The electronic spectrum of pyrrole

JOURNAL OF CHEMICAL PHYSICS VOLUME 111, NUMBER 2 8 JULY 1999 The electronic spectrum of pyrrole Ove Christiansen a) and Jürgen Gauss Institut für Physikalische Chemie, Universität Mainz, D-55099 Mainz, Germany John F. Stanton Institute for Theoretical Chemistry, Departments of Chemistry and Biochemistry, The University of Texas at Austin, Austin, Texas 78712 Poul Jo rgensen Department of Chemistry, Århus University, DK-8000 Århus, Denmark Received 22 December; accepted 14 April 1999 The electronic spectrum of pyrrole has been investigated by performing calculations using a hierarchy of coupled-cluster models consisting of CCS, CC2, CCSD, and CC3. Basis-set effects have been investigated by carrying out calculations using correlation-consistent basis sets augmented with functions especially designed for the description of Rydberg states. Oscillator strengths, excited state dipole moments, and second moments of the electronic charge distributions have been used to characterize the electronic transitions and final states. Structures and vibrational frequencies have been calculated for a few selected states, and the importance of distinguishing between vertical and adiabatic transitions for accurate comparison with experiment has been emphasized. The experimental spectrum has been scrutinized in the relevant energy region, and the accuracy of recent calculations CASPT2, MRMP, ADC 2 has been critically reexamined. 1999 American Institute of Physics. S0021-9606 99 30326-3 I. INTRODUCTION The electronic spectrum of pyrrole has been the subject of many theoretical 1 12 and experimental 13 24 studies. The interest in pyrrole is not surprising, considering that its heterocyclic five-membered ring appears as a fundamental unit in many important biological molecules. Furthermore, its electronic spectrum is nontrivial and contains many unresolved details. Various types of experiments have been applied to extract information on the spectrum, 13 24 but a detailed assignment is extremely difficult due to the presence of two low-lying Rydberg series giving rise to many transitions in the same energy region which also overlap with valence transitions. Calculations have previously been presented for pyrrole using a variety of theoretical methods: self-consistent field SCF and configuration interaction CI, 4 multireference configuration interaction MRCI, 5 the Rayleigh Schrödinger B K method, 6,7 symmetry adapted cluster-ci CSAC-CD, 8 complete active space self-consistent field CASSCF and CASPT2, 9 CASSCF, MRMP, and MCQD, 10 and most recently ADC 2 11 and density functional theory DFT. 12 A number of semiempirical calculations have also been reported. 1 3 There are many inconsistencies among the results obtained in these studies, even when focusing only on the most recent ones. For example, Nakano et al., 10 in their multireference perturbation theory calculations, obtain results that are very similar to those obtained in the CASPT2 calculations of Serrano-Andres et al. 9 for almost all electronic states, but the two sets of calculations differ by 0.51 ev for the vertical excitation energy of the lowest valence state of 1 B 2 symmetry. Furthermore, in the ADC 2 study of a Electronic mail: ove@mulliken.chemie.uni-mainz.de Trofimov and Schirmer, 11 valence excitation energies are reported that are 0.3 0.7 ev higher than those obtained in the CASPT2 and MRMP studies. The above-mentioned methods ADC 2, CASPT2, and MRMP all rely on a treatment of dynamic electron correlation that is based on second-order perturbation theory. According to our experience, second-order correlation methods may easily involve errors of the aforementioned magnitudes i.e., up to 0.7 ev. Discrepancies of this size have also been found when such calculations were compared to highly correlated coupled-cluster results as, for example, in case of furan, 25,26 s-tetrazine, 27 benzene, 28 30 and ethylene. 31,32 The large number of studies on pyrrole indicates that its electronic spectrum has developed into a benchmark example for theoretical studies of excited states. It appears that none of the recent studies can be expected to be much more accurate that the others. This, together with the fact that a number of problems still exist in connection with the assignment of the experimental spectrum, has motivated us to undertake a more detailed study of the electronic states of pyrrole using the CC response methodology. 33 Within the CC framework, a hierarchy of models has been developed that allows for systematic improvement with respect to the treatment of electron correlation in investigations of electronic transitions. The models in the CC hierarchy are denoted CCS, CC2, CCSD, CC3.... 34 The CC response methodology that is used to describe electronic transitions is a simple extension of standard ground-state CC theory to the treatment of electronic transitions. 33 In the aforementioned hierarchy, the CC3 method is the first method that includes approximately the effect of connected triple excitations. However, due to its iterative nature the CC3 model can be quite expensive, and a noniterative triples correction to CCSD excitation energies, CCSDR 3, has 0021-9606/99/111(2)/525/13/$15.00 525 1999 American Institute of Physics

526 J. Chem. Phys., Vol. 111, No. 2, 8 July 1999 Christiansen et al. been developed. The CCSDR 3 model provides excitation energies of accuracy similar to CC3 but at significantly lower cost. 35,36 Note that the popular noniterative triples correction method for ground states, CCSD T, 37 does not have any rigorous analog in response theory calculations of excited states. The CCSDR 3 method should therefore not be thought of as related to the CCSD T method, but simply as a very accurate noniterative triples correction method for calculation of electronic transition energies. 35,36,38 The hierarchy of CC methods gives excitation energies correct through increasing order in the electron correlation, but more importantly, also through increasing completeness in the cluster expansion. Therefore, results of increasing accuracy can be expected when going to higher levels in the CC hierarchy. Indeed, comparison with exact full configuration-interaction FCI results in finite but reasonable basis sets shows about a factor of three reduction in the error at each step in the CCS, CC2, CCSD, and CC3 or CCSDR 3 sequence. 