arXiv:2609.20780
Asymptotically Good Quantum Locally Testable Codes
William Gay, Fernando Granha Jeronimo
Abstract
AI Summary
The paper addresses whether quantum local testability can coexist with constant rate, linear distance, and bounded check weight and qubit incidence. It constructs families of binary CSS codes with $N\to\infty$, parameters $[[N,\Omega(N),\Omega(N)]]$, and constant-soundness testers. Soundness uses quantum Hamming distance: the normalized check Hamiltonian satisfies $\mathsf H\succeq(\sigma/N)\mathsf D_{\mathscr Q}$ for a constant $\sigma>0$. The construction uses the middle degree of a four-dimensional cubical complex, extending the Dinur–Lin–Vidick framework. Arithmetic quotients of products of trees provide Ramanujan bounds for parallel-face walks without losses depending on the quotient size. Equivariant projective Reed–Solomon codes supply compatible incidence maps despite noncommuting geometric labels. A local-code theorem establishes uniform product expansion on unequal norm-one evaluation sets, including dual codes and coefficient-field extensions. Product expansion gives local filling inequalities; face-walk mixing then yields cohomological distance and syndrome-repair bounds. A double-complex argument transfers these estimates to homology, while an Euler-characteristic calculation supplies positive rate. Restriction of scalars produces qubit codes with constant parameter losses. A fixed instantiation has rate at least $2^{-5}$, relative distance at least $2^{-2^{18}}$, soundness at least $2^{-2^{20}}$, and check weight and qubit incidence at most $2^{2^{17}}$. Thus the stated guarantees are asymptotic, with very large locality bounds and very small distance and soundness bounds. The paper explicitly distinguishes this result from the quantum PCP conjecture: good qLTCs alone do not provide the reductions that conjecture requires.
ChatGPT 6 Pro was used as an editorial assistant.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.