J . Phys. Chem. 1993, 97. 107-110
107
Proton Affinities of NH3, HzO, and HF and Their Anions: A Quest for the Basis-Set Limit Using the Dunning Augmented Correlation-Consistent Basis Sets Janet E. Del Bene Department of Chemistry, Youngstown State University, Youngstown, Ohio 44555 Received: July 24, 1992
Proton affinities of NH3, H 2 0 , H F , NH2-, OH- and F- have been computed using many-body Maller-Plesset perturbation theory with the Dunning correlation-consistent polarized split-valence basis sets (cc-pVXZ, where X = D for double-, T for triple-, Q for quadruple-, and 5 for quintuple-split) and these same basis sets augmented with diffuse functions (aug-cc-pVXZ) with one set of diffuse functions for each “1” value in the original basis. Diffuse functions on non-hydrogen atoms are required to lower computed proton affinities and reduce the basis set superposition error, but diffuse functions on hydrogen atoms are not necessary. Computed proton affinities with the diffuse-truncated basis sets (aug’-cc-pVXZ) appear to converge, so that aug’-cc-pVQZ (and aug‘cc-pV5Z) proton affinities may be regarded as basis-set limit values. The aug’-cc-pVTZ basis set without the counterpoise correction is recommended for future studies of proton affinities in larger molecules. Computed MP4/aug’-cc-pVTZ and coupled-cluster CCSD+T(CCSD)/aug’-cc-pVTZ proton affinities are compared with experimental data.
Introduction
A goal of modern quantum chemistry is the consistent, reliable calculation of molecular properties and interaction energies with an accuracy comparable to that which can be measured experimentally. In order to achieve this goal it is necessary to critically evaluate the methodology used for the calculations. This evaluation should be done in terms of two criteria, one internal and the other external. The internal criterion refers to convergence of the computed property with respect to further extensions of the methodology. The external criterion is comparison with experimental data. Computed interaction energies in general, and proton affinities (PAS) in particular, depend on four factors: (1) the level of calculation, Le., Hartree-Fock or some correlated level of theory; (2) the basis set used for the calculations; (3) the geometries of reactants and products; and (4) the method of obtaining vibrational frequencies. The focus of this paper is a critical evaluation of the basis-set dependence of computed proton affinities, with the aim of establishing a basis-set limit. For this study, the correlation-consistent polarized split-valence basis sets of Dunning (cc-pVXZ, where X = D for double-, T for triple-, Q for quadruple-, and 5 for quintuple-split),’ and these same basis sets augmented with diffuse functions,2will be used. These basis sets were chosen because they systematically expand the polarization space as the valence space is expanded. This study parallels a recent study of the performance of the Dunning basis sets for the determination of the stabilities of hydrocarbons and carbocations.3 In the present work, the Dunning basis sets will be evaluated in calculations of correlated proton affinities of small molecules (NH3, H20,and HF) and their corresponding anions (NHz-, OH-, and F-). The aims of this investigation are to determine whether the computed proton affinities of these species converge with respect to basis set size and to identify a particular basis which has achieved convergence to within an acceptable tolerance and which is feasible for use in future studies of proton affinities of larger molecules and anions. Metbod of Calculation The optimized structures of NH4+, H30+, FH2+, NH3, H20, HF, NHz-, and OH- had been determined previously4at correlated second-order many-bod y Maller-Plesset perturbation theorys-lo (MBPT2 = MP2) using the split-valence plus polarization
6-31+G(d,p) basis set augmented with diffuse s and p functions on non-hydrogen at0ms.lI-1~ MP2/6-3 l+G(d,p) vibrational frequencies were computed to ensure that these structures are equilibrium structures with no imaginary frequencies and to evaluate zero-point and thermal vibrational energies. Singlepoint calculations were then carried out on these optimized structures a t second (MP2) and full fourth-order (MP4SDTQ = MP4) many-body Maller-Plesset perturbation theory using the correlation-consistent polarized split-valence basis sets (ccpVXZ) of Dunning.’ cc-pVDZ is a valence double-split basis set with first polarization functions on all atoms (d functions on first-row atoms and p functions on hydrogens). cc-pVTZ is a valence triple-split basis with two sets of first polarization functions and a single set of second polarization functions. cc-pVQZ further splits the valence space and adds an additional set of first and second polarization functions, as well as a set of third polarization functions. The largest basis set, cc-pVSZ, has a quintuple-split valence space, with four sets of d functions, three sets off functions, two sets of g functions, and a set of h functions on non-hydrogen first (long) row elements, and four sets of p functions, three sets of d’s, two sets of f s , and a set of g functions on hydrogens. Recently, the DZ, TZ, and QZ basis sets have been augmented with additional diffuse functions,2 one set for each value of the quantum number “1” which appears in cc-pVXZ. These augmented sets are denoted aug-cc-pVXZ. Since diffuse functions have not yet been developed for the cc-pV5Z basis, the diffuse functions used with this basis in this work have exponents which are one-third of the smallest exponent for a given “1” value in the original basis. A summary description of the contracted Dunning basis sets is given in Table I. Details may be found in the original and in ref 3. papers by Because of the sizeof the larger basis sets and thecomputational problems which they pose, some work has been done to determine the effect of basis set truncation. In previous studies it has been found that diffuse s functions on hydrogens in basis sets such as 6-31++G(d,p) and 6-31 l++G(2d,2p) have little effect on computed hydrogen bond energies and proton affinities and could be omitted.15 Therefore, the effect of omitting the large set of diffuse functions on hydrogens from the Dunning basis sets has also been investigated. These truncated basis sets are designated aug‘-cc-pVXZ, which indicates that aug-cc-pVXZ has been used on non-hydrogen atoms and cc-pVXZ on hydrogens. The question of the basis-set superposition error (BSSE)16 in
0022-3654 I58 12097-0107%04.00/0 0 1993 American Chemical Societv
108
TABLE I: The Dunning cc-pVXZ and aug-cc-pVXZ Basis Sets* sets of CC-PVDZ CC-PVTZ CC-PVQZ cc-pv5z aug-cc-pVDZ aug-cc-pVTZ aug-cc-pVQZ aug-cc-pV5Z a
Del Bene
The Journal of Physical Chemistry, Vol. 97, No. 1, 1993
s
p
d
312 413 514 615 413 514 615 716
211 312 413 514 312 413 514 615
110 211 312 413 210 312 413 514
f
h
g
110 211 312
110 211
110
210 312 413
210 312
210
total ao’s 14/05 30114 55/30 91/55 23/09 46/23 80146 127180
X/Y: X refers to N , 0, or F, and Y refers to H.
computed proton affinities has also been addressed in this work. This error, which results from the use of an incomplete basis and which should disappear as the basis approaches completeness, has been evaluated using the counterpoise correction of Boys and Bernardi.I6 Of course, this correction is only a crude estimate of the BSSE. Because of the dramatically increasing size of the larger Q Z and 5 Z basis sets, it was not feasible to do all calculations at all levels of theory. How this task was reduced will be outlined below. Although it is not the purpose of this paper to examine in detail the correlation contribution to computed proton affinities, comparison with experimental data required that correlated proton affinities also be obtained using an infinite-order correlation method. Thus, correlated electronic energies have been computed using the coupled-cluster method with a noniterative estimate of thedominant effect of triple excitations [CCSD+T(CCSD)] . 1 7 J 8 Geometry optimizations, calculations of vibrational frequencies, and correlated single-point calculations using the DZ and TZ basis sets were done using the Gaussian 90 system of computer ~ r 0 g r a m s . l ~Correlated single-point calculations with the Q Z and 5Z basis sets and CCSD+T(CCSD) calculations were done using the ACES2 program.20 All calculations were performed on the Cray Y-MP8/864 computer at the Ohio Supercomputer Center. Results and Discussion cc-pVXZ versus pug-cc-pVXZPAS. Table I1 lists the computed proton affinities of NH3, H20, HF, NH2-, OH-, and F- a t MP2 and MP4 with cc-pVXZ and aug-cc-pVXZ basis sets. It is apparent from these data that the computed proton affinities of the anions NH2-, OH-, and F- are strongly dependent on the presence of diffuse functions. Thus, aug-cc-pVDZ proton affinities are lower by about 18 kcal/mol than proton affinities computed with the cc-pVQZ basis set, which is a much larger basis. Even the proton affinity of F-obtained with the cc-pV5Z basis is 9 kcal/mol higher than that computed with aug-cc-pVDZ. The overestimation of proton affinities with the cc-pVXZ basis sets is also evident for the neutral molecules NH3, H20, and HF, although the differences between cc-pVXZ and aug-cc-pVXZ PAS are not as great. Nevertheless, cc-pVQZ proton affinities are significantly greater than aug-cc-pVTZ proton affinities despite the fact that this latter basis is smaller.