36,39 With the inclusion of triple excitations in CC3 or CCSDR 3, a mean maximum error of about 0.03 0.1 ev is obtained for transitions that qualitatively can be described as single-electron excitations and when there are not very significant nondynamical correlations in the ground state. The electronic states of pyrrole around its ground-state equilibrium geometry satisfy these criteria. Since the CC response methodology is size extensive, results of similar quality can be expected for small and large systems. Furthermore, due to implementation of integral-direct strategies, 40,28,26 large one-electron basis sets can be used which is important for the treatment of excited states in particular, Rydberg states where extended basis sets are often required for reliable results. For these reasons, the CC methodology is well-suited for a benchmark study of the excitation energies of pyrrole. Similar studies have been previously reported for other prototype organic compounds. 25 29,31 The accuracy found in large basis-set CC calculations has been consistent with the benchmark calculations when comparison is made with accurate experimental 0 0 transitions and geometrical relaxation and vibrational effects are taken into account, rather than considering only the vertical excitation energy. All previous ab initio calculations of pyrrole have solely involved determination of vertical excitation energies. The only exception is the ADC 2 study of Trofimov and Schirmer, 11 where a so-called linear coupling model was used to investigate the effect of nuclear motion on the excitations. However, vertical excitation energies as reported in many calculations represent an idealized theoretical construct that does not correspond to a measurable electronic transition. Even if the maximum intensity peak position is used as an approximate experimental vertical excitation energy, the difference to the true vertical excitation energy might be quite significant. Differences between vertical and adiabatic excitation energies may be significant in pyrrole. In a very recent study, Trofimov et al. 41 calculated the vibronic structure of the photoelectron band for the 2 A 2 and 2 B 1 pyrrole cation states, and found differences between 0 0 and vertical ionization energies of 0.22 and 0.16 ev, respectively. Effects of the same order of magnitude can be expected for the related Rydberg states of pyrrole. For the valence states, a less systematic change can be expected and the effects may be even larger. Calculations of excited state structures and vibrational frequencies are possible in CC theory using analytical excited-state gradient techniques. 42,43 For these reasons, we have investigated the structure and vibrations for a few selected states of pyrrole. II. COMPUTATIONAL CONSIDERATIONS In this study, we analyze the lowest electronic states of pyrrole by reporting calculations for more than 30 states. These include the four lowest-lying valence states two of 1 B 2 symmetry and two of 1 A 1 symmetry, the 1a 2 nl Rydberg transitions for n 3,4 and l 0,1,2, as well as a number of 1a 2 5l and 2b 1 nl Rydberg transitions. In a first series of calculations, we investigate the convergence in the hierarchy of CC models and compare with the performance of the CASPT2 method. We therefore use the same experimental geometry 44 and the same atomic natural orbit ANO basis set as in the CASPT2 study. 9 For the molecule-centered basis functions, we use the same center as Serrano-Andres et al. 9 This center was obtained as the average between the centers of the charge centroids of the 2 A 2 and 2 B 1 cations and is positioned 1.453 277 a.u. from the N atom on the symmetry axis. Second, a basis-set study is performed at the CCSD level using the ANO basis set and the cc-pvdz and cc-pvtz basis sets 45 augmented with both atom- and moleculecentered diffuse functions. The atom-centered diffuse basis functions are of (sp/s) and (spd/sp) type for the cc-pvdz and cc-pvtz, respectively, with exponents taken from Ref. 46 i.e., the aug-cc-pvxz exponents. A series of moleculecentered diffuse basis functions is constructed according to the recipe of Kaufmann et al. 47 We choose to use a (7s7p7d) set giving the D 7 and T 7 basis sets, which consist of 183 and 348 basis functions, respectively. In accordance with our previous experience, such basis sets should be well converged with respect to the number of Rydberg functions for at least the n 3,4,5 Rydberg states of s,p,d type. 25 The 1s core electrons were frozen in all vertical excitation energy calculations. Third, structures and vibrational frequencies are calculated for a few selected excited states. In the geometry optimizations and vibrational frequency calculations, we use the double-zeta basis of Ref. 48 with polarization functions on all atoms 49 exponents H: p(0.7), C: d(0.654), N: d(0.8), and diffuse functions for the heavy atoms 50 C: s(0.023), p(0.021),d(0.015), N: s(0.028),p(0.025),d(0.015). This rather diffuse atom-centered basis should be able to describe reasonably well both the lowest valence and Rydberg states, while at the same time making geometry optimizations for several electronic states feasible. We denote this basis set as DZPR. All electrons were correlated in these calculations. Vertical excitation energy calculations were performed using a local version of the DALTON program 51 including the option of integral-direct CC calculation of electronic transition energies and properties. 28,26 The calculated one-electron

J. Chem. Phys., Vol. 111, No. 2, 8 July 1999 Pyrrole spectrum 527 TABLE I. Vertical excitation energies in ev for pyrrole obtained in coupled-cluster calculation in comparison with previous theoretical results. a Symmetry CCS CC2 CCSD CCSDR 3 CC3 CASSCF b CASPT2 b Assignment 1 A 1 7.59 6.53 6.55 6.43 6.37 7.22 5.92 Valence 6.70 6.62 6.80 6.78 6.77 6.54 6.54 1a 2 3d xy 7.40 6.77 6.97 6.95 6.94 6.71 6.65 2b 1 3p x 8.06 7.36 7.61 7.61 7.60 7.38 7.36 2b 1 3d xz 8.30 8.10 8.22 8.12 8.07 10.22 7.46 Valence 8.86 8.40 8.45 8.32 8.33 1a 2 4d xy 9.61 8.88 9.08 9.07 9.06 8.29 7.88 2b 1 4d xz 10.14 9.50 9.64 9.61 9.60 2b 1 4p x 1 B 1 5.88 5.75 5.88 5.86 5.85 5.72 5.85 1a 2 3p y 6.67 5.86 6.04 6.01 5.99 5.36 5.97 2b 1 3s 6.48 6.33 6.49 6.47 6.47 6.24 6.40 1a 2 3d yz 7.25 6.55 6.76 6.74 6.72 6.41 6.62 2b 1 3p z 7.97 7.10 7.33 7.32 7.31 6.90 7.32 2b 1 3d a1 7.83 7.20 7.40 7.39 7.37 6.97 7.39 2b 1 3d a1 1 B 2 5.96 5.89 6.01 5.97 5.98 5.95 5.78 1a 2 3p x Mix. 6.62 6.55 6.67 6.63 6.63 9.46 6.00 Valence Mix. 6.87 6.89 6.96 6.89 6.91 6.61 6.53 1a 2 3d xz Mix. 8.14 7.40 7.66 7.67 7.66 7.44 7.43 2b 1 3d xy 8.21 8.08 8.23 8.19 8.17 8.03 7.72 1a 2 4d xz 9.45 8.40 8.50 8.41 8.39 Valence 1 A 2 5.28 5.05 5.17 5.12 5.10 5.08 5.08 1a 2 3s 5.96 5.77 5.91 5.87 5.86 5.64 5.83 1a 2 3p z 6.44 6.31 6.47 6.44 6.43 6.15 6.42 1a 2 3d a1 6.54 6.41 6.55 6.52 6.50 6.23 6.51 1a 2 3d a1 7.48 6.65 6.88 6.86 6.84 6.53 6.77 2b 1 3p y 7.93 7.14 7.38 7.37 7.36 7.02 7.31 2b 1 3d yz a Basis set and geometry as in the CASSCF and CASPT2 study of Ref. 9. b CASSCF and CASPT2 results from Ref. 9. properties dipole moments and second moments of the electronic charge distribution correspond to the unrelaxed energy derivatives. Geometry optimizations and vibrational frequency calculations have been performed using ground- and excited-state analytic gradient techniques 42,43 as implemented in a local version of the ACESII program package. 