A team of researchers has shown that when random regular graphs are combined with a fixed gadget graph using the zig‑zag product, the spectrum of the new graph settles to a predictable limit. The limiting spectrum is exactly that of the zig‑zag product of the infinite regular tree with the gadget, and the largest eigenvalue (the spectral radius) approaches the sharp threshold identified by Cohen, Cohen, and Maor.

The proof relies on concepts from free probability and operator theory. By modelling the random permutations that define the graph product as random unitary tuples, the authors prove strong convergence of these tuples. They then use matrix‑valued polynomial techniques to show that the spectra of the finite graphs converge in Hausdorff distance to the limiting spectrum.

A concrete quantitative result follows. In the independent permutations model, the probability that the spectral radius of the new graph exceeds the limiting value Λ_H by an amount ε drops like O(1/(n ε⁹)), where n is the number of vertices. This upper‑tail bound uses estimates from recent work on concentration of measure for random matrices.

The findings also reveal limitations of common benchmarks. For graphs whose degree distribution is not constant, the fourth moment of the limiting spectral measure exceeds that predicted by free‑probability models. Thus, the free‑probability benchmark does not describe typical random irregular configuration graphs.

Overall, the work provides a rigorous yardstick for spectral performance of randomly constructed zig‑zag products, with explicit error bounds that hold for any gadget graph and depend only on its size and degree.