{"id":"6cab6ac9-2828-488e-a38e-5143b52bc259","arxiv_id":"2508.21717","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For l = binom(k+2,3), the tangent space of Hilb^l(A^3) is maximized uniquely at the monomial ideal m_k, proving the 3D Briancon-Iarrobino Conjecture.","lead":"These authors prove a 1978 conjecture about the most singular point of the Hilbert scheme of points in three-dimensional space, for numbers of points that are tetrahedral. They sharpen a prior upper bound with a combinatorial correction and show the known candidate ideal is exactly the unique maximizer.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Corollary 2.2, the lower bound t(I_i,I_j) ≥ binom(j-i+1,2), is stated without proof and is load-bearing in the final inequality of Theorem A; the central claim is not fully supported as written.","rationale":"I reviewed the proof of Theorem A and the surrounding lemmas. The paper's strategy is sound: it refines Ramkumar-Sammartano by accounting for ghost vectors through Lemma 1.14, Corollary 1.15, and the zero-vector claim, and the algebra in Theorem B and Corollary 1 checks out. The central risk is not the overall plan but the unproved Corollary 2.2; every numerical example in the text satisfies the inequality, often with equality, so I found no counterexample. However, since the inequality is exactly what produces the second correction term in the final chain, the proof as written is incomplete. The reader's conditional verdict is appropriate: I do not see a reason to reject or to accept outright. The missing proof of Corollary 2.2, and to a lesser extent the surjectivity step in Lemma 1.5, must be supplied before the proof can be considered complete.","tokens_in":11574,"tokens_out":29979,"duration_ms":270320,"concrete_test":"Use Macaulay2 (already acknowledged by the authors) to enumerate all Borel-fixed monomial ideals in k[x,y,z] with m1≤4 and, for every 0≤i<j≤m1, compute t(I_i,I_j) and compare with binom(j-i+1,2). The first nontrivial case is m1=4, i=0, j=2, where the bound requires t(I_0,I_2)≥3. A violation would disprove Corollary 2.2 and hence Theorem A; if the enumeration confirms the bound, the outstanding issue is the missing proof, which the authors should supply.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Corollary 2.2 is asserted without proof and is used essentially in the proof of Theorem A. In the displayed chain, Corollary 1.15 gives hom(I_i,S/I_j) ≤ l_i + l_j - t(I_i,I_j); to obtain the second subtraction of binom(m1+2,4), the proof needs sum_{i<j} t(I_i,I_j) ≥ sum_{i<j} binom(j-i+1,2), i.e. exactly Corollary 2.2. If that inequality fails, the final estimate is only T(I) ≤ (2m1+1)l - binom(m1+2,4), which is not the claimed sharp bound and does not force equality at m_k. The statement is labelled a corollary of Lemma 2.1, but no derivation is shown, and Lemma 2.1 concerns minimal generators γ of I and shifts α with α_x<0, whereas t(I_i,I_j) is defined solely from the y,z slices and the height of I_j; the bridge between them is not exhibited. The only concrete verification offered is Example 1.17, which checks equality for m_k, the extremal ideal, while Example 1.16 shows strict inequality in one instance. Neither establishes the general lower bound. Thus the central claim rests on an unproved combinatorial inequality.