The presence of diffuse functions also leads to a significant decrease in the basis set superposition error as estimated by the counterpoise correction. This is dramatically illustrated in Figure 1, where plots of the corrected and uncorrected proton affinities are given for F-, and again in Figure 2, which provides similar plots for HF. These two graphs clearly indicate the importance of including diffuse functions in the basis set, both for lowering computed proton affinities and for reducing the basis-set superposition error. In the remaining sections of this paper, discussion will be limited to proton affinities obtained with aug-cc-pVXZ basis sets. Convergence of aug-cc-pVXZ PAS at MP2 versus MP4. As noted earlier, one purpose of this paper is to examine the convergence properties of proton affinities computed at a correlated level of theory with the Dunning basis sets. It is computationally more expedient to do this work at MP2 rather than MP4. However, this may only be justified if it can be shown that the changes in computed proton affinities with extension of the basis set at MP2 are similar to the changes at MP4. Table I1 provides the data necessary to make this determination. These data show that changes in computed proton affinities at MP2 and MP4 are the same to within 0.1 kcal/mol with extension of the basis set. Using F-as an example, the change in the computed proton affinity upon extension of the basis from aug-cc-pVDZ to aug-cc-pVTZ is 3.3 kcal/mol at MP2, and 3.2 kcal/mol a t MP4. Extension from aug-cc-pVTZ to aug-cc-pVQZ results in 0.2 and 0.1 kcal/mol increases at MP2 and MP4, respectively. The last extension, from aug-cc-pVQZ to aug-cc-pV5Z leads to decreases in the computed MP2 and MP4 proton affinities of 0.3 and 0.2 kcal/mol, respectively. Thesedata justify using the MP2 level of theory for subsequent analyses of basis-set performance. At this point it should be emphasized that this conclusion is not meant to imply that MP2 is the appropriate level of correlation to be used for computation of proton affinities in general. It is true that for the simple neutral molecules and their anions which are included in this work, MP2 and MP4 proton affinities may be similar. However, this is not necessarily the case for species with multiple bonds or strained ring systems. pug-cc-pVXZversus aut-cc-pVXZ PAS. It has been observed in previous studies that diffuse functions on hydrogens have little effect on interaction energies.Is Since there are a large number of diffuse functions on hydrogen in the Dunning basis sets (one set for each ‘1” value in the basis), calculations with these basis sets would be much more feasible if the diffuse