Crouzeix conjecture

For every integer n1n\ge 1, every matrix ACn×nA\in\mathbb{C}^{n\times n}, and every polynomial pC[z]p\in\mathbb{C}[z], one has p(A)22supzW(A)p(z)\|p(A)\|_2\le 2\sup_{z\in W(A)}|p(z)|, where W(A)={xAx:xCn, x2=1}W(A)=\{x^*Ax:x\in\mathbb{C}^n,\ \|x\|_2=1\} is the numerical range of AA.

Progress summary

Solved

A newly posted manuscript claims to prove the conjecture, but independent verification has not yet appeared.

Crouzeix's conjecture asserts that every square complex matrix AA and polynomial pp satisfy p(A)2supzW(A)p(z)\|p(A)\|\le 2\sup_{z\in W(A)}|p(z)|. It is attributed to Michel Crouzeix and has remained open in its general form.

Known results

  • Crouzeix proved the general bound 11.0811.08.
  • Crouzeix and Palencia improved it to 1+21+\sqrt{2}.
  • The conjecture is proved for normal matrices and all 2×22\times2 matrices.
  • Further proofs cover selected tridiagonal, nilpotent, shift-compression, and spectrally separated classes.

August 2026 claimed proof

Jin Shanmu's GitHub repository provides a standalone LaTeX proof and reports audit results. Public descriptions attribute the result to a 1616-hour autonomous run of GPT-5.6 Sol; an arXiv paper cites Jin's updated preprint as a solution. The claim remains unverified because the proof has no formal peer-review confirmation or independent verification in the retrieved sources.

Current status (as of August 2026): A claimed general proof exists and is attributed to Jin Shanmu with GPT-5.6 Sol, but the conjecture remains mathematically unconfirmed pending independent verification.

Sources
Sources & referencesView supporting material

Primary source

GitHub

Additional references

Solutions 0

No solutions have been posted yet.