File description

In src/, you can find the following files.

• GUGA_diag.py: Evaluates the diagonal H matrix element of a CSF.
• gen_spinfree_rdm.py: Generates one- and two-body RDMs of a CSF.
• IntegralClass.py: Reads in an FCIDUMP file.
• runmolcas.py: Interfaces to run CSF-ROHF using OpenMolcas RASSCF module.
• settings.py: To run runmolcas.py, you need to set your OpenMolcas build directory in this file.

Set OpenMolcas build directory

To use the interface, you need to set your OpenMolcas build directory in settings.py.

Tutorial: SSG-CSF-ROHF Optimized Orbitals

This tutorial provides a step-by-step guide to obtaining SSG-CSF-ROHF optimized orbitals using the Fe dimer example from Reference [1].

Using the Python interface in this repository requires:

• OpenMolcas for orbital optimization
• NECI for PT2-RDM generation

Note: The stochastic SplitGAS (SSG) method [2] is currently only available in an internal development version of NECI. Please contact us to request access.

All files used in this tutorial are located in example/SSG-CSF-ROHF_tutorial.

---

We begin with a set of localized and sorted high-spin ROHF orbitals Fe2S2.SortOrb. The Fe 3d orbitals are site-separated: the first five orbitals belong to Fe 1 and the last five orbitals belong to Fe 2. Step 3 for PT2-RDMs generation is done with the SSG implementation [2] in NECI and the other steps are donw with OpenMolccas.

Step 1: CSF-ROHF Orbital Optimization

Directory: 1_CSF-ROHF/

Run:

python ../../../src/runmolcas.py Fe2S2 1 1 1 1 1 2 2 2 2 2

Fe2S2 is the input file name without extension. The integers define the step-vector elements of the target CSF (|uuuuu ddddd>). This command executes an OpenMolcas calculation via the Python interface and it produces single-CSF-SCF variationally optimized orbitals. For subsequent steps (2 and 4), we use the optimized orbitals from the final iteration before canonicalization (Fe2S2.IterOrb.25).

Step 2: FCIDUMP Generation

Directory: 2_dumpgen/

Run an OpenMolcas calculation using Fe2S2.inp. The integral file is fed into NECI for PT2-RDM generation (step 3).

Step 3: PT2-RDM Generation

Directory: 3_RDMgen/

Run NECI with the replica trick (dneci) using input. p1.list defines the P space of the SSG calculation. This tep creates PT2-RDM files (DMAT.1, PSMAT.1, and PAMAT.1) that are used in the next step.

Note: This step requires the internal development version of NECI with SSG support.

Step 4: SSG-CSF-ROHF Orbital Optimization

Directory: 4_SSG-CSF-ROHF/

Run:

python ../../../src/runmolcas.py -u Fe2S2 1 1 1 1 1 2 2 2 2 2

The -u flag instructs the script to use RDM files from the working directory. This produces SSG-CSF-ROHF orbitals and energy: RASSCF energy for state 1 -5092.73230830. The energy is found in Table 3 of Reference [1].

Note: The Python script supplies a fake energy to OpenMolcas to allow the iteration to proceed when the -u flag is used. This value appears in the output but should not be used for analysis.

Other Example

Go to example/ and $ python ../src/runmolcas.py N2 1 1 1 2 2 2. N2 is the Molcas input filenmae without extension, 1 1 1 2 2 2 is the CSF you use for the ROHF optimization in the step-vector format. This executes Molcas with N2.inp and feed RDMs and the RDM energy to Molcas for every RASSCF iteration.

Output

The python interface creates additional output files:

• .pylog logs the input parameters and the activites the interface does during the SCF iterations.
• .iterdata contains only the SCF iteration data of the calculation, as the OpenMolcas output contains additional information when external RDMs are used (e.g., "echo $your_RDM_Energy ..." in between every iteration).

Note: For SSG-CSF-ROHF calculations (Step 4 of the tutorial above), please only use RASSCF_energy (5th column) for the energy of the calculation. RDM_Energy (last column) prints fake energies fed into OpenMolcas just to proceed iterations with the external RDM mode.

References

• [1] Maru Song, Luca Bonfirraro, Ignacio Fdez. Galván, Roland Lindh, and Giovanni Li Manni, "Spin-Adapted Restricted Open-Shell Hartree-Fock and Its Dynamic Correlation Extension", Accepted for publication in J. Chem. Theory Comput. 2026, https://doi.org/10.1021/acs.jctc.6c00379
• [2] Luca Bonfirraro, Oskar Weser, Maru Song, and Giovanni Li Manni, "Stochastic-SplitGAS: A Quantum Monte Carlo Multi-Reference Perturbation Theory Based on the Imaginary-Time Evolution of Effective Hamiltonians", J. Chem. Theory Comput. 2025, 21, 24, 12523–12544, https://doi.org/10.1021/acs.jctc.5c01270