Efficient representation of fully many-body localized systems using tensor networks (Dr. T. Wahl)
- Date: Mar 15, 2017
- Time: 11:30 AM - 12:30 PM (Local Time Germany)
- Speaker: Dr. Thorsten Wahl
- University of Oxford, Department of Physics
- Room: Herbert Walther Lecture Hall
- Host: MPQ, Theory Division
This makes tensor network states ideally suited to
represent such fully many-body localized systems. Building on the ansatz
proposed in Phys. Rev. B 94, 041116(R) (2016), I will present a tensor network
that is able to capture the full set of eigenstates of such MBL systems
efficiently: For a given system size, local observables can be approximated
with an error that decreases as an inverse polynomial of the computational
cost, which is an exponential improvement over the previous ansatz. If the
system size is increased, the computational cost needs to grow only linearly
with the system size in order to keep the accuracy fixed. The technique turns
out to be highly accurate deep in the localized regime and maintains a surprising
degree of accuracy in predicting certain local quantities even in the vicinity
of the dynamical phase transition. Finally, the power of the technique is
demonstrated on systems of 72 sites, where clear signatures of the phase
transition can be seen.