pyscfad.scf.hf.SCF.energy_grad#

SCF.energy_grad(dm0=None, mode='rev')[source]#

Computing energy gradients w.r.t AO parameters.

In principle, MO response is not needed, and it is sufficient to compute the gradient of the eigen decomposition with the converged density matrix. But this function is implemented as to trace the SCF iterations to show the difference between unrolling for loops and implicit differentiation.

Parameters:

dm0array, optional

Input density matrix.

modestring, default=’rev’

Differentiating using the forward or reverse mode.

Returns:

molpyscfad.gto.Mole

Mole object that contains the gradients.

Notes

The attributes of the SCF instance will not be modified. This function only works with the JAX backend.

Deprecated since version 0.2.0: This function will be deprecated in PySCFAD 0.2.0.