52 In addition, structures of the 2 A 2 and 2 B 1 states of the pyrrole cation were studied using the equation-of-motion coupled cluster method for ionized states EOMIP- CCSD. 53 55 Geometries were optimized using the associated analytic gradient implementation. 56 III. ELECTRON CORRELATION EFFECTS In Table I we give the vertical excitation energies as obtained with the different CC models using the ANO basis set. In Table II oscillator strengths and various excited-state properties are reported. The response eigenvectors are analyzed in terms of their weight of single- and double-electron promotions in the electronic transitions. The resulting percentages of single excitation contributions in the solution vectors, %T 1, indicate that all excitations are primarily single-electron excitations. All states have %T 1 95 with the 2 1 A 1 state, as the only exception having %T 1 91. The accuracy obtained in excitation energies is inversely related to the size of contributions of higher than single excitations. However, states with %T 1 larger than about 90 are relatively unproblematic in CC response calculations, and fast convergence in the hierarchy of CC models is thus expected for all excited states for pyrrole. The effect of triple excitations as found in CCSDR 3 compared to CCSD range from 0.01 to 0.12 ev. The largest effect is found for the 2 1 A 1 state, as expected from the slightly larger contribution due to double excitations, but the effect of triples is not much smaller for the other valence states. The CC3 excitation energies are close to the noniterative CCSDR 3 as also found in previous studies; thus we will, in the remaining part of this study, exclusively use the less expensive CCSDR 3 method. The excited states are assigned qualitatively as valence or Rydberg (1a 2 /2b 1 3l) on the basis of the calculated properties dipole moments and second electronic moments and oscillator strengths. The same procedure also defines how the CC results are compared with the corresponding CASSCF and CASPT2 results. To represent the convergence in the CC calculations, as well as to facilitate comparison with the CASSCF and CASPT2 results, we plot in Fig. 1 the deviations in the excitation energies of the individual models relative to the CCSDR 3 results. The fast convergence of the CCS, CC2, CCSD, CCSDR 3 sequence and the large %T 1 values lead us to conclude that highly accurate results are obtained in the CCSDR 3 calculations. In previous FCI benchmark calculations, 39,36,35 we have found a reduction of the CCSD error of a factor of three upon inclusion of triple excitations in the CCSDR 3 and CC3 methods. This indicates that the remaining correlation error in CCSDR 3 is quite small, i.e.,

528 J. Chem. Phys., Vol. 111, No. 2, 8 July 1999 Christiansen et al. TABLE II. Oscillator strengths, dipole moments a.u., and electronic second moment a.u. orthogonal to the molecular plane for pyrrole obtained in coupled-cluster calculation in comparison with previous theoretical results. a Oscillator strengths Dipole moments x 2 Symmetry CCS CC2 CCSD CASSCF b CCSD CASSCF b CCSD CASSCF b Assignment 1 1 A 1 0.745 0.695 26 26 Valence 1 A 1 0.0045 0.0006 0.0018 0.0195 0.417 0.451 28 34 Valence 0.0001 0.0000 0.0002 0.0010 0.098 0.054 80 84 1a 2 3d xy 0.0071 0.0222 0.0225 0.0127 0.446 0.966 69 69 2b 1 3p x 0.0408 0.0144 0.0265 0.0047 0.511 0.854 78 78 2b 1 3d xz 0.0308 0.1763 0.2670 0.3261 0.131 1.115 55 39 Valence 0.5984 0.3234 0.2202 0.921 53 1a 2 4d xy 0.0510 0.0384 0.0248 0.0079 1.742 0.077 74 68 2b 1 4d xz 0.0034 0.0006 0.0000 0.845 66 2b 1 4p x 1 B 1 0.0410 0.0277 0.0273 0.0239 1.052 2.012 38 38 1a 2 3p y 0.0004 0.0076 0.0090 0.0006 2.735 2.872 37 38 2b 1 3s 0.0050 0.0084 0.0088 0.0117 0.956 1.442 38 41 1a 2 3d yz 0.0394 0.0293 0.0302 0.0161 2.457 3.062 41 43 2b 1 3p z 0.0023 0.0000 0.0003 0.0001 1.604 1.885 44 45 2b 1 3d a1 0.0104 0.0047 0.0052 0.006 0.212 62 80 2b 1 3d a1 1 B 2 0.0720 0.0619 0.0692 0.0399 0.816 0.987 64 68 1a 2 3p x Mix. 0.0273 0.0085 0.0242 0.0011 2.510 0.833 59 32 Valence Mix. 0.0925 0.1172 0.0765 0.1253 0.847 1.253 56 83 1a 2 3d xz Mix. 0.0044 0.0045 0.0057 0.0005 0.150 0.026 81 85 2b 1 3d xy 0.0009 0.0150 0.0077 0.0043 1.244 1.073 75 73 1a 2 4d xz 0.3371 0.2392 0.2182 1.170 40 Valence 1 A 2 2.673 3.671 36 36 1a 2 3s 1.702 2.000 40 42 1a 2 3p z 1.291 1.804 40 43 1a 2 3d a1 1.404 1.369 66 81 1a 2 3d a1 0.876 1.354 39 40 2b 1 3p y 0.748 1.466 38 40 2b 1 3d yz a Basis set and geometry as in the CASSCF and CASPT2 study of Ref. 9. b CASSCF and CASPT2 results from Ref. 9. CASPT2 excitation energies were used in calculating the oscillator strengths together with CASSCF transition dipole moments. a rather conservative estimate is 0.15 ev, but for most states significantly smaller. Implicitly, this also provides some measure of the accuracy for the other CC models as well as the CASSCF and CASPT2 calculations of Serrano-Andres et al. 9 For the following discussion, the small errors in the CCSDR 3 results are not significant. It is evident from Fig. 1 that rather large errors occur when the CCS model is used to calculate excitation energies. As a consequence, this approach often gives the wrong ordering of states. CC2 reduces the errors of CCS significantly, and actually tends to give excitation energies lower than those from CCSDR 3. Figure 1 shows that the CASSCF results are rather erratic, with values ranging from 1.7 to 2.8 ev with respect to CCSDR 3. CASPT2 corrections to the CASSCF excitation energies are accordingly quite significant up to 3.46 ev in magnitude. Nonetheless, these results are still quite far from CCSDR 3 in some cases. The CASPT2 results are in all cases below CCSDR 