The MOLCAS program CCSDT is really a shell script which calls sequentially to a set of three programs, which compute Coupled-Cluster Singles Doubles, CCSD, and Coupled-Cluster Singles Doubles and Non-iterative Triples Correction CCSD(T) wave functions for restricted single reference both closed- and open-shell systems. The set is composed by three modules: program CCSORT performs a reorganization of the integrals and the reference function from previous runs; program CCSD computes the CCSD wave function and energy allowing for different forms of spin adaptation, and program CCT3 computes the perturbative triples correction for the CCSD wave function in the different approaches explained in section ccsdt (in users guide) of the user's guide.
There are two possibilities to run the programs. One is to use the command molcas run ccsdt $Input, where the three inputs for the three programs must be placed in file $Input. Other possibility is run the programs sequentially: molcas run ccsort, molcas run ccsd, and molcas run cct3 with their respective input files. The final possibility is to use AUTOMOLCAS. In any case the programs are run sequentially: first CCSORT, second CCSD, and, if required, CCT3.
In addition to the ONEINT and ORDINT integral files, the Comenius codes require the JOBIPH file containing the reference wave function (remember that it is not possible to compute open-shell systems with the SCF program) and the transformed two-electron integrals produced by the MOTRA module and stored in the TRAINT file.
Previously to execute the CCSORT module, wave functions and integrals have to be prepared. First, a RASSCF calculation has to be run in such a way that the resulting wave function has one single reference. In closed-shell situations this means to include all the orbitals as inactive and zero active electrons. Keyword CLOSed must be then used in the CCSORT module input. Keyword CANOnical must be used in the RASSCF input to activate the construction of canonical orbitals and the calculation of the CI-vectors on the basis of the canonical orbitals. After that the MOTRA module has to be run to transform the two-electron integrals using the molecular orbitals provided by the RASSCF module. If the LUMOrb is used in the MOTRA input it will be necessary to run a previous RASREAD program using the option CANOnical in the RASREAD input. Otherwise, the JOBIPH from the RASSCF calculation can be used directly by MOTRA using the JOBIph option in the MOTRA input. Frozen or deleted orbitals can be introduced in the transformation step by the proper options in the MOTRA input.
The section of the MOLCAS output corresponding to the CC programs is self explanatory. The default CCSORT output simply contains the wave function specifications from the previous RASSCF calculation, the orbital specifications, and the diagonal Fock matrix elements and orbital energies. The default CCSD output contains the technical description of the calculation, the iterations leading to the CCSD energy, and the five largest amplitudes of each type, which will help to evaluate the calculation. The default CCT3 output contains the description of the employed method (from the three available) to compute perturbatively the triple excited contributions to the CC energy, the value of the correction, and the energy decomposition into spin parts.
Figure 3.13 contains the input files required by the seward, scf, rasscf, motra, ccsort, ccsd, and cct3 programs to compute the ground state of the HF cation. molecule, which is a doublet of symmetry. A more detailed description of the different options included in the input of the programs can be found in section ccsdt (in users guide) of the user's guide. This example describes how to calculate CCSD(T) energy for HF(+) cation. This cation can be safely represented by the single determinant as a reference function, so one can assume, that CCSD(T) method will be suitable for its description.
The calculation can be divided into few steps:
This is an open shell case, so it is suitable to choose CCSD(T) method as it is defined by Watts et al. [3]. Since CCSD amplitudes, produced by previous CCSD run are partly spin adapted and denominators are produced from the corresponding diagonal Fock matrix elements, final energy is sometimes refered the as SA1 (see [4]).
A suitable shell script to run these calculations can be found at the end of section cct3 (in users guide) of the user's guide.
Figure 3.13. Sample input containing the files required by the seward, scf, rasscf, motra, ccsort, ccsd, and cct3 programs to compute the ground state of the HF cation.
&SEWARD &END Title HF molecule Nopack Symmetry X Y Basis set F.ano-l...3S2P1D. F 0.00000 0.00000 1.73300 End of basis Basis set H.ano-l...2S1P. H 0.00000 0.00000 0.00000 End of basis End of input
&SCF &END Title HF molecule Occupied 3 1 1 0 End of input
&RASSCF &END Title HF(+) cation Canonical Symmetry 1 Spin 2 nActEl 1 0 0 Inactive 2 1 1 0 Ras2 1 0 0 0 LumOrb End of input
&MOTRA &END Title HF(+) cation JobIph Frozen 1 0 0 0 End of input
&CCSORT &END Title HF(+) cation CCT Frozen 1 0 0 0 End of input
&CCSD &END Title HF(+) cation Iterations 50 Denominators 2 Shift 0.2,0.2 Accuracy 1.0d-7 Adaptation 1 Extrapolation 5,4 End of input
&CCT3 &END Title HF(+) cation Triply 3 Denominators 0 End of input
RASSCF calculates the HF ionized state by removing one electron from the orbital in the first symmetry. Do not forget to use keyword CANONICAL.
Since this is the CCSD(T) calculation we need to prepare required files for both CCSD and CCT3 programs. Therefore CCT keyword is used.
In the CCSD run, the number of iterations is limited to 50. Denominators will be formed using orbital enegies. (This corresponds to the chosen spin adaptation.) Orbitals will be shifted by 0.2 au, what will accelerate the convergence. However, final energy will not be affected by the choosen type of denominators and orbital shifts. Required accuracy is 1.0d-7 au for the energy. T2 DDVV class of CCSD amplitudes will be spin adapted. To accelerate the convergence, DIIS procedure is exploited. It will start after 5th iteration and the last four iterations will be taken into account in each extrapolation step.
In the triples step the CCSD(T) procedure as defined by Watts et al. [3] will be performed. Corresponding denominators will be produced using diagonal Fock matrix elements.