dqmc-analyze

edwnh's avatarfrom edwnh

Extract physical observables with error estimates from completed DQMC simulations. Use when computing density, double occupancy, spin correlations, structure factors, or any measured quantity from simulation data.

⭐16stars🔀5forks📁View on GitHub🕐Updated Dec 31, 2025⬇️2downloads

When & Why to Use This Skill

This Claude skill automates the extraction and analysis of physical observables from Determinant Quantum Monte Carlo (DQMC) simulation data. It streamlines the post-processing of HDF5 binary files to calculate critical physical quantities—such as fermion sign, density, and spin correlations—while providing rigorous statistical error estimates (mean and standard error) essential for scientific validation.

Use Cases

  • Automated Physical Property Extraction: Quickly compute density, double occupancy, and Green's functions from completed simulation directories without manual data parsing.
  • Statistical Reliability Validation: Automatically check if the fermion sign error bars are within acceptable limits to ensure the simulation doesn't suffer from a severe sign problem.
  • Batch Result Aggregation: Efficiently collect and summarize data from multiple simulation runs across different parameter regimes for comparative study.
  • Derived Quantity Calculation: Use extracted correlators to derive complex physical metrics like magnetic moment squared or zero-frequency susceptibilities.
  • Simulation Troubleshooting: Identify failure modes such as missing observables or insufficient statistics by analyzing error bar magnitudes and file structures.
namedqmc-analyze
descriptionExtract physical observables with error estimates from completed DQMC simulations. Use when computing density, double occupancy, spin correlations, structure factors, or any measured quantity from simulation data.

Analyze Results

Inputs

  • Directory containing bin_*.h5 files (completed simulations)
  • Observable names (see table below)

Outputs

  • Dictionary with parameters and (mean, stderr) tuples for each observable

Procedure

Basic analysis:

from dqmc_util import analyze_hub

data = analyze_hub.get("data/run/", "sign", "den", "zzr")

print(f"sign = {data['sign'][0]:.4f} +/- {data['sign'][1]:.4f}")
print(f"density = {data['den'][0]:.4f} +/- {data['den'][1]:.4f}")

Available observables:

Name Description Requires
sign Fermion sign -
den Density -
docc Double occupancy -
gr, gk Green's function (real/k-space) -
nnr, nnq Density correlator / structure factor -
zzr, zzq Spin-z correlator / structure factor -
xxr Spin-x correlator -
swq0 S-wave pair structure factor -
nnrw0, zzrw0 Zero-freq susceptibilities period_uneqlt > 0
dwq0t D-wave pair susceptibility period_uneqlt > 0

Collect from multiple directories:

import os

def collect_results(base_dir, observables):
    results = []
    for subdir in sorted(os.listdir(base_dir)):
        path = os.path.join(base_dir, subdir)
        if os.path.isdir(path):
            try:
                results.append(analyze_hub.get(path + "/", *observables))
            except Exception as e:
                print(f"Skipping {path}: {e}")
    return results

Compute derived quantities:

# Magnetic moment squared from spin correlator
path = "data/run/"
data = analyze_hub.get(path, "zzr")
mz2 = 4 * data["zzr"][0][0, 0]       # [0] = mean, shape (Ny, Nx)
mz2_err = 4 * data["zzr"][1][0, 0]   # [1] = stderr

Validation

  • Errorbar on sign is significantly less than mean. Otherwise, sign problem is too severe.
  • Errorbars on observable are reasonable (not >> mean)

Failure Modes

Symptom Cause Recovery
KeyError for observable Observable not computed Check period_uneqlt setting
"No files found" Wrong path or no bin_*.h5 Verify directory structure
Large error bars Insufficient statistics Run more sweeps or bins