3. Considering the CASPT2 results in more detail, it can be seen that the CASPT2/CCSDR 3 difference for states of symmetry A 2 and B 1 which all have pure Rydberg character is very small typically of the order 0.1 ev. For the region of the spectrum where Rydberg and valence 1 A 1 and 1B 2 states both occur, much larger differences are found. For the highest-lying states reported in the CASPT2 calculations, we find it difficult even to correlate the results to those obtained in this research. The maximum discrepancy between CASPT2 and CCSDR 3 is reduced from 1.8 to 0.7 ev if the CASPT2 excitation energy for one particular state that resulting from the 2b 1 4d xz transition is related differently to the CC excitation energies. However, this is only possible at the expense of further increasing the already significant disagreement with CCSD results for the calculated excitedstate properties and oscillator strengths. We note that the basis set used in these calculations actually does not provide a reliable description of this state see the later discussion. For the states of 1 A 1 and 1 B 2 symmetry, the CASSCF dipole moments are in particular quite different from the CCSD results. We note in passing that in the CC response theory calculations, the various symmetry classes are treated in an equivalent manner and, therefore, all results are expected to have roughly similar accuracy. On the contrary, the CASSCF/CASPT2 calculations of Serrano-Andres et al. 9 involve different choices of active spaces for the symmetry classes containing only Rydberg states ( 1 B 1 and 1 A 2 ) and

J. Chem. Phys., Vol. 111, No. 2, 8 July 1999 Pyrrole spectrum 529 FIG. 1. Deviation in pyrrole vertical excitation energies ev for CASSCF, CASPT2, CCS, CC2, and CCSD relative to CCSDR 3. All results have been obtained with the basis set and geometry given in Ref. 9. those containing both Rydberg and valence states ( 1 A 1 and 1 B 2 ). In a recent review, it was stated that the errors in calculated CASPT2 excitation energies are smaller than 0.2 ev except for a few cases. 57 In another very recent CASPT2 work, it was stated that a 0.4 ev error for the V state of ethylene is the largest deviation found for CASPT2 in a large number of studies. 58 The present set of results suggests that previous estimates of the accuracy of CASPT2 for excitation energies may be somewhat optimistic. This is consistent with other comparisons with CC calculations. 25 31 To be more specific, we give in Fig. 2 a similar comparison of CCSDR 3 with CCS, CC2, CCSD, CASSCF, and CASPT2 for furan. See Refs. 9, 26, and 25 for details of the calculations. The same geometry and basis set is used for all models. Overall, the comparison for furan is similar to that for pyrrole. IV. BASIS-SET EFFECTS In Table III excitation energies, oscillator strengths, and one-electron properties for pyrrole as obtained in the CCSD/ D 7 calculations are given. Included also are triples corrections to the excitation energies CCSDR 3 -CCSD obtained with the D 7 basis set and basis-set corrections found at the CCSD level between results obtained with the D 7 and T 7 basis sets. We see that the D 7 basis is able to describe many more states below 8.0 ev than is possible with the ANO basis set. For example, we see in the 1 A 1 symmetry class that there are more states coming in between the two lowest excited valence states. The numerical effects on the excitation energies when going from the ANO to the D 7 basis for the low-lying states are of the order of 0.05 ev, but larger effects are found for the higher-lying states where the additional flexibility of the Rydberg basis in particular is crucial. Improving the atom-centered basis set by going from D 7 to T 7, we find a systematic increase in the excitation energy FIG. 2. Deviation in furan vertical excitation energies ev for CASSCF, CASPT2, CCS, CC2, and CCSD relative to CCSDR 3. All results have been obtained with the basis set and geometry given in Ref. 9. of 0.15 0.21 ev for almost all Rydberg states. The magnitude of the effects is similar to that found in calculations on small molecules and from other applications and is expected to decrease by roughly a factor of three at each step in the cc-pvxz, X D,T,Q,..., series of basis sets extended with sufficient diffuse functions note that D 7 represents an augmented cc-pvdz set and T 7 an augmented cc-pvtz set. Thus, for the large T 7 basis set, errors of about 0.1 ev may persist for the Rydberg states. The valence states exhibit a smaller basis-set dependence a change of 0.05 ev or less when going from D 7 tot 7. We note that the different basis-set dependence for valence and Rydberg states has consequences for the ordering of states in symmetry 1 A 1. The results in Table III also indicate that there are significant valence Rydberg mixings in the three lowest states of 1 B 2 symmetry. These states have similar diffuseness and oscillator strengths, and also show a less pronounced basis-set dependence than other, more pure, Rydberg states. A Rydberg/valence distinction is often very rewarding for interpretation, and therefore has been made here despite the mixings. It should be said here that such a distinction is not inherent to nature, which provides only the spectra themselves and not a simple interpretation. On the basis of the D 7 basis-set triples corrections and the T 7 results, reliable estimates are given in Table III for the vertical excitation energies. Additivity is assumed in obtaining these results. A rather conservative estimate of the final accuracy is about 0.2 ev. The basis-set trends indicate that the vertical excitation energy for the Rydberg states is probably underestimated by roughly 0.1 ev. On the other hand, the basis set and correlation contributions for the 2 1 A 1 valence state indicate that estimates for this state are probably on the high side. We note that we have solely concentrated on Rydberg states of s, p, and d nature, and have made no attempts to describe f and higher angular momentum states.

