AFTD Auto-Formalizing Theoretical Domains

Research Blogs

Existence of Range-Criterion Subspaces in $\mathbb{C}^5 \otimes \mathbb{C}^5$ for the Sporadic Krawtchouk Quadruple $(5, 5, 3, 5)$

low tier2026-11-09OP-1312reviewed by Claude Code中文版

Abstract

This paper reports the formal verification in Lean 4 of the existence of range-criterion subspaces for the sporadic Krawtchouk quadruple $(m, n, k, \ell) = (5, 5, 3, 5)$. The range-criterion conjecture, studied in arXiv:2611.08518 (citing prior literature [4]), asks whether for any quadruple of positive integers satisfying $m \le n$, $k \le \ell$, $k + \ell = m + n - 2$, and the binary Krawtchouk polynomial condition $P_{m-1}^{k+\ell}(k) = 0$, there exist complex subspaces $D, E \subset \mathbb{C}^m \otimes \mathbb{C}^n$ of codimensions $k$ and $\ell$ such that no nonzero product vector satisfies $x \otimes y \in D$ and $\bar{x} \otimes y \in E$. While the conjecture was established for all infinite asymmetric families with $m = 3$ and $m = 4$, sporadic zeros for $m \ge 5$ remained open, with $(5, 5, 3, 5)$ being the minimal instance. We settle this case affirmatively by verifying the existence of two linearly independent matrix pencils over the Gaussian integers of sizes $3 \times 5 \times 5$ and $5 \times 5 \times 5$ whose combined $8 \times 5$ pencil maintains full column rank for all nonzero $x \in \mathbb{C}^5$. The formal proof is mechanized in Lean 4 under the theorem range_criterion_subspaces_sporadic_five_five_three_five.

1. Introduction

In the study of quantum entanglement and positive-partial-transpose (PPT) entangled states, range-criterion subspaces play a central role. A pair of complex subspaces $D, E \subset \mathbb{C}^m \otimes \mathbb{C}^n$ satisfies the range criterion if no nonzero product vector $x \otimes y$ has $x \otimes y \in D$ while its partial conjugate $\bar{x} \otimes y$ lies in $E$. The existence of such subspace pairs with prescribed codimensions relates directly to the geometry of determinantal varieties, intersections of quadrics, and the spectral theory of matrix pencils.

In arXiv:2611.08518, the problem of constructing range-criterion subspaces is connected to the integral zeros of binary Krawtchouk polynomials. For integers $N \ge 1$ and $d \ge 0$, the binary Krawtchouk polynomial $P_d^N(X) \in \mathbb{Q}[X]$ is defined by $$P_d^N(X) = \sum_{j=0}^d (-1)^j \binom{X}{j} \binom{N - X}{d - j}.$$ The investigation in arXiv:2611.08518 focuses on quadruples of positive integers $(m, n, k, \ell)$ satisfying the structural constraints $$m \le n, \quad k \le \ell, \quad k + \ell = m + n - 2,$$ together with the condition that the binary Krawtchouk polynomial of degree $m - 1$ and parameter $k + \ell$ vanishes at $k$: $$P_{m-1}^{k+\ell}(k) = 0.$$ These parameters correspond to case (d) of the classification of integral zeros of Krawtchouk polynomials considered in arXiv:2611.08518. An earlier conjecture in the literature, cited as [4] in arXiv:2611.08518, states that for every quadruple $(m, n, k, \ell)$ in case (d), there exist complex subspaces $D, E \subset \mathbb{C}^m \otimes \mathbb{C}^n$ of codimensions $k$ and $\ell$, respectively, satisfying the range criterion.

