Weak kk-metric dimension formula for Hamming graphs Kn□KmK_n\square K_m

About 1 year old · traced to

Let Kn□KmK_n\square K_m be the Cartesian product of complete graphs, with n≥3n\ge 3, m≥n+1m\ge n+1, and 3≤k≤2n3\le k\le 2n. The weak kk-metric dimension formula asserts that

wdim⁡k(Kn□Km)={m⌈k2⌉,if k is even,m⌈k2⌉−1,if k is odd.\operatorname{wdim}_{k}(K_n\square K_m)= \begin{cases} m\left\lceil\frac{k}{2}\right\rceil,& \text{if }k\text{ is even},\\[0.2cm] m\left\lceil\frac{k}{2}\right\rceil-1,& \text{if }k\text{ is odd}. \end{cases}

This conjecture is motivated by computational results, an integer-linear-programming bound, and formulas established for related parameter ranges; the source notes that a proof may require lengthy considerations similar to those used for the corresponding theorem.

References

Primary source

Elena Fernandez, Sandi Klavzar, Dorota Kuziak, Manuel Muñoz-Marquez and Ismael G. Yero, “On the weak k-metric dimension of Hamming graphs”, arXiv:2505.19642 (2025).

Progress summary

Refreshed
Claimed progress

The rectangular formula remains unproved in the literature, but a reader-submitted argument claims to establish it and has not been checked.

Fernández, Klavžar, Kuziak, Muñoz-Márquez, and Yero proposed the formula in 2025 for n≥3n\ge 3, m≥n+1m\ge n+1, and 3≤k≤2n3\le k\le 2n. Their paper presents it as a conjecture supported by computation and integer-linear programming.

Known results

  • The square case Kn□KnK_n\square K_n is proved for n≥3n\ge 3 and 2≤k≤2n2\le k\le 2n (Fernández et al., 2025).
  • For the rectangular case, the original paper gives computational and optimization evidence but no proof.

Community submission (unverified), August 21, 2026

A submitted proof argues that the stated formula holds throughout the rectangular range, using row and column layer counts and a construction covering k=3k=3. This is a substantive claim of a complete proof, but it has not been independently verified.

Current status (as of September 2026): The square case is settled, while the rectangular formula has only an unverified submitted proof and remains mathematically unconfirmed.

Sources

Solutions 1

ProofThis solution needs a summarySee full solutionHide full solution

The weak metric dimension of rectangular Hamming graphs

Let n≥3n\ge3, m≥n+1m\ge n+1, and 3≤k≤2n3\le k\le2n be integers. Then

wdim⁡k(Kn□Km)={m⌈k/2⌉,k even,m⌈k/2⌉−1,k odd.\begin{aligned} &\operatorname{wdim}_k(K_n\square K_m)\\ &\quad=\begin{cases} m\lceil k/2\rceil,&k\text{ even},\\ m\lceil k/2\rceil-1,&k\text{ odd}. \end{cases} \end{aligned}

This proves Conjecture 6.1 of Fernández, Klavžar, Kuziak, Muñoz-Márquez and Yero, On the weak kk-metric dimension of Hamming graphs. The layer-count identities used below are those of their Proposition 4.2; their Proposition 4.1 also shows that aligned pairs suffice when k≥4k\ge4. We give the identities explicitly and construct a set attaining the required lower bound, including when k=3k=3.

1. Distances and layer counts

Write the vertices of G=Kn□KmG=K_n\square K_m as (i,j)(i,j), where 0≤i<n0\le i<n and 0≤j<m0\le j<m. The distance between two vertices is the number of coordinates in which they differ. For a subset S⊆V(G)S\subseteq V(G), put

aij=1{(i,j)∈S},ri=∑j=0m−1aij,cj=∑i=0n−1aij.\begin{aligned} a_{ij}&=\mathbf1_{\{(i,j)\in S\}},\\ r_i&=\sum_{j=0}^{m-1}a_{ij},\\ c_j&=\sum_{i=0}^{n-1}a_{ij}. \end{aligned}

For distinct vertices u,vu,v, define

ΔS(u,v)=∑w∈S∣d(u,w)−d(v,w)∣.\Delta_S(u,v)=\sum_{w\in S}|d(u,w)-d(v,w)|.

By definition, SS is weak kk-resolving exactly when every such sum is at least kk.

