Fourier Spectrum of Noisy Quantum Algorithms

TL;DR

This work analyzes the power of noisy quantum computation models that lie below full quantum universality by quantifying how acceptance probabilities distribute across Fourier coefficients. Using a novel Matrix Decomposition Lemma, the authors bound level-$\ell$ Fourier growth for DQC$_k$, $\tfrac{1}{2}\mathsf{BQP}$ and $\mathsf{BQP}$, revealing that noise patterns modulate growth in model-specific ways and enabling oracle separations such as 2-Forrelation in DQC$_1$ and 3-Forrelation in $\tfrac{1}{2}\mathsf{BQP}$. These bounds yield hierarchy results and new oracle separations between the models, illustrating a refined landscape between $\mathsf{BQP}$ and $\mathsf{BPP}$. The work also points to broader applications to near-term noisy devices and raises questions about tightness and extensions to other intermediate models like NISQ and IQP.

Abstract

Quantum computing promises exponential speedups for certain problems, yet fully universal quantum computers remain out of reach and near-term devices are inherently noisy. Motivated by this, we study noisy quantum algorithms and the landscape between $\mathsf{BQP}$ and $\mathsf{BPP}$. We build on a powerful technique to differentiate quantum and classical algorithms called the level-$\ell$ Fourier growth (the sum of absolute values of Fourier coefficients of sets of size $\ell$) and show that it can also be used to differentiate quantum algorithms based on the types of resources used. We show that noise acting on a quantum algorithm dampens its Fourier growth in ways intricately linked to the type of noise. Concretely, we study noisy models of quantum computation where highly mixed states are prevalent, namely: $\mathsf{DQC}_k$ algorithms, where $k$ qubits are clean and the rest are maximally mixed, and $\frac{1}{2}\mathsf {BQP}$ algorithms, where the initial state is maximally mixed, but the algorithm is given knowledge of the initial state at the end of the computation. We establish upper bounds on the Fourier growth of $\mathsf{DQC}_k$, $\frac{1}{2}\mathsf{BQP}$ and $\mathsf{BQP}$ algorithms and leverage the differences between these bounds to derive oracle separations between these models. In particular, we show that 2-Forrelation and 3-Forrelation require $N^{Ω(1)}$ queries in the $\mathsf{DQC}_1$ and $\frac{1}{2}\mathsf{BQP}$ models respectively. Our results are proved using a new matrix decomposition lemma that might be of independent interest.

Figures (7)

• Figure 1: A $d$-query $\mathsf{DQC}_{k}$ algorithm. The initial state on the first $n+w$ qubits is maximally mixed.
• Figure 2: A $d$-query $\tfrac{1}{2}\mathsf{BQP}$ algorithm. The initial state on the first $n+w$ qubits can be thought of as maximally mixed, or as the pure state $2^{-(n+w)/2}\sum_{I\in\{0,1\}^{n+w}}\ket{I}$; the resulting circuits are equivalent.
• Figure 3: Fourier growth of acceptance probability of algorithms with $d=\mathrm{poly}\log N$ queries.
• Figure 4: A simple example of a $d$-query $\mathsf{DQC}_{1}$ algorithm. The initial state on the first $n$ qubits is maximally mixed.
• Figure 5: A $d$-query $\mathsf{BQP}$ algorithm.

Theorems & Definitions (40)

• Definition 1.1: Signed Fourier Growth
• Definition 1.2: Fourier Growth
• Definition 1.3: $k$-Forrelation function
• Definition 1.4: $k$-Forrelation problem with parameter $\varepsilon=\Theta(1/\log^k N)$
• Theorem 1.5
• Theorem 1.6
• Theorem 1.7
• Corollary 1.8
• Corollary 1.9
• Corollary 1.10