Mrcc can be used as a standalone code, but interfaces have been developed to the Cfour, Columbus, Dirac, Molpro, Orca, and Psi quantum chemistry packages, the xTB and Mopac semi-empirical programs as well as the Amber molecular dynamics (MD) code. Mrcc, in standalone mode, can currently be used for single-point energy calculations, evaluation of first-order properties, geometry optimizations, and harmonic vibrational frequency calculations with the standard nonrelativistic Hamiltonian and effective core potentials. The interfaces enable the calculation of further molecular properties as well as several other features, such as the use of relativistic Hamiltonians and MCSCF orbitals or quantum mechanics/molecular mechanics (QM/MM) simulations. If Mrcc is used together with the aforementioned quantum chemistry packages, the integral, property integral, HF, MCSCF, and CPHF calculations, the integral and density-matrix transformations, etc. are performed by these program systems. Transformed MO (property) integrals are passed over to Mrcc, which carries out the correlation calculation and returns unrelaxed MO density matrices if necessary.

In the following we describe the use of the Cfour, Columbus, Dirac, Molpro, xTB, Mopac, and Amber interfaces and the features that they enable. For a complete list of available features, see Sect. 6. See also the description of keyword iface. For the Orca and Psi interfaces see the manual of these packages.


Most of the implemented features are available via the Cfour interface using RHF, ROHF, and UHF orbitals: single-point energy calculations, geometry optimizations, first-, second-, and third-order property calculations, electronic excitation energies, excited-state and transition properties, diagonal Born–Oppenheimer correction (DBOC) calculations. Most of the properties implemented in Cfour are also available with Mrcc. The interface also enables the use of several relativistic Hamiltonians.

The Cfour interface is very user-friendly. You only have to prepare the input file ZMAT for Cfour with the keyword CC_PROG=MRCC, and run Cfour. The Mrcc input file is then written automatically and Mrcc is called directly by Cfour, and you do not need to write any input file for Mrcc. Most of the features of Mrcc can be controlled by the corresponding Cfour keyword, see Cfour’s homepage at If you use the Cfour interface, you can safely ignore the rest of this manual.

You also have the option to turn off the automatic construction of the Mrcc input file by giving INPUT_MRCC=OFF in the Cfour input file ZMAT. However, it is only recommended for expert users.


Single-point energies, equilibrium geometries, ground- and excited-state first-order properties, and transition moments can be computed with RHF, ROHF, and MCSCF reference states using the Columbus interface. Evaluation of harmonic vibrational frequencies is also possible via numerically differentiated analytical gradients.

Running Mrcc with Columbus requires three additional programs, colto55, coldens, and runc_mrcc, which are available for Columbus licensees from the authors of Mrcc upon request.

To use this interface for single-point energy calculations first prepare input files for Columbus using the colinp script. It is important to set a calculation in the input file which requires a complete integral transformation (e.g., CISD and not just MCSCF). Execute Columbus. If you do not need the results of the Columbus calculations, you can stop them after completing the integral transformation. Run the colto55 program in the WORK directory created by Columbus. This will convert the Columbus integral files to the Mrcc format. Prepare input file MINP for Mrcc as described in Sect. 11. Run dmrcc as described in Sect. 9. It is recommended to execute first some inexpensive calculation (e.g., CISD) with Mrcc and compare the HF and CISD energies in order to test your input files.

For property calculations create the Columbus and the Mrcc input files. In the Columbus input set the corresponding MRCI property calculation. Copy the Mrcc input file MINP to the WORK directory of Columbus. If the directory does not exist, create it. Then execute the runc_mrcc script.


The interface to the Dirac code enables four-component relativistic calculations with the full Dirac–Coulomb Hamiltonian and several approximate variants thereof. Single-point energy calculations are possible with all CC and CI methods implemented in Mrcc using Kramers-paired Dirac–Fock orbitals. First-order property (unrelaxed) calculations are available with iterative CC and CI methods. See Refs. 109 and 67 for more details.

If you use Dirac, you should first prepare input files for Dirac (see It is important to run a full integral transformation with Dirac (see the description of the MOLTRA keyword in Dirac’s manual), and to use Abelian symmetry (that is, the D2h group and its subgroups). Execute the pam script saving the MRCONEE and MDCINT files, e.g., running it as
pam --get="MRCONEE MDCINT" --inp=Y.inp --mol=X.mol
where X.mol and Y.inp should be replaced by your input files as appropriate. Then run the dirac_mointegral_export interface program, which generates the files needed by Mrcc. It also creates a sample input file MINP for Mrcc, which contains the input for a closed-shell CCSD calculation. If you intend to run another type of calculation, please edit the file as described in Sect. 11. Please also note that you may need to modify the occupation vector under the refdet keyword (see the description of the keyword on page 12), and you should set hamilton=x2c if exact 2-component Hamiltonians are used. Then run dmrcc as described in Sect. 9.

