arXiv:2610.14259
Quantum Signal Processing Phase Factorization via Logarithmic Derivatives
Christopher Peng
Abstract
Quantum signal processing (QSP) requires the efficient computation of phase factors associated with a polynomial transformation. One existing stable factorization method avoids direct root finding but relies on Prony's method to recover a polynomial from weighted power sums of its roots (arXiv:2202.02671). We present a modification of this approach that eliminates Prony's method by using the logarithmic derivative of an associated polynomial. Because the logarithmic derivative has the same residue at every root after accounting for multiplicity, its Fourier coefficients directly yield the unweighted power sums of the roots inside the unit disk. The Newton--Girard identities then recover the coefficients of the corresponding polynomial without computing its roots. The resulting algorithm has $O(d\log^2 d)$ computational complexity and simplifies the factorization procedure of the original method. In numerical experiments on Hamiltonian simulation, eigenstate filtering, matrix inversion, and the Fermi--Dirac operator, our implementation computes phase factors in seconds with maximum errors on the order of $10^{-12}$.
AI Summary
The paper addresses classical computation of quantum signal processing (QSP) phases for a prescribed polynomial transformation. It modifies a root-free factorization pipeline by replacing Prony reconstruction with logarithmic derivatives and Newton–Girard identities. For a real degree-$d$ target $a$ of definite parity, it sets the auxiliary polynomial $b=0$ and forms $h(z)=z^{2d}(1-a(z)^2)$. A contour-integral argument shows that negative Fourier coefficients of $h'/h$ give unweighted power sums of the $2d$ roots inside the unit disk, counting multiplicities. Newton–Girard recovery, implemented through a truncated power-series exponential, constructs the complementary polynomial in $O(d\log d)$ operations without root finding. Existing divide-and-conquer phase stripping gives overall complexity $O(d\log^2 d)$. The complementary polynomial coincides with that of an existing FFT-based logarithmic spectral-factorization method. Double-precision tests cover Hamiltonian simulation, eigenstate filtering, matrix inversion, and Fermi–Dirac transformations, with targets scaled to $\|a\|_\infty=0.3$ and approximately $40d$ Fourier samples. Reported reconstruction errors are below $3\times10^{-12}$; the largest matrix-inversion case has degree $51{,}633$ and takes about 3.5 seconds. These errors are sampled estimates using only 21 evaluation points, not certified uniform bounds; polynomial-approximation costs and errors are reported separately. The paper leaves rigorous numerical-error analysis and behavior near $\|a\|_\infty=1$ unresolved. Scaling reduces circuit success probability. Comparisons use previously published timings on different hardware, and speedups may largely reflect faster phase stripping rather than the Prony replacement.
Error found: The abstract’s reported maximum errors around 10^-12 are based on reconstruction checks at only 21 points per instance, so they do not establish the maximum error over the full interval.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 12 pages, 5 figures. Code available at https://github.com/chessbear18/QSP-Phase-Factorization-via-Logarithmic-Derivatives