Quantum Complexity of Matrix Functions Probed with Four Functions

Key Takeaways

• Classical simulation works for poly-degree polynomials when matrix has O(log n) Pauli terms
• Monomials become easy classically, but Chebyshev, time‑evolution, inverse stay BQP‑hard
• Sparse vs Pauli access models change complexity classification dramatically
• Hierarchy establishes clear ranking of difficulty among four key functions
• Findings guide algorithm designers to target genuine quantum advantage zones

Pulse Analysis

Estimating functions of large matrices lies at the heart of many quantum algorithms, from Hamiltonian simulation to quantum linear systems. While these tasks are BQP‑complete in the general sparse‑access model, the new analysis reveals that the structure of the input—specifically, how the matrix is presented—can dramatically shift the computational landscape. By focusing on Pauli‑sparse representations, where only O(log n) Pauli terms are non‑zero, the authors prove that any polynomial of degree poly(n) can be reproduced with classical resources, effectively collapsing a broad class of quantum‑heavy problems into tractable ones.

The paper goes further to rank four canonical functions—monomials, Chebyshev polynomials, time‑evolution, and matrix inversion—by their relative difficulty. In regimes where monomials are efficiently computable classically, the other three remain resistant, retaining their BQP‑hard status. This hierarchy provides a nuanced map for researchers: it pinpoints exactly which function families still demand quantum hardware and which can be off‑loaded to classical processors without sacrificing performance. The distinction between sparse and Pauli access models emerges as a decisive factor, underscoring the importance of data encoding in algorithmic complexity.

For industry practitioners, these insights translate into clearer investment decisions. Companies developing quantum‑accelerated linear‑algebra tools can now identify problem instances where a quantum advantage is provable, avoiding costly over‑engineering on tasks that are already classically efficient. Moreover, the catalog of efficient classical algorithms serves as a benchmark for validating quantum hardware, offering a practical pathway to certify quantum speedups. As the field matures, such rigorous complexity characterizations will be essential for aligning research priorities with real‑world computational gains.