The near-ultraviolet spectrum of CS2 is studied thoroughly using high level
CISTDQ and EOM-CCSD theory with DZP and TZ2P(f) basis sets. Potential ener
gy curves of the first seven singly excited states with respect to the SCS
bond angle have been predicted. The relative energy ordering of the excited
states as established at the equilibrium geometries using EOM-CCSD theory
is: (X) over tilde (1)Sigma(g)(+) < (a) over tilde B-3(2) < (b) over tilde
(3)A(2)(R) < (A) over tilde (1)A(2) < (c) over tilde B-3(2) < (B) over tild
e B-1(2)(V) < (d) over tilde (3)A(2) < (C) over tilde (1)A(2), which is at
odds with previous theoretical work. The bent (B) over tilde B-1(2) state a
rises from the Renner-Teller splitting of a linear la, state into B-1(2) an
d (1)A(2) states. In disagreement with previous experimental predictions th
at the (1)A(2) state is the lower lying component of this splitting, it is
firmly established from the high level results in this research that the (C
) over tilde (1)A(2) state is the higher lying component of this Renner-Tel
ler pair. Equilibrium geometries are determined and shown to agree qualitat
ively with experiment, save for the (B) over tilde B-1(2) state. From an an
alysis of the complex rotational structure of the V band the experimental g
eometry of the (B) over tilde B-1(2) state was determined to be slightly be
nt (r(e) = 1.544 Angstrom, theta(e) = 160 degrees), whereas the methods emp
loyed in this research predict a significantly bent geometry (r(e) = 1.632
Angstrom, theta(e) = 132.1 degrees at EOM-CCSD/TZ2P(f)). Relative transitio
n energies, T-e, are predicted, and EOM-CCSD calculations are shown to agre
e well with available experimental data, and show the assignment of the V a
nd R bands to be correct while the T band is correctly assigned to the A 1A
2 State and not the lower lying Renner-Teller component of the (1)Delta(u)
state. The high level CI methods, although yielding qualitatively correct p
otential energy curves, do not provide quantitatively accurate transition e
nergies, and it is clear that even an exhaustive treatment of the eight val
ence pi electrons is not entirely satisfactory, while EOM-CCSD is establish
ed as an accurate method for the prediction of excited state potential curv
es.