J. Phys. Chem. 1995, 99, 4000-4007
4000
Ab Initio Computation of 77SeNMR Chemical Shifts with the IGLO-SCF, the GIAO-SCF, and the GIAO-MP2 Methods? M. Buhl* and W. Thiel Organisch-Chemisches Institut, Universitat Zurich, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland
U. Fleischer" and W. Kutzelnigg Lehrstuhl fur Theoretische Chemie, Ruhr- Universitat Bochum, 0-44780 Bochum, Germany Received: August IO, 1994; In Final Form: January 3, 1995@ Ab initio geometries and S e NMR chemical shifts have been computed for Me*Se, HzSe, MeHSe, EtSeH, Se(SiH&, SeF4, SeF6, SeOF2, HzCSe, MezCSe, SeCO, CSe2, and selenophene (C4hSe). The mean absolute deviations from experimental 6(77Se) data are 97, 76, and 78 ppm at the IGLO-SCF/II//Expt., GIAO-SCF, and GIAO-MP2/962+(d)//MP2/962(d) levels, respectively, over a range of 2800 ppm. For selenium chemical shift calculations, diffuse functions have to be included in the basis set in order to describe the Se lone pairs properly, which provide the largest localized valence M O contributions to each S e shielding. A 6(77Se) value of ca. -40 ppm is predicted for gaseous EtSeH. Both SCF and MP2 methods fail to reproduce the cY('~S~) chemical shift of Se42+which requires a multireference description.
Introduction Accurate calculations of magnetic properties can add to the unrivaled versatility of NMR spectroscopy as a structural and analytical tool in chemistry. With modem theoretical methods,' the computation of NMR chemical shifts for many nuclei has become routinely possible. The most widely used methods to date are GIAO (gauge including atomic orbitals),2 IGLO (individual gauge for localized orbital^),^ and LORG (localized orbitals, local rigi in).^ Implementations of so-called direct versions of the IGL05 and the GIA06 algorithms allow the treatment of rather large systems. All these methods are based on coupled Hartree-Fock (CHF) theory and permit computations of magnetic properties at the SCF level only. Recent developments focused on the inclusion of electron correlation which has been introduced via second-order perturbation theory in the SOLO7 and GIAO-MBPT(2) (or GIAO-MP2)8 approaches, and recently, via third-order (MP3)9s10and partial fourth-order perturbation theory (SDQ-MP4).I0 Whereas these approaches mainly take care of dynamical correlation, multiconfigurational based methods as MC-IGLO" and MC-GIAO1* are more appropriate for the description of static correlation effects. Very recently, a combination of the IGLO method with density functional theory, DFPT,13 has been employed for computation of NMR chemical shielding tensors. DFPT apparently accounts for correlation effects, too, albeit in a way not yet fully understood. The aforementioned methods, GIAO and IGLO in particular, have been extensively applied to compute chemical shifts of various nuclei in numerous classes of c o m p o u n d ~ . ~Con~J~ siderable experience has been accumulated regarding the accuracy of relative chemical shifts, computed at a given level of theory. In favorable cases, experimental 6 values can be reproduced within a few ppm already at the SCF/DZ level (Le., employing double zeta basis), e.g., for saturated hydrocarbon^,^^ most carbo~ations,'~ and polyhedral boranes.16 For the latter two, the dependence of the computed chemical shifts on the geometry employed has proven to be a valuable structural tool, termed ab initio/IGLO/NMR methodI7 (or ab initio/GIAO/
'
This paper is dedicated to Prof. Dr. P. v. R. Schleyer on the occasion of his 65th birthday. e- Abstract published in Advance ACS Abstracts, February 15, 1995.
