dqmc-analyze

edwnh 的头像作者 edwnh

从完成的 DQMC 模拟中提取具有误差估计的物理可观察量。在计算密度、双占位、自旋相关性、结构因子或从模拟数据中测量的任何量时使用

📖 中文导读:从完成的 DQMC 模拟中提取具有误差估计的物理可观察量。在计算密度、双占位、自旋相关性、结构因子或从模拟数据中测量的任何量时使用 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: dqmc-analyze description: 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.

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