Abstract
The tunneling decay event of a metastable state in a fully connected quantum spin model can be simulated efficiently by path-integral quantum Monte Carlo (QMC) [Isakov et al., Phys. Rev. Lett. 117, 180402 (2016)]. This is because the exponential scaling with the number of spins of the thermally assisted quantum tunneling rate and the Kramers escape rate of QMC are identical [Jiang et al., Phys. Rev. A 95, 012322 (2017)], a result of a dominant instantonic tunneling path. In Isakov et al., it was also conjectured that the escape rate in open-boundary QMC is quadratically larger than that of conventional periodic-boundary QMC; therefore, open-boundary QMC might be used as a powerful tool to solve combinatorial optimization problems. The intuition behind this conjecture is that the action of the instanton in open-boundary QMC is a half of that in periodic-boundary QMC. Here, we show that this simple intuition—although very useful in interpreting some numerical results—deviates from the actual situation in several ways. Using a fully connected quantum spin model, we derive a set of conditions on the positions and momenta of the end points of the instanton, which remove the extra degrees of freedom due to open boundaries. In comparison, the half-instanton conjecture incorrectly sets the momenta at the end points to zero. We also found that the instantons in open-boundary QMC correspond to quantum tunneling events in the symmetric subspace (maximum total angular momentum) at all temperatures, whereas the instantons in periodic-boundary QMC typically lie in subspaces with lower total angular momenta at finite temperatures. This leads to a lesser-than-quadratic speedup at finite temperatures. The results provide useful insights in utilizing open-boundary QMC to solve hard optimization problems. We also outline the generalization of the instantonic tunneling method to many-qubit systems without permutation symmetry using spin-coherent-state path integrals.
- Received 24 August 2017
DOI:https://doi.org/10.1103/PhysRevA.96.042330
©2017 American Physical Society