Preprint disproves Bapat-Sunder eigenvalue conjecture over the reals with 32-by-32 counterexample
A preprint by Logan Chalmers disproves the Bapat-Sunder eigenvalue conjecture over the real numbers with a rank-4, 32-by-32 integer counterexample. The result closes a question posed in 1986 and, together with an earlier complex counterexample, settles the conjecture negatively over both fields. It also connects the failure to a 36-mode quantum-optics interferometer involving 32 photons.
A preprint by Logan Chalmers says the Bapat-Sunder eigenvalue conjecture fails over the real numbers. The 1986 question asked whether the permanent of a positive semidefinite Hermitian matrix, which has real eigenvalues and appears in covariance and quantum descriptions, is always the largest eigenvalue of a related matrix F_A built from minors. After Drury gave a complex counterexample in 2018, he conjectured the claim still held for real positive semidefinite matrices. Chalmers reports a rank-4, 32-by-32 integer Gram matrix and an integer vector whose Rayleigh quotient exceeds the permanent, disproving that final real case.
The evidence is a constructed counterexample rather than an experiment. Chalmers gives a rank-4 Gram matrix, built from inner products, with integer entries and an integer test vector; its Rayleigh quotient, a standard eigenvalue test, is larger than the matrix permanent, the exact inequality the conjecture forbids. The paper says the verification uses integer arithmetic, avoiding floating-point uncertainty. Together with Drury's earlier complex counterexample, the result settles the Bapat-Sunder question negatively over both the real and complex fields. It also states that counterexamples exist for every matrix order n at least 32.
The conjecture sat at the junction of matrix theory and combinatorics, where permanents count weighted permutations and eigenvalues summarize a matrix's behavior. Long-standing conjectures often become trusted scaffolding for proofs, algorithms and bounds; a counterexample forces those assumptions to be rechecked. The paper's quantum-optics application shows the same algebra can describe interference in multi-photon systems, linking an abstract inequality to probabilities in a 36-mode interferometer. The result is a preprint, so the mathematical community has not yet peer-reviewed the construction.
The finding matters first to mathematicians working on matrix inequalities, because a hypothesis open since 1986 is now false. It also matters to researchers using linear algebra in computation and quantum information, where permanents and eigenvalues appear in complexity and interference calculations. In the paper's example, a 36-mode orthogonal interferometer with 32 photons in identical-variance Gaussian wave packets has a delay direction in which every sufficiently small nonzero delay raises the chance that all photons are detected in four designated output modes.
Mathematicians and computational researchers lose a long-held assumption about matrix eigenvalues and permanents, while quantum-optics researchers gain a concrete interferometer setup where tiny photon delays change multi-photon detection probabilities.
Sources (1)
- [1]Original source(https://arxiv.org/abs/2611.23269)