jesterTOV.inference.likelihoods.chieft.ChiEFTLikelihood

class ChiEFTLikelihood(low_filename=None, high_filename=None, nb_n=100)

Bases: LikelihoodBase

Likelihood function enforcing chiral EFT constraints on the nuclear EOS.

This likelihood evaluates how well a candidate equation of state agrees with theoretical predictions from chiral effective field theory in the low-density regime (0.75 - 2.0 n_sat). The chiEFT calculations provide a band of allowed pressures at each density; EOSs within the band receive higher likelihood, while those outside are penalized proportional to their deviation.

The likelihood is computed as an integral over density of a penalty function that assigns: - Weight 1.0 for pressures within the chiEFT band - Exponential penalty for pressures outside the band (slope β = 6/(p_high - p_low))

This formulation smoothly incorporates theoretical uncertainties while strongly disfavoring unphysical EOSs.

Parameters:

• low_filename (str | Path | None, optional) – Path to data file containing the lower boundary of the chiEFT allowed band. The file should have three columns: density [fm⁻³], pressure [MeV/fm³], energy density [MeV/fm³] (only first two are used). If None, defaults to the Koehn et al. (2025) low band in the package data.

• high_filename (str | Path | None, optional) – Path to data file containing the upper boundary of the chiEFT allowed band. Same format as low_filename. If None, defaults to the Koehn et al. (2025) high band in the package data.

• nb_n (int, optional) – Number of density points for numerical integration of the penalty function. More points provide better accuracy but increase computation time. Default is 100, which provides good balance for typical applications.

Variables:

• n_low (Float[Array, " n_points"]) – Density grid for lower bound in units of n_sat (saturation density = 0.16 fm⁻³)

• p_low (Float[Array, " n_points"]) – Pressure values for lower bound in MeV/fm³

• n_high (Float[Array, " n_points"]) – Density grid for upper bound in units of n_sat

• p_high (Float[Array, " n_points"]) – Pressure values for upper bound in MeV/fm³

• EFT_low (Callable[[Float | Float[Array, "..."]], Float | Float[Array, "..."]]) – Interpolation function returning lower bound pressure at given density

• EFT_high (Callable[[Float | Float[Array, "..."]], Float | Float[Array, "..."]]) – Interpolation function returning upper bound pressure at given density

• nb_n (int) – Number of density points used for integration

Notes

The penalty function f(p_sample, p_low, p_high) is defined as:

\[\begin{split}f(p) = \begin{cases} 1 - \beta(p - p_{high}) & \text{if } p > p_{high} \\ 1 & \text{if } p_{low} \leq p \leq p_{high} \\ 1 - \beta(p_{low} - p) & \text{if } p < p_{low} \end{cases}\end{split}\]

where β = 6/(p_high - p_low) controls the penalty strength.

The integration is performed from 0.75 n_sat (lower limit of chiEFT validity) to nbreak (where the CSE extension begins, if present).

See also

REXLikelihood

Nuclear radius constraints from PREX/CREX experiments

Examples

Create a chiEFT likelihood with default data:

>>> from jesterTOV.inference.likelihoods import ChiEFTLikelihood
>>> likelihood = ChiEFTLikelihood(nb_n=100)
>>> log_like = likelihood.evaluate(params, data={})

__init__(low_filename=None, high_filename=None, nb_n=100)

Methods

__init__([low_filename, high_filename, nb_n])
evaluate(params)	Evaluate the log-likelihood for chiEFT constraints.

Attributes

data	The data for the likelihood.
model	The model for the likelihood.
n_low
p_low
n_high
p_high
EFT_low
EFT_high
nb_n
n_max_nsat

EFT_high: Callable[[Float | Float[Array, '...']], Float | Float[Array, '...']]

EFT_low: Callable[[Float | Float[Array, '...']], Float | Float[Array, '...']]

evaluate(params)

Evaluate the log-likelihood for chiEFT constraints.

Parameters:

params (dict[str, Float | Array]) – Dictionary containing EOS quantities from the transform. Required keys: - “n” : Baryon number density grid (geometric units, i.e. fm⁻³ × utils.fm_inv3_to_geometric) - “p” : Pressure values on density grid (geometric units, i.e. MeV/fm³ × utils.MeV_fm_inv3_to_geometric) - “nbreak” : Breaking density where CSE begins (fm⁻³, physical units)

Return type:

Float

Returns:

Float – Natural logarithm of the likelihood. Higher values indicate better agreement with chiEFT predictions. The value is normalized by the integration range so that perfect agreement gives log L ≈ 0.

Notes

All density arithmetic inside this method is performed in units of n_sat (saturation density = 0.16 fm⁻³). Input quantities are converted at the start:

• nbreak [fm⁻³] → nbreak / 0.16 [n_sat]

• n [geometric] → n / fm_inv3_to_geometric / 0.16 [n_sat]

• p [geometric] → p / MeV_fm_inv3_to_geometric [MeV/fm³]

The integration runs from 0.75 n_sat (lower limit of chiEFT validity) to min(nbreak, n_max_nsat) [n_sat] using nb_n equally spaced points. The upper limit is capped at n_max_nsat (2.0 n_sat for the default Koehn et al. 2025 bands) to avoid integrating over the flat constant extrapolation that jnp.interp produces beyond the data range, which has no physical meaning. If nbreak falls below 0.75 n_sat the upper limit is clamped to 0.75 + 1e-8 n_sat to prevent a degenerate integration range.

n_high: Float[Array, 'n_points']

n_low: Float[Array, 'n_points']

n_max_nsat: float

nb_n: int

p_high: Float[Array, 'n_points']

p_low: Float[Array, 'n_points']