Algorithm Zoo · Simulation
Qubitization / Quantum Signal Processing
Also known as: QSP, QSVT-based simulation
First described: Low, Chuang (qubitization); Gilyén, Su, Low, Wiebe (QSVT framework), 2017
The problem
Implement e^{-iHt} with optimal query complexity in t and ε.
Embed H into a block of a larger unitary (block encoding), then use QSP to apply a polynomial f(H) — here f(x) = e^{-ixt} approximated by a Jacobi-Anger expansion. Achieves optimal Õ(t + log(1/ε)) query complexity, an additive-vs-multiplicative improvement over Trotter.
Best classical
Exponential
Quantum complexity
Õ(t · ||H|| + log(1/ε)) queries to block encoding
Our verdict
The asymptotic champion for Hamiltonian simulation, but a primitive — block encoding is where the engineering happens. The QSVT framework unifies most algorithms (HHL, AE, Hamiltonian simulation) under one polynomial-transformation lens; in 2026 it's the right mental model for new algorithm design.
When to use it
- When you have a clean block encoding (matrix oracle access) of H.
- For high-precision simulation: error ε enters logarithmically, not polynomially.
- Inside larger algorithms that need polynomial transformations of matrices (HHL, ground-state preparation).
When not to use it
- When the block encoding is too expensive to construct (constants dominate).
- NISQ — needs deep, error-corrected circuits.
Classical baseline
Same as Trotter — exact classical diagonalization is exponential. Tensor-network methods remain strong for low-entanglement instances.
Hardware cost
Adds 1-2 ancilla qubits per block-encoding level plus the cost of preparing the block encoding. Practical only post-FTQC.
Key papers
- Hamiltonian Simulation by Qubitization ↗
Low, Chuang · 2019 · Quantum
- Quantum Singular Value Transformation and Beyond ↗
Gilyén, Su, Low, Wiebe · 2019 · STOC
Deep-dive tutorials
Last verified: 2026-05-24