NMR). For the heavier first- and second-row nuclei, larger basis sets (typically of singly or doubly polarized triple zeta quality) have to be used in order to achieve comparable accuracy. In many cases, particularly for nuclei involved in multiple bonds, inclusion of electron correlation in the chemical shift calculations is i m p ~ r t a n t . ~ - For ' ~ example, the error with respect to experiment for the carbonyl 613C value in acetone is reduced from ca. 10 to 1 ppm in going from GIAO-SCF to G I A O - M P ~ I ~ Compared Z ~ ~ . ~ ~to their lighter first-row counterparts, second-row nuclei appear to be somewhat less "well behaved", as far as the computational requirements (basis sets) are concerned. Higher levels are usually required to achieve comparable accuracy. For even heavier elements, relativistic effects on chemical shifts may also become important. 1 8 s i 9 Ab initio calculations of NMR chemical shifts for third-row elements are relatively scarce. A few applications have been reported for Ga (also employing GIAO-MP2),20 Ge, Se, and Br c o m p o ~ n d s . ~Within ~ - ~ ~ this period, 77Se is probably the most important nucleus, since its spin l/2 and its sufficient natural abundance make it readily observable with modem NMR technique^.^^ Therefore, and because of the importance of Se compounds in organic chemistry,26many experimental d(77Se) data are On the other hand, only a few ab initio computations of Se magnetic shieldings have been reported. Tossell and Lazzeretti employed the CHF method to calculate the shielding constants of H2Se and SeF6,22while Jokisaari, Lazzeretti, and Pyykko used the same method to compute the chemical shift anisotropy Aa(Se) in CSe2.23 GIAO-SCF computed Se chemical shifts in HzSe and CH3SeH (relative to MezSe) by Ellis et al. differed up to 135 ppm from e ~ p e r i m e n t . ~ ~ The description of these molecules is much improved with the '~~ DFFT approach (maximum deviation ca. 50 ~ p m ) ,suggesting that electron correlation is important for d(Se) even in completely "normal" molecules without any unusual bonding. We now report ab initio chemical shift calculations at SCF (IGLO and GIAO) and at correlated levels (GIAO-MP2) for a larger set of selenium containing compounds (comprising multiply bonded and hypervalent Se), which cover a d(Se) chemical shift range of ca. 2800 ppm. Indeed, substantial correlation effects on the computed d(Se) values are revealed (up to 300 ppm). Concerning the overall performance, however, the improvement of GIAO-MP2 over GIAO-SCF or IGLO is
0022-3654/95/2099-4000$09.00/0 0 1995 American Chemical Society
Computation of 77SeNMR Chemical Shifts
J. Phys. Chem., Vol. 99, No. 12, 1995 4001 1.565
1.458 1.471 1.455 I
C
1.698 1.755 1.766
1.601
1.587 /.576(4)
:j;H / 92.8 91.8 91.6 90.9
95.3 95.0 95,51201
1.708 2 6SCI?I
I."",\&,
11.3 b 91.2 91.4 92.2f.5)
104.8 105.7 105.3 lOd.8(.5)
1.732 1.759
ccsc:110.0 109.1
109.0 108.7f2) 1,969 1.972
1.461 1.475 1.457 /.467(4)
1.460
1.474 1.456
[MeCHSel
121.7 121.5 /2/.3(5)
1.523 1,521 1.525(4) H
H
9 C2"
96.4 95.7 95.3 93.5(6)
4a Cs
1.723 1.710 1.704 1.71N.I)
4b c1 165.2 167.5
1.127 1.179 1.176 1.154f.1)
1.342
se=c=o 11 c,, 1.679 1.702 1.702 1.692(2) Se= C
= Se
Figure 1. Geometries of selenium compounds 1-13 including key geometric parameters from ab initio optimizations and from experiment (in iruiics).
