This package provides lightweight, low-level access to neutron reflectometry kernels in jax.
The provided kernels are all jit and grad compatible and easy to integrate with your favourite optimisation or uncertainty quantification library for fast parameterisation of stratified reflectometry models. Several examples for popular libraries are available in the examples folder.
Reflectometry models are created by creating reflectivity kernels. The call to the refjax.kernel function takes three arguments.
kernel = refjax.kernel(model, smearing, jit=True)Currently implemented models:
refjax.models.abelesAbeles matrix formalism for stratified media
Currently implemented smearing kernels
refjax.smear.no_smearNo smearing kernel (default)refjax.smear.constant_gaussianConstant gaussian smearing at every Q valuerefjax.smear.pointwise_gaussianPointwise Gaussian approximation to the true smearing kernel (requires FWHM resolution measurements)
Once a kernel has been created, it is called with the model parameters and Q (and FWHM values for smeared kernels).
R = kernel(params, Q, FWHM) # fast beacuse jitCheck out the examples folder for more advanced usage and examples of parameter identification schemes.
Having trouble? Feature request? Raise an issue on the GitHub page.