arXiv:2610.02157
Learning Many-Body Hamiltonians Using a Local Probe
Suying Liu, Zitai Xu, Alexey V. Gorshkov, Xiaodi Wu, Yu-Xin Wang, Zhi-Yuan Wei
Abstract
Hamiltonian learning provides a systematic framework for reconstructing unknown quantum dynamics. However, existing protocols typically assume direct measurement access to the entire system. With fast single-qubit control and a connected reference backbone, we show that a single measurable qubit suffices to learn all $O(N)$ independent parameters of a bounded-degree two-body Hamiltonian on $N$ qubits at the Heisenberg limit. Crucially, our protocol uses robust SWAP gates synthesized by quantum signal processing, enabling coherent transfer of states evolving under distant Hamiltonian parameters to the measurable qubit. This transfer requires no prior calibration of the Hamiltonian parameters of the intermediate links. A parallel learning architecture achieves total query time $\widetilde{O}(N)$ for an $N$-qubit chain, while retaining Heisenberg-limited precision scaling. On the chain, these scalings match the fundamental precision and information-propagation lower bounds up to logarithmic factors. The framework further extends to arbitrary bounded-degree interaction graphs. Our results establish a scalable route to learning an extensive number of many-body Hamiltonian parameters through only a local measurement interface.
AI Summary
Can an extensive many-body Hamiltonian be learned when only one qubit can be measured and reset? This paper considers general nearest-neighbor two-body Hamiltonians on $N$-qubit chains, containing $12N-9$ unknown Pauli coefficients, with fast single-qubit control throughout the register. It assumes known coefficient bounds and a specified reference interaction on every bond whose unknown strength satisfies $J_-\leq |g_i|\leq J_+$. The protocol learns every coefficient to additive error $\varepsilon$, jointly with probability at least $1-\delta$. Quantum signal processing constructs SWAP channels uniformly over the promised coupling range, without learning coupling magnitudes or signs, in interaction time $O(J_-^{-1}\log(1/\eta))$ for diamond-norm error $\eta$. These gates initialize the register through endpoint resets and transport phase records encoded in parallel using Hamiltonian reshaping. Taking $\eta=O(N^{-2})$ keeps readout biases below a constant threshold compatible with robust multiscale phase estimation. Writing $L=\log(2\Lambda/\varepsilon)$, total native-Hamiltonian evolution time is $O[\varepsilon^{-1}\log(N/\delta)+(N/J_-)\log N\,(\log(N/\delta)+L)L]$. At fixed precision and confidence this is $O(N\log^2N)$; precision dependence is Heisenberg limited. Chain lower bounds are $\Omega(1/\varepsilon)$ and $\Omega(N/\Lambda)$. The construction extends to known bounded-degree graphs with a reference spanning tree. Chains with common-axis reference interactions require only independently controlled periodic sublattices. The resource model neglects pulse, measurement, and reset durations and assumes arbitrarily fast Hamiltonian reshaping. Finite control rates, decoherence, and optimal geometry-dependent complexity remain open; irregular-graph routing still requires individual addressing.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 32 pages, including 1 figure and 12 pages of Supplementary Material