530 J. Chem. Phys., Vol. 111, No. 2, 8 July 1999 Christiansen et al. TABLE III. Excitation energies ev, oscillator strengths, dipole moments a.u., and second electronic moments a.u. for pyrrole. Excitation energies Osc. st. x 2 y 2 z 2 c Symmetry CCSD/D 7 T a B b Estimate CCSD/D 7 Assignment 1 1 A 1 0.762 26 141 140 1 A 1 6.53 0.11 0.05 6.37 0.0007 0.384 30 146 142 Valence 6.73 0.03 0.17 6.87 0.0001 0.008 90 197 158 1a 2 3d xy 6.89 0.02 0.17 7.04 0.0215 0.406 78 159 155 2b 1 3p x 7.33 0.03 0.18 7.48 0.0029 0.039 286 396 223 1a 2 4d xy 7.53 0.01 0.16 7.68 0.0333 1.048 108 181 200 2b 1 3d xz 7.60 0.02 0.20 7.78 0.0000 0.029 669 777 353 1a 2 5d xy 8.00 0.09 0.00 7.91 0.2906 0.101 164 250 214 Valence 7.79 0.03 0.18 7.94 0.0023 0.096 878 988 421 1a 2 6d xy 7.89 0.00 0.21 8.10 0.0310 0.836 243 211 209 2b 1 4p x 8.11 0.04 0.18 8.25 0.0318 0.038 534 635 314 1a 2? 8.20 0.02 0.0644 3.946 274 221 363 2b 1 4d xz 1 B 1 5.82 0.03 0.17 5.95 0.0245 0.970 37 180 156 1a 2 3p y 5.97 0.03 0.18 6.12 0.0059 2.827 37 158 165 2b 1 3s 6.43 0.04 0.16 6.55 0.0070 1.184 38 180 181 1a 2 3d yz 6.67 0.03 0.18 6.82 0.0253 3.055 42 162 178 2b 1 3p z 7.02 0.03 0.18 7.17 0.0050 2.453 96 353 215 1a 2 4p y 7.22 0.03 0.18 7.37 0.0001 2.824 86 324 318 1a 2 4d yz 7.25 0.02 0.18 7.41 0.0004 2.183 51 192 191 2b 1 3d a1 7.33 0.03 0.16 7.47 0.0080 0.277 59 185 155 2b 1 3d a1 7.45 0.03 0.0031 4.784 227 747 356 1a 2 5p y 7.55 0.03 0.0002 5.339 196 655 632 1a 2 5d yz 1 B 2 5.96 0.05 0.13 6.04 0.0617 0.759 69 151 153 1a 2 3p x Mix. 6.61 0.06 0.02 6.57 0.0325 3.478 69 151 171 Valence Mix. 6.86 0.06 0.10 6.90 0.0764 3.099 89 158 193 1a 2 3d xz Mix. 7.08 0.04 0.16 7.20 0.0026 1.225 236 204 207 1a 2 4p x 7.33 0.03 0.18 7.48 0.0070 0.387 283 220 395 1a 2 4d xz 7.46 0.03 0.21 7.64 0.0005 1.363 653 343 348 1a 2 5p x 7.59 0.01 0.19 7.77 0.0073 0.399 186 228 259 2b 1 3d xy 7.60 0.02 0.20 7.78 0.0002 2.912 591 330 685 1a 2 5d xz... and more Rydberg states... 8.38 0.07 0.2204 1.072 102 183 176 valence 1 A 2 5.12 0.07 0.15 5.20 6.869 37 151 166 1a 2 3s 5.83 0.05 0.16 5.94 5.738 41 157 176 1a 2 3p z 6.40 0.04 0.15 6.51 3.752 39 191 169 1a 2 3d a1 6.48 0.05 0.14 6.57 4.538 66 151 171 1a 2 3d a1 6.76 0.03 0.19 6.92 12.557 89 202 238 1a 2 4s 6.81 0.03 0.19 6.97 5.986 53 187 182 2b 1 3p y 7.01 0.03 0.18 7.16 6.493 121 224 303 1a 2 4p z 7.19 0.03 0.18 7.34 1.089 186 320 212 1a 2 4d a1 7.22 0.04 0.18 7.36 16.749 117 289 357 1a 2 4d a1 7.31 0.01 0.18 7.48 3.708 41 187 191 2b 1 4p y 7.34 28.554 277 298 509 1a 2 5s 7.45 5.933 317 386 601 1a 2 5p z 7.53 4.538 477 618 380 1a 2 5d a1 7.55 28.147 259 596 730 1a 2 5d a1 a T CCSDR 3 -CCSD in the D 7 basis. b B CCSD/T 7 -CCSD/D 7. c Origin placed at the center of mass. V. VIBRATIONAL AND GEOMETRICAL RELAXATION CONTRIBUTIONS The optimized geometries for the ground- and lowestexcited states of 1 B 1, 1 B 2, and 1 A 2 symmetry are reported in Table IV as obtained in CCS, CC2, and CCSD calculations using the DZPR basis set. In Table V harmonic vibrational frequencies are given for the ground state of pyrrole, while Table VI reports the shifts in the vibrational frequencies upon excitations into the lowest states of 1 B 1, 1 B 2, and 1 A 2 symmetry. Included also in Tables IV VI are available experimental results for structures and vibrational frequencies. We first note that the DZPR basis set is rather small for the accurate calculation of structures and vibrational frequencies. However, our primary interest is not so much on accurate absolute values, but rather to obtain estimates of geometry relaxation and zero-point vibrational energy effects in the electronic transition. The electronic states considered here were all found to

J. Chem. Phys., Vol. 111, No. 2, 8 July 1999 Pyrrole spectrum 531 TABLE IV. Geometrical parameters bond lengths in Å and angles in deg for the ground state and the three lowest electronically excited states and the two lowest ionized states of pyrrole. a CCS Exp. b (1 1 A 1 ) 1 1 A 1 1 1 B 1 1 1 B 2 1 1 A 2 1 2 A 2 1 2 B 1 r NC 2 ) 1.370 1.366 1.350 1.348 1.338 r NH 0.996 0.994 0.999 0.998 1.009 r C 2 C 3 ) 1.382 1.365 1.426 1.431 1.433 r(c 2 H 2 ) 1.076 1.073 1.070 1.071 1.072 r(c 3 H 3 ) 1.077 1.074 1.070 1.071 1.073 NHC 2 ) 125.2 125.2 125.2 125.1 125.0 NC 2 C 3 ) 107.7 108.2 108.2 108.0 108.3 NC 2 H 2 ) 121.5 121.3 122.3 122.2 121.7 C 2 C 3 H 3 ) 125.5 125.9 124.4 124.5 124.7 CC2 r NC 2 ) 1.370 1.381 1.366 1.364 1.355 r NH 0.996 1.013 1.021 1.020 1.054 r C 2 C 3 ) 1.382 1.395 1.443 1.447 1.448 r C 2 H 2 ) 1.076 1.085 1.090 1.089 1.091 r C 3 H 3 ) 1.077 1.086 1.086 1.086 1.088 NHC 2 ) 125.2 124.8 125.1 125.1 125.2 NC 2 C 3 ) 107.7 107.4 108.0 107.8 108.4 NC 2 H 2 ) 121.5 121.2 121.8 121.7 120.6 C 2 C 3 H 3 ) 125.5 125.4 124.2 124.4 124.9 CCSD r NC 2 ) 1.370 1.380 1.364 1.362 1.351 1.363 1.384 r NH 0.996 1.010 1.017 1.016 1.049 1.018 1.023 r C 2 C 3 ) 1.382 1.388 1.440 1.447 1.448 1.443 1.378 r C 2 H 2 ) 1.076 1.085 1.088 1.087 1.089 1.088 1.084 r C 3 H 3 ) 1.077 1.086 1.085 1.085 1.087 1.086 1.087 NHC 2 ) 125.2 125.0 125.1 125.0 125.1 125.2 123.3 NC 2 C 3 ) 107.7 107.8 108.1 107.8 108.4 108.2 106.5 NC 2 H 2 ) 121.5 121.2 121.9 121.7 120.7 121.7 121.5 C 2 C 3 H 3 ) 125.5 125.7 124.3 124.5 124.9 124.7 127.1 a All results obtained with the DZPR basis described in the text. b Reference 44. possess C 2v minima. The most significant structural change upon electronic excitation is an increase in C 2 C 3 bond distances. This occurs because the a 2 orbital from which the primary electronic promotion takes place is bonding between C 2 and C 3. The CC2 and CCSD ground-state vibrational frequencies are significantly lower than the SCF CCS vibrational frequencies. We observe that the experimental frequencies for symmetry a 1 and b 2 are always lower than the CCSD results, while with one exception the opposite is true for symmetry b 1 and a 2. The deviations between the calculated harmonic and the experimental frequencies are due to a combination of anharmonicity, basis-set inadequacies, and neglect of triple excitations. In Table VI, we also include a few experimental results for frequency shifts that can be extracted from the experimental absorption spectra in the 5.8 6.0 ev region see the next section for a more detailed discussion of the assignments. The experimental shifts relative to the experimental ground-state frequencies are in fair agreement with the calculated shifts. The most striking feature is probably the large decrease in the a 1 CH stretching frequencies for the 1 A 2 state. This large decrease alone contributes about 0.11 ev to the difference between vertical and 0 0 excitation energy for this state. In Table VII vertical and adiabatic excitation energies are given in the DZPR basis; the former are given with respect to both the experimental and the corresponding optimized ground-state geometries. The adiabatic excitation energies are obtained by subtracting the total electronic energies at the optimized geometries. Furthermore, we include in Table VII the zero-point vibrational contributions to the 0 0 excitation energies, and differences between the vertical excitation energies and the 0 0 excitation energies. We see that significant geometrical relaxation effects 0.1 0.3 ev are found. Zero-point vibrational effects are somewhat smaller but particularly for the 1 1 A 2 state still significant. The differences between the vertical excitation energies and the 0 0 transition energy are found to be of the order of 0.1 0.4 ev. This is in fair agreement with the 0.22

