Source-linked AI summary
An integer determinant approach to the length of Type II codes and the weights of shadows
Jae-Hyun Baek, Jon-Lark Kim
TL;DR
The paper addresses how to prove that binary Type II code lengths are divisible by 8 without relying on Gleason’s theorem. It uses an integer Gram-matrix determinant congruence and gives two character-theoretic proofs. These arguments yield the length restriction and the shadow-weight congruence, with the Arf invariant identified as the obstruction.
Problem
The classical proof that binary Type II code lengths are divisible by 8 uses Gleason’s theorem, which determines more about weight enumerators than this restriction requires.
Source: §1 · p. 2Method
An integer Gram-matrix determinant congruence modulo 8 proves a row-weight theorem that yields the length and shadow results; two character-theoretic proofs provide alternatives.
Source: §1 · p. 2Results
Takeaways & Limitations
The arguments identify the Arf invariant of the quadratic form on even-weight vectors as the shared obstruction.
Source: §1 · p. 2In the determinant proof, Step 0 cannot be dropped: k≡3 mod 4 also satisfies 3k+2k≡1 mod 8.
Source: §2 · p. 5
Abstract
from arXiv · showhide
It is well known that the length $n$ of a binary Type II code is divisible by $8$. The classical proof relies on Gleason's theorem on weight enumerators. Motivated by a question of Vera Pless, we give elementary proofs of this fact that avoid invariant theory. Our main tool is a congruence modulo $8$ for the determinant of the integer Gram matrix $AA^{\mathsf T}$ of a binary orthogonal matrix $A$, obtained directly from the Leibniz expansion. It gives a short proof of a theorem of Janusz: the number of rows of weight congruent to $3 \pmod 4$ in a binary orthogonal matrix is divisible by $4$. From this single statement we derive both the length restriction for Type II codes and the theorem of Conway and Sloane that every vector in the shadow of a binary self-dual code of length $n$ has weight congruent to $n/2 \pmod 4$. We also give two short character-theoretic proofs of the length restriction and explain how all of these arguments detect the Arf invariant of a quadratic form.
1 Introduction
The paper replaces Gleason-based proofs of the Type II length restriction with an elementary route motivated by Vera Pless’s question. It connects a determinant congruence and Janusz’s row-count theorem to the length restriction and the shadow-weight theorem, alongside character-theoretic proofs and an Arf-invariant interpretation.
1.2 Notation and preliminaries
- The preliminaries define self-duality and doubly-evenness, and note that a Type II code contains the all-ones vector.Self-duality gives dim C=n/2; even weights imply 1∈C⊥=C.Source: §1 · p. 1, §1.2 · p. 3
1 Introduction
- A binary Type II code exists exactly at lengths divisible by 8, making this the basic arithmetic constraint in the theory.Direct sums of n/8 copies of the extended Hamming code establish existence at every such length.Source: §1 · p. 1
- The standard proof uses Gleason’s theorem [6] to determine the full ring of possible weight enumerators, more than the length restriction requires.The elementary approach is motivated by Vera Pless’s question about proving the restriction without that machinery.Source: §1 · p. 2
- An integer Gram-matrix determinant congruence modulo 8 proves Janusz’s result on rows of weight 3 mod 4 and yields both the Type II length restriction and the shadow-weight theorem [4].Every vector in the shadow has weight congruent to n/2 mod 4.Source: §1 · p. 2
- Two character-theoretic proofs use a Gauss sum and the MacWilliams identity; the arguments detect the Arf invariant of wt(x)/2 mod 2 on even-weight vectors.The paper presents these alongside an elementary, unified route to the results.Source: §1 · p. 2
1.1 Related work
- Earlier approaches include invariant theory [11], spinor groups and Witt–Arf theory [13], and Janusz’s group-order computation [8]; the determinant computation gives a different proof of his known result.The paper also notes a related determinant-modulo-8 computation for a Z4-lifting problem, which addresses a different question.Source: §1.1 · pp. 2–3
2 First proof: an integer Gram determinant
The proof reduces a Type II code to a binary orthogonal matrix and uses its integer Gram matrix modulo 8. This determinant argument proves the length restriction and yields Janusz’s divisibility result for row weights.
