[Paper Review] Exponentially tighter bounds on limitations of quantum error mitigation

Yihui Quek, Daniel Stilck França+3 more|arXiv (Cornell University)|Oct 20, 2022

Quantum Computing Algorithms and ArchitectureComputer Science57 references22 citations

TL;DR

This paper establishes exponentially tighter bounds on the sample complexity required for quantum error mitigation, showing that even shallow circuits require superpolynomial samples in the worst case. By framing error mitigation as a statistical inference problem, the authors prove that noise-induced state scrambling occurs at exponentially smaller depths than previously thought, severely limiting the scalability of current mitigation techniques for large-scale quantum devices.

ABSTRACT

Quantum error mitigation has been proposed as a means to combat unwanted and unavoidable errors in near-term quantum computing without the heavy resource overheads required by fault tolerant schemes. Recently, error mitigation has been successfully applied to reduce noise in near-term applications. In this work, however, we identify strong limitations to the degree to which quantum noise can be effectively `undone' for larger system sizes. Our framework rigorously captures large classes of error mitigation schemes in use today. By relating error mitigation to a statistical inference problem, we show that even at shallow circuit depths comparable to the current experiments, a superpolynomial number of samples is needed in the worst case to estimate the expectation values of noiseless observables, the principal task of error mitigation. Notably, our construction implies that scrambling due to noise can kick in at exponentially smaller depths than previously thought. They also impact other near-term applications, constraining kernel estimation in quantum machine learning, causing an earlier emergence of noise-induced barren plateaus in variational quantum algorithms and ruling out exponential quantum speed-ups in estimating expectation values in the presence of noise or preparing the ground state of a Hamiltonian.

Motivation & Objective

• To rigorously analyze the fundamental limits of quantum error mitigation in near-term quantum devices.
• To identify worst-case scenarios where error mitigation fails due to excessive sampling requirements.
• To extend prior bounds by incorporating circuit width (n) and non-unital noise, such as T1 decay.
• To unify and analyze major error mitigation protocols—virtual distillation, ZNE, PEC, CDR—under a common theoretical framework.
• To demonstrate that exponential sample costs emerge even at constant circuit depth, challenging the scalability of current mitigation strategies.

Proposed method

• Formulates error mitigation as a statistical inference problem using classical descriptions of noiseless circuits and noisy measurement outcomes.
• Distinguishes between weak mitigation (estimating expectation values) and strong mitigation (sampling from noiseless states).
• Introduces a framework based on quantum channel purity and contraction coefficients to quantify state distinguishability under noise.
• Derives bounds on success probability of virtual distillation using channel parameters qn and rn, showing exponential decay unless channels are unitary or replacer channels.
• Analyzes non-unital noise (e.g., T1 decay) and shows that such noise causes exponentially small success probabilities in mitigation protocols.
• Applies information-theoretic tools, including quantum estimation theory and contraction coefficients, to derive universal lower bounds on sample complexity.

Experimental results

Research questions

• RQ1 What is the worst-case sample complexity required for error mitigation in large-scale quantum circuits?
• RQ2 How does circuit width (n) and noise type (non-unital vs. unital) affect the scalability of error mitigation?
• RQ3 At what circuit depth does noise-induced scrambling make error mitigation infeasible?
• RQ4 Can existing protocols like zero-noise extrapolation or probabilistic error cancellation achieve sub-exponential scaling in general cases?
• RQ5 To what extent do current error mitigation schemes approach the theoretical limits imposed by noise and statistical inference?

Key findings

• Even at constant circuit depth, error mitigation requires a superpolynomial number of circuit runs in the worst case.
• The success probability of virtual distillation decays exponentially with system size unless the noise channel is unitary or a replacer channel.
• Non-unital noise, such as T1 decay, causes exponentially small success probabilities in mitigation protocols, even for shallow circuits.
• Noise-induced scrambling emerges at exponentially smaller depths than previously believed, limiting the feasible regime for error mitigation.
• The bounds rule out exponential quantum speed-ups in estimating expectation values or preparing ground states under noisy conditions.
• The results imply that current error mitigation schemes are nearly optimal in general settings and must be redesigned to exploit special circuit structures to achieve efficiency.