Join GitHub today
ECP Convergence Failures #1338
Yet another issue rescued from the the forums. The user reports using conda's 1.2.1 to attempt to converge a molecule with ECPs and observes a spread in deltaE between reactant and product of over 1000 kcal/mol, depending on basis set.
Suspecting some catastrophic case of landing on an excited state, I ran a modified version of the input file on the cluster, using 1.3a2.dev122. I doubled the memory allocation and replaced the attempt to unpickle with simply calling the energy with
So the ECP integrals won't even converge for me, and the SCF converges to garbage for our user. Because this is a transition metal, we can't immediately rule out excited state problems, but
User's Input File: