Usage

Installation

To install SU2xSU2 using pip run

(.venv) $ pip install SU2xSU2

Its is recommended to work in a virtual environment.

Examples

import numpy as np
from SU2xSU2.SU2xSU2 import SU2xSU2
from SU2xSU2.calibrate_paras import calibrate
import SU2xSU2.analysis as analysis
import SU2xSU2.plotting as plotting

### basic model creation and simulation call signature ###
# measuring and plotting the correlation function is used in this example but the structure is identical for other observables
model_paras = {'D':2, 'L':40, 'a':1, 'ell':5, 'eps':1/5, 'beta':0.6} # define lattice and integration parameters as well as model parameter beta
# find number of integration steps and their size (under the constraint that their product is 1) to get an acceptance rate in the interval [0.6, 0.75]
paras_calibrated = calibrate(model_paras, accel=True)
# make a model with the calibrated parameters
model = SU2xSU2(**paras_calibrated)

# define the simulation parameters, what observables should be measures and where the chain is stored
# shows how own measurement functions can be passed
def func(phi, pi):
    '''
    Example measurement function which must take the field and momentum configuration as its arguments.
    '''
    L = phi.shape[0]
    O = np.sum(phi) / L**2 # observable O with no physical meaning
    return O

sim_paras = {'M':500, 'burnin_frac':0.5, 'accel':True,
            'measurements':[model.ww_correlation_func, func], 'ext_measurement_shape':[()], 'chain_paths':['corfunc_test', 'new_observable'],
            'chain_state_dir':'corfunc_test/chain_state/'}

# run simulation
model.run_HMC(**sim_paras)

# find ensemble average of the measurement chain and make plot
avg, err = analysis.get_avg_error(np.load('corfunc_test.npy'))
analysis.get_corlength(avg, err, 'corfunc_processed')
plotting.correlation_func_plot('corfunc_processed.npy', 'plots/corfunc.pdf')

# optionally can continue the previous chain
sim_paras = {'M':500, 'burnin_frac':0.0, 'accel':True,
            'measurements':[model.ww_correlation_func], 'chain_paths':['corfunc_test_continue'],
            'starting_config_path':'corfunc_test/chain_state/config.npy', 'RNG_state_path':'corfunc_test/chain_state/RNG_state.obj',
            'chain_state_dir':'corfunc_test/chain_state/'}
model.run_HMC(**sim_paras)


### compute internal energy density and plot it to compare it to coupling expansions ###
betas = np.linspace(0.1, 4, 10)
D, L = 2, 16
analysis.internal_energy_coupling_exp(betas, D, L, 5000, 0.1, chaindata_pathbase='energy_data/', simdata_path='energy.txt', plot_path='energy_exp.pdf')


### mass over lambda ratio ###
# value pairs which largely avoid finite size effects
D = 2
Ls = [40, 40, 64, 64, 64, 96, 96, 160, 160, 224, 400, 512, 700]
betas = np.array([0.6, 0.6667, 0.7333, 0.8, 0.8667, 0.9333, 1.0, 1.0667, 1.1333, 1.2, 1.2667, 1.3333, 1.4])
analysis.mass_lambda(betas, D, Ls, 1e5, 0.02)


### acceleration mass grid search ###
xi = 7.93
masses = 1 / np.linspace(0.1*xi, 3*xi, num=10, endpoint=True)
analysis.acceleration_mass_search(1e4, 0.05, 1, 2, 96, xi, masses)


### critical slowing down ###
# assuming the data produced in 'analysis.mass_lambda' was stored at the default locations
analysis.critical_slowingdown(1e5, 0.05)