SKILL.md
Generate Simulation Files
Inputs
| Input | Required | Default | Description |
|---|---|---|---|
Nx, Ny |
No | 8 | Lattice dimensions |
U |
No | 6.0 | On-site interaction strength |
mu |
No | 0.0 | Chemical potential |
tp |
No | 0.0 | Next-nearest neighbor hopping |
dt |
No | 0.1 | Imaginary time step |
L |
No | 40 | Number of time slices (beta = L*dt) |
nsweepwarm |
No | 200 | Warmup sweeps |
nsweepmeas |
No | 2000 | Measurement sweeps |
Nfiles |
No | 1 | Number of independent bins |
prefix |
No | sim |
Output path prefix |
period_uneqlt |
No | 0 | Unequal-time period (0=disabled) |
Outputs
{prefix}0.h5,{prefix}1.h5, ... - HDF5 simulation files{prefix}.h5.params- common params HDF5 file
Procedure
Option A: CLI (simple cases)
dqmc-util gen Nx=6 Ny=6 U=4.0 mu=0.0 dt=0.1 L=50 Nfiles=4 prefix=data/run/binOption B: Python (parameter scans)
from dqmc_util import gen_1band_hub
gen_1band_hub.create_batch(
prefix="data/run/bin", Nfiles=4,
Nx=6, Ny=6, U=4.0, mu=0.0, tp=0.0, dt=0.1, L=50
)Choosing dt and L for a target temperature T:
import numpy as np
T = 0.25 # target temperature
U = 4.0
beta = 1.0 / T
dt = min((0.05/U)**0.5, beta / 10) # Trotter error constraint
L = int(np.ceil(beta / dt / 5) * 5) # L must be divisible by n_matmul and period_eqlt (defaults: 5)
dt = beta / L # recalculate exact dtValidation
- Files created:
ls {prefix}_*.h5 {prefix}.h5.params - Trotter error check:
U * dt^2 <= 0.05
Failure Modes
| Symptom | Cause | Recovery |
|---|---|---|
| "File exists" error | Previous files not cleaned | Remove old files or use overwrite=1 argument |
| ValueError: L must be divisible ... | L must be multiple of nmatmul and periodeqlt | Adjust value of L and dt, keeping their product (beta) constant |
Do Not
- Overwrite existing simulation files without confirmation
- Ignore the Trotter error bound (
U * dt^2 <= 0.05)