arXiv:2609.38513
Copy-scarce learning of ground states
Kaito Wada, Nobuyuki Yoshioka
Abstract
Probing multiple properties of ground states is central to understanding quantum matter, yet conventional readout can be prohibitively costly because it consumes many independently prepared copies. Here, in copy-scarce regimes, we establish optimal trade-offs between the number of initial ground-state copies and elapsed Hamiltonian-evolution time required for estimating multiple observables or reconstructing a full classical description of a unique gapped ground state. We derive a unifying inequality that quantifies how well ground states can be distinguished using initial copies and controlled dynamics, even when all copies may be consumed. It yields worst-case lower bounds on elapsed Hamiltonian time that any protocol must incur to perform either task. We construct parallel catalytic readouts that return all initial copies nearly unchanged while attaining these bounds up to poly-logarithmic factors. Our protocols require no coherent preparation circuit or its inverse, and their elapsed Hamiltonian time depends only logarithmically on the inverse of the desired return error. Thus, optimal copy--dynamics trade-offs can be attained together with catalytic return. Numerical benchmarks further show substantial reductions in end-to-end elapsed Hamiltonian time relative to a task-specialized copy-consuming protocol, even under an optimistic preparation-cost model. Our results reveal a copy--dynamics resource landscape connecting copy-scarce learning to copy-only endpoints realized by known protocols for shadow tomography and pure-state tomography.
AI Summary
The paper asks how controlled Hamiltonian dynamics can replace fresh copies when learning a unique gapped ground state, and whether optimal readout can return the supplied states nearly unchanged. Its black-box model supplies $P$ ground-state copies, controlled forward and backward evolution, and observable block encodings, but no coherent preparation circuit. Cost is elapsed Hamiltonian time with $P$ registers evolving in parallel, not total evolution across registers. For $M$ bounded observables, a catalytic protocol achieves $T=\widetilde O(\Delta^{-1}[\Gamma_{\rm obs}/(P\varepsilon)+1])$, where $\Gamma_{\rm obs}=\|\sum_jO_j^2\|^{1/2}\leq\sqrt M$. Full tomography to trace-distance error $\eta$ takes $\widetilde O(\Delta^{-1}[d/(P\eta)+1])$. Both return all supplied copies with controllable error entering time only logarithmically. Worst-case lower bounds allowing copy consumption match these scalings, up to polylogarithms, for high-precision observable estimation with $M\leq d$, $\varepsilon\lesssim M^{-1/2}$ and $P\lesssim\sqrt M/\varepsilon$, and tomography with $P\lesssim d/\eta$. The lower bounds use an adversary-style inequality combining distinguishability from initial copies and dynamics. Protocols combine energy filtering, joint multidimensional phase kickback and GHZ-enhanced parallel sensing; tomography additionally uses nearly unbiased estimation and matrix concentration. The complete trade-off remains unresolved at intermediate copy counts and lower precision. Numerical Hamiltonian-time comparisons against specialized copy-consuming measurements include preparation costs but exclude other gates and classical processing, and use an extrapolated dissipative-preparation baseline. Gate and ancilla costs, stronger collective-measurement comparators, and extensions beyond unique ground states remain open.
Central claim may not hold: The abstract overstates the scope of optimal multi-observable copy–dynamics trade-offs: the established matching bounds require a high-precision regime, while low-precision copy-scarce estimation remains unresolved. The supplementary tomography lower-bound proof incorrectly replaces a trace-norm error bound of 2η by η, although correcting this factor leaves the asymptotic theorem intact.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.