not spectacular. The strongly deshielded cyclic Sed2+,and to a lesser extent cyclic ieNSeNSk2+, turn out to be a challenge to theory as both SCF and MP2 based methods fail to reproduce the experimental Se chemical shifts. After submission of this paper, an independent GIAO-SCF and GIAO-MP2 study of HzSe, CH3SeH, and MezSe was published by Magyarfalvi and Pulay,61who reached the same conclusions regarding the importance of correlation effects in these molecules. In addition, we became aware of another recent theoretical study of Se NMR chemical shifts.62
Methods, Basis Sets, and Geometries Chemical shieldings have been computed employing the IGLO-SCF,3 as well as the GIAO-SCP and correlated GIAOMP28 methods, as implemented in the DIGL05 and Aces IIz7 program packages, respectively. Relativistic corrections have not been included. The following contracted basis sets have been used: 641(d): Se: Binning and Curtiss' [6s4pld] contraction28of Dunning's (14sl lp5d) primitive setz9 augmented with one set of d polarization functions (Q = 0.315); C,N,O,F,Si: 6-31G*?6 H: Huzinaga (4s) contracted to [ 2 ~ ] . ~ l 962(d): Se: same as 641(d) decontracted to [9s6p2d]; C,N,O,F: TZP, i.e., Dunnings [5s3p] augmented with one set of polarization functions (Q = 0.654, 0.8, 0.85, and 1.0 for C,N,O, and F, respectively). Si: McLean-Chandler's [6s5p] valence TZ basis3zbaugmented with one set of polarization functions (Q = 0.4). H: TZP, i.e., Huzinaga [5s] contracted to (3s),32aaugmented with one set of p polarization functions (a,= 0.7). 962+(d): same as 962d, augmented with one set of diffuse s- and p-functions on Se (a,= a, = 0.022).28
DZd:Se: ( 1 3 ~ 9 p 5 dcontracted )~~ to [8s6p2d] augmented with one set of d polarization functions (Q = 0.35). C: H ~ z i n a g a ~ ~ (7s3p) contracted to [4s2p]. H: Huzinaga (3s) contracted to PSI.
1I:Se: ( 1 5 ~ l l p 7 d 2 f contracted )~~ to [ 10s9p5d2fl (exponents of polarization functions: Q = 0.39,0.14, af= 0.35, 1.40). C, N, 0, F: H ~ z i n a g a(9s5p) ~ ~ contracted to [5s4p], augmented with one set of d polarization functions (Q = 1.0 each). Si: Huzinaga (1 ls7p) contracted to [7s6p], augmented with two sets of d polarization functions (Q = 0.35, 1.4). H: Huzinaga (5s) contracted to (3s), augmented with one set of p polarization functions (ap= 0.65).3c d(Se) chemical shifts are reported relative to MezSe, the common experimental standard. In the IGLO calculations, core and valence orbitals were localized separately. The Se d-orbitals were localized together with the core; when they were localized together with the valence MO's, the resulting absolute shieldings differ somewhat, but the overall trends are not affected.35 In particular, the relative chemical shifts and the anisotropies agree within a few ppm for both localization schemes. Geometries have been fully optimized36in the given symmetry using the Gaussian 9237program at the SCF/641(d) level. The minimum character of each stationary point has been verified by frequency calculations followed by reoptimizations at the correlated MP2/641(d) and MP2/962(d) levels. Experimental geometries have been taken from standard compendia.38 Five Cartesian d and seven Cartesian f-functions have been used in geometry optimizations and in the IGLO calculations, whereas the present GIAO-MP2 code required use of the full set of Cartesian functions. Results of chemical shift calculations are denoted "Level of calculation // geometry employed".
Biihl et al.
4002 J. Phys. Chem., Vol. 99, No. 12, 1995 TABLE 1: Basis Set Effects on Computed Se Absolute Shieldings u in MezSe and HzSe“ method basis on Se/C/H a(Me&)b o(H$3e)b 6(HzSe)’ IGLO-SCF DZdlDZdlDZd IVIVIId
II+f/IIAI
2248.3 1912.6 2036.6 1907.5 1897.1
2387.6 2163.7 2258.1 2138.3 2138.7
-129 -251 -222 -231 -242
GIAO-SCF 641(d)/6-31G*/DZ 641+(d)/6-3lG*/DZ 962(d)/TZP/DZPg 962+(d)/TZP/DZP 962+(d)/TZP/TZP 962+(f,d)h/TZP/TZP II+f/IIAI
2119.7 2065.0 1925.3 1881.9 1892.7 1879.3 1889.2
2218.7 2229.2 2102.0 2118.5 2134.5 2144.4 2135.6
-99 -164 -177 -237 -242 -265 -246
GIAO-MP2 641(d)/6-31G*/DZ 641+(d)/6-31G*/DZ 962(d)/TZP/DZP 962+(d)/TZP/DZP 962+(d)/TZP/TZP 962+(f,d)h/TZP/TZP II+f/II/II
2136.9 2075.4 1907.2 1859.0 1867.8 1868.9 1889.3
2346.2 2357.9 2181.6 2200.0 2204.6 2217.9 2228.4
-209 -293 -274 -341 -337 -349 -339
1”
experiment
-345 (gas)’ -226 (liquidy’
MP2/641(d) geometries except where otherwise noted. Isotropic absolute shielding in ppm, i s . , with respect to the bare nucleus. Se chemical shift of H2Se relative to MeZSe. Experimental geometries. e Same as basis 11, but without f functions. /Same as basis I1 augmented with one set of s and p functions (a, = a, = 0.022).g DZP basis for H same as in 641(d),augmented with one set of p polarization functions (a, = 1.0). hSame as basis 962+(d) augmented with one set of f functions (af= 0.4). Reference 24. j Reference 25.
Results and Discussion Geometries. The geometries of compounds 1-13 of this study are displayed in Figure 1, including key geometrical parameters from ab initio optimizations and from experiment. The SCF/641(d) geometries are in good agreement with experiment, as reported by Binning and Curtiss for third-row compounds in general.28 Two general trends are apparent when going to higher levels of theory: Inclusion of electron correlation at the MP2/641(d) level significantly increases most bond lengths with respect to the SCF values (up to ca. 0.04 8, for SeF4), while basis set improvement at MP2 generally results in a bond length decrease (compare MP2/641(d) and MP2/962(d) entries in Figure l), albeit to a lesser extent. At our best level, MP2/962(d), most computed parameters agree well with experiment, usually within 0.01 8, and 0.5”.At that level, SeF bond lengths are somewhat overestimated, up to ca. 0.036 A for SeOF2. Also, the CO separation in SeCO is computed ca. 0.02 A too long. For compounds of first-row elements, ab initio geometries (optimized at sufficiently high levels of theory) often perform better in chemical shift calculations than do experimental s t r u c t u r e ~ This . ~ ~ may ~ ~ ~not ~ always be the case; for a consistent description, however, the GIAO-MP2 computations employed ab initio geometries optimized at essentially the same level. IGLO calculations have been performed employing both experimental and ab initio geometries. Basis Set Effects. Absolute Se isotropic shieldings 0, computed with various methods and basis sets, are collected in Table 1 for Me2Se (1) and H2Se (2). Compound 1 is the experimental standard for relative 6(Se) chemical shifts,25while 2 is one of the very few selenium compounds for which the accurate chemical shift is known in the gas phase (6 = -344.8 PPm).24 With increasing basis set quality, the a(Se) values appear to converge to more or less the same values with IGLO-SCF and
