Skip to content

Determine the sharp odd denominator for noise-optimal four-qubit Clifford decoding - #7802

Open
apoursh wants to merge 1 commit into
CrocSwap:mainfrom
apoursh:codex/clifford-dyadic-noise-floor
Open

apoursh wants to merge 1 commit into
CrocSwap:mainfrom
apoursh:codex/clifford-dyadic-noise-floor

Conversation

@apoursh

@apoursh apoursh commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

For the normalized four-qubit Clifford tensor map at width 13, the least common odd denominator among all exact Gaussian-rational left inverses attaining the optimal Euclidean operator norm is 470,624,175. At every width f ≥ 13 the required odd denominator is the reduced numerator of the smallest Gram eigenvalue. With dyadic coefficients on a 2⁻ˢ grid, the optimal excess in squared noise norm is Θ_f(4⁻ˢ).

A primitive vector in the least eigenspace proves necessity. One correction using the reciprocal least eigenvalue, followed by dyadic corrections supported away from the bottom eigenspace, proves attainment. The proof also shows why both Gram endpoints carry from width 4k−1 to 4k. It builds on the splitting in #7799, Ali Romman's spectra and canonical obstruction in #7770, #7772, #7788, #7789 and #7792, and Nick Petalas's projection data from #7747.

The replay checks 81 width-13 levels, 1,701 projected neutral-sector equalities, exact precision-grid examples, and independent dense calculations on the 24-element one-qubit alphabet. The constants in the precision result depend on the fixed matrix. A practical compact norm-optimal decoder and the exact denominator of the canonical inverse itself remain open.

Canonical base e54523517d0f03e58352e647c403a5af3e525a57. All four registered commands passed under CPython 3.12.14, both locally and from the read-only, Git-free export of the published commit. The package passed the repository’s committed-byte, file-mode, source-history and export checks.

python3.12 -B scripts/research_package.py --package research/clifford-dyadic-noise-floor

Prepared by apoursh with substantial OpenAI Codex assistance. Maintainer scientific review is open. The selected conditional multiplication exponent is unchanged. Proof at the submitted revision · Research claim

Hosted validation. The shared onboarding test failed before package selection. The CI guide still references #6030 while the selected record requires make frontier7000-verify. The identical failure occurs with the upstream test and guide. The package-specific checks above passed locally and on the exported commit. Hosted job · Existing navigation repair #7276

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant