Efficient Implementation For Frozen-core Gradient Calculations
There are three options for frozen-core gradient calculations in CFOUR.
The first one (FCGRADNEW=OFF) is to zero out the amplitudes of frozen-core orbitals in an all-electron calculation. This implementation is not efficient but it is generally applicable for both first and second derivatives in CFOUR. This is the option available in the public version and have been extensively used and shown to be robust.
The second one (FCGRADNEW=NEW) has been implemented as described in (L. Cheng and J. Gauss, J. Chem. Phys. 134, 244112 (2011)). It has been implemented for calculating analytic gradients for closed-shell systems. It works for RHF-based calculations of first-order properties and geometrical gradients, including various CC models as well as EOMCC methods. This option is recommended for frozen-core gradient calculations for closed shell systems.
The third one (FCGRADNEW=ON) is based on an interface that expands the amplitudes of a frozen-core calculations to the dimension of the all-electron case to be used in density calculations. This implementation is efficient at the coupled-cluster step, and it works for both RHF and UHF based gradient calculations. This option has been set to be default. It currently does NOT work for second derivatives.
Second option FCGRADNEW=NEW