Error getting 1rdm after KCCSD calculation

Hello,

Can you please help me resolve the following issue: I am trying to obtain a 1rdm after performing a KCCSD calculation for a single k point that is not the gamma point, as shown below (PySCF v1.7.6) :

import numpy
from   pyscf.pbc import gto,scf,cc,mp
cell = gto.M(a       = numpy.eye(3)*3.5668,
             atom    = '''Mg 0 0 0''',
             basis   = 'gth-dzv',
             pseudo  = 'gth-pade',
             verbose = 4)

kpts = numpy.ones((1,3))*0.1
mf   = scf.KRHF(cell,kpts).density_fit()
mf.kernel()
c = mp.KMP2(mf)
c.kernel()
c.make_rdm1()

c = cc.KCCSD(mf)
c.kernel()
c.make_rdm1()

The above code was able to produce the KMP2 based 1 body RDM, but fails when called for KCCSD with the following error:

Traceback (most recent call last):
  File "pyscf_1rdm_error.py", line 20, in <module>
    c.make_rdm1()
  File ".../lib/python3.8/site-packages/pyscf/cc/ccsd.py", line 1139, in make_rdm1
    if l1 is None: l1, l2 = self.solve_lambda(t1, t2)
  File ".../lib/python3.8/site-packages/pyscf/cc/ccsd.py", line 1079, in solve_lambda
    ccsd_lambda.kernel(self, eris, t1, t2, l1, l2,
  File ".../lib/python3.8/site-packages/pyscf/cc/ccsd_lambda.py", line 51, in kernel
    imds = fintermediates(mycc, t1, t2, eris)
  File ".../lib/python3.8/site-packages/pyscf/cc/ccsd_lambda.py", line 81, in make_intermediates
    nocc, nvir = t1.shape
ValueError: too many values to unpack (expected 2)

Is there a way to resolve this error and obtain a 1RDM for a KCCSD calculation at a single k point which is not the gamma point, please?

Thank you very much for the help!

[yangcal]

Hello,

We currently don’t have lambda solvers for pbc.cc and make_rdm1 is only implemented for kuccsd as Xiao mentioned. Please note that we just found a bug in kuccsd_rdm (l1, l2 should be approximated by (t1.conj(), t2.conj()) rather than (t1, t2), this will be fixed after PR: https://github.com/pyscf/pyscf/pull/1043).

[xwang862]

I just found out that @yangcal has implemented 1-RDM for KUCCSD, which should work for your case (if you are not restricted to RCCSD). Here is a sample input:

import numpy
from   pyscf.pbc import gto,scf,cc,mp
cell = gto.M(a       = numpy.eye(3)*3.5668,
             atom    = '''Mg 0 0 0''',
             basis   = 'gth-dzv',
             pseudo  = 'gth-pade',
             verbose = 7)

kpts = numpy.ones((1,3))*0.1
mf   = scf.KRHF(cell,kpts).density_fit()
mf.kernel()
c = mp.KMP2(mf)
c.kernel()
c.make_rdm1()

c = cc.KUCCSD(mf)
ecc, t1, t2 = c.kernel()

from pyscf.pbc.cc.kuccsd_rdm import make_rdm1
dm1a,dm1b = make_rdm1(c, t1, t2, l1=None, l2=None)

Note that I didn’t provide lambdas (l1, l2). They will be approximated later by (t1, t2) amplitudes, which I think is a fair treatment in many cases. @yangcal can probably share more insights.