GIAO-SCF, namely to ca. 1890 and 2140 ppm for 1 and 2, respectively. These values are in line with earlier calculations, e.g., 2163 ppm for 2 employing the CHF method,22and, more recently, 1932 and 2133 ppm for 1 and 2, respectively, employing GIAO-SCF.24 Hence, the SCF limit for &Se) of H2Se appears to be near ca. -250 ppm. The effect of electron correlation on the isotropic o(Se) in 1 is quite small,40 less than 1 ppm with the best basis (compare corresponding GIAO-SCF and GIAO-MP2 values in Table 1). For 2, GIAO-MP2 increases the selenium shielding by ca. 90 ppm with respect to the GIAO-SCF value, resulting in d(Se) of ca. -340 ppm, which is in excellent agreement with the gasphase value, -345 ~ p m and , ~is ~close to the DFPT value of ca. -400 ~ p m . ’ ~ ~ In their recent GIAO-SCF and GIAO-MP2 study of 1,2, and 3, Magyarfalvi and Pulay amve at similar conclusions.61 They used SCF geometries and more decontracted and extended basis sets on Se (up to 8slOp5dlf). While their reported absolute, isotropic Se shieldings appear to be slightly larger than our data in Table 1, e.g., 1923 (1922) and 2181 (2276) ppm for 1and 2, respectively (MP2 values in parentheses), their computed relative 6(Se) chemical shifts of 2 [-259 (SCF) and -354 ppm (MP2)] are close to our values. When examining the basis set dependence of the GIAO results in Table 1 in more detail, it appears that the inclusion of diffuse functions is far more important than that of f-functions, particularly for relative shifts. For example, 6(Se) of 2 changes by ca. -60 ppm in going from 962(d) to 962+(d) basis, while the corresponding difference between the 962+(d) and 962+(f,d) results is only 10 ppm (MP2 values in Table 1; GIAO-SCF data are similar). The IGLO results are much less sensitive to the inclusion of diffuse functions: 6(Se) of 2 changes by ca. - 10 ppm (compare IGLO-SCF/II and IGLO-SCF/II+ entries in Table 1). Obviously, IGLO basis I1 (which is considerably larger than 962(d)) is sufficiently diffuse in the valence region. In this context it should be noted that the 962(d) basis is not of triple zeta quality for Se. Rather, it was designed “analogous to the ... 6-311G* basis sets for first- and second-row atoms”.28 Apparently, diffuse functions are needed to improve the not fully saturated 962(d) basis, rather than to describe “real” diffuse orbitals as in, e.g., carbanions. In conjunction with GIAO-MP2, the 962+(d) basis appears to be the best compromise between accuracy of the results and computational effort for Se chemical shifts. In the following, this is tested for a larger set of componds which span a much wider chemical shift range. Relative Se Chemical Shifts. The relative Se chemical shifts of compounds 2-13, computed at various levels and employing different sets of geometries, are summarized in Table 2. There is some degree of variation in the theoretical 6(Se) values (compare entries in each row in Table 2), but in most cases this variation is moderate, typically a few percent of the large chemical shift range covered (ca. 2800 ppm). The graphical display of the IGLO-SCF/II//Expt., GIAO-SCF, and GIAOMP2/962+(d)//MP2/962(d) results in Figure 2 illustrates that all methods generally perform comparably well, with mean absolute deviations from experiment of 97, 76, and 78 ppm, respectively. The degree of agreement between the computed and experimental 6(Se) data is comparable to that found for sulfufl’” (“scaled’, of course, with the larger range of Se shieldings). It thus appears that the relativistic corrections to a(Se) are more or less the same for the systems studied, Le., the differences of the relativistic corrections are of the same order as the deviations typically found for other reasons (basis set dependence, medium effects).
Computation of 77Se NMR Chemical Shifts
J. Phys. Chem., Vol. 99, No. 12, 1995 4003
TABLE 2: Relative Se Chemical Shifts” IGLO-SCF compound 1 MezSe 2 H2Se 3 MeSeH 4a EtSeH (anti) 4b EtSeH (gauche) 5 Se(SiH3)z 6 SeF4 7 SeF6 8 SeOFz 9 HZC=Se 10 MezC-Se 11 Se=C=O 12 Se=C=Se 13
WI
I1//
Expt. 0 -251 - 104 -74 25 -629 885 525 1284 277 1
MPZ641(d)
-419 438 542
0.
