Reservoir engineering turns dissipation from a problem into a tool. Instead of shielding a system from its environment, one designs the coupling to an engineered environment so that the dynamics relaxes the system into a desired state, and keeps it there. Because the target is an attractor of the dynamics, it is reached from any initial state and recovered after perturbations, without measurement or active feedback.

The idea

For Lindblad dynamics, \(\dot\rho = \mathcal L(\rho)\), the long-time state is a steady state, \(\mathcal L(\rho_{ss}) = 0\). If the jump operators \(L_k\) and the Hamiltonian are designed so that a target pure state \(|\psi\rangle\) is a dark state — annihilated by every \(L_k\) and an eigenstate of \(H\) — and the dynamics has no other steady state, every initial state converges to \(|\psi\rangle\). This makes dissipation a deterministic, self-correcting preparation protocol. The approach was proposed for trapped ions, whose motion can be laser-cooled into an engineered bath [1].

How far it reaches

  • Many-body states. Any pure multipartite state can be prepared as the unique steady state of a suitable dissipative process [2]. Driven, dissipative cold-atom systems can be steered into given many-body states and phases [3].
  • Computation. Dissipation alone can perform universal quantum computation and prepare complex states such as ground states of frustration-free Hamiltonians [4].
  • Control theory. For Markovian dynamics, the invariance and attractivity of subspaces have been characterized [5]. Any generic state can be made globally asymptotically stable by constructively designing the Lindblad generator [6].

In experiments

  • Atomic ensembles. Engineered dissipation generated steady-state entanglement between two macroscopic atomic ensembles at room temperature, maintained for 0.04 s [7].
  • Trapped ions. An open-system quantum simulator with up to five qubits implemented engineered dissipative dynamics [8]. Unitary operations combined with engineered dissipation produced and stabilized an approximate Bell state [9].
  • Superconducting circuits. Tailoring the microwave noise in a cavity relaxes a qubit to an arbitrary chosen superposition [10]. Two qubits were autonomously stabilized in a Bell state [11]. Engineered two-photon loss confines a cavity field to the manifold of cat states [12], the basis of dissipatively protected cat qubits [13]. See Bosonic & continuous-variable control and Control for quantum error correction.

Reviews cover engineered dissipation across quantum information science [14] and as a resource in superconducting circuits [15].

Reservoir engineering and control

Reservoir engineering is control at the level of the generator of the dynamics: drives, couplings and auxiliary lossy modes are chosen to shape \(\mathcal L\). Optimal control can design those drives, or speed up the convergence to the steady state; see Control of open quantum systems. Compared with measurement-based feedback, autonomous schemes need no fast classical processing. Feeding quantum outputs back coherently — see Linear quantum systems & coherent feedback — sits in between. The same ideas underpin autonomous quantum error correction [16].

Limitations

  • Engineering the right jump operators needs strong, well-controlled couplings to auxiliary dissipative modes.
  • Steady states are reached only asymptotically, and the convergence rate (the dissipative gap) can be small, especially in many-body systems.
  • Unwanted, natural dissipation competes with the engineered one and limits the attainable fidelity.