Fermionic RNNs and NQS quasiparticle dispersions

RNN implementation to simulate the bosonic and fermionic 2D $t-J$ model, or its general form, the $t$-XXZ model, using pytorch, see (our paper). The code for the ground state search of bosonic and fermionic systems and an run.py (run_sr.py) file to run it (with minSR) can be found here. All data shown in the paper is provided here.

In order to run the code, run.py (or run_sr.py and stoch_reconfig.py), helper.py, observables.py,localenergy.py and model.py have to be in the same directory. Furthermore, you need to create a folder for the results, e.g. mkdir 4x4_qubits/open for a system with 4x4 sites and open boundaries. You can run the code by calling e.g.

run.py -Nx 4 -Ny 4 -den 1 -t 3 -Jp 1 -Jz 1 -boundsx 0 -boundsy 0 -load 0 -antisym 0 -hd 50 -sym 0

for a bosonic $N_x\times N_y=4\times 4$ square lattice system with open boundaries (boundsx=boundsy=0), $t=3$, $J_{\pm}=J_{z}=1$ and hidden dimension $h_d=50$.

Furthermore, the one-hole dispersions can be calculated by enforcing a target momentum $k_\mathrm{target}$ by adding this constrain to the cost function:

The RNN implementation is based on Hibat-Allah et al. (2020), see also their Git repository.