n// MP2/962(d)
GIAO-SCF 641(d)l/ 962+(d)// MP2/641(d) MP2/962(d)
0
0
0
0
-231 -99 -38 36 -652 1023 551 1415 2768 1897 -373 397 554
-226 -118 -56 16 -657 922 551 1402 2770 1893 -368 402 559
-99 -43 -8 55 -629 699 255 1171 255 1 1856 -353 399 5 18
-270 -130 -64 11 -656‘ 966 577 1464 2909 1943 -364 446 589
GIAO-MP2 641(d)N 962+(d)// MP2/641(d) MP2/962(d) 0 0 -209 -362 -102 -180 -58 - 106 8 -22 -664 -700‘ 872 1175 384 727 1041 1365 2394 2869 1956 2200 -495 -532 89 121 5 17 598
expb 0 -345‘ -lSd 42 -666 1083 610 1378 2131f -447 299 605
In ppm relative to MezSe (see Table 1 for the absolute shieldings; c7 values employing 962+(d) basis and the MP2/962(d) geometry are 1897.6 and 1874.5 ppm with GIAO-SCF and GIAO-MP2, respectively, and 1912.5 ppm with IGLO-SCFIII). Experimental values (liquid or solution) from ref 25, except where otherwise noted. Gas phase, from ref 24; solution value -226 ppm. Gas phase, from ref 24; solution value -1 16 ppm. e DZP basis on hydrogens. ftBuzC=Se. &Se) [ppm] IGLO-SCFflWExpt.
f
/
2oool
2500
1500
-5,,i/ d
-500
5 0 0 1000 150020002500
0
N S e ) [ppm] GIAO-SCF/MP2/962+(d) 2500 2000 1500 1000
&Sc) [ppml GIAO-MPUMPU962+(d)
//MW962(d)
J
5ooj 0
/
5 O O v
S(”Se) [ppm] Expt. -500
0
500 1000 150020002500
Figure 2. Plots of theoretical (IGLO-SCF, GIAO-SCF, and GIAOMP2) vs experimental 6(77Se)NMR chemical shifts. The ideal slope = 1 is shown in each case. Linear regression analyses afford slopes of 0.90, 0.93, and 1.05, respectively. The effect on the computed Se chemical shifts of employing geometries from different sources can be studied comparing the
IGLO-SCF/II results for the experimental, MP2/641(d), and MP2//962(d) geometries in Table 2 (fist three coloumns). The 6(Se) values obtained with the ab initio geometries are shifted downfield with respect to those employing experimental structures, but only slightly so: the corresponding difference rarely exceeds 30-40 ppm. Two notable exceptions are the fluorinecontaining species 6 and 8, with geometry effects on 6(Se) of ca. 130 ppm, probably due to the overestimation of the SeF bond lengths at MP2.42 As expected, the performance of IGLO-SCF and GIAO-SCF methods is very similar (compare third and fifth coloumns in Table 2); the corresponding differences are again in the 30-40 ppm range (exception: HzC=Se, 9). The effects of electron correlation, Le., going from GIAO-SCF to GIAO-MP2, are substantially larger (compare fifth and seventh coloumn in Table 2) and can amount to more than 300 ppm (for CSe2). In most cases, electron correlation increases the Se shielding (Le., resulting in an upfield shift of the computed 6 values). For hypervalent SeF4 and SeF6, however, the selenium is more deshielded at MP2 than at SCF levels. With 962+(d) basis, Mp2 seems to “overshoot” the correlation effect on 6(Se): going from GIAO-SCF to GIAO-MP2 reverses the sign of the deviation from experiment. As mentioned above and illustrated in Figure 2, both methods perform comparably well in terms of mean absolute deviations. In the following, the results for some individual compounds are discussed in more detail. EtSeH. The computed Se chemical shifts of trans-ethylselenol (4a) and gauche-ethylselenol (4b) differ appreciably, 6(Se) = -106 and -22 ppm, respectively [GIAO-MP2/962+ (d)//MP2/962(d) level]. At that level, 4b is more stable than 4a by 0.6 kcal/mol. Neglecting entropy effects, this would correspond to an equilibrium mixture 4a:4b of roughly 1:3 at room temperature. The correspondingly averaged 6(Se) value, -43 ppm, is shielded considerably with respect to the experi~ HzSe and MeSeH, mental value in solution, +42 ~ p m . 2As~ for however, the experimental gas-phase chemical shift for 4 (which is not known) is expected to be lower than the solution data and would thus probably in better accord with the computed value. Clearly, a determination of the gas-phase Se NMR spectrum of 4 would be desirable. and the potential Both rotamers are present in the gas energy for rotation around the C-Se bond has been deduced from microwave spectroscopic data. In accord with the ab initio
