Permanent-of-Gaussians Conjecture

Unsolved ID op_af68b033a88ef934 Last edited 14 September 2026
Edit

Problem

Is the following Gaussian permanent estimation task \(\#\mathrm P\)-hard under randomized polynomial-time oracle reductions? For \(n\geq1\), let \(X\in\mathbb C^{n\times n}\) have independent entries with density \(\pi^{-1}e^{-|z|^2}\). Given \(0<\epsilon,\delta<1\), output \(z(X)\in\mathbb C\) satisfying

\begin{equation} \operatorname{Per}X=\sum_{\pi\in S_n}\prod_{j=1}^n X_{j,\pi(j)}, \qquad \Pr[|z(X)-\operatorname{Per}X|\leq\epsilon|\operatorname{Per}X|]\geq1-\delta . \tag{1} \end{equation}

Here \(S_n\) is the set of permutations of \(\{1,\ldots,n\}\). Probability in Eq. (1) includes the matrix and the estimator’s randomness. Cost is measured in the binary input length, \(1/\epsilon\), and \(1/\delta\). Use a binary truncation precise enough to alter the target by at most \(\epsilon|\operatorname{Per}X|/2\) except with probability \(\delta/2\). Polynomially many bits in \(n,\log(1/\epsilon),\log(1/\delta)\) per entry suffice.

Equivalently, ask for the same hardness of additive squared-permanent estimation:

\begin{equation} \Pr[|\widehat p-|\operatorname{Per}X|^2|\leq\epsilon n!]\geq1-\delta . \tag{2} \end{equation}

For Eq. (2), use a rounded, truncated matrix whose squared permanent differs from the original by at most \(\epsilon n!/2\) except with probability \(\delta/2\). The input precision and reduction cost obey the same polynomial bounds. Here \(\#\mathrm P\) counts accepting paths of nondeterministic polynomial-time machines.

Source

Aaronson and Arkhipov state the multiplicative Permanent-of-Gaussians Conjecture as Conjecture 5 in arXiv Section 1.2.3. Their Theorem 3 uses the additive squared-permanent task, and Theorem 7 relates the two under permanent anticoncentration [AA13]. The finite-precision convention makes the binary input model explicit.

Progress

  • 2010–2013: Aaronson–Arkhipov prove exact Gaussian-permanent \(\#\mathrm{P}\)-hardness at success fraction \(3/4+1/\operatorname{poly}(n)\), and hardness for estimating \(|\operatorname{Per}(X)|^2\) with exponentially small additive error (arXiv Theorems 60 and 62) [AA13].

  • 2021: Bouland–Fefferman–Landau–Liu obtain additive tolerance \(e^{-4n\log n-O(n)}\) for \(|\operatorname{Per}(X)|^2\), with any constant failure probability below \(1/4\), using \(\mathrm{BPP}^{\mathrm{NP}}\) reductions (proof of Corollary 1) [BFLL22].

  • 2023 (unpublished): Krovi’s tighter total-variation estimate improves this tolerance to \(e^{-2n\log n-O(n)}\), as reported and explained by Bouland et al. (Section 1.2.1) [BDFH25].

  • 2025 (real ensemble): for \(R\sim\mathcal{N}(0,1)^{n\times n}\), Bouland et al. prove \(\#\mathrm{P}\)-hardness at additive error \(n!e^{-O(n^\gamma)}\) for \(|\operatorname{Per}(R)|^2\), for every fixed \(\gamma>0\), with constant success probability, assuming permanent anticoncentration (Theorem 1, Section 4; \(\mathrm{BPP}^{\mathrm{NP}}\) reduction) [BDFH25].

  • Koehler and Leung’s July 2026 preprint proves permanent anticoncentration for both real and complex Gaussian matrices (Theorem 1.1) [KL26]. Together with Aaronson–Arkhipov Theorem 7, it establishes polynomial-time equivalence of Eqs. (1) and (2) [AA13]. The binary-rounding convention follows from Gaussian tail bounds, polynomial continuity, and the lower-tail estimate. Anticoncentration does not itself prove counting hardness.

  • Koehler–Leung Theorem 4.1 also proves Bouland et al.’s Conjecture 6 on gently perturbed Gaussian permanents, with arbitrary deterministic shifts and, more generally, arbitrary independent random perturbations. This removes the perturbation-size and bounded-matrix restrictions in that conjecture. It supplies anticoncentration, not the complex-Gaussian counting-hardness conclusion asked for here [KL26], [BDFH25].

Comment

The complex-Gaussian hardness conjecture remains open. Its additive squared-permanent formulation is represented by Eq. (2) in this same canonical record. The equivalence now uses the resolving anticoncentration preprint of Koehler and Leung, rather than an unproved hypothesis [KL26].

The 2025 tolerance \(n!e^{-O(n^\gamma)}\) applies to real Gaussian matrices and falls short of \(n!/\operatorname{poly}(n)\). The real anticoncentration assumption in that theorem is supplied by the 2026 preprint. Its result still uses \(\mathrm{BPP}^{\mathrm{NP}}\) reductions and does not prove the complex-Gaussian conjecture. Bouland et al., Appendix E, leave the required complex square-method argument unproved [BDFH25].

References

[AA13]
S. Aaronson and A. Arkhipov, "The computational complexity of linear optics," Theory of Computing 9(4), 143–252 (2013).(2010).DOIarXiv
[BFLL22]
A. Bouland, B. Fefferman, Z. Landau, and Y. Liu, "Noise and the frontier of quantum supremacy," in FOCS 2021, 1308–1317 (2022).DOIarXiv
[BDFH25]
A. Bouland, I. Datta, B. Fefferman, and F. Hernandez, "Exponential improvements to the average-case hardness of BosonSampling," in FOCS 2025, 912–933 (2025). The cited Krovi improvement is attributed there to a November 2023 personal communication.DOIarXiv
[KL26]
F. Koehler and P. K. Leung, "Anticoncentration of the Permanent in Ginibre Ensembles," arXiv preprint (July 2026).DOIarXiv

Page edit log

  • Record created
  • Last edited
  • Revisions3

View the full history on GitHub

Your contribution is welcome!

Found progress, a correction, or a resolution? Edit this record on GitHub and open a pull request, or report an update with the primary sources. The proposal page explains the available submission route; see the contribution guide for details.

Cite this page

“Permanent-of-Gaussians Conjecture,” Quantum Information and Quantum Computation Open Problem Zoo (QIQCOP Zoo), ID op_af68b033a88ef934, accessed 2026-09-24.

Use the Cite button above for BibTeX and the permanent link.

Cite this problem

Please also cite the primary sources listed under References. Cite this page for the statement, status, and stable identifier.

BibTeX

@incollection{qiqcop_op_af68b033a88ef934,
  title = {Permanent-of-Gaussians Conjecture},
  booktitle = {Quantum Information and Quantum Computation Open Problem Zoo (QIQCOP Zoo)},
  year = {2026},
  howpublished = {\url{https://qiqc-op.com/problem/op_af68b033a88ef934/}},
  note = {Stable ID op_af68b033a88ef934; status: Unsolved; accessed 2026-09-24}
}

Plain text

“Permanent-of-Gaussians Conjecture,” Quantum Information and Quantum Computation Open Problem Zoo (QIQCOP Zoo), https://qiqc-op.com/problem/op_af68b033a88ef934/, ID op_af68b033a88ef934, accessed 2026-09-24.

Share this problem

Permanent link

Identifiers

op_af68b033a88ef934
01M20DFNK3W29RR12PZ50N40AH