functions could be omitted. The extent to which diffuse functions on hydrogens alter proton affinities may be seen from Table 111, which reports MP2 proton affinities with the aug-cc-pVXZ basis sets and with the diffuse-truncated aug’-cc-pVXZ basis sets which have diffuse functions on N, 0, and F but not on H. For the anions the difference between aug-cc-pVDZ and aug‘-cc-pVDZ PASvaries from 0.2 kcal/mol for OH- to 0.7 kcal/mol for F.With the TZ, QZ, and 5 Z basis sets this difference does not exceed 0.3 kcal/ mol. For the neutral molecules, the difference between aug-ccpVXZ and aug‘-cc-pVXZ proton affinities also does not exceed 0.3 kcal/mol. These data justify the use of truncated aug‘-ccpVXZ basis sets for the investigation of the convergence of
TABLE 11: Computed MP2 and MP4 Proton Affinities (kcal/mol) Using the cc-DVXZand aup-cc-pVXZ Basis Sets NH2OHF NH3 H20 HF basis
MP2
DZ TZ QZ 5z aug-DZ aug-TZ aug-QZ aug-5Z
450.0 433.2 424.2
MP4 451.4 434.5 425.5
407.1 408.5
408.7 410.0
441.0 421.1 410.0
MP2
MP4 442.8 422.9 411.7
391.5 393.2 393.4
392.3 393.9 394.0
MP2
MP4
MP2
418.2 399.3 388.1 379.1 370.1 373.4 373.6 373.3
420.3 401.5 389.9 380.5 370.3 373.5 373.6 373.4
219.0 213.8 211.7
MP4 220.1 214.6 212.7
MP2 179.6 174.2 171.9
MP4 180.7 175.0 172.6
MP2 128.1 123.9 122.0
MP4 129.1 125.1 122.6
210.3 210.7
211.1 211.5
168.9 170.2
169.7 170.9
117.8 120.5 120.6
118.5 121.1 121.1
The Journal of Physical Chemistry, Vol. 97, No. 1, 1993 109
Proton Affinities of NH3, H 2 0 , HF, NH2-, OH-, and F-
TABLE III: Computed MP2 Proton Affinities (kcal/mol) Using the aug-cc-pVXZ and aug'-cc-pVXZ Basis Sets' basis NHI OHFNH3 H20 DZ TZ QZ
407.1/401.6 408.51408.7 1408.1
5z
391.51391.7 393.21393.3 393.41393.5
310.1 1369.4 373.41373.1 373.61373.4 373.31373.2
210.3/210.4 210.7/210.7 1210.7
HF 1 17.81 1 17.5 120.51120.3 120.61120.5
168.91169.2 170.21170.0 1170.2
a Entries are X/Y. X values were computed with the aug-cc-pVXZ basis set on all atoms. Y values were computed with the aug-cc-pVXZ basis on the non-hydrogen atom and the cc-pVXZ basis without diffuse functions on hydrogens.
TABLE IV MP2 Counterpoise-Corrected Proton Affinities (kcal/mol) with the aug'-cc-pVXZ Basis Sets' basis NHzOHFNH3 H2O ~
HF
~~
DZ TZ QZ 5z a
407.61405.2 408.71407.4 408.71407.9
391.71389.8 393.31392.1 393.51392.1
369.41368.0 373.11311.7 373.41372.5 373.21372.7
2 10.41208.4 210.71209.8 210.71210.2
169.21167.6 170.0/169.1 170.21169.4
117.511 16.3 120.311 19.3 120.511 19.9
Entries are X/Y. X values are the uncorrected proton affinities. Y values are the counterpoise-corrected values.
130
-
128
-
420
h
c
2 s &
0
E 126 :
410
:
-8 Y
400
r c, .-
c
c
3
E
ac
390
c
0
0
c,
2
c,
0,
a
124-
.-r c,
-
120
-
118
-
n
'.
