We investigate the relationship between the lexicographic product of graphs and their multi-word-representation number. Although the lexicographic product of two word-representable graphs need not itself be word-representable, a precise characterization has not previously been established. We provide a complete characterization, showing that for word-representable graphs $G_1$ and $G_2$, the lexicographic product $G_1 \circ G_2$ is word-representable if and only if $G_2$ is a comparability graph. For lexicographic powers, we prove that $G^{[k]}$ is word-representable if and only if $G$ is a comparability graph. The multi-word-representation number $μ$ for lexicographic powers and products satisfies the following bounds. If $G$ is a non-comparability graph, then $μ(G^{[k]}) \le k$, whereas if $G$ is the union of two comparability graphs, then $μ(G^{[k]}) = 2$. More generally, for graphs $G_1$ and $G_2$ with $μ(G_1) = k_1$ and $μ(G_2) = k_2$, the lexicographic product $H = G_1 \circ G_2$ satisfies the upper bound $μ(H) \le k_1 + k_2$. This bound is tight, with equality $μ(H) = k_1$, when $k_1 \ge k_2$ and $G_2$ is the union of $k_1$ comparability graphs. Moreover, if $G_1$ and $G_2$ are minimal non-word-representable graphs, then $μ(G_1 \circ G_2) \le 3$. Finally, we study the function $τ(n)$, which measures the size of the largest word-representable induced subgraph guaranteed in every $n$-vertex graph. By constructing extremal graphs via lexicographic powers, we establish a sublinear upper bound, showing that $τ(n) \le n^{0.86}$ for sufficiently large $n$.

翻译：暂无翻译