Source code for scripts.clock.correlation_comparison

"""
Comparison of spin-spin correlation functions G(r) for the q-state Clock model.
Analyzes correlation behavior in ordered, quasi-ordered, and disordered phases.

The q=6 clock model has two Kosterlitz-Thouless transitions at T1 ≈ 0.68
and T2 ≈ 0.92 (Challa and Landau 1986; Tomita and Okabe 2002 obtain 0.7014
and 0.9008), yielding three distinct correlation regimes:
long-range order below T1, algebraic (quasi-long-range) order between T1 and T2,
and exponential decay above T2.
"""
from __future__ import annotations

import argparse
import logging
from typing import Any

import matplotlib.pyplot as plt
import numpy as np

from scripts.clock._model_choice import add_clock_model_arguments, resolve_clock_model
from utils.observables import (
    CorrelationPoint,
    fit_correlation_exponent,
    fit_correlation_length,
    measure_correlation_point,
)
from utils.plotting import ensure_results_dir, save_plot
from utils.system import parallel_sweep, parse_args_compat, setup_logging

# Approximate transition temperatures of the discrete q=6 clock model
# (Challa and Landau 1986, as quoted by Tomita and Okabe 2002).
T1_CLOCK6: float = 0.68
T2_CLOCK6: float = 0.92

# Anisotropy of the continuous model when --continuous is given without
# --aniso; it matches the ClockSimulation constructor default used before the
# discrete model became the default.
_DEFAULT_CONTINUOUS_ANISO = 1.0


