Configuration

Configuration Setup

The SPLIT orchestrator runs by ingesting two distinct JSON configuration files:

1. emri_config.json: Controls the physical properties and the signal injection parameters.

   • Defines fundamental parameters including m1, m2, a, p0, e0, and starting orientations, phases.

   • Sets up the trajectory slice settings, dictating the total time T in years, number of blocks Nblocks, desired network SNR scaling, frequency bounds, and prior tolerance allowances sigma_prior.

   • Model Specification: Controls the waveform generator used for data injection (data_model) and the template used for inference (analysis_model). These natively accept standard FEW string identifiers (e.g., "FastKerrEccentricEquatorialFlux"). You can also inject purely custom environmental/glitch waveforms via code while recovering with vacuum GR. The user can also supply additional arguments for the custom waveform model using "add_args".

   • Response Wrapper: Allows toggling the LISA instrument response simulation via the "response" boolean flag. If false, the pipeline defaults to evaluating raw strain using the Long-Wavelength Approximation (LWA) sensitivity curve.

2. sample_config.json: Manages the Eryn MCMC sampler mechanics.

   • Assigns static and evolving parameters that should be inferred through static_params and evolving_params, respectively.

   • Assigns the set of fixed evolving parameters through fixed_evolving and automatically assigns any remaining model parameters to fixed_static.

   • Configures Eryn’s EnsembleSampler method with nwalkers, ntemps, nsteps, probability thresholds for the custom moves, etc. Allows for dynamically linked resuming output chains via new_filename/old_filename.

See emri_config.json and sample_config.json for working examples.

emri_config.json reference

Key	Type	Description
m1, m2	float	Primary and secondary masses (solar masses).
a	float	Dimensionless spin of the primary.
p0, e0	float	Initial semi-latus rectum and eccentricity.
xI0	float	Initial cosine of the orbital inclination.
dist	float	Luminosity distance (Gpc).
qS, phiS	float	Sky-location polar angles (radians).
qK, phiK	float	Spin-orientation polar angles (radians).
Phi_phi0, Phi_r0, Phi_theta0	float	Initial orbital phases.
T	float	Total observation time (years).
Nblocks	int	Number of loosely-coherent blocks. Set to 1 for a fully coherent analysis.
dt	float	Sampling cadence (seconds).
fmin, fmax	float	Frequency-band edges used for likelihood (Hz).
SNR_target	float	Target network SNR; the injection is rescaled to match.
sigma_prior	float | list	Tolerance(s) controlling the per-block Student-t prior width.
data_model	str	FEW string identifier of the waveform model used for injection.
analysis_model	str	FEW string identifier of the template used for inference.
add_args	dict	Extra keyword arguments forwarded to a custom waveform model.
response	bool	If true, apply the LISA ResponseWrapper; if false, use the LWA sensitivity curve on raw strain.

sample_config.json reference

Key	Type	Description
static_params	list[str]	Parameter names to be sampled as static (shared across all blocks).
evolving_params	list[str]	Parameter names to be sampled as evolving (per-block).
fixed_evolving	list[str]	Evolving parameters held fixed at their injected values.
nwalkers	int	Number of MCMC walkers per temperature. Minimum recommended: 4 × ndim.
ntemps	int	Number of parallel-tempering temperatures.
nsteps	int	Number of sampler iterations.
burn	int	Number of burn-in iterations before the main run.
jitter	float	Gaussian scatter applied when initializing walkers around the truth.
custom_bounds	dict	Optional per-parameter (low, high) overrides of the default prior bounds.
move_probs	dict	Selection probabilities for each custom MCMC move.
new_filename	str	Output HDF5 backend filename.
old_filename	str | null	If non-null, resume from this backend instead of starting fresh.

Tip: Any model parameter listed in neither static_params, evolving_params, nor fixed_evolving is automatically appended to fixed_static and held at its injected value.