arXiv:2610.01878
Floquet-Universal Hamiltonian Simulation
Emilio Onorati, Harriet Apel, Michael M. Wolf, Toby Cubitt
Abstract
Analogue and digital Hamiltonian simulation represent two distinct approaches to making one quantum system replicate the physical properties of another, with consequences ranging from practical applications to complexity theory and quantum gravity. They are rooted in two distinct regimes of quantum dynamics: time-independent Hamiltonians for the former; fully controllable time-dependent quantum circuits in the latter. Time-periodic Hamiltonians represent an intermediate regime. In this work, we establish a theory of Floquet simulation where we use periodically driven Hamiltonians to synthesise time-independent ones. We show that a set of interactions $S$ can Floquet-simulate every Hamiltonian in the Lie algebra $\mathrm{Lie}(S)$, and consequently we give a complete and constructive characterisation of Floquet-universal Hamiltonians that are able to produce any target Hamiltonian. Notably, our construction uses only $O(1)$ local interaction strengths and ratios thereof, avoiding the impractical multi-scale local interaction strengths often required in analogue simulation via time-independent Hamiltonians. Instead, the construction requires a range of driving frequencies which scales polynomially with system size for important classes such as $k$-local lattice Hamiltonians. This has immediate complexity-theoretic implications, including numerous BQP-completeness and QMA-hardness results for natural physical properties and quantities studied in Floquet physics.
AI Summary
The paper asks which static Hamiltonians can be synthesised using smooth periodic driving with finitely many Fourier components, and whether this can avoid the interaction-strength hierarchies of static analogue simulation. It formulates Floquet simulation through operator-norm approximation of a target by a rescaled effective Floquet Hamiltonian, allowing simulation over multiple driving periods. For an interaction set $\mathcal S$ whose generated Lie algebra is perfect, the construction approximates every Hamiltonian in $\mathrm{Lie}(\mathcal S)$. Consequently, generating the full $\mathfrak{su}(D)$ algebra characterises Floquet universality. All nonzero elementary Fourier amplitudes can have magnitudes within a factor of two. For general Lie-polynomial decompositions, the frequency bound grows exponentially in a polynomial of the maximum degree $K$, so universality alone does not imply efficiency. For fixed-locality lattice targets, or targets with constant-chromatic-number intersection graphs, the paper gives polynomial-resource simulation with $O(1)$ amplitudes and polynomially bounded frequencies and multi-period runtime. The proof chooses integer frequency resonances to cancel lower Magnus orders. Residue and shuffle identities establish that frequency-dependent rational functions span the required multilinear Lie coefficients; the canonical projection onto the free Lie algebra then yields target terms. Frequency dilation calibrates their weights, while Magnus bounds control errors. Commuting-family decompositions enable frequency reuse for local targets. The stated qualifications include the perfect-Lie-algebra assumption, potentially exponential frequencies for unrestricted targets, and runtime overhead from weak effective interactions. The construction provides asymptotic resource guarantees rather than experimental demonstrations or optimised driving parameters.
Central claim may not hold: The abstract claims simulation throughout Lie(S), but the formal universality construction establishes this only when Lie(S) is perfect. The final universality theorem approximates NT H_F rather than the N H_F required by the simulation definition, leaving an unaddressed factor of the driving period. The general efficiency proof assumes without justification that a local target's membership in Lie(S) guarantees a bounded-degree, bounded-support synthesis from the available interactions. The spectral corollary incorrectly deduces individual eigenvector closeness from operator-norm closeness without spectral-gap assumptions and omits the N-period quasienergy scaling. The definition of admissible frequency sets forbids mixed-vector complete resonances that necessarily exist whenever two frequency vectors are present, contradicting the filtering construction's requirements.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 60 pages