How to get 2 body density matrix after a dft calculation?

Hi all, after running a dft calculation, it is possible to use the make_rdm1() method to get the reduced 1 body density matrix, but I cannot seem to get the two body density matrix too - so get_rdm2() does not seem to be implemented. Can anyone help me in getting this 2 body density matrix after a dft calculation?

[maxnus]

This transformation you show for the 1-RDM is not correct, you need to consider the AO-overlap matrix which you can get from mf.get_ovlp():

dm1_ao = mf.make_rdm1()
r = np.dot(mf.mo_coeff.T, mf.get_ovlp())
dm1_mo = np.linalg.multi_dot((r, dm1_ao, r.T))

For the 2-DM you can use the numpy.einsum function:

dm2_ao = mf.make_rdm2()
r = np.dot(mf.mo_coeff.T, mf.get_ovlp())
dm2_mo = numpy.einsum("abcd,ia,jb,kc,ld->ijkl", dm2_ao, r, r, r, r)

Note however that in the MO basis the DMs are very simple and diagonal, and can be build directly:

dm1_mo = np.zeros((norb, norb))
for i in range(nocc):
    dm1_mo[i,i] = 2

and

dm2_mo = np.zeros((norb, norb, norb, norb))
for i in range(nocc):
    for j in range(nocc):
        dm2_mo[i,i,j,j] += 4
        dm2_mo[i,j,j,i] -= 2

This is all for restricted spin-symmetry

[maxnus]

I actually submitted a PR for the HF classes recently here If this gets accepted, the make_rdm2 method should also be available for DFT, via inheritance. In the meantime you can check out the additions in the PR to see how to build it from the 1-DM. Note that in practice you probably don’t need to build the 2-DM in a mean-field method, but simply contract your operator with the product of 1-DMs directly.