz order; if that property does not yield divisors of exactly the required norm, Theorems 7 and 8 fail, while the sup-norm conclusion separately leans on the unproved assertion that Proposition 1.13's hypothesis holds for almost all choices of the first column $a_1$.","fun_headline_variants_meta":{"raw":{"variants":["4D Gaussian orthonormal extension needs no condition","No condition needed to complete a Gaussian 4D icube","Gaussian 4D: every vector has an equal-length orthogonal completion","Unconditional extension for Gaussian 4D vectors"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000887,"raw_usage":{"total_tokens":3815,"prompt_tokens":919,"completion_tokens":2896,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":535,"completion_tokens_details":{"reasoning_tokens":2828}},"tokens_in":535,"tokens_out":2896,"duration_ms":24977,"temperature":1.0,"reasoning_tokens":2828,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T05:51:10.881858+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A finite computer search in the order $A = \\{r + sj : r, s \\in \\mathbb{Z}[i]\\}$ for a rational prime $p$ and an element $t$ with $p \\mid |t|^2$ but no left divisor of $t$ of norm $p$ would settle the load-bearing Lemma 3.6, since the proof splits into cases according to whether $|t|^2 < p^2$ and claims every such configuration factors. For the sup-norm application, one can compute the density of primitive vectors $a_1$ with $|a_1|^2 = |\\ell_1\\ell_2|^2$, with $a_2$ chosen as in display (16), for which all prime divisors of $d_2(A_0)$ have norm $4k+1$ and are pairwise non-conjugate; if that density fails to approach 1, the lower bound $\\#S_4(\\ell_1,\\ell_2) \\gg L^6$ collapses.","supporting_citations":[],"review_version":1}