4004 J. Phys. Chem., Vol. 99, No. 12, 1995 result, 4b is indicated to be lower in energy than 4a, albeit only by ca. 0.2 kcal/m01.~~ Se(SiH&. In the 1986 gas-phase electron diffraction analysis of disilyl selenide (5),44it has been noted that a C2 symmetric model with the SiH3 groups rotated by 30" fit the experimental data slightly better than the expected C2" model. As the differences between the C2 and CzVmodels were too small to be conclusive, the latter was adopted in the final refinement. Interestingly, 5 is not a minimum in C2, symmetry at MP2/ 641(d): one imaginary frequency is computed (41i cm-l). In the C2 symmetric minimum, the silyl groups are rotated by ca. 10" [MP2/641(d) level]. The energies of the CzV and the C2 forms, however, are essentially the same, suggesting free rotation of the SiH3 groups. As the other structural parameters are very similar for CzVand C2, the computed &Se) values, -694 and -705 ppm, respectively [GIAO-MP2/962+(d)//MP2/641(d)], are very close. Reoptimization of the C2 form at the higher MP2/962(d) level slightly decreases the rotation angle of the silyl groups (to ca. 8"). For MezSe (l),no other minimum besides the C2" structure could be located. SeF6. Selenium hexafluoride with Se in its highest coordination number 6 has been the subject of several ab initio s t ~ d i e s . For ~ ~ SeF6, , ~ ~ the absolute Se chemical shielding o(Se) has been deduced experimentally by means of measurements of T1 relaxation times.47 The primary result is the spin-rotation constant from which the paramagnetic contribution OP can be calculated. The total shielding o is obtained adding the diamagnetic part ad which can be estimated using the theoretically calculated shielding of the free atom. Thus, a o(Se) value in SeF6 of 1438 f 64 ppm has been obtained. This value, however, is explicitly corrected for the relativistic effect on d which was estimated to be ca. 300 ppm for @free atom).18b,c Our GIAO-MP2/962+(d)//MP2/962(d) computed o, 1148 ppm (GIAO-SCF: 1320):* is 300 ppm smaller than the experimental value and is thus in good agreement with the not explicitly corrected data. However, this accord may be to a certain extent fortuitous: the ab initio value is probably too low, since the relative Se chemical shift of SeF6 is computed too strongly deshielded at that level (GIAO-MP2 727, experiment 610 PPm).49 R2C=Se. The substituent effect on the computed Se chemical shift in going from seleno formaldehyde (9, R = H) to dimethyl selenoketone (10, R = CH3) is substantial, ca. 1000 and 670 ppm at GIAO-SCF and GIAO-MP2/962+(d)//MP2/962(d) levels, respectively. The di-tert-butyl derivative is known e ~ p e r i m e n t a l l ythe , ~ ~Se chemical shift of which (2131 ppm) compares quite well to the best GIAO-MP2 value for 10, 2200 ppm (Table 2). For selenoacetaldehyde CH3CH=Se, which has been observed in the gas phase,516(Se) = 2391 ppm is predicted [GIAO-MP2/962+(d) //MP2/641 (d)]. Tensor Components. The principal values of Se shielding tensors have been measured by means of solid-state NMR for several compound^,^^,^^ including Me2Se (1) and H2Se(2). Experimental and ab initio data in Table 3 agree only qualitatively, with deviations up to 130 ppm. The computation of chemical shieldings with a given accuracy is inherently more demanding for the principal components than for isotropic values, since possible errors in the former may be averaged or may even cancel in the latter. On the other hand, the medium effect on the isotropic Se chemical shift of 2 (relative to 1) is substantial, cf. -345 (gas phase) vs -226 ppm (liquid), and may be important for the tensor components in the solid state as well. Hence, quantitative agreement of experimental (solid state) and theoretical values (gas phase) could not be expected.