380
122
370 116
DZ
cc-pv TZ QZ
52
DZ
aug-cc-pV TZ QZ 52
Figure 1. Basis-set dependence of the proton affinity of F: -,computed without the counterpoise correction; - -, computed with the counterpoise correction.
computed proton affinities with basis set extension, and it is these diffuse-truncatedbasis sets which will beconsidered subsequently in this paper. ConvergenceofPAS at MPZ/aug'-cc-pVXZ. Table IV presents the computed MP2 proton affinities using the aug'-cc-pVXZ basis sets with and without the counterpoise correction. The data for all species are complete through aug'-cc-pVQZ, and aug'-cc-pV5Z results are presented for F-. The proton affinities computed without the counterpoise correction do indeed appear to be converging with extension of the basis set. Convergence has not yet been obtained at a&-cc-pVDZ, since there is a change of more than 1 kcal/mol upon extension of the basis set from DZ to T Z for all anions and HF. However, extension from T Z to QZ leads to changes in computed proton affinities which do not exceed 0.3 kcal/mol. The similarity of T Z and QZ values and the small difference between Q Z and 5Z PAS for F-suggest that aug'-cc-pVTZ proton affinities are probably within 0.3 kcal/mol of converged values. The basis set superposition error as estimated by the counterpoise correction shows a greater basis set dependence than the
! cc-pv augcc-pV DZ
TZ
QZ
DZ
TZ
QZ
Figure 2. Basis-set dependenceof the proton affinity of HF: -,computed without the counterpoise correction; - -, computed with the counterpoise correction.
proton affinity itself. Thus, the trends observed for F- and H F in Figures 1 and 2, respectively, namely, that the counterpoisecorrected proton affinities show a greater dependence on basisset size and that the corrected PAS approach the uncorrected values with extension of the basis set, are also evident for the remaining neutral and anionic species. These trends are also consistent with trends in hydrogen bond energies reported recently.2' At aug'-cc-pVTZ the basis-set superposition error as estimated by the counterpoise correction has been reduced to a little more than 1 kcal/mol for the anions and about 1 kcal/mol for the neutral species. At aug'-cc-pVQZ, the estimated BSSE is less than 1 kcal/mol for the anions and about 0.5 kcal/mol for the neutral molecules. At aug'-cc-pV5Z the BSSE for F-is 0.5 kcal/mol. Since the cunterpoise correction provides only a crude estimate for the BSSE,and since uncorrected proton affinities are essentially the same at T Z and QZ, aug'-cc-pVTZ proton affinities without the counterpoise correction may indeed approach the basis-set limit of computed proton affinities. At this point it is appropriate to compare the proton affinities computed in this work with those of previous studies. Of the
110 The Journal of Physical Chemistry, Vol. 97, No. 1 , 1993
TABLE V: Computed and Experimental Proton Affinities (kcal/mol)' -AN9' MP4 electronic PAS
this work
ref 22 MP4
this work MP4
ref 22
MP4
MP4
CCSD+ T(CCSD)
exptb
NH2 OHF-
412.4 397.8 377.6
410.2 394.1 373.1
404.3 391.2 372.9
401.5 387.2 368.1
403.0 389.7 370.5
403.8 390.8 371.3
NH, H20 HF
211.6 171.3 121.4
211.5 170.8 120.8
204.0 203.2 165.1 163.8 116.9115.8
203.7 164.2 116.2
204.0 166.5 117
values from ref 22 are based on MP4/6-3 1 1 ++G(3df,3pd) electronic energies; - A N 9 8 values from this work are based on MP4 and CCSD+T(CCSD) electronic energies with the aug'-cc-pVTZ basis set. Experimental data taken from ref 26. @
Del Bene MP4 PAS. There are two points which emerge. First, the good agreement between theory and experiment obtained in ref 22 resulted in part from some cancellation of basis set and correlation effects on computed proton affinities, particularly for the anions. Second, anions are notoriously difficult to describe well. Further work needs to be done on the sensitivity of anion energies to the method of evaluating the electron correlation contribution to these energies. Acknowledgment. These calculations were made possible through a generous grant of computer time from the Ohio Supercomputer Center, whose support is gratefully acknowledged. Thanks are also due to Dr. Thom Dunning and his group for providing the cc-pV5Z basis set and the diffuse functions for the DZ, TZ, and Q Z basis sets prior to publication and to Dr. Rod Bartlett and his group for access to and assistance with theACES2 program.