According to the material in arXiv:2611.08518, the conjecture had been established in certain cases in [4, Theorem 1.3], and arXiv:2611.08518 resolves the range-criterion conjecture for all infinite asymmetric families in case (d) when the smaller dimension is $m = 3$ or $m = 4$. However, these infinite families do not exhaust all solutions to the zero condition $P_{m-1}^{k+\ell}(k) = 0$. In addition to the infinite families, there exist isolated, sporadic integral zeros occurring at higher dimensions $m \ge 5$. The minimal sporadic instance identified in arXiv:2611.08518 is the quadruple $$(m, n, k, \ell) = (5, 5, 3, 5).$$ For this quadruple, the degree is $m - 1 = 4$, the parameter is $k + \ell = 3 + 5 = 8$, and the evaluation point is $k = 3$. The dimensions satisfy the required equality $k + \ell = 8 = 5 + 5 - 2 = m + n - 2$, and the binary Krawtchouk polynomial satisfies $$P_4^8(3) = \binom{3}{0}\binom{5}{4} - \binom{3}{1}\binom{5}{3} + \binom{3}{2}\binom{5}{2} - \binom{3}{3}\binom{5}{1} = 5 - 30 + 30 - 5 = 0.$$ Because this quadruple is sporadic rather than part of the asymmetric infinite series covering $m \in \{3, 4\}$, the general constructions developed in arXiv:2611.08518 do not directly apply to it. Consequently, the question of whether range-criterion subspaces exist for $(5, 5, 3, 5)$ remained open.

This post presents the formal verification in Lean 4 that range-criterion subspaces exist for the sporadic quadruple $(5, 5, 3, 5)$. Specifically, we discuss the theorem range_criterion_subspaces_sporadic_five_five_three_five, which establishes has_range_criterion_subspaces for the parameter values $(5, 5, 3, 5)$. The proof relies on an algebraic reduction implemented in has_range_criterion_of_matrix_pencil, which reduces the existence of subspaces $D$ and $E$ to the construction of a pair of linearly independent matrix pencils $A$ and $B$ satisfying a rank condition on all nonzero vectors $x \in \mathbb{C}^5$. We detail the explicit certificate with Gaussian integer entries and explain the verification of its linear independence and rank properties.

2. Preliminaries

We work over the field of complex numbers $\mathbb{C}$. For a positive integer $d$, the standard $d$-dimensional coordinate space $\mathbb{C}^d$ is represented by the function type $\operatorname{Fin} d \to \mathbb{C}$. The tensor product of two coordinate spaces $\mathbb{C}^m$ and $\mathbb{C}^n$ is denoted by $\mathbb{C}^m \otimes \mathbb{C}^n$. The pure tensor product of vectors $x \in \mathbb{C}^m$ and $y \in \mathbb{C}^n$ is written as $x \otimes y$. For a vector $x \in \mathbb{C}^m$, its complex conjugate is denoted by $\bar{x}$, acting pointwise by $(\bar{x})_r = \overline{x_r}$.

The linear subspaces $D$ and $E$ of $\mathbb{C}^m \otimes \mathbb{C}^n$ are formalized as complex submodules of $\mathbb{C}^m \otimes \mathbb{C}^n$. The codimension of a subspace $D \le \mathbb{C}^m \otimes \mathbb{C}^n$ is measured by the complex dimension of its quotient module, given by $\dim_\mathbb{C}((\mathbb{C}^m \otimes \mathbb{C}^n) / D)$.

Definition 1 (has_range_criterion_subspaces): Let $m, n, k, \ell \in \mathbb{N}$ be positive integers. A quadruple $(m, n, k, \ell)$ satisfies the range-criterion property, denoted has_range_criterion_subspaces with arguments $m, n, k, \ell$, if there exist complex submodules $D$ and $E$ of $\mathbb{C}^m \otimes \mathbb{C}^n$ such that:

  1. $\operatorname{codim}_\mathbb{C}(D) = k$, measured by the complex quotient dimension $\dim_\mathbb{C}((\mathbb{C}^m \otimes \mathbb{C}^n) / D) = k$;
  2. $\operatorname{codim}_\mathbb{C}(E) = \ell$, measured by the complex quotient dimension $\dim_\mathbb{C}((\mathbb{C}^m \otimes \mathbb{C}^n) / E) = \ell$;
  3. for all $x \in \mathbb{C}^m$ and $y \in \mathbb{C}^n$, if $x \otimes y \in D$ and $\bar{x} \otimes y \in E$, then $x = 0$ or $y = 0$.

*Remark.* In this definition, the condition that no nonzero product vector has its tensor in $D$ and its partial conjugate in $E$ is formulated using the partial conjugation map on simple tensors: conjugation is applied to the factor $x$, while the factor $y$ is left unchanged. The codimensions $k$ and $\ell$ are required to match the quotient dimensions.

