Brown–Colbourn conjecture on all-terminal reliability roots

About 9 years old · traced to

Let G=(V,E)G=(V,E) be a connected graph, and let Rel(G;q)\mathrm{Rel}(G;q) denote its all-terminal reliability polynomial in the edge-failure probability qq. A complex number zz is an all-terminal reliability root if Rel(G;z)=0\mathrm{Rel}(G;z)=0. Brown–Colbourn conjecture. If zz is a root of Rel(G;q)\mathrm{Rel}(G;q), then

∣z∣≤1.|z|\leq 1.

Equivalently, all all-terminal reliability roots lie in the unit disk. The conjecture is false in general, although it holds for series-parallel graphs; counterexamples were found by Sokal and Royle.

References

Primary source

Jason Brown and Lucas Mol, “On the roots of all-terminal reliability polynomials”, arXiv:1703.10566 (2017).

Progress summary

Refreshed
Claimed solved

Published counterexamples show that the conjecture is false, although it remains true for series-parallel graphs.

Jason Brown and Charles Colbourn formulated the conjecture in 1992. It asserted that all-terminal reliability roots lie in the unit disk; this assertion is now known to fail for general graphs.

Known results

  • Wagner proved the conjecture for series-parallel graphs.
  • Royle and Sokal reported both univariate and multivariate counterexamples; the smallest cited example comes from K4K_4 with two opposite edges replaced by bundles of six edges, producing a root of modulus approximately 1.00171.0017.
  • Brown and collaborators later found simple-graph roots with larger modulus.
  • Buys proved in 2026 that reliability roots of connected simple graphs are dense in the unit disk, while boundedness for connected multigraphs remains open.

2004 counterexamples; 2017 quantitative strengthening

Royle and Sokal's paper, revised April 2, 2004, established that the original conjecture is false. The 2017 paper On the roots of all-terminal reliability polynomials reported a stronger upper-bound framework and examples whose root modulus was nearly three times farther outside the unit disk than previously known examples.

Current status (as of September 2026): The conjecture for general connected graphs is disproved, and the series-parallel case is settled positively; related questions, including boundedness for connected multigraphs, remain open.

Sources

Solutions 0

No solutions have been posted yet.