many studies which have been published, perhaps the most extensive and systematic is the study of DeFrees and McLeanZ2 References and Notes who computed MP4 proton affinities for a groupof small molecules and anions including the ones investigated here. These authors ( I ) Dunning, T. H., Jr. J . Chem. Phys. 1989, 90, 1007. employed the triple-split 6-3 1 l++G(3df,3pd) basis set23.24for (2) Kendall, R. A.; Dunning, T. H., Jr.; Harrison, R. J. J . Chem. Phys. 1992, 96, 6796. their calculations. (The diffuse functions with this basis set are limited to a single set of s and p orbitals which share the same (3) Del Bene, J. E.; Aue, D. H.; Shavitt, I. J. Am. Chem. SOC.1992,114, 1631. exponents for heavy atoms and a singles function for hydrogen.) (4) Del Bene, J. E.; Shavitt, I. J . Phys. Chem. 1990, 94, 5514. For the neutral molecules, uncorrected MP4/augf-cc-pVTZ PAS ( 5 ) Bartlett, R. J.; Silver, D. J. J . Chem. Phys. 1975,62,3258; 1976,64, are slightly lower than the uncorrected MP4/6-3 11++G(3df, 1260, 4578. 3pd) PASof DeFrees and McLean by 0.1 kcal/mol for NH3,0.5 (6) Binkley, J. S.; Pople, J. A. Int. J . Quantum Chem. 1975, 9, 229. kcal/mol for H20, and 0.6 kcal/mol for HF. Greater differences (7) Pople, J. A.; Binkley, J. S.; Seeger, R. Inf. J . Quantum Chem., are found for theanions, with the MP4/augf-cc-pVTZ PAS being Quantum Chem. Symp. 1976, IO, 1. lower than MP4/6-31 l++G(3df,3pd) PAS by 2.2 kcal/mol for (8) Krishnan, R.; Pople, J. A. Int. J . Quantum Chem. 1978, 14, 91. NH2-,3.7 kcal/mol for OH-, and 4.5 kcal/mol for F-. DeFrees (9) Purvis, G.D.; Bartlett, R. J. J . Chem. Phys. 1978, 14, 561. and McLean converted their electronic proton affinities ~ o - A H ~ ~ ( I O ) Bartlett, R. J.; Purvis, G.D. Inr. J . Quantum Chem. 1978, 14, 561. values by standard techniques25 and compared their results with (11) Hariharan, P. C.; Pople, J. A. Theor. Chim. Acra. 1973, 28, 213. experimental data. Their -AWS8values and values obtained in (12) Dill, J. D.; Pople, J. A. J. Chem. Phys. 1975, 62, 2921. this work are reported in Table V. Differences in the computed G.W.; Clark, T.; Chandrasekhar, J.; Schleyer, P. v. R. (13) Spitznagel, - A H 2 9 8 values reflect differences in electronic energies, as well J . Comput. Chem. 1982, 3, 363. as differences in zero-point and thermal vibrational energies, the (14) Clark, T.; Chandrasekhar, J.; Spitznagel, W. G.;Schleyer, P. v. R. latter resulting from the use of different geometries. Also given J. Comput. Chem. 1983, 4 , 294. in Table V are experimental proton affinities taken from the (1 5 ) Del Bene, J. E. J . Comput. Chem. 1985,6,296; 1989,5,603;J . Phys. critical compilation and evaluation of Lias et a1.26 In their paper, Chem. 1988, 92, 2874. DeFrees and McLean speculated as to the reason that generally (16) Boys, S. F.; Bernardi, F. Mol. Phys. 1970, 19, 553. good agreement was obtained between their computed values (17) Purvis, G.D., 111; Bartlett, R. J. J . Chem. Phys. 1982, 76, 1910. and experimental data. Theone explanation which they dismissed (18) Urban, M.; Noga, J.; Cole, S. J.; Bartlett, R. J. J . Chem. Phys. 1985, 83, 404 1. was the possibility