scqubits.core.diag.esys_jax_dense#

scqubits.core.diag.esys_jax_dense(matrix, evals_count, **kwargs)[source]#

Diagonalization based on jax’s (dense) jax.scipy.linalg.eigh function. Both evals and evecs are returned.

If available, different backends/devics (e.g., particular GPUs) can be set though jax’s interface, see https://jax.readthedocs.io/en/latest/user_guides.html

Note, that jax’s documentation is inconsistent, and eigvals and/or subset_by_index seems not to be implemented. Hence, here we calculate all the eigenvalues and eigenvectors, but then only return the requested subset.

Parameters:

  • matrix – ndarray or qutip.Qobj to be diagonalized

  • evals_count – how many eigenvalues/vectors should be returned

  • kwargs – optional settings that are passed onto the diagonalization routine

Return type:

Union[Tuple[ndarray, ndarray], Tuple[ndarray, QutipEigenstates]]

Returns:

a tuple of eigenvalues and eigenvectors. Eigenvectors are Qobjs if matrix is a Qobj instance