MRCC interface

MRCC is a very powerful wavefunction-theory QM code, with special support for higher-order coupled cluster methods (CCSDT, CCSDT(Q), CCSDTQ etc.) as well as local correlation methods (LNO-CCSD(T)). Compiled binaries are available that makes the program easy to set up. ASH can be used to drive MRCC calculations in a convenient way by using the MRCCTheory class. The interface is flexible as you provide a multi-line string containing a MRCC-inputfile. Energies and gradients are supported in the ASH interface. Note that while MRCC has analytic gradients for coupled cluster methods this only applies to pure CC methods (CCSD,CCSDT, CCSDTQ), not approximate methods like CCSD(T).

QM/MM pointcharge-embedding (energies and gradients) is also supported in the MRCC interface so MRCCTheory can be used to create a QMMMTheory object. However, note that pointcharge gradients are only available for HF, DFT and MP2 methods.

MRCCTheory class:

class MRCCTheory:
    def __init__(self, mrccdir=None, filename='mrcc', printlevel=2,
                mrccinput=None, numcores=1, parallelization='OMP-and-MKL'):

Keyword	Type	Default value	Details
`mrccdir`	string	None	Path to MRCC directory.
`printlevel`	integer	2	Printlevel
`mrccinput`	string	None	MRCC input as a multi-line string
`numcores`	integer	1	Number of CPU cores MRCC will use
`parallelization`	string	'OMP-and-MKL'	Type of parallelization used. Options: 'OMP', 'OMP-and-MKL' or 'MPI'
`filename`	string	'mrcc'	Name of MRCC inputfile

Finding the MRCC program

To use the interface you need to have MRCC-binaries already installed. You can then either provide the path to the MRCC binaries using the mrccdir keyword or alternatively ASH will try to find the binaries (searches for the dmrcc binary) automatically in PATH.

Using the interface

You need to provide the mrccinput keyword which should be a multi-line string containing the MRCC syntax. Other keywords are optional.

Note

Do NOT provide geometry, charge or spin information in the mrccinput string. This information is handled automatically by ASH.

Parallelization

Parallelization of MRCC can be based on OpenMP, BLAS MKL libraries or MPI. The MPI parallelization is currently not supported by ASH. The OpenMP and MKL parallelization is controlled by setting the numcores keyword.

Note

Expert note: ASH sets ccsdthreads and ptthreads options to numcores.

Examples

CCSDT(Q) singlepoint calculation:

from ash import *

#Add coordinates to fragment
HF_frag=Fragment(xyzfile="hf.xyz", charge=0, mult=1)

#Defining MRCCTheory object
mrccinput="""
basis=def2-SVP
calc=CCSDT(Q)
mem=9000MB
scftype=UHF
ccmaxit=150
"""
MRCCcalc = MRCCTheory(mrccinput=mrccinput, numcores=2)

result=Singlepoint(theory=MRCCcalc,fragment=frag)

Note

If you are running MRCC on multiple fragments that might be either closed-shell or open-shell it might be convenient to leave out scftype=RHF/scftype=UHF and let MRCC choose this. MRCC may otherwise complain.

MRCC CCSDT density calculation and visualization:

MRCC allows you to not only calculate the CCSDT energy but also the CCSDT density. You can request this by adding dens=1 to the MRCC inputstring which will result in a file "CCDENSITIES" that will contain the CCSDT density. MRCC also produces a MOLDEN file, hower, this file contains the SCF orbitals.

To get the CCSDT density we can use functionality built-into Multiwfn (See Multiwfn interface for details) to create correlated natural orbitals. You can use the ASH function convert_MRCC_Molden_file that interfaces to Multiwfn to create the correct Molden-file. The Moldenfile created by the function is called: mrccnew.molden

Once you have the correct Moldenfile you can use it on its own or use the multiwfn_run function (See Multiwfn interface for details) to perform various types of WF and density analysis or plot surfaces. See also Electronic structure analysis .

from ash import *

numcores=8

#Add coordinates to fragment
frag=Fragment(databasefile="hf.xyz", charge=0, mult=1)

#Defining MRCCTheory object
mrccinput="""
basis=def2-SVP
calc=CCSDT
mem=9000MB
scftype=UHF
ccmaxit=150
dens=1
"""
MRCCcalc = MRCCTheory(mrccinput=mrccinput, numcores=numcores)

#Run MRCC
result=Singlepoint(theory=MRCCcalc,fragment=frag)

#Files produced by MRCC job above: MOLDEN and CCDENSITIES

#Now calling convert_MRCC_Molden_file to create a Molden-file with correlated natural orbitals
convert_MRCC_Molden_file(mrccoutputfile=f"{MRCCcalc.filename}.out", moldenfile="MOLDEN", mrccdensityfile="CCDENSITIES")