Improved robustness of quantum supremacy for random circuit sampling

Yasuhiro Kondo, Ryuhei Mori, Ramis Movassagh

Abstract

Motivated by the recent experimental demonstrations of quantum supremacy, proving the hardness of the output of random quantum circuits is an imperative near term goal. We prove under the complexity theoretical assumption of the non-collapse of the polynomial hierarchy that approximating the output probabilities of random quantum circuits to within $\exp(-\Omega(m\log m))$ additive error is hard for any classical computer, where $m$ is the number of gates in the quantum computation. More precisely, we show that the above problem is $\#\mathsf{P}$-hard under $\mathsf{BPP}^{\mathsf{NP}}$ reduction. In the recent experiments, the quantum circuit has $n$-qubits and the architecture is a two-dimensional grid of size $\sqrt{n}\times\sqrt{n}$. Indeed for constant depth circuits approximating the output probabilities to within $2^{-\Omega(n\log{n})}$ is hard. For circuits of depth $\log{n}$ or $\sqrt{n}$ for which the anti-concentration property holds, approximating the output probabilities to within $2^{-\Omega(n\log^2{n})}$ and $2^{-\Omega(n^{3/2}\log n)}$ is hard respectively. We made an effort to find the best proofs and proved these results from first principles, which do not use the standard techniques such as the Berlekamp--Welch algorithm, the usual Paturi's lemma, and Rakhmanov's result.

Introduction

To validate claims of "quantum supremacy," it is essential to demonstrate that a quantum device performs a task fundamentally intractable for classical computers. Random Circuit Sampling (RCS)—where a quantum computer executes a circuit of random gates and samples from the resulting distribution—is a primary candidate for this demonstration.

A significant hurdle in complexity theory is establishing average-case hardness: proving that calculating the output of most random circuits is difficult, rather than just a few "worst-case" instances. This paper strengthens the theoretical foundation of quantum supremacy by significantly expanding the "robustness" of these hardness claims. Specifically, it addresses how much additive error a classical approximation algorithm can tolerate before the problem ceases to be $\#\mathsf{P}$-hard (the hardest class of counting problems).

Methods

The authors derive their results from first principles, moving away from standard tools like the Berlekamp–Welch algorithm or Paturi’s lemma, which they found restrictive or unstable in the presence of noise. Their methodology centers on two primary innovations:

  1. The Cayley Path: This mathematical path allows for interpolation between a "worst-case" hard circuit and an "average-case" random circuit. It maintains unitary properties throughout the transition, facilitating a robust reduction.
  2. Lagrange Interpolation: To handle polynomial extrapolation, the authors utilize Lagrange polynomials. This approach provides tighter error bounds for extrapolation compared to previous techniques.

The core of the proof is a reduction algorithm $\mathcal{R}$ that utilizes an NP-oracle to solve $\#\mathsf{P}$ problems, as detailed in Algorithm 1.

Algorithm 1: The reduction algorithm $\mathcal{R}$ utilizing an NP-oracle $W$ to approximate circuit output probabilities.

Complexity Reductions and Results

The primary contribution of this work is the proof that approximating the output probability $| \langle 0 | C | 0 \rangle |^2$ to within an additive error of $2^{-\Omega(m \log m)}$ is $\#\mathsf{P}$-hard. This represents a "super-polynomial" improvement over previous robustness bounds, making the hardness claim significantly more resilient to classical approximation attempts.

The logical flow of this reduction, specifically for Theorem 2, is illustrated below. It shows how any $\#\mathsf{P}$ problem can be reduced to computing output probabilities on average, which in turn relates to the hardness of Random Circuit Sampling.

The reduction chain for Theorem 2, establishing $\#\mathsf{P}$-hardness under $\mathsf{BPP}^{\mathsf{NP}}$ reductions.

Key results for specific $n$-qubit circuit depths include:

  • Constant depth: Hardness within $2^{-\Omega(n \log n)}$.
  • Depth $\log n$: Hardness within $2^{-\Omega(n \log^2 n)}$.
  • Depth $\sqrt{n}$: Hardness within $2^{-\Omega(n^{3/2} \log n)}$.

These results hold for a significant fraction of all possible quantum circuits ($3/4 + 1/\text{poly}(n)$ under the $\mathsf{BPP}^{\mathsf{NP}}$ assumption). Furthermore, the authors demonstrate that this hardness persists even in the neighborhood of the identity circuit.

Conclusion and Future Work

While this work provides a more robust mathematical "moat" for quantum supremacy, several challenges remain. The current proof focuses on the hardness of calculating probabilities; bridging the gap to the hardness of the sampling task itself requires the "anti-concentration" property. Additionally, while the robustness bound has seen a massive leap, reaching the $2^{-n}/\text{poly}(n)$ threshold required to fully cover experiments like Google’s Sycamore remains a final goal. Future research may also apply the Cayley path and Lagrange interpolation techniques to other models, such as BosonSampling.

Summary updated

Abstract

Motivated by the recent experimental demonstrations of quantum supremacy, proving the hardness of the output of random quantum circuits is an imperative near term goal. We prove under the complexity theoretical assumption of the non-collapse of the polynomial hierarchy that approximating the output probabilities of random quantum circuits to within $\exp(-\Omega(m\log m))$ additive error is hard for any classical computer, where $m$ is the number of gates in the quantum computation. More precisely, we show that the above problem is $\#\mathsf{P}$-hard under $\mathsf{BPP}^{\mathsf{NP}}$ reduction. In the recent experiments, the quantum circuit has $n$-qubits and the architecture is a two-dimensional grid of size $\sqrt{n}\times\sqrt{n}$. Indeed for constant depth circuits approximating the output probabilities to within $2^{-\Omega(n\log{n})}$ is hard. For circuits of depth $\log{n}$ or $\sqrt{n}$ for which the anti-concentration property holds, approximating the output probabilities to within $2^{-\Omega(n\log^2{n})}$ and $2^{-\Omega(n^{3/2}\log n)}$ is hard respectively. We made an effort to find the best proofs and proved these results from first principles, which do not use the standard techniques such as the Berlekamp--Welch algorithm, the usual Paturi's lemma, and Rakhmanov's result.
BibTeX
@inproceedings{kondo2021improved,
  author = {Yasuhiro Kondo and Ryuhei Mori and Ramis Movassagh},
  title = {{Improved robustness of quantum supremacy for random circuit sampling}},
  booktitle = {FOCS},
  year = {2021},
}
Figure