Integer Gram matrix
- The proof treats A as a 0/1 integer matrix and sets B = AA^T, whose diagonal entries are odd and off-diagonal entries even.Self-duality gives A A^T = I_k over F_2; the diagonal entries of B are row weights.Source: §2 · p. 3
Determinant modulo 8
- Modulo 8, each moved index contributes an even off-diagonal factor, so terms whose permutations move at least three indices are divisible by 8 and vanish; only identity and transpositions remain.Source: §2 · p. 4
Congruence and consequences
- The total entry sum of B is evaluated from the odd column weights of A, then compared with the determinant congruence to obtain the row-count divisibility.For Type II codes, the resulting condition gives 4 | k and hence 8 | n; for general A in O_k, it gives 4 | a.Source: §2 · pp. 4–5
First proof
- 8 | n for a binary Type II code, by reducing its generator matrix [I_k | A] to rows of A of weight 3 mod 4 and showing 4 | k.The doubly-even condition makes each generator row’s weight 1 + wt(r_i) divisible by 4, and k = n/2.Source: §2 · pp. 3–4
Janusz’s theorem
- For A in O_k, the number of rows of weight 3 mod 4 is a multiple of 4, recovering Janusz’s result by a determinant computation.The paper presents this as a new proof of a known result, not a new theorem.Source: §2 · p. 5
3 Second proof: a quadratic form and a Gauss sum
A quadratic form on the even-weight vectors converts the doubly-even condition into a question about a Gauss sum. A totally singular Lagrangian makes the form’s Arf invariant zero, yielding the length restriction.
Quadratic form
- The even-weight vectors carry a quadratic form q(x) = wt(x)/2 mod 2, with the standard inner product as its polar form.The doubly-even condition is exactly q|_C = 0.Source: §3 · p. 6
Arf invariant
- The quotient C/⟨1⟩ is a totally singular Lagrangian, so the corresponding Gauss sum is positive and the Arf invariant is zero.The form descends to E/⟨1⟩, a nondegenerate space of dimension n − 2.Source: §3 · p. 7
Length restriction
- For n divisible by 4, the Arf invariant of the form on E is n/4 mod 2; its vanishing forces 8 | n.The proof uses q(x + 1) = q(x) to relate the sum over E to the quotient-space sum.Source: §3 · p. 7
4 Third proof: the MacWilliams identity at one point
The third proof evaluates the MacWilliams identity at the single point (1, i), rather than using the full weight-enumerator framework. The resulting phase condition forces the Type II length to be divisible by 8.
Single-point evaluation
- Evaluating the MacWilliams identity at (1, i) reduces the weight-enumerator relation to an equality involving (1 + i)^n and 2^(n/2).Every codeword has weight divisible by 4, so its contribution at the selected point is 1.Source: §4 · p. 7
Consequence
- The equality (1 + i)^n = 2^(n/2) forces the phase e^(inπ/4) to equal 1, hence 8 | n.This proves the Type II length restriction.Source: §4 · p. 7
5 An application: the shadow of a self-dual code
For a binary self-dual code, an explicit coset representative describes its shadow. The proof shows every shadow vector has weight congruent to n/2 mod 4, and the residue also yields Type II neighbors in the stated setting.
Shadow weight theorem
- Every vector in the shadow of a binary self-dual code of length n has weight congruent to n/2 mod 4.The proof uses an explicit shadow representative and Proposition 2.2 rather than the usual shadow weight enumerator.Source: §5 · p. 8
Explicit representative
- The representative s is constructed from row-weight residues, and its weight is k − a, where a counts rows of A of weight 3 mod 4.Proposition 2.2 gives 4 | a, so wt(s) ≡ k mod 4.Source: §5 · p. 8
All shadow vectors
- For each c in C, wt(s + c) ≡ wt(s) mod 4, extending the representative’s residue to every shadow vector.The proof uses |s ∩ c| mod 2 = s · c = q(c).Source: §5 · p. 8
Type II neighbors
- When C is not doubly-even and n is divisible by 8, the two codes C0 + ⟨s⟩ and C0 + ⟨s + c⟩ are distinct Type II codes of length n.Here c is any element of C \ C0.Source: §5 · p. 8
6 Concluding remarks
The concluding remarks contrast the determinant proof with the character-theoretic proofs and connect the length restriction to the Arf invariant. They also relate the restriction to even unimodular lattices.
6 Concluding remarks
- The first proof replaces the computation of with a determinant congruence and uses neither characters nor weight enumerators.Its modulo-8 information comes from det B being an odd square.Source: §6 · p. 9
- The length restriction is detected directly by the first proof’s integer determinant congruence, while the second interprets it through the Arf invariant.The obstruction is the Arf invariant of q on E/⟨1⟩, which must vanish for a Type II code to exist.Source: §6 · p. 9
- The second and third proofs are essentially the same computation, with and without the weight enumerator, and the third uses T1 and T2 once each at one point.The scalar (T1T2)^3 = e^(iπ/4)I and homogeneity of degree n force e^(inπ/4) = 1.Source: §6 · p. 9
- Under Construction A, a Type II code of length n yields a positive definite even unimodular lattice of rank n, making the length restriction its coding analogue.The rank of an even unimodular lattice is divisible by 8.Source: §6 · p. 9
Use of Generative-AI tools declaration
The authors disclose using Claude Opus 5.5 to help draft parts of the text and proofs and perform numerical checks; they state that they verified the mathematical contents and take responsibility for them.
Use of Generative-AI tools declaration
- Claude Opus 5.5 was used to help draft parts of the text and proofs and perform numerical checks; the authors state they verified the mathematical contents and take full responsibility.Source: Use of Generative-AI tools declaration · p. 9