Liu–Siemons conjecture on walk-matrix equivalence

For any two graphs GG and HH of order nn, with adjacency matrices AGA_G and AHA_H, let e\mathbf e be the all-one vector and define WG=[e,AGe,…,AGn−1e]W_G=[\mathbf e,A_G\mathbf e,\ldots,A_G^{n-1}\mathbf e] and WH=[e,AHe,…,AHn−1e]W_H=[\mathbf e,A_H\mathbf e,\ldots,A_H^{n-1}\mathbf e]. The Liu–Siemons conjecture asserts that WG=WHW_G=W_H implies that GG and HH are isomorphic.

References

Primary source

arXiv

Additional references

Progress summary

Refreshed
Claimed solved

A new unrefereed preprint claims the conjecture is false and gives counterexamples in every sufficiently large size, but the claim has not been independently checked.

The Liu–Siemons conjecture concerns when graphs with equivalent walk matrices must be structurally equivalent. Chaochao Zhu and Qin Yue claim a complete disproof and replacement by a switching classification.

Known results

  • If a graph’s walk matrix has rank at least n−1n-1, its adjacency matrix is determined by that walk matrix; in this range, equality corresponds to graph isomorphism.

Recent preprint

Chaochao Zhu and Qin Yue claim a complete characterization at corank two via reciprocal WQH switching, no examples through order 99, and connected counterexamples for every n≥10n \ge 10. The source is an unrefereed preprint, so this disproof and classification remain unverified.

Current status (as of September 2026): The conjecture is claimed false, with a claimed structural classification and connected examples for every n≥10n \ge 10, but the preprint’s results remain unverified.

Sources

Solutions 0

No solutions have been posted yet.