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a proof of the 1978 Briançon-Iarrobino Conjecture for the Hilbert scheme Hilb^l(A^3) when l is a tetrahedral number. The strategy is to refine the Ramkumar-Sammartano upper bound for the tangent space by introducing a correction term, obtaining T(I) ≤ (2m1+1)l − 2 binom(m1+2,4) for every Borel-fixed ideal of colength l. The authors then prove that the function ψ(m1) defined by this bound is strictly increasing, and they combine this with the known value T(m_k) = ψ(k) to conclude that the maximum singularity occurs at [m_k]. The paper also derives a conjectural necessary condition for maximal singularity. The main technical ingredients are a decomposition I = ⊕ x^i I_i, a bound hom(J,S/J') ≤ l + l' − t(J,J'), and a purported lower bound t(I_i,I_j) ≥ binom(j-i+1,2).","tokens_in":11838,"tokens_out":12160,"duration_ms":101403,"significance":"If the proof is completed, the result resolves a major conjecture in the study of Hilbert schemes of points, open since 1978, and it supplies a clean explanation of why the maximum singularity occurs at the tetrahedral ideal m_k. The overall strategy is attractive, and the arithmetic identity ψ(k) = T(m_k) checks out. The paper is clearly organized and the examples in Section 1 are helpful. However, the central new combinatorial inequality, Corollary 2.2, is stated without proof and is used essentially in the proof of Theorem A; this is a load-bearing gap in the current version.","major_comments":[{"comment":"The inequality binom(j-i+1,2) ≤ t(I_i,I_j) is asserted without proof, and it is exactly the term that makes the final bound sharp. In the proof of Theorem A, after applying Corollary 1.15, the displayed chain passes through ≤ (2m1+1)l − Σ_{0≤i<j≤m1-1} binom(j-i+1,2) − binom(m1+2,4). This step is legitimate only if Corollary 2.2 holds. The text offers no derivation: Lemma 2.1 concerns shifts α with α_x<0 and minimal generators γ, whereas t(I_i,I_j) is defined from the yz-slices and the height of I_j, and the bridge between these statements is not exhibited. Example 1.17 verifies equality for the extremal ideal m_k only, and Example 1.16 exhibits strict inequality in another case; neither establishes the general lower bound. Since this is the load-bearing novelty of the paper, the proof of Theorem A is incomplete as written. Please supply a complete proof of Corollary 2.2.","section":"Section 2, Corollary 2.2"},{"comment":"The well-definedness of g is not established. The proof defines g: B^{J,J'}_n → N^2 \\ \\tilde J by (U,α) ↦ (γ_y, −α_z−1) and must show (γ_y, −α_z−1) ∉ \\tilde J. The text says that because U lies in the upper half-plane, γ_z+α_z+1>0, hence γ_z > −α_z−1, 'and so f(U,α)/∈ J'. This conclusion does not follow: \\tilde J is upward closed, so the inequality γ_z > −α_z−1 gives no information about whether the smaller-height point (γ_y, −α_z−1) lies in \\tilde J. One needs an argument using the maximality of γ_y (and minimality of γ_z among ties) to rule out (γ_y, −α_z−1) ∈ \\tilde J; no such argument is supplied. Since the bijection B^{J,J'}_n ≅ N^2 \\ \\tilde J underlies the exact count in Corollary 1.9 and hence the bound in Lemma 1.5, this gap also affects the proof of Theorem A.","section":"Section 1, Lemma 1.5, Claim 2"}],"minor_comments":[{"comment":"In the proof of Claim 2, the displayed statement 'and so f(U,α)/∈ J' should read 'and so g(U,α) ∉ \\tilde J'; the function being defined is g, not f.","section":"Section 1, proof of Lemma 1.5"},{"comment":"The proof of the Claim is very terse: from (2.1) and (2.2) it asserts that the relevant components form a single bounded connected component and a single unbounded component, but the details of this implication are not spelled out. Please expand this step.","section":"Section 2, proof of Theorem A"},{"comment":"The derivative computation is correct, but the positivity step would be clearer if written as follows: for fixed k and ∆ ≥ 0, the displayed expression is decreasing in m1, so its minimum occurs at m1 = k, where it equals 3k^2 + 5k + 1 + 12∆ > 0.","section":"Section 2, proof of Theorem B"},{"comment":"The phrase 'xy−z-octant' is nonstandard; consider replacing it with 'the first octant of