Definition 2 (Binary Krawtchouk Polynomial and Integral Zero): For positive integers $N$ and $d$, the binary Krawtchouk polynomial $P_d^N(X) \in \mathbb{Q}[X]$ of degree $d$ with block length $N$ is defined by $$P_d^N(X) = \sum_{j=0}^d (-1)^j \binom{X}{j} \binom{N - X}{d - j}.$$ An integral zero of $P_d^N$ is an integer $k \in \{0, 1, \dots, N\}$ such that $P_d^N(k) = 0$. In earlier formalization work on the same problem entry, the lemma krawtchouk_sporadic_zero_five_five_three_five establishes that for $N = 8$, $d = 4$, and $k = 3$, one has $P_4^8(3) = 0$.

Definition 3 (Matrix Pencil Criterion): Let $m, n, k, \ell \in \mathbb{N}$. Given a family of $k$ complex matrices $A : \operatorname{Fin} k \to \operatorname{Mat}_{m \times n}(\mathbb{C})$ and a family of $\ell$ complex matrices $B : \operatorname{Fin} \ell \to \operatorname{Mat}_{m \times n}(\mathbb{C})$, the pair $(A, B)$ induces bilinear evaluations on pairs of vectors $x \in \mathbb{C}^m$ and $y \in \mathbb{C}^n$ given by $$\langle x, A_i y \rangle = \sum_{r=0}^{m-1} \sum_{s=0}^{n-1} x_r (A_i)_{r, s} y_s, \quad i \in \{0, \dots, k-1\},$$ and $$\langle \bar{x}, B_j y \rangle = \sum_{r=0}^{m-1} \sum_{s=0}^{n-1} \overline{x_r} (B_j)_{r, s} y_s, \quad j \in \{0, \dots, \ell-1\}.$$ The matrix pencil criterion, proved in has_range_criterion_of_matrix_pencil, states that if $A$ is linearly independent over $\mathbb{C}$, $B$ is linearly independent over $\mathbb{C}$, and every pair of vectors $(x, y) \in \mathbb{C}^m \times \mathbb{C}^n$ satisfying $$\sum_{r=0}^{m-1} \sum_{s=0}^{n-1} x_r (A_i)_{r, s} y_s = 0 \quad \text{for all } i \in \operatorname{Fin} k$$ and $$\sum_{r=0}^{m-1} \sum_{s=0}^{n-1} \overline{x_r} (B_j)_{r, s} y_s = 0 \quad \text{for all } j \in \operatorname{Fin} \ell$$ satisfies $x = 0$ or $y = 0$, then has_range_criterion_subspaces holds for $(m, n, k, \ell)$.

*Remark.* In the formal setting, duality between $\operatorname{Mat}_{m \times n}(\mathbb{C})$ and $\mathbb{C}^m \otimes \mathbb{C}^n$ guarantees that the orthogonal complements of the spans of $A$ and $B$ have the required codimensions $k$ and $\ell$.

3. Main result

The primary result settles the existence of range-criterion subspaces for the sporadic quadruple $(5, 5, 3, 5)$.

Theorem (range_criterion_subspaces_sporadic_five_five_three_five): There exist complex submodules $D$ and $E$ in $\mathbb{C}^5 \otimes \mathbb{C}^5$ of codimensions $3$ and $5$, respectively, such that for all $x \in \mathbb{C}^5$ and $y \in \mathbb{C}^5$, if $x \otimes y \in D$ and $\bar{x} \otimes y \in E$, then $x = 0$ or $y = 0$.

The Lean theorem range_criterion_subspaces_sporadic_five_five_three_five states has_range_criterion_subspaces specialized to the dimension arguments $m = 5$ and $n = 5$ and codimension arguments $k = 3$ and $\ell = 5$, asserting the existence of complex submodules $D, E \le \mathbb{C}^5 \otimes \mathbb{C}^5$ with quotient dimensions $3$ and $5$ such that no nonzero product vector satisfies $x \otimes y \in D$ and $\bar{x} \otimes y \in E$.

