Driven-Dissipative Ground State Preparation: Mixing Time and Randomness

Dissipative protocols offer a route to ground state preparation that does not require a large initial overlap with the ground state. To move beyond time independent Lindbladians and randomized Lindbladians sampled from a fixed distribution, we study driven-dissipative protocols in which the generator varies over time. Our construction uses a family of randomized Lindbladians whose jump operators are obtained by averaging the unitary evolution of random matrices under a fixed target Hamiltonian. We show that empirical averages of these Lindbladians drive arbitrary initial density matrices to low energy states. The resulting mixing time bounds depend on the Hamiltonian only through its eigenvalue distribution and remain valid even for cold starts.