If j≠ℓj\ne\ell, then

ΔS((i,j),(i,ℓ))=cj+cℓ.(1)\Delta_S((i,j),(i,\ell))=c_j+c_\ell. \tag{1}

Indeed, precisely the selected vertices in columns jj and ℓ\ell contribute, each by one. Similarly, for i≠hi\ne h,

ΔS((i,j),(h,j))=ri+rh.(2)\Delta_S((i,j),(h,j))=r_i+r_h. \tag{2}

When both coordinates differ, the exact identity is

ΔS((i,j),(h,ℓ))=ri+rh+cj+cℓ−2aiℓ−2ahj.(3)\begin{aligned} &\Delta_S((i,j),(h,\ell))\\ &\quad=r_i+r_h+c_j+c_\ell\\ &\qquad-2a_{i\ell}-2a_{hj}. \end{aligned} \tag{3}

To see this, the selected endpoints (i,j)(i,j) and (h,ℓ)(h,\ell) contribute two each. A selected vertex in exactly one of the four indicated layers contributes one. The two cross-corners (i,ℓ)(i,\ell) and (h,j)(h,j) contribute zero, although the sum of the four layer counts counts each twice. All remaining vertices contribute zero.

2. The lower bound

Put s=⌈k/2⌉s=\lceil k/2\rceil. Any weak kk-resolving set satisfies cj+cℓ≥kc_j+c_\ell\ge k for every two distinct columns, by (1). Let a=min⁡jcja=\min_j c_j.

If k=2sk=2s and a≥sa\ge s, then ∣S∣≥ms|S|\ge ms. If a<sa<s, every other column contains at least 2s−a2s-a selected vertices, so

∣S∣≥a+(m−1)(2s−a)=ms+(m−2)(s−a)≥ms.\begin{aligned} |S|&\ge a+(m-1)(2s-a)\\ &=ms+(m-2)(s-a)\ge ms. \end{aligned}

If k=2s−1k=2s-1 and a≥sa\ge s, then again ∣S∣≥ms|S|\ge ms. Otherwise write a=s−1−ba=s-1-b with an integer b≥0b\ge0. Every other column contains at least s+bs+b selected vertices. Hence

∣S∣≥s−1−b+(m−1)(s+b)=ms−1+(m−2)b≥ms−1.\begin{aligned} |S|&\ge s-1-b+(m-1)(s+b)\\ &=ms-1+(m-2)b\ge ms-1. \end{aligned}

Thus both asserted values are lower bounds.

3. A cyclic construction attaining the bound

Choose column sizes q0,…,qm−1q_0,\ldots,q_{m-1} as follows:

qj={s−1,k odd and j=0,s,otherwise.q_j= \begin{cases} s-1,&k\text{ odd and }j=0,\\ s,&\text{otherwise}. \end{cases}

Define cumulative offsets and their total by

T0=0,Tj+1=Tj+qj,L=Tm.\begin{gathered} T_0=0,\qquad T_{j+1}=T_j+q_j,\\ L=T_m. \end{gathered}

In column jj, select the rows with residues

Tj,Tj+1,…,Tj+1−1(modn).(4)T_j,T_j+1,\ldots,T_{j+1}-1\pmod n. \tag{4}

This defines an actual subset of the graph: since 2≤s≤n2\le s\le n, each column size is between 11 and nn, so no column repeats a row. Its column counts are exactly cj=qjc_j=q_j, and its cardinality is L=msL=ms for even kk and L=ms−1L=ms-1 for odd kk.

Across all columns, the integers used in (4) are exactly 0,1,…,L−10,1,\ldots,L-1. Each row therefore occurs either ⌊L/n⌋\lfloor L/n\rfloor or ⌈L/n⌉\lceil L/n\rceil times. Moreover, m≥n+1m\ge n+1 gives L≥nsL\ge ns: this is immediate for even kk, while for odd kk,

L=ms−1≥(n+1)s−1≥ns.\begin{aligned} L=ms-1&\ge(n+1)s-1\\ &\ge ns. \end{aligned}

Consequently every row satisfies ri≥s≥2r_i\ge s\ge2. Every pair of columns satisfies cj+cℓ≥kc_j+c_\ell\ge k, since all column sizes are ss, except possibly one of size s−1s-1 when k=2s−1k=2s-1.

Equations (1) and (2) now give the required inequality for pairs sharing a coordinate: column pairs sum to at least kk, and row pairs sum to at least 2s≥k2s\ge k. For pairs differing in both coordinates, (3) gives

ΔS((i,j),(h,ℓ))≥4+k−4=k,\begin{aligned} &\Delta_S((i,j),(h,\ell))\\ &\quad\ge4+k-4=k, \end{aligned}

because the two rows contain at least four selected vertices in total, the two columns contain at least kk, and each cross-corner indicator is at most one. Thus the constructed set is weak kk-resolving for every stipulated parameter, including k=3k=3 and k=2nk=2n. Its cardinality equals the lower bound, proving the formula. □\square