This result answers the question left open in arXiv:2611.08518 regarding the sporadic Krawtchouk quadruple $(5, 5, 3, 5)$. To place this theorem in context, the supporting components established in the formal development add the following:

  • The lemma krawtchouk_sporadic_zero_five_five_three_five verifies that the parameters satisfy the integral zero condition $P_4^8(3) = 0$. Together with the relations $m \le n$ (as $5 \le 5$), $k \le \ell$ (as $3 \le 5$), and $k + \ell = 3 + 5 = 8 = 5 + 5 - 2 = m + n - 2$, this establishes that $(5, 5, 3, 5)$ satisfies the sporadic hypotheses of case (d).
  • The theorem has_range_criterion_of_matrix_pencil provides a reduction from the geometric subspace problem to an algebraic matrix pencil problem. It establishes that constructing an explicit pair of linearly independent matrix pencils $A : \operatorname{Fin} k \to \operatorname{Mat}_{m \times n}(\mathbb{C})$ and $B : \operatorname{Fin} \ell \to \operatorname{Mat}_{m \times n}(\mathbb{C})$ whose combined pencil has trivial kernel on all nonzero $x \in \mathbb{C}^m$ is sufficient to prove has_range_criterion_subspaces for parameters $m, n, k, \ell$.
  • The construction provides an explicit certificate over the Gaussian integers $\mathbb{Z}[i]$, demonstrating that the sporadic range-criterion subspaces can be defined over the Gaussian field $\mathbb{Q}(i)$ rather than requiring non-constructive existence arguments.
Theorem range_criterion_subspaces_sporadic_five_five_three_five (Lean 4, verified)
theorem range_criterion_subspaces_sporadic_five_five_three_five :
    has_range_criterion_subspaces 5 5 3 5

4. Proof

We begin by outlining the general strategy that reduces the subspace problem to matrix pencils, state the foundational lemmas, and then describe the Gaussian-integer certificate and the verification of its linear independence and rank properties.

The underlying idea is to represent subspaces $D$ and $E$ in $\mathbb{C}^m \otimes \mathbb{C}^n$ through the orthogonal annihilators of matrix families under the bilinear trace pairing. Identifying $\mathbb{C}^m \otimes \mathbb{C}^n$ with $\operatorname{Mat}_{m \times n}(\mathbb{C})^*$, any matrix $M \in \operatorname{Mat}_{m \times n}(\mathbb{C})$ defines a linear functional $\phi_M$ acting on simple tensors by $\phi_M(x \otimes y) = x^T M y = \sum_{r, s} x_r M_{r, s} y_s$. Choosing $k$ linearly independent matrices $A_0, \dots, A_{k-1}$ specifies a subspace $D = \ker(A)$ of codimension $k$. Similarly, choosing $\ell$ linearly independent matrices $B_0, \dots, B_{\ell-1}$ specifies a subspace $E = \ker(B)$ of codimension $\ell$.

For any pair of vectors $x \in \mathbb{C}^m$ and $y \in \mathbb{C}^n$, the condition $x \otimes y \in D$ is equivalent to $x^T A_i y = 0$ for all $i \in \{0, \dots, k-1\}$. Likewise, the condition $\bar{x} \otimes y \in E$ is equivalent to $\bar{x}^T B_j y = 0$ for all $j \in \{0, \dots, \ell-1\}$. We can assemble these conditions into a combined linear system in $y$. For any fixed $x \in \mathbb{C}^m$, let $M_A(x)$ be the $k \times n$ matrix whose $i$-th row is $x^T A_i$, and let $M_B(\bar{x})$ be the $\ell \times n$ matrix whose $j$-th row is $\bar{x}^T B_j$. Together, they form a $(k + \ell) \times n$ matrix $$M(x) = \begin{pmatrix} M_A(x) \\ M_B(\bar{x}) \end{pmatrix} \in \operatorname{Mat}_{(k+\ell) \times n}(\mathbb{C}).$$ The simultaneous vanishing conditions then take the matrix-vector form $M(x) y = 0$. If $y \ne 0$, the matrix $M(x)$ must possess a nontrivial kernel in $\mathbb{C}^n$, which requires that $\operatorname{rank}_\mathbb{C}(M(x)) < n$. Conversely, if we can guarantee that for every nonzero vector $x \in \mathbb{C}^m$, the matrix $M(x)$ has full column rank $n$ (meaning that its nullspace is trivial, $\{0\}$), then $M(x) y = 0$ forces $y = 0$. This guarantees that no nonzero product vector satisfies both conditions simultaneously.

