Hamiltonian structure learning without known interaction terms

Determine whether a k-local Hamiltonian H acting on n qubits can be learned from access to its real-time evolution unitaries e^{-iHt} without any prior knowledge of which interaction terms E_a may be present (i.e., recover both the interaction structure and the corresponding coefficients).

Background

Prior work on Hamiltonian learning typically assumes the interaction terms E_a are known and only the interaction strengths λ_a are unknown. This restricts applicability because the structure must be specified in advance, and existing techniques rely on that knowledge to design measurements or decoupling schemes.

The authors frame this as the structure learning problem for Hamiltonians, analogous to structure learning in graphical models, and highlight that prior techniques cannot handle even a single unspecified long-range interaction.

References

This work leaves a basic question unresolved: Can we learn a local Hamiltonian without prior knowledge about what interactions are allowed?

— Structure learning of Hamiltonians from real-time evolution  (2405.00082 - Bakshi et al., 2024) in Section 1 (Introduction), Question 1 (label: question:central-open-q)

For general nearest-neighbor Hamiltonians with Pauli $Y$ terms, conjugation by diagonal unitaries and time reversal $H \mapsto -H*$ are both invisible to solely $Z$-basis measurements. Are these transformations the only generic ambiguities within a fixed local family?

— Out-of-control Hamiltonian Learning  (2610.06709 - Gong et al., 5 Oct 2026) in Section 6, Outlook, paragraph “Beyond IXZ Hamiltonians”

But what changes if we allow uniform control during the time evolution? More precisely, suppose that state preparation remains a uniform product state and measurement is still performed in a common basis, but between evolution segments we may apply global pulses $U_j{\otimes n}$ with $U_j\in\mathrm{SU}(2)$. Perhaps this additional control can be used to reconstruct the generic nearest-neighbor Hamiltonians of \Cref{thm:uniform} with total evolution time $\widetilde O({-1}\poly(n,R,\varsigma{-1},\delta{-1}))$, polynomial classical post-processing time, and only polynomially many pulses.

— Out-of-control Hamiltonian Learning  (2610.06709 - Gong et al., 5 Oct 2026) in Section 6, Outlook, paragraph “Heisenberg scaling with interleaved global control”