Quantum Signal Processing
Applying a polynomial transformation to the eigenvalues of a block-encoded operator by interleaving it with tunable single-qubit rotations.
The single-qubit picture
Quantum signal processing (QSP) starts from a signal rotation W(x) that rotates about one axis by an angle set by a hidden parameter x. Sandwiching W(x) between single-qubit rotations by chosen phases phi_0, phi_1, ..., phi_d produces a unitary whose matrix element is a degree-d polynomial P(x) in x. By selecting the phases you sculpt P.
What polynomials are achievable
For a sequence of d+1 phases, the top-left entry is a complex polynomial P(x) of degree at most d with matching parity, subject to |P(x)|^2 + |Q(x)|^2 = 1 on the domain. Theorems (Low, Yoder, Chuang) characterize exactly which P are realizable, and efficient classical routines compute the phases for a target P.
From one qubit to operators
Replacing the scalar signal W(x) with a qubitized walk operator W lifts the polynomial from a number to a function of the Hamiltonian: the QSP sequence applies P(H/alpha) inside the block-encoding. This lift from scalar to operator is the quantum singular value transformation.
- Hamiltonian simulation: approximate e^(-i alpha t x) by a Jacobi-Anger polynomial expansion in x.
- Matrix inversion: approximate 1/x for linear-systems solving.
- Eigenstate filtering: approximate a step function to project onto a spectral window.
- Amplitude amplification: a special low-degree case of QSP.
Why QSP is a unifying framework
Nearly every quantum algorithm that transforms a spectrum, simulation, phase estimation, search, linear systems, and more, can be expressed as choosing a polynomial and applying it via QSP or QSVT. The cost is the polynomial degree, which for smooth target functions grows only logarithmically in the required precision.
The Jacobi-Anger expansion e^(-i tau cos theta) = sum_k i^k J_k(tau) e^(i k theta), truncated at degree O(tau + log(1/epsilon)), is exactly the polynomial that gives optimal Hamiltonian simulation, which is why QSP and qubitization together achieve the theoretical best.