532 J. Chem. Phys., Vol. 111, No. 2, 8 July 1999 Christiansen et al. TABLE V. Comparison of calculated harmonic and experimental vibrational frequencies in cm 1 for the ground state of pyrrole. a Symmetry SCF CCS CC2 CCSD Exp. b a 1 957.8 877.3 892.5 881 1096.7 1034.8 1040.5 1016 1149.5 1099.0 1100.7 1074 1250.0 1162.6 1180.4 1144 1533.8 1439.9 1452.7 1382 1629.6 1502.7 1532.0 1467 3414.5 3295.2 3290.4 3129 3437.6 3317.3 3314.1 3145 3933.4 3704.8 3744.4 3531 b 1 487.4 451.8 439.8 474 673.9 626.8 628.6 601 814.1 689.8 715.2 721 960.0 758.5 802.7 826 b 2 937.5 856.7 872.0 865 1148.1 1057.1 1074.3 1048 1234.6 1163.6 1172.6 1134 1416.0 1298.5 1320.4 1287 1574.1 1472.4 1496.9 1422 1717.4 1554.1 1600.6 1530 3402.8 3282.8 3277.6 3129 3431.2 3310.4 3306.8 3145 a 2 667.3 593.9 603.2 618 798.0 605.2 656.8 710 981.6 785.6 820.3 869 a All results obtained with the DZPR basis described in the text. b Experimental results as quoted in Ref. 62. ev difference between vertical and 0 0 ionization found by Trofimov et al. 41 for the 2 A 2 cation state. Rydberg states related to a particular cation state often behave similarly, as expected from qualitative arguments. Our results do not fully support this expectation for these low-lying states of pyrrole, as significant variations are seen among these. At least for those states, the concept of a Rydberg state is an idealized construct. Mixings with other states will cause variations between the different states see, for example, the discussion in the previous section concerning the mixings between the lowest states of 1 B 2 symmetry, as well as a low-lying 3s Rydberg state is less ideal than a higher-lying d Rydberg state, for example. Our results show that consideration of geometry relaxation contribution and of differences in zero-point vibrations is important when aiming for accuracy of a few tenths of an ev in the prediction of excitation energies. This also holds for Rydberg states and is in line with experience from previous calculations for furan and benzene. 25,28 However, optimization of structures and calculation of vibrational frequencies for a large number of electronic states is computationally a rather demanding task. For the higher lying states including the valence states we have therefore refrained from geometry optimization and vibrational frequency calculations. Despite the variations found for the low-lying states investigated above, we will use instead the differences between vertical and adiabatic ionization energies for the related cation states as a crude estimate for these effects. Since this approximation is crude anyway, we will not include zero-point vibrational effects. The data for the ionization potentials are included in Table VII. Thus, for other Rydberg states than the three considered explicitly above, we assume an effect of 0.19 ev for Rydberg states originating from excitation from 1a 2, and 0.15 ev for Rydberg states originating from excitation out of 2b 1. These values are quite close to the differences between 0 0 and vertical ionization potentials reported by Trofimov et al. 41 0.22 ev for 1a 2 ionization, and 0.16 ev for 2b 1 ionization. In Table VIII, our best estimates are given for vertical excitation energies, corrections for geometrical relaxation, and zero-point vibrational corrections, as well as 0 0 transition energies obtained by simple addition of these contributions. The uncertainty in all of these corrections as well as the fact that they are not always additive certainly increases the uncertainty in the final values. However, the given 0 0 energies may serve as rough estimates of the 0 0 excitation energies, though a slightly lower accuracy than for the vertical excitation energies which have 0.1 0.2 ev accuracy should be expected. VI. COMPARISON WITH EXPERIMENT AND PREVIOUS CALCULATIONS The accuracy of the different contributions and the final 0 0 transition energies of the excitation energies listed in Table VIII has been discussed in the previous sections. We emphasize that significantly larger basis sets and more accurate correlation methods have been used than in previous studies. In addition, we have also considered geometrical and vibrational effects which have been largely neglected in the past. We will now compare our final results with experiment and other calculations. Unfortunately, we must conclude that it is still difficult to give definite conclusions about details in the pyrrole spectrum due to the large number of close-lying electronic states. In view of the extensive nature of the present computational study compared to previous theoretical investigations, this may be considered somewhat disappointing. However, there are so many electronic states within a modest energy regime that there almost always will be some lines that can be interpreted to fit reasonably well to any computational results. Cautionary and critical comments towards the assignment of the pyrrole spectra have been given by Robin 24 in a similar spirit. Trofimov and Schirmer 11 and Robin 24 have reviewed many different assignments of the features in the electronic spectrum of pyrrole, and we refer to these two papers for historical details and references. In the following subsections, we only discuss issues concerning assignments of the spectrum where our calculations are able to provide new perspectives. A. The 1a 2 3s transition A weak feature reported at about 5.22 ev has been argued to be due to a transition to a Rydberg state. 24 The present calculations also indicate that this feature might be due to the vibrational structure of the dipole-forbidden 1a 2

