Computational Options
Here, we provide sample code to get started with running any of the computational options available in CCpy. Before executing any correlated CC/EOMCC steps, the Hartree-Fock mean field solution and transformed one- and two-electron integrals must be provided via an external source. Currently, this can be done in one of three ways:
PySCF
CCpy is fully interfaced with PySCF and can build the Driver object out of a PySCF mean field. This approach is arguably the most convenient one since the Hartree-Fock calculation can be run using PySCF within the same script.
from pyscf import gto, scf
from ccpy import Driver
# Set up geometry for symmetrically stretched H2O
WATER = [["O", (0.0, 0.0, -0.0180)],
["H", (0.0, 3.030526, -2.117796)],
["H", (0.0, -3.030526, -2.117796)]]
# Create PySCF molecule object
mol = gto.M(
atom=WATER,
basis="cc-pvdz",
charge=0,
spin=0,
symmetry="C2V",
cart=True,
unit="Bohr",
)
# Create PySCF RHF mean field object
mf = scf.RHF(mol)
# Run RHF
mf.kernel()
# Now set up the CCpy driver object using the PySCF mean field
driver = Driver.from_pyscf(mf, nfrozen=1)
# Print the system information
driver.system.print_info()
GAMESS
CCpy can read system information and extract transformed one- and two-electron integrals from a completed
GAMESS calculation by passing in the locations of the GAMESS output logfile (gms_logfile) and
companion FCIDUMP (gms_fcidump). GAMESS can be used to generate FCIDUMP files for RHF and ROHF
calculations using the runtyp=fcidump option.
Important Note: The FCIDUMP option in GAMESS has a bug for high-spin ROHF references. The number of electrons (NELEC) and number of unpaired electrons (MS2) will have incorrect values upon output. Therefore, to use FCIDUMP files corresponding to ROHF references, one should manually change the NELEC and MS2 fields to their proper values.
from ccpy import Driver
# Set up CCpy driver object using GAMESS logfile (gms_logfile) and FCIDUMP (gms_fcidump)
driver = Driver.from_gamess(gms_logfile,
gms_fcidump,
nfrozen=0)
# Print the system information
driver.system.print_info()
FCIDUMP
The most general way to pass in information about a mean field is through a single FCIDUMP file. For ROHF
references, it is a good idea to specify the appropriate canonicalization scheme so that the molecular orbital
energies are output correctly (the ROHF single-particle energies are not actually used in correlated computations,
so in practice, this is optional). If no canonicalization is provided, CCpy will default to using Guest-Saunders,
which is most common, but this is not always the correct choice. For example, when using ROHF FCIDUMP files generated
with GAMESS under default settings, the correct canonicalization scheme is Roothaan.
Important Note: Currently, CCpy does not use the spatial point group symmetry information contained in the FCIDUMP file. This will be changed soon.
from ccpy import Driver
# Set up CCpy driver object using an FCIDUMP file
driver = Driver.from_fcidump(fcidump, nfrozen=0, charge=0, rohf_canonicalization="Roothaan")
# Print the system information
driver.system.print_info()
CCD
Sample code
from ccpy import Driver
# Run CCD calculation
driver.run_cc(method="ccd")
CCSD
Sample code
from ccpy import Driver
# Run CCSD calculation
driver.run_cc(method="ccsd")
CCSD(T)
Sample code
from ccpy import Driver
# Run CCSD calculation
driver.run_cc(method="ccsd")
# Run CCSD(T) triples correction to CCSD energetics
driver.run_ccp3(method="ccsdpt")
CC3
Sample code
from ccpy import Driver
# Run CC3 calculation
driver.run_cc(method="cc3")