the (x,y,z) lattice' or similar wording.","section":"Section 1, Definition 1.1"}],"recommendation":"major_revision","confidential_remarks":"The main concern is the unproved Corollary 2.2. If the authors can supply a complete proof, the result appears significant and likely correct. Note that [40] is by the second named author and is used for two key external facts, the equality T(m_k)=ψ(k) and the uniqueness of m_k; this is acceptable but should be clearly flagged for the editor and readers. I recommend major revision rather than rejection because the issue is a missing proof rather than a demonstrated counterexample."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this paper claims a proof of the 1978 Briançon–Iarrobino conjecture in three dimensions for tetrahedral l. The architecture is clean and mostly convincing, but one load-bearing lemma is stated without proof, and that makes the main theorem unsupported as written. The result is probably true, and the gaps look repairable, but this submission is not ready for acceptance.\n\nWhat's new: The Ramkumar–Sammartano decomposition is refined with an explicit error term t(I_i,I_j) that subtracts ghost vectors, converting an upper bound off by 4/3 into the sharp bound. The monotonicity lemma for ψ is simple and effective. That's a genuine new idea.\n\nWhere the paper does well: The structure is transparent. Theorem A is a clean inequality; Theorem B is a short calculus check; Corollary 1 follows. The introduction is historically thorough. The examples are helpful, especially Examples 1.16 and 1.17 showing equality and strictness.\n\nSoft spots: Corollary 2.2 is asserted without any argument. It supplies the lower bound t(I_i,I_j) ≥ binom(j-i+1,2) for i<j, and that lower bound is exactly what converts the sum of hom(...) terms into the final sharpened bound. Without a proof, the central chain in Theorem A stops. The lemma is labelled a corollary of Lemma 2.1, but the connection is not shown—Lemma 2.1 concerns the full ideal and shifts with negative x-coordinate, while t(I_i,I_j) is about the slices. There is no derivation, only a checking of the extremal case in Example 1.17. That's not enough. Also, in the proof of Lemma 1.5, Claim 2, the well-definedness of g has a confusing step: the text says 'γ_z > -α_z-1, and so f(U,α) ∉ ~J' — presumably meaning g(·) — and the implication is not explained. That's minor but needs fixing.\n\nIs it circular? No. Corollary 2 uses [40] for uniqueness of m_k, which is independent. The external computations from [5] and [37] are standard.\n\nVerdict: The proof is probably fixable. The missing proof of Corollary 2.2 is the key. If a referee can fill the gap or the authors supply it, the paper becomes a major result. As written, it's a strong claim with a hole.\n\nRecommendation: Send to peer review. A serious referee should sit down with Lemma 2.1 and see whether the corollary follows from it or needs a separate argument. This paper deserves referee time.","headline":"The main idea is likely right, but a load-bearing lemma is unproved; the paper needs a real referee and a repaired proof before it can be accepted.","tokens_in":12382,"tokens_out":6706,"would_cite":true,"duration_ms":56352,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14C05","13D10","13P10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that for every tetrahedral number $l = \\binom{k+2}{3}$, the Hilbert scheme $\\mathrm{Hilb}^{l}(\\mathbb{A}^{3})$ is most singular at the point cut out by the $k$-th power of the maximal ideal, resolving the 1978…","keywords":["Hilbert scheme of points","Briançon–Iarrobino conjecture","tangent space dimension","Borel-fixed ideal","monomial ideal","tetrahedral number","maximal singularity","colength"],"falsifier":"Take the