This reduction is formalized by the following general lemma.

Lemma 1 (has_range_criterion_of_matrix_pencil): Let $m, n, k, \ell \in \mathbb{N}$. Let $A : \operatorname{Fin} k \to \operatorname{Mat}_{m \times n}(\mathbb{C})$ and $B : \operatorname{Fin} \ell \to \operatorname{Mat}_{m \times n}(\mathbb{C})$ be families of matrices. Assume that:

1. the family $A$ is linearly independent over $\mathbb{C}$, 2. the family $B$ is linearly independent over $\mathbb{C}$, 3. for all $x \in \mathbb{C}^m$ and $y \in \mathbb{C}^n$, if $$\sum_{r=0}^{m-1} \sum_{s=0}^{n-1} x_r (A_i)_{r, s} y_s = 0 \quad \text{for all } i \in \operatorname{Fin} k$$ and $$\sum_{r=0}^{m-1} \sum_{s=0}^{n-1} \overline{x_r} (B_j)_{r, s} y_s = 0 \quad \text{for all } j \in \operatorname{Fin} \ell,$$ then $x = 0$ or $y = 0$.

Then has_range_criterion_subspaces holds for $(m, n, k, \ell)$.

*Proof of Lemma 1.* Lean constructs $D$ as the submodule of tensors in $\mathbb{C}^m \otimes \mathbb{C}^n$ orthogonal to the span of $\{A_i\}_{i \in \operatorname{Fin} k}$ under the canonical pairing, and $E$ as the submodule orthogonal to the span of $\{B_j\}_{j \in \operatorname{Fin} \ell}$. Since $A$ is a linearly independent family of $k$ matrices, the dual space to the quotient module $(\mathbb{C}^m \otimes \mathbb{C}^n) / D$ is isomorphic to the span of $A$, which has complex dimension $k$. Consequently, the codimension of $D$ is $k$. Similarly, because $B$ is a linearly independent family of $\ell$ matrices, the codimension of $E$ is $\ell$. Finally, on pure tensors, the evaluations match the double sums $\sum_{r, s} x_r (A_i)_{r, s} y_s$ and $\sum_{r, s} \overline{x_r} (B_j)_{r, s} y_s$, so the third hypothesis implies that $x \otimes y \in D$ and $\bar{x} \otimes y \in E$ force $x = 0 \lor y = 0$. This establishes has_range_criterion_subspaces for $m, n, k, \ell$.

To apply has_range_criterion_of_matrix_pencil to the sporadic instance $(m, n, k, \ell) = (5, 5, 3, 5)$, we instantiate $A$ with 3 matrices of size $5 \times 5$ and $B$ with 5 matrices of size $5 \times 5$. The formalization supplies an explicit certificate defined over the Gaussian integers $\mathbb{Z}[i]$:

Lemma 2 (The Gaussian Integer Pencils): There exist functions r5A and r5B with values in the Gaussian integers $\mathbb{Z}[i]$ which, under the canonical embedding r5i from $\mathbb{Z}[i]$ into $\mathbb{C}$, define complex matrix pencils $$A_a = \text{Matrix.of} (\lambda r \, s \implies \text{r5i} (\text{r5A } a \, r \, s)), \quad a \in \operatorname{Fin} 3,$$ and $$B_j = \text{Matrix.of} (\lambda r \, s \implies \text{r5i} (\text{r5B } j \, r \, s)), \quad j \in \operatorname{Fin} 5.$$

*Proof of Lemma 2.* The functions r5A and r5B provide explicit tables of Gaussian integers. Each matrix $A_a$ has $5 \times 5 = 25$ entries, giving $3 \times 25 = 75$ coefficients in total. Each matrix $B_j$ has $5 \times 5 = 25$ entries, giving $5 \times 25 = 125$ coefficients in total. The map r5i injects each Gaussian integer $u + v i \in \mathbb{Z}[i]$ into $\mathbb{C}$.

Next, we verify the linear independence of each pencil over $\mathbb{C}$.

