Validation Gallery

OpenQuantumSim should earn trust by matching simple analytic limits and established reference implementations. The current validation gallery covers a closed-form qubit decay problem, a damped Jaynes-Cummings comparison against QuTiP, and the command used for direct QuTiP speed comparisons.

Analytic Qubit Decay

A two-level atom with spontaneous emission has excited-state population

\[P_e(t) = \exp(-\gamma t).\]

The corresponding OpenQuantumSim model is intentionally small enough to inspect line by line:

import numpy as np
import openquantumsim as oqs

gamma = 0.35
atom = oqs.SpinSpace(0.5, label="atom")
H = 0.0 * oqs.sigmaz(atom)
collapse = np.sqrt(gamma) * oqs.sigmam(atom)
excited = oqs.basis(atom, "up")
rho0 = oqs.ket2dm(excited)
P_excited = oqs.Operator(oqs.ket2dm(excited), atom, "P_excited")
times = np.linspace(0.0, 6.0, 61)

result = oqs.mesolve(
    H,
    rho0,
    times,
    c_ops=[collapse],
    e_ops=[P_excited],
    options=oqs.Options(rtol=1e-9, atol=1e-11),
)

expected = np.exp(-gamma * times)
max_error = np.max(np.abs(result.expect[0].real - expected))

This case is covered by tests/physics/test_qubit_decay.py.

Jaynes-Cummings Against QuTiP

The Jaynes-Cummings validation uses the same cavity-atom Hamiltonian and collapse operators in OpenQuantumSim and QuTiP, then compares cavity photon number and atomic excited-state population.

Install the optional validation dependency, then run:

python -m pip install -e ".[validation]"
python scripts/validate_jaynes_cummings_qutip.py \
    --cavity-dim 5 \
    --time-points 81 \
    --t-final 8.0 \
    --tolerance 5e-7 \
    --output-csv runs/jaynes_cummings_qutip_validation.csv

The script exits nonzero if either observable exceeds the requested tolerance. A typical successful run prints:

max photon-number delta: ...
max excited-population delta: ...
tolerance: 5.000e-07
Jaynes-Cummings QuTiP validation: passed

This case is covered by tests/physics/test_jaynes_cummings_qutip.py when QuTiP is installed.

QuTiP Speed Benchmark

Use the benchmark harness to compare OpenQuantumSim and QuTiP on the same models and time grids:

python benchmarks/bench_vs_qutip.py \
    --case qubit \
    --case jc5 \
    --time-points 101 \
    --t-final 6.0 \
    --repeats 3 \
    --output-json runs/bench_vs_qutip.json

The report includes median wall time, minimum wall time, speedup relative to QuTiP, and the maximum expectation-value deviation from the QuTiP reference. For reproducible benchmark numbers, record the CPU model, Julia thread count, Python version, Julia version, OpenQuantumSim commit, and QuTiP version next to the generated JSON.

The current published benchmark snapshot is in Performance Benchmarks.