layers $I_i$ of every Borel-fixed ideal of small colength, compute $t(I_i, I_j)$ directly from the monomial sets, and check whether $t(I_i, I_j) \\geq \\binom{j-i+1}{2}$ holds in all cases; one pair with a smaller value, or one Borel-fixed ideal whose tangent-space dimension exceeds $\\psi(m_1)$, would refute the central claim.","tokens_in":11358,"feed_emoji":"📐","tokens_out":10912,"duration_ms":98862,"temperature":0.7,"pith_summary":"The Hilbert scheme $\\mathrm{Hilb}^{l}(\\mathbb{A}^{3})$ parameterizes ideals of colength $l$ in $\\mathbb{C}[x,y,z]$ — geometrically, configurations of $l$ points with multiplicities. The paper proves a 1978 conjecture stating that when $l$ is a tetrahedral number, $l = \\binom{k+2}{3}$, the largest possible dimension of the tangent space is attained at the ideal $(x,y,z)^{k}$. The proof refines a known upper bound by subtracting an explicit correction term, then shows the resulting function of the smallest pure exponent is strictly increasing. At the conjectured ideal the corrected upper bound equals the known tangent-space dimension, so the bound is sharp. The same argument gives a necessary condition for maximal singularity: the smallest pure exponent must be $k$.","feed_headline":"A 50-year guess about the most singular Hilbert scheme is proved","feed_subtitle":"For tetrahedral numbers of points in 3-space, the maximum tangent-space dimension occurs at the ideal $(x,y,z)^k$.","key_machinery":"The argument works with the decomposition $I = \\bigoplus_i x^{i} I_i$, where each $I_i$ is a monomial ideal in $\\mathbb{C}[y,z]$; this is the filtration introduced in [37]. For two such ideals $J, J'$, the paper proves $\\hom(J, S/J') = l + l' - \\operatorname{Card}(B_n^{J,J'} \\setminus A_n^{J,J'})$, where the subtracted term counts 'ghost vectors', connected components that lie in the upper half-plane but are unbounded and hence do not contribute to the tangent space. A computable lower bound for that term is $t(J, J')$, the number of monomials of $S/J'$ sitting at or above the height of $J'$. The new input is Corollary 2.2, which asserts $t(I_i, I_j) \\geq \\binom{j-i+1}{2}$ for the layers of any Borel-fixed ideal; summing these triangular lower bounds produces the term $2\\binom{m_1+2}{4}$ that makes the earlier upper bound sharp.","core_discovery":"The central result is Theorem A: for any 0-dimensional Borel-fixed ideal $I = (x^{m_1}, y^{m_2}, z^{m_3}, \\text{mixed generators})$ of colength $l = \\binom{k+2}{3}+\\Delta$ with $0 \\leq \\Delta \\leq \\binom{k+2}{2}-1$, the tangent-space dimension satisfies $T(I) \\leq (2m_1+1)l - 2\\binom{m_1+2}{4}$. Theorem B states that the right-hand side, viewed as a function $\\psi(m_1)$, is strictly increasing. Since Lemma 2.3 forces $m_1 \\leq k$ in the tetrahedral range, one gets $T(I) \\leq \\psi(k)$. A direct computation gives $\\psi(k) = \\binom{k+2}{2}\\binom{k+1}{2} = T(m_k)$, where $m_k = (x,y,z)^{k}$. Hence the tangent space at $[m_k]$ is at least as large as at any other point, and the maximum singularity of $\\mathrm{Hilb}^{l}(\\mathbb{A}^{3})$ occurs at $[m_k]$.","pith_inferences":["A reader who wants to verify the paper will find that Corollary 2.2 is the only step stated without proof; a direct combinatorial proof of $t(I_i, I_j) \\geq \\binom{j-i+1}{2}$ would make the argument self-contained, and a counterexample would destroy the theorem.","The same ghost-vector correction is a natural template for higher dimensions: for $\\mathrm{Hilb}^{l}(\\mathbb{A}^{N})$ with $l = \\binom{k+N-1}{N}$, the expected extremal ideal is the power $(x_1,\\dots,x_N)^{k}$, and the correction terms should be higher binomial coefficients, a direction the paper does not pursue.","The monotonicity of $\\psi$ suggests that just above a tetrahedral number