1 3
Top Curr Chem (Z) (2018) 376:24
the energy ranges of 2DUV experiments increases dramatically with the size of the
(multi-)chromophoric system under consideration, requiring the use of large stateaveraging procedures to converge the multi-configurational wave functions, with a
consequent further increase in computational cost. To cope with the computational
feasibility issues of multi-configurational treatments, we have generally adopted the
RASSCF methodology, as its flexibility makes it possible either to improve the electronic wave function of the largest feasible CASSCF computation, usually involving
the full-valence π-orbital space, or to reduce the overall computational cost by limiting the CAS treatment to a relatively small active space (RAS2).
Indeed, with the RASSCF approach, extra virtual orbitals can be included in
the active space where a restricted number of electrons are allowed (RAS3 space),
increasing the number of electronic excitations and expanding the multi-configurational character of the electronic wave functions relative to what is already accounted
for in a full-valence RAS2 space. This type of RAS scheme can be employed to
benchmark S 0 → S N vertical excitations of chromophore monomeric units. To this
end, we have generally attempted to effect a direct comparison between vertical
excitations in the Franck–Condon (FC) region, i.e. at the GS equilibrium geometries, and experimentally recorded cross-sections in a vacuum, when available. The
choice of the gas phase reflects the need for comparison in the absence of environmental effects that would both complicate the computational modeling strategy and
reduce the experimental spectral resolution (as line broadening is larger in solution),
and is aimed at assessing the prediction of the overall distribution of the absorption
bands in the whole UV range, keeping in mind that vibronic effects [70] are not
accounted for in this study.
Benzene and phenol molecules are the chromophores related to the Phe and Tyr
amino acid aromatic side chains (see Fig. 2). As shown in Fig. 4, CASSCF//PT2
computations of the S 0 → S N vertical excitations (in the FC region) indicate that
the full-valence active spaces of benzene and phenol, i.e. CAS(6,6) and CAS(8,7),
respectively, are not sufficiently accurate when compared with the values of experimental cross-sections in a vacuum [58]. Instead, increasing the active space by
allowing single and double excitations to additional extravalence RAS3 orbitals enables an appropriate description of the dynamic σ–π polarization and significantly
improves accuracy, reaching convergence with eight extra orbitals and quantitative
agreement with experiments. It is worth mentioning that these computations have
been performed applying a computational recipe to eliminate Rydberg contamination in the valence states after adding a set of diffuse and uncontracted basis functions in the center of charge of the molecules (ANO-L-aug) [71, 72], which has been
further used to simulate dimers of amino acid side chains and other chromophores.
While the S 0 → S N excitations do not cover all possible excitations involved in
a 2DUV experiment, because the S N → S M excitations would strongly contribute
as ESA signals in the 2D spectra, the energy range considered in this benchmark
study involves quite large TEs (with energies up to 11 eV from the GS). In fact,
considering a one-color 2DUV experiment with pump and probe frequencies centered at the S 0 → S 1 transition energy (4.5–4.7 eV), the S M excited states considered here already comprise those that could be involved as ESA signals (i.e. the
S 1 → S M transitions, with M < 9) and are properly described at the RASSCF//PT2
77
Reprinted from the journal
Top Curr Chem (Z) (2018) 376:24
the energy ranges of 2DUV experiments increases dramatically with the size of the
(multi-)chromophoric system under consideration, requiring the use of large stateaveraging procedures to converge the multi-configurational wave functions, with a
consequent further increase in computational cost. To cope with the computational
feasibility issues of multi-configurational treatments, we have generally adopted the
RASSCF methodology, as its flexibility makes it possible either to improve the electronic wave function of the largest feasible CASSCF computation, usually involving
the full-valence π-orbital space, or to reduce the overall computational cost by limiting the CAS treatment to a relatively small active space (RAS2).
Indeed, with the RASSCF approach, extra virtual orbitals can be included in
the active space where a restricted number of electrons are allowed (RAS3 space),
increasing the number of electronic excitations and expanding the multi-configurational character of the electronic wave functions relative to what is already accounted
for in a full-valence RAS2 space. This type of RAS scheme can be employed to
benchmark S 0 → S N vertical excitations of chromophore monomeric units. To this
end, we have generally attempted to effect a direct comparison between vertical
excitations in the Franck–Condon (FC) region, i.e. at the GS equilibrium geometries, and experimentally recorded cross-sections in a vacuum, when available. The
choice of the gas phase reflects the need for comparison in the absence of environmental effects that would both complicate the computational modeling strategy and
reduce the experimental spectral resolution (as line broadening is larger in solution),
and is aimed at assessing the prediction of the overall distribution of the absorption
bands in the whole UV range, keeping in mind that vibronic effects [70] are not
accounted for in this study.
Benzene and phenol molecules are the chromophores related to the Phe and Tyr
amino acid aromatic side chains (see Fig. 2). As shown in Fig. 4, CASSCF//PT2
computations of the S 0 → S N vertical excitations (in the FC region) indicate that
the full-valence active spaces of benzene and phenol, i.e. CAS(6,6) and CAS(8,7),
respectively, are not sufficiently accurate when compared with the values of experimental cross-sections in a vacuum [58]. Instead, increasing the active space by
allowing single and double excitations to additional extravalence RAS3 orbitals enables an appropriate description of the dynamic σ–π polarization and significantly
improves accuracy, reaching convergence with eight extra orbitals and quantitative
agreement with experiments. It is worth mentioning that these computations have
been performed applying a computational recipe to eliminate Rydberg contamination in the valence states after adding a set of diffuse and uncontracted basis functions in the center of charge of the molecules (ANO-L-aug) [71, 72], which has been
further used to simulate dimers of amino acid side chains and other chromophores.
While the S 0 → S N excitations do not cover all possible excitations involved in
a 2DUV experiment, because the S N → S M excitations would strongly contribute
as ESA signals in the 2D spectra, the energy range considered in this benchmark
study involves quite large TEs (with energies up to 11 eV from the GS). In fact,
considering a one-color 2DUV experiment with pump and probe frequencies centered at the S 0 → S 1 transition energy (4.5–4.7 eV), the S M excited states considered here already comprise those that could be involved as ESA signals (i.e. the
S 1 → S M transitions, with M < 9) and are properly described at the RASSCF//PT2
77
Reprinted from the journal