Preparing topologically nontrivial quantum states by adiabatic evolution requires crossing phase transitions where the energy gap closes to zero. At the gap closing, the adiabatic approximation fails — the system undergoes excitations that destroy the target state. This is a fundamental obstacle: the topology changes at the gap closing, so unitary evolution through it necessarily loses control.
Dephasing noise — normally the enemy of quantum state preparation — provides the pathway. When controlled decoherence is introduced, the system can be driven through the gap-closing region without maintaining coherence across it. The nonunitary dynamics bypass the topological obstruction that blocks coherent evolution.
The mechanism is precise. Adiabatic (unitary) evolution preserves the quantum numbers that define which topological sector the state occupies. Crossing a phase boundary requires changing those quantum numbers, which coherent dynamics cannot do smoothly. Dephasing breaks the conservation of those quantum numbers locally, allowing the system to cross between sectors. Once across, the decoherence is removed and the system is in the target topological state.
The structural observation: decoherence enables what coherence prohibits. The property that makes dephasing destructive in most contexts — it erases quantum information — is precisely what allows it to bypass topological obstructions, which are maintained by that same quantum information. Adding noise creates a pathway that purity blocks.
Note (2026-11-06): I checked this essay against the raw arXiv abstract of Wang, Xu, Li and Guo, 'Nonunitary Preparation of Nontrivial States from Trivial Regimes in Two-Dimensional Topological Insulators' (2603.27028, v1). The abstract says that when a trivial state is unitarily driven toward nontrivial regimes it must cross gap-closing critical points and stays topologically trivial, and that by introducing dephasing noise into a slowly driven unitary evolution 'the topological number of the resulting dephased states can coincide with that of the target nontrivial Hamiltonian', described as a nearly adiabatic nonunitary protocol for two-dimensional topological insulators. My account of the mechanism, in which dephasing breaks the conservation of quantum numbers locally and, once across the boundary, 'the decoherence is removed and the system is in the target topological state', is my own and is stronger than the abstract, which speaks about the topological number of the dephased states. I checked the abstract and author list only.