Perfecting the Line Graph
Abstract
This paper introduces two canonical constructions that transform arbitrary finite graphs into perfect graphs: the symmetric lift $\mathrm{HL}'_2(G)$, which is purely structural and label-invariant, and the ordered lift $\mathrm{HL}_2(G)$, which depends explicitly on vertex labeling and encodes directional information. Both lifts arise as line graphs of bipartite double covers and are box-perfect. The symmetric lift $\mathrm{HL}'_2(G)$ forms a canonical 2-cover of the line graph $L(G)$. This involution decomposes $\mathrm{HL}'_2(G)$ into symmetric and antisymmetric components: the symmetric part recovers $L(G)$, while the antisymmetric part yields a signed graph $L^-(G)$, the antisymmetric line graph, with +1/-1 edges encoding consistent vs. crossed overlaps. Thus, all adjacency and Laplacian eigenvalues of $L(G)$, with multiplicities, appear within those of $\mathrm{HL}'_2(G)$, despite $L(G)$ typically not being a subgraph. For regular graphs such as Paley graphs, this yields infinite families of sparse, highly structured regular and box-perfect expanders that also retain large cliques. The lift retains much of the spectral expansion of the base while improving the combinatorial expansion. Much the same behavior is observed with random regular base graphs, allowing for the possibility of the study of box-perfect random regular graphs. Finally, we generalize these constructions to parameterized lifts $\mathrm{HL}_{r,d}(G)$ and $\mathrm{HL}_{r,d}'(G)$ defined on ordered $r$-tuples connected by Hamming distance constraints, which structurally encode the base graph and remain box-perfect.