J. Chem. Phys., Vol. 111, No. 2, 8 July 1999 Pyrrole spectrum 533 TABLE VI. Shift in cm 1 in vibrational frequencies upon electronic excitation. a 1 1 B 1 1 1 B 2 1 1 A 2 Symmetry CCS CC2 CCSD CCS CC2 CCSD CCS CC2 CCSD Exp. 5.864 ev b Exp. 5.818 ev b a 1 6.8 6.2 6.1 10.9 8.5 9.0 7.1 17.7 13.6 1 880 31 850 58.4 42.0 45.8 52.6 40.5 41.8 38.7 33.7 31.9 39 1055 10.6 10.0 8.4 5.2 18.3 13.6 15.7 22.0 11.5 2.0 7.5 10.4 19.0 4.9 7.6 17.1 18.4 22.3 39.2 49.3 28.3 67.7 62.3 60.7 57.2 68.7 63.3 28 1410 41.4 37.7 34.6 47.6 54.5 46.6 57.3 52.0 50.8 33 1500 57.3 35.4 17.4 60.2 0.4 20.0 32.5 443.8 385.4 47.6 0.0 11.0 57.1 4.0 13.6 16.1 69.3 53.4 48.3 96.4 85.7 41.1 80.0 70.4 315.0 418.5 451.9 b 1 49.2 26.5 58.8 45.1 10.9 21.2 33.9 24.8 59.3 43.4 54.5 72.7 55.7 43.2 53.7 52.2 29.0 41.9 9.8 13.8 13.9 44.5 12.8 3.7 0.5 10.4 15.6 64.2 87.1 85.1 167.2 6.6 2.9 51.6 95.5 90.2 b 2 182.9 101.7 99.4 286.7 124.6 157.3 117.9 119.9 139.5 285 580? 76.1 52.6 62.7 244.3 59.7 91.9 77.0 69.0 84.2 106.1 117.9 143.4 124.2 135.9 162.2 57.1 130.1 142.4 110.7 7.2 8.3 102.6 27.7 37.3 26.1 52.4 61.6 142.8 13.8 21.3 9.3 155.0 157.1 133.7 164.9 164.9 84.1 47.1 27.9 20.5 66.5 84.4 99.4 87.8 102.2 50.3 37.4 25.7 60.0 10.5 25.2 38.6 28.5 13.0 31.9 5.7 6.2 60.8 4.4 7.2 25.8 36.6 26.3 a 2 119.0 116.0 112.0 131.5 151.8 137.7 116.9 114.6 110.8 152.7 188.2 168.7 105.7 112.1 94.6 159.4 202.3 180.8 53.8 84.9 81.5 31.6 70.5 66.0 50.1 76.0 75.0 Sum 165.1 80.4 113.4 87.9 456.4 484.5 303.5 1164.0 1129.7 a All results obtained with DZPR basis set described in the text. b Experimental results from Ref. 17. The experimental vibrational shift is calculated from the total excited-state vibrational frequency given in parenthesees and obtained from the vibrational transition compared to the expected 0 0 transition energy and the corresponding experimental ground-state frequency listed in Table V. 3s transition. In particular, we may safely conclude from our calculations that there is no other singlet state in this energy region. B. The 5.7 6.6 ev energy region A firm assignment of the 1a 2 3p states in the experimental spectra has proven difficult. The 2 1 B 2 valence state has generally been assumed to be the major source of intensity for the absorption in the 5.7 6.6 ev region. However, Trofimov and Schirmer recently suggested that the transition to the 2 1 B 2 valence state is outside this energy region, and that the intensity should instead be ascribed primarily to the strong 1a 2 3p x (1 1 B 2 ) Rydberg transition. Our calculations show that the vertical excitation energy for the 2 1 B 2 valence state is at the high end of the considered energy interval. It is somewhat lower than predicted by the ADC 2 calculations, 11 and thus could, in principle, contribute to the intensity, in particular as the excited-state potentical energy surface for this state is so far unknown. However, the oscillator strength for the transition to the 2 1 B 2 valence state is smaller than that for the 1a 2 3p x (1 1 B 2 ) Rydberg transition, and our calculations therefore lend some support to the suggestion of Trofimov and Schirmer. Nevertheless, more detailed studies of the relevant potential energy surfaces, etc. are probably required to be conclusive concerning the contributions from various electronic states. In the vacuum UV spectra, 17,24 there is almost certainly a 0 0 transition found at 5.864 ev, and most probably another at 5.818 ev. The 5.864 ev transition is the more intense. Other nearby peaks are best interpreted as part of the vibrational structure of these two transitions. The observed vibrational structure fits reasonably well with the calculated frequency shifts see Table VI. The final calculated 0 0 transition energies for the two states are within 0.04 ev of each other, as are the experimental results. We find it reasonable to conclude that the two transitions can be described as 1a 2 3p x (1 1 B 2 ) and 1a 2 3p y (1 1 B 1 ). The calculated oscillator strengths point to describing the 5.86 ev transition as 1a 2 3p x (1 1 B 2 ). We note that the in this case explicitly calculated difference between the vertical and the 0 0 transition actually interchanges the order of the states. The interchanged order does, in fact, not agree with the assignment proposed on the basis of the oscillator strengths, but it must be said that a 0.04 ev difference in 0 0 transition energy is

534 J. Chem. Phys., Vol. 111, No. 2, 8 July 1999 Christiansen et al. TABLE VII. Shift in ev in electronic excitation energies and ionization potentials due to geometrical relaxation and zero-point vibration. CCS Vert. grs.g a Vert. exp g. b Adia. Vert.- Adia. c 0-vib. Vert.- 0 0 c 1 1 B 1 5.98 5.87 5.75 0.24/0.12 0.02 0.26/0.14 1 1 B 2 6.13 6.00 5.84 0.29/0.16 0.01 0.30/0.17 1 1 A 2 5.40 5.28 5.09 0.31/0.19 0.04 0.35/0.23 CC2 1 1 B 1 5.63 5.65 5.52 0.11/0.13-0.01 0.10/0.12 1 1 B 2 5.84 5.85 5.67 0.17/0.18 0.06 0.23/0.24 1 1 A 2 4.95 4.97 4.74 0.21/0.23 0.14 0.35/0.37 CCSD 1 1 B 1 5.76 5.78 5.66 0.10/0.12-0.01 0.09/0.11 1 1 B 2 6.00 5.96 5.78 0.22/0.18 0.06 0.28/0.24 1 1 A 2 5.11 5.08 4.85 0.26/0.23 0.14 0.40/0.37 1 2 A 2 8.08 8.06 7.89 0.19/0.17 1 2 B 1 8.81 8.89 7.89 0.15/0.23 a Vertical excitation energy/ionization potential at the optimized ground-state geometry. b Vertical excitation energy/ionization potential at the experimental ground-state geometry. c A/B with A relative to vertical excitation energy/ionization potential at the optimized ground-state geometry; B relative to the vertical excitation energy/ionization potential at the experimental ground-state geometry. probably outside the resolution provided by our 0 0 transition energy estimates. The 2b 1 3s transition is predicted at higher energy than the 1a 2 3p x ( 1 B 2 ) transition and with a much smaller intensity. The same applies for the transition to the V(2 1 A 1 ) state and three 1a 2 3d Rydberg transitions. The relatively prominent and broad band at 6.224 ev has been assigned to a2b 1 3stransition by Derrick et al. 16 Bavia et al. 17 suggested that it is a transition to a vibrationally excited level of the 1a 2 3p x (1 1 B 2 ) Rydberg state. An alternative explanation of this and some other peaks at slightly higher energies could invoke a transition to the V(2 1 B 2 ) state. This transition should be significantly more intense than the 1a 2 3d and 2b 1 3s transitions, and has a predicted vertical excitation energy of 6.57 ev. Furthermore, for the valence state there might easily be geometrical relaxation and zero-point vibrational effects of considerable magnitude, making this state contribute with significant intensity in the 6.2 6.5 ev region and even below. We note that for the similar state in furan, estimates of relaxation effects of the order of 0.3 ev have been made. Concerning the Rydberg states, we note that in the multiphoton ionization MPI experiments of Cooper et al., peaks at 5.994, 6.024, and 6.175 ev were assigned to 1a 2 3p, but with some uncertainty. 