Scientific-package integration - honest evaluation¶
This document records which third-party scientific packages were evaluated for integration into qgrav, and the reasoning for each decision. It exists because a request to "integrate QuTiP / Qiskit / GEANT4 / LAMMPS" should be answered honestly: only one of those four genuinely fits light-pulse atom-interferometric gravimetry, and saying so is more useful than building four integrations of wildly varying value.
Summary¶
| Package | Domain | Fit to qgrav | Decision |
|---|---|---|---|
| QuTiP | Open quantum systems / Lindblad dynamics | Strong | Integrated (v1.2.1) |
| Qiskit | Gate-based quantum computing | Weak (novelty only) | Deferred / optional |
| GEANT4 | Radiation & particle transport (Monte Carlo) | None for the core problem | Documented, not built |
| LAMMPS | Classical molecular dynamics | None | Documented, not built |
QuTiP - integrated (v1.2.1)¶
Why it fits. qgrav's core physical object is a driven two-level (effective
Raman) system whose internal state evolves under Rabi dynamics, with optional
decoherence (spontaneous emission, dephasing). That is exactly what QuTiP is
built for. QuTiP integrates the Schrödinger equation (sesolve) and the Lindblad
master equation (mesolve) numerically - a completely independent code path
from qgrav's closed-form 2×2 propagator matrix.
What we built. src/qgrav/validation/qutip_crosscheck.py:
- aisim_rabi_population vs qutip_rabi_population vs analytic_rabi_population
over a grid of (Ω_eff, δ, τ). Result: qgrav's matrix matches the closed form
to ~3×10⁻¹⁶ (machine precision) and QuTiP's numerical integrator agrees to
~1.6×10⁻⁶. This is the independent-simulator validation the v1.0
honest-assessment asked for (caveat C14, partial): a second, unrelated
engine reproduces qgrav's Raman dynamics.
- qutip_spontaneous_emission_loss (Lindblad mesolve with a decay collapse
operator) gives an order-of-magnitude cross-check on the analytic
spontaneous_emission_loss_probability.
Status. Optional dependency (pip install qgrav[qutip]); tests gated with
pytest.importorskip("qutip") so the core suite is unaffected when QuTiP is
absent.
Qiskit - deferred (novelty, not physics)¶
Why it does not fit (yet). Qiskit models gate-based quantum computation on qubit registers. A light-pulse atom interferometer is not a gate circuit; the three-pulse Mach-Zehnder is a continuous-time Rabi/free-evolution sequence on a momentum-state ladder, not a discrete gate sequence on logical qubits.
One could map the MZ sequence onto a 1-qubit circuit (π/2 → √X-like, free evolution → Rz, π → X) purely for pedagogy or visualization, but it would compute nothing the AISim/QuTiP path does not already give, and it risks implying a quantum-computing capability the project does not have. Deferred as an optional outreach/teaching feature, not a v1.2 deliverable.
GEANT4 - not applicable¶
GEANT4 is a Monte-Carlo toolkit for the passage of particles (and radiation) through matter - detectors, dosimetry, high-energy physics. The only conceivable touch-point with an atom gravimeter is modelling stray-light or cosmic-ray backgrounds in the detection region, which is a third-order systematic far below anything qgrav currently models and is not part of the gravimetric phase measurement at all. Integrating GEANT4 would be a large, C++-toolkit-bridging effort with no path to the published-reference validation bar. Not built; documented here so the decision is on record.
LAMMPS - not applicable¶
LAMMPS is a classical molecular-dynamics engine for dense systems of interacting classical particles (materials, biomolecules). Cold-atom interferometry operates on a dilute, ultracold quantum gas where the relevant dynamics are single-atom quantum evolution and (at most) mean-field BEC physics - neither of which LAMMPS addresses. The only superficial overlap (the classical MOT-loading cloud) is already captured by qgrav's thermal-ensemble sampling. Not built.
If the scope changes¶
- A Qiskit circuit view of the MZ for teaching could be added as
qgrav/visuals/circuit_view.py(optional dep) without affecting the physics. - BEC-source modelling (for instruments like the HUST BEC gravimeter) would be a better use of effort than any of GEANT4/LAMMPS, and would reuse the existing ensemble/source layer rather than a new external engine.