For relativistic property calculations define the corresponding operator in the Dirac input file (see the description of the PROPERTIES and MOLTRA keywords in Dirac’s manual). Then execute the pam script as
pam --get="MRCONEE MDCINT MDPROP" --inp=Y.inp --mol=X.mol
and edit the MINP file, in particular, set the dens keyword (page 12). The CC property code currently does not work with double-group symmetry, and you need to turn off symmetry for CC property calculations, that is, set symm=off in the MINP file. Finally run dmrcc.


With Molpro single-point energy calculations are available using RHF, UHF, ROHF, and MCSCF orbitals. The interface also enables the use of Douglas–Kroll–Hess Hamiltonians as well as effective core potentials.

The Molpro interface is very user-friendly. You only have to prepare the input file for Molpro with a line starting with the mrcc label followed by the corresponding keywords, and run Molpro. The Mrcc input file is then written automatically and Mrcc is called directly by Molpro, and you do not need to write any input file for Mrcc. Most of the features of Mrcc can be controlled by the corresponding Molpro keywords. If you use Molpro, you also have the option to install Mrcc with the makefile of Molpro.

For a detailed description of the interface point your browser to the Molpro User’s Manual at and then click “34 The MRCC program of M. Kallay (MRCC)”.

If you use the Molpro interface, you can safely ignore the rest of this manual.


The Amber/Mrcc interface enables QM/MM and other multiscale calculations. The interface of the Amber MD code and Mrcc is based on the work of Götz et al. [35], which facilitates the integration of QM codes into Amber as external modules. The detailed description of the Amber/Mrcc interface is documented in Ref. 47. Currently for the separation of the QM and MM subsystems only the link atom approach is supported.

With the Amber/Mrcc program the projection-based embedding techniques, namely the projector-augmented operator of Manby and Miller [84] and our Huzinaga-operator [48] approaches, can also be employed for the QM region. The multilevel approach based on our local correlation methods is also supported [48]. The latter approaches enables the embedding of wave function theory (WFT) or density functional theory (DFT) methods into lower-level WFT or DFT methods and also the combinations thereof (DFT-in-DFT, WFT-in-DFT, WFT-in-WFT, and WFT-in-WFT-in-DFT). Consequently, with the Amber interface you can also define three (QM/QM/MM) or four (QM/QM/QM/MM) layers for the multilevel calculations. At the present stage of development, the multilevel methods with three or four layers are only available for single point energy calculations. The traditional multilevel approach, “Our own n-layered integrated molecular orbital and molecular mechanics” (ONIOM) of Morokuma et al. [86], is also supported with two QM layers by our in-house interface for single point, geometry optimization, and (Born–Oppenheimer) molecular dynamics calculations. The unofficial Mrcc module, which supports ONIOM calculations, is available on user-request.

Using the RISM modules of Amber, the embedded cluster reference interaction site model (EC-RISM) method of Kast et al. [kloss2008quantum, kast2010prediction] was implemented as a solvation model [RISM]. These calculations are controlled by Mrcc, and Amber is used for the 3D-RISM part of the calculations. Based on the integral equation theory, 3D-RISM can provide solvation thermodynamics and distribution functions like molecular dynamics (MD), but in a few minutes, while its accuracy is also satisfying. EC-RISM is a QM/MM-like combination of 3D-RISM and QM to describe systems in solutions. See keyword rism for further details.

To use the Amber/Mrcc interface you need a properly installed version of Amber (version 2017 or later), see the Amber homepage at The detailed description of the usage of the interface is well documented in the Amber manual, see section “ AMBER/MRCC” in the manual of Amber21.


The ONIOM method [86] can be also applied with the xTB program [xtbprog], which implements the density functional tight binding (DFTB) methods of Grimme and co-workers [GFN1, GFN2]. These theoretical models are parameterized for geometries, vibrational frequencies, and non-covalent interactions (GFN), and the necessary parameters are available for most of the periodic system. In the ONIOM implementation, single-point and gradient calculations are supported with the mechanical embedding approach, single-point calculations with various point charge models can be utilized for electronic embedding, including the Mulliken charges of the GFN-xTB models, and single-point and gradient calculations can be carried out with electronic embedding if the point charges are user-defined and independent of the geometry. Note that the use of the force-field version of the GFN family (GFN-FF) [GFN-family] and the implicit solvent models of xTB [GFN-ALPB] is also supported in the ONIOM framework. Finally, the GFN1-xTB and GFN2-xTB can also be used for initial guess calculations as it is described in Ref. guess.


The Mrcc program is also interfaced to the Mopac program of Stewart [MOPAC2016], which allows various semi-empirical molecular orbital methods, for instance, the PM6 and PM7 versions of the neglect of diatomic differential overlap (NDDO) approach, to be used with the mechanical embedding version of the ONIOM implementation. Single-point and gradient calculations are supported along with the COSMO implicit solvent model of the Mopac program.