22 We note that these assignments do not agree too well with our calculations and the assignments discussed above for the 1a 2 3p transitions. The observed peaks could instead be interpreted as components of the 2b 1 3s transition predicted in this energy region. Though we cannot give definite assignments for all transitions in this energy region, we hope that the present calculations might be helpful for future experimental investigations of this part of the electronic spectra. C. The 6.6 7.1 ev region According to our calculations, we should expect within this energy regime contributions from the three 2b 1 3p transitions, the 1a 2 4s transition, and two 1a 2 3d transitions. Three 1a 2 4p transitions should be expected at the high end of this energy region, or even at slightly higher energies. An MPI transition at 6.668 ev 22 has been tentatively assigned to a d Rydberg series, which does in fact agree with the predicted energies for the 1a 2 3d transitions. Similarly, an s label has been attached to peaks at 6.779 ev in the one-photon absorption spectrum 17 and to 6.782 ev MPI transition, 22 which is also in agreement with the predicted excitation energy for 1a 2 4s. However, the 1a 2 4s transition is symmetry forbidden while the 1a 2 3d xz transition is predicted to have a rather large oscillator strength. An assignment of the relatively strong one-photon absorption at 6.779 ev to the 1a 2 3d xz transition would also be in reasonable agreement with the calculations, since the predicted 0 0 excitation energies may easily be underestimated by around 0.1 ev. Alternative assignments for the abovementioned peaks might be possible, considering the error bounds of the calculations. We do not believe that the present or any previous calculations can reliably resolve the detailed assignments, but we do believe that it is safe to conclude that the 1a 2 3d, 1a 2 4s, and 2b 1 3p transitions are responsible for the structure in the 6.6 7.1 ev region. The 1a 2 4p states are expected to be significantly less split than the corresponding 3p states, and this is also found in the calculations. The calculated splittings are in good agreement with the MPI results of Cooper et al., 22 who reported excitations at 7.080, 7.099, and 7.121 ev. The abso-

J. Chem. Phys., Vol. 111, No. 2, 8 July 1999 Pyrrole spectrum 535 TABLE VIII. Comparison of calculated and experimental excitation energies. a This Work estimated vertical and 0 0 transitions Previous calculated vertical excitation energies Vert. corr. 0 0 Osc. Assign. Exp. b ADC 2 c CASPT2 d MRMP e MCQD e SAC-CI f 5.20 0.37 4.83 3s( 1 A 2 ) 5.22 5.03 5.08 4.92 4.91 5.20 5.94 0.19 5.75 3p z ( 1 A 2 ) 5.71 5.83 5.74 5.74 5.95 5.95 0.11 5.84 0.025 3p y ( 1 B 1 ) 5.82 5.68 5.85 5.81 5.80 5.85 6.04 0.24 5.80 0.062 3p x ( 1 B 2 ) 5.86 5.86 5.78 5.87 5.88 5.54 6.12 0.15 5.97 0.006 3s ( 1 B 1 ) 5.77 5.97 5.70 5.65 6.13 6.37 0.001 2V( 1 A 1 ) 6.66 5.92 5.98 6.01 6.63 6.51 0.19 6.32 3d a1 ( 1 A 2 ) 6.25 6.42 6.38 6.37 6.55 0.19 6.36 0.007 3d yz ( 1 B 1 ) 6.27 6.40 6.45 6.44 6.57 0.033 V( 1 B 2 ) 6.2 6.4? 6.71 6.00 6.51 6.51 6.63 6.57 0.19 6.38 3d a1 ( 1 A 2 ) 6.37 6.51 6.44 6.43 6.82 0.15 6.67 0.025 3p z ( 1 B 1 ) 6.43 6.62 6.48 6.50 6.87 0.19 6.68 0.001 3d xy ( 1 A 1 ) 6.668? 6.54 6.54 6.62 6.64 6.88 6.90 0.19 6.71 0.076 3d xz ( 1 B 2 ) 6.668?,6.779? 6.48 6.53 6.61 6.62 6.20 6.92 0.19 6.73 4s( 1 A 2 ) 6.782?,6.779? 6.55 6.97 0.15 6.82 3p ( 1 y A 2 ) 6.51 6.77 6.70 6.65 6.85 7.04 0.15 6.89 0.022 3p x ( 1 A 1 ) 6.42 6.65 6.52 6.52 6.50 7.16 0.19 6.97 4p z ( 1 A 2 ) 7.080 6.78 7.17 0.19 6.98 0.005 4p y ( 1 B 1 ) 7.099 6.78 7.20 0.19 7.02 0.003 4p x ( 1 B 2 ) 7.121 6.88 7.34 0.19 7.15 4d a1 ( 1 A 2 ) 6.98 7.36 0.19 7.17 4d a1 ( 1 A 2 ) 6.99 7.37 0.19 7.18 0.000 4d yz ( 1 B 1 ) 7.00 7.41 0.15 7.26 0.000 3d a1 ( 1 B 1 ) 6.97 7.32 7.14 7.13 7.47 0.15 7.32 0.008 3d a1 ( 1 B 1 ) 7.09 7.39 7.23 7.20 7.48 0.19 7.29 0.003 4d xy ( 1 A 1 ) 7.35? 22 7.12 7.48 0.19 7.29 0.007 4d xz ( 1 B 2 ) 7.35? 22 7.14 7.72 7.48 0.15 7.33 4p ( 1 A 2 )... and more Rydberg states... 7.91 0.291 V( 1 A 1 ) 7 8 24 7.52 7.46 7.48 7.51 7.20... and still more... 8.31 g 0.220 V( 1 B 2 ) a Assignments are given with symmetries in parentheses; primed symbols refer to Rydberg excitations from the 2b 1 orbital and unprimed symbols refer to excitations from the 1a 2 orbital. b Experimental results are assumed to be 0 0 transitions except when given in parenthesis. See the text for discussion and references. c ADC 2 results from Ref. 11. d CASPT2 results from Ref. 9. e MRMP and MCQD results from Ref. 10. f SAC-CI results from Ref. 8. g CCSDR 3 /D 7 results without T 7 basis-set correction. lute predicted 0 0 excitation energies are too low by 0.15 ev, which would correspond to a relatively large but still acceptable error for the theoretical predictions. We note that 4p in the nomenclature of Cooper et al. is 4p x, 4p is 4p y, and 4p is 4p z. In agreement with this, the dipole-allowed 7.099 and 7.121 ev transitions are also found in one-photon absorption spectra, 17 while peaks corresponding to the dipole-forbidden 4p z transition have not been reported. D. Beyond 7.1 ev Beyond 7.1 ev, a broad intense absorption overlapped with a few sharp peaks is found in the experimental spectrum. 24 Considering the calculated oscillator strengths, it is expected that a significant portion of the one-photon absorption intensity beyond 7 ev is due to transitions to the higher-lying valence state of 1 A 1 symmetry. We note that for the similar state in furan, estimates of relaxation effects of the order of 0.8 ev or larger have been made. Geometrical relaxation effects may also bring the higher-lying valence state of 1 B 2 symmetry below the first ionization potential of about 8.21 ev. There is a fundamental difference between our calculations and those of Trofimov and Schirmer. These authors explain the one-photon spectrum in this region primarily in terms of absorption to three valence states. Their ADC 2 calculations predict the vertical excitation energies for the lowest-lying valence 1 A 1 and 1 B 2 states to be only 0.5 and 0.8 ev lower, respectively, than the higherlying 1 A 1 valence state, while we find for the same energy differences values of 1.5 and 1.3 ev, respectively. Therefore, the explanation given by Trofimov and Schirmer for the electronic spectrum in this energy region appears to be incorrect. A rather high intensity is found in the complete 7.1 8