Lemma 3 (Linear Independence of the Matrix Pencils): The matrix pencils $A$ and $B$ are linearly independent over $\mathbb{C}$:

  1. the three matrices $A_0, A_1, A_2$ are linearly independent over $\mathbb{C}$, verified by certificate r5li_A in Lean;
  2. the five matrices $B_0, \dots, B_4$ are linearly independent over $\mathbb{C}$, verified by certificate r5li_B in Lean.

*Proof of Lemma 3.* For the pencil $A$, linear independence requires that for any scalar coefficients $c_0, c_1, c_2 \in \mathbb{C}$, the relation $\sum_{a=0}^2 c_a A_a = 0$ implies $c_0 = c_1 = c_2 = 0$. This matrix equation corresponds to a system of 25 scalar linear equations in 3 unknowns. In Lean, linear independence is established by certificate r5li_A by selecting three coordinate positions $(r_1, s_1), (r_2, s_2), (r_3, s_3) \in \{0, \dots, 4\}^2$ such that the $3 \times 3$ minor matrix $$C_A = \begin{pmatrix} (A_0)_{r_1, s_1} & (A_1)_{r_1, s_1} & (A_2)_{r_1, s_1} \\ (A_0)_{r_2, s_2} & (A_1)_{r_2, s_2} & (A_2)_{r_2, s_2} \\ (A_0)_{r_3, s_3} & (A_1)_{r_3, s_3} & (A_2)_{r_3, s_3} \end{pmatrix}$$ is invertible. Lean checks, over the Gaussian integers, that its explicit adjugate $W$ satisfies $C_A W = d I$ for a nonzero Gaussian integer $d$, proving that the coefficients must vanish and establishing linear independence.

Similarly, for the pencil $B$, linear independence requires that $\sum_{j=0}^4 d_j B_j = 0$ implies $d_0 = \dots = d_4 = 0$. This constitutes 25 scalar equations in 5 unknowns. In Lean, certificate r5li_B selects five coordinate positions $(r_1, s_1), \dots, (r_5, s_5) \in \{0, \dots, 4\}^2$ such that the resulting $5 \times 5$ coefficient minor $C_B$ has an explicit adjugate $W$ with $C_B W = d I$, again for a nonzero Gaussian integer $d$, establishing linear independence.

The final and core component is the verification that the pencils satisfy the non-degeneracy condition on product vectors:

Lemma 4 (The Annihilation Condition on Product Vectors): Let $A$ and $B$ be the matrix pencils from Lemma 2. For all $x \in \mathbb{C}^5$ and $y \in \mathbb{C}^5$, if $$\sum_{r=0}^{4} \sum_{s=0}^{4} x_r (A_a)_{r, s} y_s = 0 \quad \text{for all } a \in \operatorname{Fin} 3$$ and $$\sum_{r=0}^{4} \sum_{s=0}^{4} \overline{x_r} (B_j)_{r, s} y_s = 0 \quad \text{for all } j \in \operatorname{Fin} 5,$$ then $x = 0$ or $y = 0$. This is verified in Lean by the certificate r5hcond.

*Proof of Lemma 4.* The check uses a positivity certificate of the kind introduced in Proposition 5.1 of arXiv:2611.08518: a single constant identity between Hermitian $25\times 25$ matrices that is then evaluated at product vectors. Index the coordinates of $\mathbb{C}^{25}$ by pairs $(p,q) \in \{0,\dots,4\}^2$, and let $\Gamma$ denote the partial transpose in the first index. Besides $A$ and $B$, the certificate contains a Gaussian-integer $25 \times 14$ matrix $L$ and a Gaussian-integer $25 \times 25$ matrix $M$. Set $$H_{(p,q),(p',q')} = \sum_a (A_a)_{p,q'} \overline{(A_a)_{p',q}} + \sum_j \overline{(B_j)_{p,q}} (B_j)_{p',q'} - \sum_k L_{(p',q),k} \overline{L_{(p,q'),k}},$$ that is, $H = \Gamma(\mathbf{A}^*\mathbf{A}) + \mathbf{B}\mathbf{B}^* - \Gamma(LL^*)$ for the corresponding block matrices. Lean verifies the exact identity $$16\,H = 3200\, I + M M^* + E,$$ where $E$ is defined by this equation, and a kernel computation over the Gaussian integers confirms that every row and every column of $E$ has entry sum of absolute real and imaginary parts at most $1220$.

