Option Matrix

This page collects the main configuration choices in one place. It is meant as a fast selection guide, not a replacement for the detailed user guide.

Grid topology

Option	Use when	Notes
periodic	the simulated domain should wrap around naturally, or you want the cleanest match to FFT-based solvers	best fit for spectral methods such as split-step-fft and etd-rk2
closed-interval	you want a bounded box with explicit edges	usually pairs most naturally with FDM-based solvers

Boundary absorption

Option	Use when	Notes
no-absorption	you want no artificial damping at the domain edge	best only when edge reflections are physically acceptable
mask	you want a simple absorbing rim near the boundary	practical default for many finite-domain runs; larger strength increases per-step damping and 0 disables it
cap	you want absorption represented as a complex absorbing potential	useful when you want boundary damping folded into the effective potential picture; strength scales the imaginary absorbing potential

Potential type

Option	Use when	Notes
zero	you want a homogeneous system without external trapping	simplest baseline for convergence and solver comparison
double-well-supergaussian	you want two localized wells with smooth but tunable walls	exposes positions, amplitudes, widths, and super-Gaussian order

Laser type

Option	Use when	Notes
uniform	the pump should be spatially homogeneous	simplest driven case
continuous-gaussian	you want a localized continuous spot	common default for a single focused pump
continuous-exp	you want a localized pump with longer radial tails than a Gaussian	matches the current implementation exactly
continuous-exp-length	you want an exponential pump whose sigma_space is the 1/e radial decay length	preferred for new length-consistent configurations
pulse-gaussian	you want repeated pulsed excitation with Gaussian spot shape	supports per-pulse strength ramping from P0 up to Pmax; choose power_definition carefully

For pulsed Gaussian studies:

power_definition	Use when	Notes
peak_amplitude	reproducing legacy fixed-spot runs	fixed P0 means fixed centre density, not fixed total dose
pulse_energy	comparing spot sizes or geometries	fixed P0 means fixed integrated pulse dose over space and time

Laser mode

Option	Use when	Notes
single	one pump definition is enough	configure directly on cfg.laser
multiple	you want several pumps in one run	define them in an external YAML file

Reservoir model

Option	Use when	Notes
single	one effective active reservoir is enough for the physics you want to resolve	cleanest baseline model
double	you need to separate inactive and active reservoir populations	adds inter-reservoir transfer timescales and more stiffness risk
quadratic-double	you model pulsed nonresonant excitation with delayed inactive-to-active feeding	uses kappa*nI^2 transfer and can retain pulse-train memory

Solver method

Option	Use when	Notes
rk4-fdm	you want the safest reference solver	best first choice for validation and comparisons
rk4-fdm-fused	you want a faster FDM-style path without changing the basic method family	good follow-up once rk4-fdm is validated
rk4-cuda	you want FDM GPU production runs on periodic or closed-interval grids	FDM production/reference; stage-coupled quadratic-double
ifrk-fft-cuda	you want selectable-order spectral integration on a periodic grid	solver.tableau selects RK4, DP5, DOP853, Verner 7, or Verner 9; stage-coupled reservoirs
ifrk4-fft-cuda	you need the historical spectral RK4 method name	alias for ifrk-fft-cuda with RK4 coefficients
split-step-fft	second-order spectral evolution on a periodic grid	Strang splitting with two reservoir half-steps; supports potentials and CAP; does not evaluate kinetic_relaxation_eta
split-step-fft-cuda	second-order spectral evolution on a periodic CUDA grid	supports built-in reservoirs; does not evaluate kinetic_relaxation_eta
etd-rk2	second-order exponential time differencing on a periodic grid	periodic only; second-order midpoint reservoir coupling; does not evaluate kinetic_relaxation_eta
ip-rk4	spectral-style RK4 reference	requires periodic; stage-coupled reservoir; for GPU production use ifrk4-fft-cuda

RK4 CUDA Laplacian

Option	Use when	Notes
five-point	you want historical rk4-cuda behavior	fastest and backward-compatible
isotropic-9pt	diagonal/axis artifacts suggest stencil anisotropy	requires square cells and should still be grid-checked

Initial seed mode

Option	Use when	Notes
legacy_positive_uniform	you need to reproduce old runs	biased positive-uniform complex seed
complex_gaussian_zero_mean	you want an unbiased stochastic seed	no spectral cutoff
filtered_complex_gaussian	you want an explicit condensate-band seed cutoff	requires init_k_cutoff_um; report/sweep the cutoff

Backend choice

Option	Use when	Notes
use_gpu = False	you want the most portable CPU baseline	best default for correctness checks
use_gpu = True	CuPy is available and you want GPU execution	validate against a CPU or FDM baseline first

Practical default path

If you want a stable starting point, the shortest conservative path is:

1. grid_type = "periodic" or closed-interval, depending on the physical boundary you want
2. potential_type = "zero" unless you explicitly need trapping
3. laser_type = "continuous-gaussian"
4. reservoir_type = "single"
5. solver.method = "rk4-fdm"
6. use_gpu = False