CCSDt
Sample code
from ccpy import Driver, get_active_triples_space
# Choose the active space for the problem. Here, we are using (2,2).
driver.system.set_active_space(nact_occupied=2, nact_unoccupied=2)
# Obtain the list of triples excitations corresponding to the CCSDt truncation (ground-state symmetry adapted)
t3_excitations = get_active_triples_space(driver.system, target_irrep=driver.system.reference_symmetry)
# Run active-space CCSDt calculation via general CC(P) solver
driver.run_ccp(method="ccsdt_p", t3_excitations=t3_excitations)
CCSDT
Sample code
from ccpy import Driver
# Run CCSDT calculation
driver.run_cc(method="ccsdt")
Alternatively, full CCSDT calculations are also available by running active-orbital-based CCSDt with full active space. The advantage of this approach is that it allows for point group symmetry-adapted CCSDT runs.
from ccpy import Driver, get_active_triples_space
# Choose the active space for the problem. Here, we are using a full active space.
driver.system.set_active_space(nact_occupied=driver.system.noccupied_alpha, nact_unoccupied=driver.system.nunoccupied_beta)
# Obtain the list of triples excitations corresponding to the CCSDt truncation (ground-state symmetry adapted)
t3_excitations = get_active_triples_space(driver.system, target_irrep=driver.system.reference_symmetry)
# Run full CCSDT calculation via general CC(P) solver
driver.run_ccp(method="ccsdt_p", t3_excitations=t3_excitations)
CC4
Note: CC4 is available for closed-shell references only!
Sample code
from ccpy import Driver
# Run CC4 calculation
driver.run_cc(method="cc4")
CCSDTQ
Note: CCSDTQ is available for closed-shell references only!
Sample code
from ccpy import Driver
# Run CCSDTQ calculation
driver.run_cc(method="ccsdtq")
EOMCCSD
Sample code
# Run the ground-state CCSD calculation
driver.run_cc(method="ccsd")
# Compute and store the CCSD similarity-transformed Hamiltonian (this will overwrite the bare integrals in driver.hamiltonian)
driver.run_hbar(method="ccsd")
# Perform an initial CI-like diagonalization to obtain guess vectors
driver.run_guess(method="cis", multiplicity=1, roots_per_irrep={"A1": 3, "B1": 2, "B2": 2, "A2": 0})
# Run the EOMCCSD calculation for the specified states. The values `state_index` map one-to-one
# with the guess vectors, so in this example,
# states 1, 2, 3 -> A1
# states 4, 5 -> B1
# states 6, 7 -> B2
driver.run_eomcc(method="eomccsd", state_index=[1, 2, 3, 4, 5, 6, 7])
EOM-CC3
Sample code
#
driver.run_cc(method="cc3")
driver.run_hbar(method="cc3")
driver.run_guess(method="cisd", roots_per_irrep={"A1": 3, "B1": 3, "B2": 2, "A2": 1}, multiplicity=1, nact_occupied=2, nact_unoccupied=4)
driver.run_eomcc(method="eomcc3", state_index=[1, 2, 3, 4, 5, 6, 7, 8, 9])
EOMCCSDT(a)*
Sample code
# Run ground-state CC calculation
driver.run_cc(method="ccsd")
# Obtain the CCSD(T)(a) similarity-transformed Hamiltonian
driver.run_hbar(method="ccsdta")
# Run EOMCCSD-like calculation using CCSD(T)(a) HBar
driver.run_guess(method="cisd", multiplicity=1, roots_per_irrep={"A1": 4, "B1": 2, "B2": 0, "A2": 2}, nact_occupied=3, nact_unoccupied=7)
driver.run_eomcc(method="eomccsd", state_index=[2, 3, 4, 5, 6, 7, 8])
# Obtain the left eigenstates for each EOMCC root
driver.run_lefteomcc(method="left_ccsd", state_index=[2, 3, 4, 5, 6, 7, 8])
# Compute EOMCCSDT(a)* excited-state corrections
driver.run_ccp3(method="eomccsdta_star", state_index=[0, 2, 3, 4, 5, 6, 7, 8])