arXiv:2609.38738
Generating random unitaries by products of conjugated Hamiltonian evolutions
Oscar Scholin, Apollonas S. Matsoukas-Roubeas, Sathyawageeswar Subramanian
Abstract
Random unitary operations are a fundamental resource for quantum information processing, yet provably generating them with experimentally available controls remains challenging. Existing approaches often require calculating higher-order operator expectation values, relying on statistical assumptions about the experimental Hamiltonian's spectrum to make these calculations tractable. We show how to implement random unitaries by a sequence of evolutions generated by a fixed Hamiltonian conjugated by a global control operator. We give a geometric argument that establishes exponential convergence to the uniform random (Haar) measure on the accessible unitary group under uniform parameter sampling, beyond a finite sequence-length threshold that we bound. We show this holds when the connected Lie subgroup generated by the conjugated Hamiltonians is closed. Native dynamics and global control thus provide a constructive route to uniform randomness, with applications to many experimental platforms including Rydberg atoms, bosons and fermions in optical lattices, transmons, and trapped ions.
AI Summary
Can native Hamiltonian dynamics with a single global control parameter generate Haar-random operations without assumptions about quantum chaos? This paper studies independent products of $F(\phi,t)=e^{i\phi A}e^{itB}e^{-i\phi A}$ for fixed Hermitian $A,B$, uniformly sampling $\phi$ and nonnegative durations from bounded domains containing intervals, including arbitrarily short durations. Let $\mathfrak h$ be the Lie algebra generated by $i e^{i\phi A}Be^{-i\phi A}$, and let $G$ be the closure of the accessible products. If the connected subgroup associated with $\mathfrak h$ is closed, the endpoint distribution becomes absolutely continuous relative to Haar measure after finitely many steps and converges exponentially in total variation. Writing $n=\dim G$ and $r$ for the real span dimension of the conjugated generators, the minimum submersion length satisfies $\lceil n/2\rceil\leq L_0\leq r+(n-r)(n-r+1)\leq d^4$. The proof constructs a full-rank product-map differential; real analyticity and the coarea formula establish absolute continuity, enabling existing compact-group random-walk theorems. For the full unitary group, convergence implies approximate designs at every fixed moment order. An additional rejection-sampling construction gives exact finite-length Haar sampling under fiber-volume and boundedness assumptions. The asymptotic rate is characterized through irreducible-representation Fourier coefficients, not bounded explicitly for concrete Hamiltonians. Physical examples span Rydberg arrays, Hubbard models, transmons, and trapped ions, but their closed-subgroup conditions and practical mixing times remain unanalysed. Application-specific design bounds, coherence-time constraints, and control errors are left open.
to improve the clarity of the manuscript throughout, in particular to make discussion of the application to experimental Hamiltonians in Appendix \Cref{apdx:examples} more pedagogical.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.