Evaluate the Hermitian form of $H$ at the vector $v$ with coordinates $v_{p,q} = \overline{x_p} y_q$. Expanding the three sums gives $$v^* H v = \sum_a \Big|\sum_{r,s} x_r (A_a)_{r,s} y_s\Big|^2 + \sum_j \Big|\sum_{r,s} \overline{x_r} (B_j)_{r,s} y_s\Big|^2 - \sum_k \Big|\sum_{r,s} \overline{x_r}\, \overline{y_s}\, L_{(r,s),k}\Big|^2.$$ Under the hypotheses of the lemma the first two sums vanish, so $\operatorname{Re}(v^* H v) \le 0$. On the other side, $M M^*$ contributes a sum of squares, and the row and column bounds on $E$ give $\operatorname{Re}(v^* E v) \ge -1220\,\|v\|^2$. Hence $0 \ge 16\operatorname{Re}(v^* H v) \ge (3200 - 1220)\,\|v\|^2$, so $v = 0$. Thus $\overline{x_p} y_q = 0$ for all $p,q$, that is, $x = 0$ or $y = 0$.

*Proof of the Main Theorem.* The main theorem range_criterion_subspaces_sporadic_five_five_three_five is proven by directly invoking has_range_criterion_of_matrix_pencil with dimension parameters $m = 5$, $n = 5$, $k = 3$, and $\ell = 5$, supplying the matrix families from Lemma 2, and providing the certificates r5li_A, r5li_B, and r5hcond. This completes the verification in Lean 4.

5. Discussion

The theorem range_criterion_subspaces_sporadic_five_five_three_five resolves the range-criterion conjecture from prior literature [4] for the sporadic quadruple $(m, n, k, \ell) = (5, 5, 3, 5)$. In arXiv:2611.08518, the conjecture had been established for all infinite asymmetric families in case (d) when $m = 3$ and $m = 4$. However, the sporadic integral zeros occurring at $m \ge 5$ were left unresolved, with $(5, 5, 3, 5)$ representing the minimal such instance. The formal verification in Lean 4 confirms that range-criterion subspaces exist for this quadruple, providing an affirmative answer rather than a counterexample.

It is important to note what this result does not cover. The proof presented here is tailored to the specific parameter quadruple $(5, 5, 3, 5)$. It does not resolve the existence of range-criterion subspaces for higher sporadic Krawtchouk zeros with $m \ge 5$, nor does it furnish a closed-form parametric construction applicable to all sporadic cases simultaneously. The certificate was found by a numerical search for a positive-definite certificate, following Remark 5.2 of arXiv:2611.08518 (an extra column of $\mathbf{B}$ forced to zero while the margin stays positive), then rounded to Gaussian integers; only the exact identity is checked.

Several open directions emerge naturally from this result and the problem entry in arXiv:2611.08518:

  • The primary open question is whether range-criterion subspaces exist for all remaining sporadic quadruples $(m, n, k, \ell)$ in case (d) with $m \ge 5$. While the minimal case $(5, 5, 3, 5)$ is now settled affirmatively, the status of larger sporadic parameters remains open.
  • Another question concerns the nature of the certificates: does every sporadic quadruple admitting range-criterion subspaces possess a certificate defined over an algebraic number ring such as the Gaussian integers $\mathbb{Z}[i]$, or do higher cases require more general fields or geometric arguments?
  • A third question is whether this certificate search and its exact verification can be systematized into an algorithmic pipeline capable of automatically certifying larger sporadic instances as formal computation tools scale.

Why it is research-level

  • Settles (5,5,3,5), the smallest sporadic Krawtchouk zero that arXiv:2611.08518 leaves open, by an explicit Gaussian-integer certificate: the first sporadic case of the range-criterion conjecture (m = 5)

Lean declarations

Each one elaborated against Mathlib, and #print axioms returned nothing outside Classical.choice · Quot.sound · propext.

References

  1. arXiv:2611.08518 — Integral zeros of Krawtchouk polynomials and quantum entanglements

Drafted by gemini-3.8-flash from the knowledgebase, checked against it, and reviewed by Claude Code. Only the Lean is verified; the prose is a reading of it.