that both their theoretical proton affinities (19) Frisch, M. J.; Head-Gordon, M.; Trucks, G. W.; Foreman, J. B.; and experimental proton affinities were systematically too large. Schlegel, H. B.; Raghavachari, K.;Robb, M. A.; Binkley, J. S.;Gonzalez,C.; The MP4/augf-cc-pVTZ proton affinities obtained in this work DeFrees, D. J.; Fox, D. J.; Whiteside, R. A.; Seeger, R.; Melius, C. F.;Baker, suggest t h a t t h e i r c o m p u t e d M P 4 / 6 - 3 11 + + G ( 3 d f , J.; Martin, R.; Kahn, L. R.; Stewart, J. J. P.; Topiol, S.; Pople, J. A. Gausqan 3pd) proton affinities, particularly for the anions, may indeed be 90; Gaussian, Inc.: Pittsburgh, PA, 1990. too large. (20) ACES2 is a new program package from the Quantum Theory Project of the University of Florida. The SCF, integral transformation, and correlation Since the computed - A H 2 9 8 PASbased on MP4/aug'-cc-pVTZ energy codes in this package were written by J. F. Stanton, J. Gauss, J. D. electronic proton affinities are lower than the experimental proton Watts, W. J. Lauderdale, and R. J. Bartlett. See, e.g., Stanton, J. F.; Gauss, affinities, particularly for the anions, does this mean that the J.; Watts, J. D.; Bartlett, R. J. J . Chem. Phys. 1991, 94,4334. The package also includes the VMOL integral program written by P. R. Taylor and J. experimental values are too high? Before making such a claim, Almldf. the adequacy of MP4for evaluating the electron correlation energy (21) Del Bene, J. E. Int. J . Quantum Chem.. Quantum Chem. Symp., in should be examined. While there are many cases for which MP4 press. energies are comparable to energies obtained with an infinite(22) DeFrees, D. J.; McLean, A. D. J. Comput. Chem. 1986, 7, 321. order electron correlation method, there are specific examples, (23) Frisch, M. J.; Pople, J. A.; Binkley, J. S. J . Chem. Phys. 1984, 80, such as N20, for which MP4 has been shown to be inadequate.2'328 3265. Therefore, proton affinities for all species have also been computed (24) Krishnan, R.; Binkley, J. S.;Seeger, R.; Pople, J. A. J . Chem. Phys. using the CCSD+T(CCSD) method with the sameaugf-cc-pVTZ 1980, 72, 650. basis set. -AH298 PAS based on these electronic energies are (25) Del Bene, J. E. In Molecular Structure and Energetics; Greenberg, A., Liebman, J. F., Eds.; Springer-Verlag Publishing Company, Inc.: Deerfield reported in Table V. It is apparent that for the neutral species, Beach, FL, 1986; Vol. I, pp 319-349. MP4 and CCSD+T(CCSD) values are similar, differing by no (26) Lias, S. G.;Bartmess, J. E.; Liebman, J. F.; Holmes, J. L.; Levin, R. more than 0.5 kcal/mol. However, the PAS of the anions change D.; Mallard, W. G.J. Phys. Chem. Ref. Data, Suppl. 1988, 17 (Suppl. I). significantly, with the CCSD+T(CCSD) values being greater (27) Frisch, M. J.; Del Bene, J. E. Inr. J . Quantum Chem., Quantum than the MP4 values by 1.5 kcal/mol for NH2-, 2.5 kcal/mol for Chem. Symp. 1989, 23, 363. OH-, and 2.4 kcal/mol for F-. As a result, the CCSD+T(CCSD) (28) Del Bene, J. E.; Frisch, M. J. Int. J . Quantum Chem., Quantum PAS are in better agreement with experimental values than the Chem. Symp. 1989, 23, 371.
