Explore Data structure from Qiskit and/or Pennylane

[NicoRenaud]

Find out what is needed for Qiskit and Peenylane to run a simple VQE

[Cmurilochem]

After a close look into qiskit-nature and pennylane source codes, I suggest the above coupling schema.

Few remarks:

1. As expected, our qc2 data format must gather (from the ase calculators) at least information about:

   • the molecule: atom types, positions, multiplicity, # of electrons, multiplicity.
   • calculation results: hf energy, nuclear repulsion energy, one body integrals (MO basis), two body integrals (MO basis). For qiskit-nature, we would additionally need MO coefficients and one-electron integrals (Fock matrix in the AO basis). Further details to be discussed later in #1

2. To couple our qc2 interface with OpenFermion/Pennylane and Qiskit-nature, I suggest developing "translators" that could convert our qc2-formatted file, say qc2_output.h5, into the corresponding accepted formats into these QC software frameworks. These are:

   • qcschema formatted files (JSON/HDF5) for Qiskit-nature (https://github.com/qiskit-community/qiskit-nature/blob/main/qiskit_nature/second_q/formats/qcschema/qc_schema.py)
   • hdf5 files with specific OpenFermion format; see, e.g., https://github.com/quantumlib/OpenFermion/blob/master/src/openfermion/chem/molecular_data.py#L719

• [ ] This would eliminate the need to construct openfermion-ASE calculators (for OpenFermion/Pennylane) and/or ASE-drivers ( for Qiskit-nature)

3. The above generated files could then be internally loaded by OpenFermion/Pennylane and Qiskit-nature and contain by built-in all we need to proceed with a (simple) VQE calculations.

I have made short "mock" programs to test the above suggestion. Further details and discussions to come...