S M Nazmuz Sakib Entanglement Rank Entropy: A Graph-Theoretic Invariant of Bipartite Entanglement Heterogeneity
S M Nazmuz Sakib
Abstract
Graph states provide a bridge between graph theory and multipartite quantumentanglement, with cut-rank, rank-width, and related parameters playing a centralrole in both structural graph theory and the characterization of stabilizer states.Building on this interplay, we introduce a new scalar invariant—the S M NazmuzSakib entanglement rank entropy—defined as the Shannon entropy of the cut-rankdistribution over vertex subsets of fixed cardinality. For a simple graph G = (V, E)of order n, and k ∈ {1, . . . , n − 1}, we consider the random variable ρG(X) whereX ⊆ V is a uniformly random k-subset, and we define the Sakib entanglementrank entropy H(k)ρ (G) as the Shannon entropy (in bits) of its distribution.Forgraph states, where ρG(X) coincides with the Schmidt rank exponent across thebipartition (X, V \ X), H(k)ρ (G) is precisely the Shannon entropy of the bipartiteentanglement (in ebits) seen by random size-k bipartitions.We formally define the S M Nazmuz Sakib entanglement rank entropy, provethat it is invariant under local complementation and hence a local-Clifford invariantfor graph states, and study a balanced version Hbalρ (G) = H(n/2)ρ(G) (for even n).We then perform a data-based analysis on explicit graph families (paths, cycles,stars, complete graphs, and 2 × m grid graphs) of size up to n = 12, computingHbalρ (G), the average cut-rank, and the full cut-rank distributions. The resultingdataset shows that, while complete and star graphs have vanishing Sakib entangle-ment rank entropy, more “distributed” structures such as paths, cycles, and gridsexhibit increasing entropy with system size, despite similar ranges of cut-rank.Motivated by these observations, we formulate the S M Nazmuz Sakib entan-glement heterogeneity hypothesis, proposing that, at fixed rank-width and averagecut-rank, a larger Sakib entanglement rank entropy correlates with higher classicalsimulation complexity of the corresponding graph states.We outline how pub-licly available graph-state orbit datasets and average cut-rank techniques can becombined with the proposed invariant to further test this hypothesis.