Minimal Supports in Hamming Graphs

• Minimal Supports are the smallest subsets where nonzero eigenfunctions exist in Hamming graphs, quantified as 2(q-1)q^(n-2) for λ1 when q>2.
• The study employs coordinate-restriction and additive function reduction techniques to rigorously classify extremal eigenfunctions as signed differences of coordinate-symbol slices.
• This result refines previous bounds and links spectral graph theory with coding theory by providing both precise metrics and a structural understanding of eigenfunctions.

The study of minimal supports in Hamming graphs concerns an extremal question about eigenfunctions of the graph $H(n,q)$: for the eigenvalue $\lambda_1=n(q-1)-q$, what is the minimum possible size of the support of a nonzero eigenfunction, and which eigenfunctions attain that minimum? The paper "Minimal supports of eigenfunctions of Hamming graphs" gives a sharp answer for $q>2$: every nonzero eigenfunction for $\lambda_1$ has support of size at least $2(q-1)q^{n-2}$, and equality holds exactly for signed differences of two coordinate-symbol slices (Valyuzhenich, 2015).

1. Hamming graphs, support, and the first nontrivial eigenspace

Let

$$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$

Thus the vertices of the Hamming graph $H(n,q)$ are $q$-ary words of length $n$. The Hamming distance between $x,y\in \Sigma_q^n$ is

$\lambda_1=n(q-1)-q$0

and the graph adjacency is

$\lambda_1=n(q-1)-q$1

Each vertex has degree $\lambda_1=n(q-1)-q$2 (Valyuzhenich, 2015).

If $\lambda_1=n(q-1)-q$3 is the adjacency matrix of $\lambda_1=n(q-1)-q$4, then a function

$\lambda_1=n(q-1)-q$5

is an eigenfunction with eigenvalue $\lambda_1=n(q-1)-q$6 when

$\lambda_1=n(q-1)-q$7

that is, for every vertex $\lambda_1=n(q-1)-q$8,

$\lambda_1=n(q-1)-q$9

The support of $q>2$0 is

$q>2$1

The spectrum recalled in the paper is

$q>2$2

The eigenvalue studied is the case $q>2$3: $q>2$4 This is the first nontrivial eigenspace below the top eigenvalue $q>2$5 (Valyuzhenich, 2015).

2. The extremal support problem and earlier bounds

The central problem is to determine the smallest possible support of a nonzero eigenfunction in the eigenspace for

$q>2$6

and to classify all extremal eigenfunctions attaining that minimum (Valyuzhenich, 2015).

The paper places this question against a previously known general lower bound: if $q>2$7 is an eigenfunction corresponding to $q>2$8, then

$q>2$9

For $\lambda_1$0, this yields only

$\lambda_1$1

The paper notes that this estimate is much weaker than the exact bound proved for $\lambda_1$2 (Valyuzhenich, 2015).

The discussion also notes related support bounds for $\lambda_1$3-valued eigenfunctions and perfect colorings from coding theory. This suggests that the support-minimization problem sits at an interface between spectral graph theory and coding-theoretic combinatorics, although the paper’s main theorem is formulated purely for eigenfunctions of Hamming graphs.

3. Exact minimum support for $\lambda_1$4

The principal result is Theorem 3. Let $\lambda_1$5, and let

$\lambda_1$6

be a nonzero eigenfunction corresponding to

$\lambda_1$7

Then

$\lambda_1$8

Moreover, equality holds if and only if

$\lambda_1$9

for some nonzero constant $2(q-1)q^{n-2}$0, some coordinates $2(q-1)q^{n-2}$1, and some symbols $2(q-1)q^{n-2}$2, where

$2(q-1)q^{n-2}$3

(Valyuzhenich, 2015).

Thus the minimum support size is exactly

$2(q-1)q^{n-2}$4

For an extremal function of this form, the support is the disjoint union of two sets:

• $2(q-1)q^{n-2}$5, of size $2(q-1)q^{n-2}$6,
• $2(q-1)q^{n-2}$7, also of size $2(q-1)q^{n-2}$8.

Hence

$2(q-1)q^{n-2}$9

(Valyuzhenich, 2015).

4. Extremal eigenfunctions as signed coordinate-slice differences

The paper isolates the extremal functions in the explicit form

$$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$0

Equivalently,

$$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$1

interpreted pointwise on $$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$2: if both equalities hold, the values cancel; if neither holds, the value is $$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$3; if exactly one holds, the value is $$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$4 (Valyuzhenich, 2015).

The paper describes these as essentially rank-1 coordinate functions: a difference of two one-coordinate indicator functions on different coordinates. Their support-minimality is therefore accompanied by a very rigid structural description.