[docs] def main() -> None: """Run the clock model correlation comparison analysis.""" parser = argparse.ArgumentParser( description='q-state Clock Model Correlation Comparison', ) parser.add_argument('--size', type=int, default=128, help='Linear lattice size L') parser.add_argument('--q', type=int, default=6, help='Number of clock states') add_clock_model_arguments(parser=parser, default_aniso=_DEFAULT_CONTINUOUS_ANISO) parser.add_argument('--steps', type=int, default=4000, help='Measurement steps') parser.add_argument('--eq-probe', type=int, default=200, help='Convergence probe chunk size') parser.add_argument('--eq-max', type=int, default=50000, help='Max equilibration steps') parser.add_argument('--interval', type=int, default=20, help='Sample interval') parser.add_argument( '--seed', type=int, default=520, help='Random seed shared by all three temperature points', ) parser.add_argument( '--t-ordered', type=float, default=None, help='Temperature below T1 (default 0.5, discrete q=6 only)', ) parser.add_argument( '--t-quasi', type=float, default=None, help='Temperature between T1 and T2 (default 0.8, discrete q=6 only)', ) parser.add_argument( '--t-disordered', type=float, default=None, help='Temperature above T2 (default 1.2, discrete q=6 only)', ) parser.add_argument('--output-dir', type=str, default='results/clock', help='Output directory') parser.add_argument('--log-file', type=str, default=None, help='Optional log file path') parser.add_argument('--verbose', action='store_true', help='Enable verbose logging') args = parse_args_compat(parser=parser) choice = resolve_clock_model( parser=parser, args=args, default_aniso=_DEFAULT_CONTINUOUS_ANISO, ) # Only q >= 5 has the intermediate algebraic phase this comparison fits. if args.q < 5: parser.error(f'--q {args.q} has no quasi-ordered phase; use q >= 5') # The default temperatures bracket the transitions of the discrete q=6 # model only. T1 and T2 shift with q and, for the continuous model, with A, # so other choices must name all three temperatures. default_model = choice.is_discrete and args.q == 6 phase_temps = (args.t_ordered, args.t_quasi, args.t_disordered) if default_model: T_ORDERED, T_QUASI, T_DISORDERED = ( default if given is None else float(given) for given, default in zip(phase_temps, (0.5, 0.8, 1.2), strict=True) ) elif any(T is None for T in phase_temps): parser.error( 'The default temperatures apply to the discrete q=6 model only; ' 'give --t-ordered, --t-quasi, and --t-disordered for this model' ) else: T_ORDERED, T_QUASI, T_DISORDERED = (float(T) for T in phase_temps) if not T_ORDERED < T_QUASI < T_DISORDERED: parser.error('Temperatures must satisfy t-ordered < t-quasi < t-disordered') log_level = logging.DEBUG if args.verbose else logging.INFO logger = setup_logging(level=log_level, log_file=args.log_file) logger.info( f'Starting clock correlation comparison ' f'(q={args.q}, {choice.variant}, L={args.size}, steps={args.steps})' ) logger.info( f'Temperatures: ordered={T_ORDERED}, quasi={T_QUASI}, disordered={T_DISORDERED}' ) common: dict[str, Any] = dict( model_cls=choice.model_cls, model_kwargs=choice.model_kwargs, size=args.size, seed=args.seed, eq_probe=args.eq_probe, eq_max=args.eq_max, meas_steps=args.steps, interval=args.interval, ) points = [ # Below T1 a discrete random start can freeze into a domain state; # stuck detection then ends the run and the ordered start is measured. # The continuous model has no such accepted stuck state. CorrelationPoint( label=label, temperature=T, detect_stuck=choice.is_discrete and label == 'ordered', **common, ) for label, T in ( ('ordered', T_ORDERED), ('quasi', T_QUASI), ('disordered', T_DISORDERED), ) ] results = { label: (r, G) for label, r, G in parallel_sweep( worker_func=measure_correlation_point, params=points, ) } r_ordered, G_ordered = results['ordered'] r_quasi, G_quasi = results['quasi'] r_disordered, G_disordered = results['disordered'] # Verify r arrays are identical (all from same lattice size). if not (np.array_equal(r_ordered, r_quasi) and np.array_equal(r_quasi, r_disordered)): raise RuntimeError( 'Radial distance arrays differ between temperature points; ' 'all three measurements must share the same lattice size.' ) r = r_ordered # Between T1 and T2 the discrete anisotropy is irrelevant at long distances # and the model behaves like XY, so the same spin-wave exponent applies; # above T2 correlations decay exponentially with a finite length. eta_quasi = fit_correlation_exponent(r=r, G=G_quasi) xi_disordered = fit_correlation_length(r=r, G=G_disordered) eta_spin_wave = T_QUASI / (2.0 * np.pi) logger.info( f'T={T_QUASI}: fitted eta = {eta_quasi:.4f} ' f'(spin-wave prediction {eta_spin_wave:.4f})' ) logger.info(f'T={T_DISORDERED}: fitted correlation length xi = {xi_disordered:.4f}') # --- Plotting --- fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(14, 6)) ax1.loglog(r[1:], G_ordered[1:], 'o-', label=f'T={T_ORDERED} (ordered, T < T₁)', alpha=0.7) ax1.loglog(r[1:], G_quasi[1:], 's-', label=f'T={T_QUASI} (quasi-ordered)', alpha=0.7) ax1.loglog( r[1:], G_disordered[1:], 'x-', label=f'T={T_DISORDERED} (disordered, T > T\u2082)', alpha=0.7, ) ax1.set_title('Log-Log Plot') ax1.set_xlabel('Distance r') ax1.set_ylabel('Correlation G(r)') ax1.legend() ax1.grid(True, which='both', ls='-', alpha=0.5) ax2.plot(r, G_ordered, 'o-', label=f'T={T_ORDERED} (ordered)', alpha=0.7) ax2.plot(r, G_quasi, 's-', label=f'T={T_QUASI} (quasi-ordered)', alpha=0.7) ax2.plot(r, G_disordered, 'x-', label=f'T={T_DISORDERED} (disordered)', alpha=0.7) ax2.set_yscale('log') ax2.set_title('Semi-Log Plot') ax2.set_xlabel('Distance r') ax2.set_ylabel('Correlation G(r)') ax2.legend() ax2.grid(True, which='both', ls='-', alpha=0.5) fig.suptitle( f'{args.q}-state Clock Model ({choice.variant}): ' f'Correlation Comparison (L={args.size})' ) output_dir: str = ensure_results_dir(directory=args.output_dir) save_plot(filename='correlation_comparison.png', directory=output_dir) # Save data for notebook consumption. npz_path = f'{output_dir}/correlation_comparison.npz' np.savez_compressed( npz_path, r=r, G_ordered=G_ordered, G_quasi=G_quasi, G_disordered=G_disordered, T_ordered=T_ORDERED, T_quasi=T_QUASI, T_disordered=T_DISORDERED, # The reference transition temperatures belong to the discrete q=6 # model and are NaN for any other choice. T1=T1_CLOCK6 if default_model else np.nan, T2=T2_CLOCK6 if default_model else np.nan, L=args.size, q=args.q, model_variant=choice.variant, steps=args.steps, eq_probe=args.eq_probe, eq_max=args.eq_max, sample_interval=args.interval, seed=args.seed, eta_quasi=eta_quasi, eta_quasi_spin_wave=eta_spin_wave, xi_disordered=xi_disordered, ) logger.info(f'Data saved to {npz_path}')
if __name__ == '__main__': main()