IP/EA-ADC up to 3rd order

[fedy9]

• added IP/EA ADC folders like adc_pp
• added type attribute to ADC matrix to distinguish between IP/EA/PP (e.g. matrix.type="ip"). Accessible wherever it's necessary
• IP/EA ADC matrices up to 3rd order yielding excitation energies
• added pole strengths attribute to ExcitedStates class in analogy to oscillator strengths
• IP/EA pole strengths up to 3rd order computing f amplitudes as intermediate
• added state_diffdm's and s2s transition dm's up to 2nd order for IP/EA
• new guess setup for IP/EA: appropriate estimate of guesses, symmetry setup, guess construction on C++ side
• generalisation of multiple PP specific routines, e.g. the singles block is selected dynamically (p/h/ph) and no longer automatically "ph"
• New and adapted error handling
• modified 'validate_state_parameters()' function in workflow to check for valid IP/EA inputs and returns a new Boolean 'is_alpha' which is None for PP and specifies whether an alpha or beta electron is attached/detached for EA/IP. Default value is "True". Added keyword 'n_doublets' specifying the doublet states for IP/EA
• spin_change calculation was updated for IP/EA case in workflow
• New enforce_spin_kind routines for IP/EA
• adapted output: during calc. (solver printing) and adaptations for the ExcitedStates methods 'describe()' and 'describe_amplitudes()'
• added total dipole moment property in ExcitedStates for all calculations (PP/IP/EA)
• Warning for state dipole moments of IP/EA states since dipole moments of charged molecules are gauge dependent

New methods on user side: ip_adc{0..3}() ea_adc{0..3}() New keywords on user side: n_doublets, is_alpha e.g. ip_adc3(scf_result, n_doublets=5, is_alpha=True)

Remarks: Energies and properties were verified against Q-Chem calculations.

Pole strengths of 3rd order were implemented for completeness. They are currently not used in the calculation since all properties are computed at most at 2nd order even for ADC(3) calculations. Hence they were not verified yet.

No simultaneous calculation of alpha and beta states to simplify the code and error handling and to keep consistency with PP calculations. If alpha and beta would be computed in the same adcc instance, intermediates could be reused. However, these intermediates are not the bottleneck of the calculation and thus recomputed for the other spin.

Quartets not yet implemented since they are pure doubles states which are inaccurate up to 3rd order.

To Do Testing procedures for IP/EA ADC Verify 3rd order pole strengths

Observed issue True for PP/IP/EA If 'kind="any"' and the user requests a significant amount of states getting somewhere near the number of total accessible states, the Davidson will quickly converge to 0-vector. This is not true if kind is specified. Example input `from pyscf import gto, scf import adcc

Run SCF in pyscf

mol = gto.M( atom='O 0.0 0.0 0.0;' 'H 0.0 0.0 0.957;' 'H 0.927 0.0 -0.240', basis='sto-3g', spin=0 ) scfres = scf.RHF(mol) scfres.conv_tol = 1e-13 scfres.kernel()

state = adcc.adc2(scfres, n_states=65) `

[fedy9]

Added ip/ea adc2x for consistency with pp basemethods

Added tests for IP/EA-ADC:

• workflow
• state_densities
• consistent properties
• guesses

Modified adcc testdata generation: the adcc hdf5 reference data have to be regenerated so the pp tests pass, since I moved "available_kinds" to the "adc" level

Modified read-in of reference data and creating MockAdcStates

Generated ip/ea reference data for h2o_sto3g and cn_sto3g Testing alpha and beta attachment/detachment for unrestricted cn_sto3g Only alpha for restricted h2o_sto3g

Removed some intermediates in ip and ea matrix.py to avoid conflicts and reuse code in pp matrix.py

Introduced a blacklist for excitation_properties that depend on the type (pp vs ip/ea)

All former tests are running (except: pcm-lr/ptlr using pyscf that fails with pyscf version 2.3.0, scf energy fails, pcm with ddx solver since not installed)