The eigenvalue verification is direct. If $$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$5, then $$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$6. Among the $$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$7 neighbors of $$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$8,

• $$\Sigma_q=\{0,1,\dots,q-1\}, \qquad V(H(n,q))=\Sigma_q^n.$$9 remain in the same $H(n,q)$0 region,
• none lie in the $H(n,q)$1 region,
• the remaining neighbors lie where $H(n,q)$2.

Therefore

$H(n,q)$3

The case $H(n,q)$4 is symmetric. If $H(n,q)$5, then $H(n,q)$6 has exactly one $H(n,q)$7 neighbor and one $H(n,q)$8 neighbor, so the sum over neighbors is $H(n,q)$9. Hence these functions indeed lie in the eigenspace $q$0 (Valyuzhenich, 2015).

5. Reduction to additive functions

A central tool is a coordinate-restriction argument. For a function

$q$1

fix a coordinate $q$2 and symbols $q$3. For $q$4, let $q$5 and $q$6 be the words obtained from $q$7 by inserting $q$8 and $q$9, respectively, into the $n$0-th coordinate, and define

$n$1

Lemma 1 states that if $n$2 is an eigenfunction of $n$3 with eigenvalue $n$4, then $n$5 is an eigenfunction of $n$6 with eigenvalue

$n$7

In the case $n$8,

$n$9

which is exactly the top eigenvalue $x,y\in \Sigma_q^n$0. Since the eigenspace for $x,y\in \Sigma_q^n$1 is one-dimensional and consists of constants, Lemma 2 yields the following: if $x,y\in \Sigma_q^n$2 is an eigenfunction for $x,y\in \Sigma_q^n$3, then for every $x,y\in \Sigma_q^n$4, the function $x,y\in \Sigma_q^n$5 is constant (Valyuzhenich, 2015).

The paper calls such functions additive. Concretely, for every coordinate $x,y\in \Sigma_q^n$6, the difference between two slices $x,y\in \Sigma_q^n$7 and $x,y\in \Sigma_q^n$8 is constant across the remaining coordinates. This is a strong structural condition and is the key to the classification.

A plausible implication is that the first eigenspace for Hamming graphs is much more rigid than a general eigenspace with the same dimension count alone might suggest: support extremality is driven by an additive slice structure, not merely by cancellation in neighbor sums.

6. Proof architecture and mathematical significance

The main combinatorial work is first carried out for additive functions with sufficiently small support, rather than for eigenfunctions directly (Valyuzhenich, 2015). The base case $x,y\in \Sigma_q^n$9 is treated separately in Lemma 3, and the remainder of the argument uses the additive condition produced by the reduction lemmas.

From the information provided, the proof strategy has three visible stages:

1. Reduction: pass from an eigenfunction for $\lambda_1=n(q-1)-q$00 on $\lambda_1=n(q-1)-q$01 to constant slice-differences on $\lambda_1=n(q-1)-q$02.
2. Additive classification: analyze additive functions under a small-support hypothesis.
3. Reconstruction of extremals: conclude that equality in the support bound occurs exactly for signed differences of two coordinate-symbol slices.

The significance of the theorem is twofold. First, it replaces the earlier general estimate

$\lambda_1=n(q-1)-q$03

with the exact value

$\lambda_1=n(q-1)-q$04

for the eigenspace of $\lambda_1=n(q-1)-q$05 (Valyuzhenich, 2015). Second, it gives a complete description of all extremizers, not merely a lower bound.

This classification is especially sharp because equality is characterized up to the obvious parameters $\lambda_1=n(q-1)-q$06, $\lambda_1=n(q-1)-q$07, and $\lambda_1=n(q-1)-q$08. In that sense, the paper solves both the quantitative and structural versions of the extremal support problem for the first nontrivial eigenvalue.

7. Relation to spectral and coding-theoretic viewpoints

The Hamming graph $\lambda_1=n(q-1)-q$09 is a standard object in algebraic combinatorics and coding theory, and the paper explicitly notes related support bounds for $\lambda_1=n(q-1)-q$10-valued eigenfunctions and perfect colorings (Valyuzhenich, 2015). Within that landscape, the result identifies the exact smallest support for eigenfunctions in the eigenspace immediately below the constant eigenspace.

The extremal functions are supported on two coordinate-symbol slices with opposite signs, and vanish elsewhere. This gives them a particularly transparent combinatorial interpretation: they detect disagreement between two one-coordinate conditions, one positive and one negative. A plausible implication is that these extremizers are the most elementary nonconstant eigenfunctions available at the $\lambda_1=n(q-1)-q$11 level.

The paper therefore establishes that, for $\lambda_1=n(q-1)-q$12, the minimal-support problem for $\lambda_1=n(q-1)-q$13 has a uniquely rigid answer: the support minimum is

$\lambda_1=n(q-1)-q$14

and the only functions attaining it are exactly the signed differences of two coordinate-symbol slices (Valyuzhenich, 2015).