arXiv:2610.01669
Dissipation-Enabled Operator-Norm Locality Bounds for Bose-Hubbard Hamiltonians
Marius Lemm, Tim Möbus
Abstract
Closed bosonic lattice systems do not generally admit operator norm Lieb-Robinson bounds uniformly in the initial state. The reason is that the information propagation velocity can grow with the local boson occupancy, which can be macroscopically large. Here, we show that local dissipation restores an operator-norm Lieb-Robinson bound. We consider the dissipative Bose-Hubbard model described by a Lindbladian operator with on-site $\ell$-photon loss, $\ell>2$. For our main result, we establish an almost-ballistic Lieb-Robinson bound. Dissipation rapidly depletes highly occupied sites, thus regularizing the state on the Sobolev-type scale of local particle moments. The resulting moment bounds diverge as time approaches zero but remain integrable near zero at sufficiently low orders, which is crucial for our argument. We extend these ideas to treat cat-code dissipation and, for initial states in the code space, we prove a local adiabatic approximation uniform in the total volume. Further consequences include local channel approximation, a thermodynamic limit, and efficient hybrid bosonic-qubit simulation. The point is that these applications are now available uniformly in the input state, as for quantum spin systems, but in contrast to closed Bose-Hubbard systems.
AI Summary
Closed Bose–Hubbard systems generally lack operator-norm Lieb–Robinson bounds uniform over initial states: arbitrarily large occupations can defeat spatial local approximation. This paper asks whether on-site dissipation restores such state-independent locality. For Bose–Hubbard dynamics with on-site $\ell$-photon loss, $\ell>2$, and loss rates uniformly bounded above and away from zero, it proves that a bounded observable $A$ supported on $X\subset\mathbb Z^D$ satisfies $\sup_{0\le t\le T}\|\tau_t(A)-\tau_t^{X[r]}(A)\|_\infty\le C|X|(1+T)\|A\|_\infty\exp[-c\log^2(2+r/(1+T))]$. Constants are independent of volume, input state, and on-site interaction and chemical-potential coefficients. At fixed accuracy, this permits an almost-ballistic radius $r=O(T^{1+\eta})$ for every $\eta>0$. The proof combines instantaneous local-moment regularization, a time-dependent occupation cutoff, and a bounded-interaction dissipative Lieb–Robinson estimate insensitive to unbounded on-site terms. Moment roots diverge as $t^{-1/(\ell-1)}$, but the integrated cutoff remains finite for $\ell>2$. Consequences include diamond-norm local channel approximation, a thermodynamic limit, and correlation-spreading bounds. An extension to quartic interactions with shifted four-photon loss yields, for code-space inputs, a volume-uniform local adiabatic error $O(\gamma^{-1+\eta})$. For $\ell>3$, a hybrid bosonic–qubit simulation handles arbitrary inputs, assuming physical loss preparation, occupation testing, and coherent encoding; circuit costs exclude these steps. Two-photon loss remains unresolved because the cutoff becomes nonintegrable at zero. A finite-velocity exponential light cone also remains open.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.