And as per quantum Bluesky tradition, here’s a meme about the work! ( @dulwichquantum.bsky.social )

These are early steps, but the results are promising. We are excited to extend the protocol to different system geometries and interaction structures. Many thanks to the whole team ( @egle-pagliaro.bsky.social , @luplaciano.bsky.social , Leonardo Zambrano, Antonio Acín) 😁

We tested the protocol with tensor-network simulations up to 128 qubits and also implemented it on IBM superconducting hardware, successfully learning a 109-qubit Hamiltonian.

A few repetitions of the Zeno reshaping + local tomography pipeline, combined with careful merging of the estimates from different patches, are sufficient to reconstruct the global Hamiltonian. In this work, we focus on 1D geometrically 2-local Hamiltonians.

Following this idea, we implement a coherence-preserving reshaping of the Hamiltonian via the quantum Zeno effect, using virtual-Z gates. Each reshaping configuration isolates different parts of system, and within each configuration, these patches can be learned in parallel via process tomography.

HL is a hard but important problem. As quantum computers and simulators grow in size and complexity, there is a need for experimentally-friendly protocols. A natural approach is to exploit the local structure of many systems and reshape the Hamiltonian so that each part can be learned locally.