the maximizer should still have smallest pure exponent $k$; the announced sequel addresses certain non-tetrahedral $l$, but the mechanism here is already visible in the inequality $T(I) \\leq \\psi(m_1)$."],"forward_implications":["For each tetrahedral $l = \\binom{k+2}{3}$, the point $[m_k]$ has tangent-space dimension exactly $\\binom{k+2}{2}\\binom{k+1}{2}$, and no point of $\\mathrm{Hilb}^{l}(\\mathbb{A}^{3})$ has a larger tangent space.","Any maximally singular point of $\\mathrm{Hilb}^{l}(\\mathbb{A}^{3})$ at a tetrahedral $l$ must have smallest pure exponent $m_1 = k$, confirming the necessary condition proposed in [40].","The upper bound $T(I) \\leq (2m_1+1)l - 2\\binom{m_1+2}{4}$ is valid for every Borel-fixed ideal whose colength lies in the interval $[\\binom{k+2}{3}, \\binom{k+3}{3}-1]$, giving a uniform estimate in that range.","Because the upper bound is increasing in $m_1$ and equals the lower bound at $m_k$, the maximal singularity problem for tetrahedral $l$ is settled exactly, not up to a multiplicative constant."],"supporting_citations":[{"why":"Supplies the decomposition $I = \\bigoplus_i x^{i} I_i$ and the tangent-space bound that this paper refines by introducing an error term.","marker":"[37]"},{"why":"States the original conjecture and gives the formula $T(m_k) = \\binom{k+2}{2}\\binom{k+1}{2}$ that fixes the lower bound.","marker":"[5]"},{"why":"Provides the necessary condition (Conjecture B) and the uniqueness lemma identifying $m_k$ as the only Borel-fixed ideal of tetrahedral colength with smallest pure exponent $k$.","marker":"[40]"}],"fun_headline_variants":["Tetrahedral Hilbert schemes: max singularity location proved","Max tangent-space dimension in 3D Hilbert schemes now known","Proof of Briancon-Iarrobino conjecture for tetrahedral Hilbert schemes","Maximum singularity of Hilb^l(A^3) pinned to (x,y,z)^k ideal"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the unproved Corollary 2.2: for the two-variable ideals $I_i$ in the decomposition of a Borel-fixed ideal, the count $t(I_i, I_j)$ is always at least the triangular number $\\binom{j-i+1}{2}$; if this inequality fails for some ideal, the paper's upper bound on $T(I)$ does not follow.","fun_headline_variants_meta":{"raw":{"variants":["Tetrahedral Hilbert schemes: max singularity location proved","Max tangent-space dimension in 3D Hilbert schemes now known","Proof of Briancon-Iarrobino conjecture for tetrahedral Hilbert schemes","Maximum singularity of Hilb^l(A^3) pinned to (x,y,z)^k ideal"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001191,"raw_usage":{"total_tokens":4913,"prompt_tokens":940,"completion_tokens":3973,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":556,"completion_tokens_details":{"reasoning_tokens":3894}},"tokens_in":556,"tokens_out":3973,"duration_ms":25542,"temperature":1.0,"reasoning_tokens":3894,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:40:01.200003+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the layers $I_i$ of every Borel-fixed ideal of small colength, compute $t(I_i, I_j)$ directly from the monomial sets, and check whether $t(I_i, I_j) \\geq \\binom{j-i+1}{2}$ holds in all cases; one pair with a smaller value, or one Borel-fixed ideal whose tangent-space dimension exceeds $\\psi(m_1)$, would refute the central claim.","supporting_citations":[{"cited_title":"Ramkumar and A","cited_arxiv_id":null,"evidence_quote":"Supplies the decomposition $I = \\bigoplus_i x^{i} I_i$ and the tangent-space bound that this paper refines by introducing an error term."},{"cited_title":"Briançon and A","cited_arxiv_id":null,"evidence_quote":"States the original conjecture and gives the formula $T(m_k) = \\binom{k+2}{2}\\binom{k+1}{2}$ that fixes the lower bound."}],"review_version":2}