Buhl et al. TABLE 3: Principal Values of Se Shielding Tensor9 tensor components compd
method
aaa
a,,
ubb
0,"
Me2Se IGLO-SCFb 1593 (-320) 2391 (478) 1754 (-158) 1913 GIAO-SCFC1611 (-289) 2370 (473) 1709 (-188) 1897 GIAO-MP2' 1670 (-205) 2293 (418) 1661 (-214) 1874 exptd -220f4 354f4 -13414 H2Se IGLO-SCF 1768 (-396) 2622 (459) 2101 (-63) 2164 GIAO-SCF 1822 (-344) 2611 (444) 2067 (-100) 2166 GIAO-MP2 1981 ( - 2 5 5 ) 2619 (383) 2108 (-128) 2236 exptd,e -241 1 5 250f5 -10 i5 tensor components comod
method
CSe2
CHF~B~ IGLO-SCFh,' IGLO-SCFhJ GIAO-SCF GIAO-MP2 exptk
UII
UI
Ad
a".
3006 3030 3008 3003 3003
714 698 686 675 1228
2293 2332 2322 2331 1875 2210 1120
1478 1475 1460 1452 1753
Absolute shieldings in ppm; in parentheses: relative to the isotropic value, uav;orientation: a,, along twofold axis, Ubb in the molecular plane, uccperpendicular to it, ql parallel to the molecular axis, ul perpendicular to it. IGLO-SCFIIVIExpt. GIAO-SCF and MP2/ 962S(d)//MP2/962(d). Note that the GIAO-MP2 tensor components U I I and u22 of Me2Se have been inadvertently switched in Table 4 of ref 61 (Magyarfalvi, G., private communication). Experimental data from ref 52c. e Original assignment of uaa and reversed. /Chemical - al. Coupled Hartree-Fock results shift anisotropy, Au = employing basis 11; ql and ul values employing a smaller [10s9p5d] basis are 3005 and 926 ppm, respectively, cf. ref 23. Geometry from ref 23 employed, d(CSe) = 1.710 A. I a and n orbitals localized together. u and ?c orbitals localized separately. Reference 23.
It is obvious, however, that the original assignment52cof the om and o b b principal axes is erroneous for 2 (original experiment +250 and -241, GIAO-MP2 -255 and +383 ppm, respectively). The assignment, based on unpublished calculations, was found to be in agreement with a subsequent calculation,21but was recently also proven erroneous by the same author.53 For CSe2, a chemical shielding anisotropy of Ao(Se) = 2212 i 120 ppm has been determined experimentally in liquid crystals.23 SCF-based calculations yield similar values (Ao = 2322,2331, and 2293 ppm with IGLO, GIAO-SCF, and CHF, respectively, cf. Table 3), whereas a substantially lower anisotropy, Ao = 1875 ppm, is computed at GIAO-MP2. Only 01,the component perpendicular to the molecular axis, is affected by electron correlation, and is distinctly less deshielded at GIAO-MP2 as compared to GIAO-SCF. It appears that the correlation effect is overestimated at MP2, as the relative Se chemical shift computed at GIAO-MP2/962+(d)//MP2/962(d), 6 = 106 ppm, is strongly upfield from the experimental value, 299 ppm (in solution). A determination of the gas-phase 6(77Se) value of CSe2 would be of interest in order to estimate medium effects on the chemical shifts in the liquid and in the liquid crystal phase. Problem Cases. The cyclic selenium-nitrogen dication
-
(SeNSeNSe)2+(14) has the highest 6(77Se)chemical shift known so far, 2434 A single signal is found for the two nonequivalent Se nuclei, suggesting a dynamic exchange process. We were interested whether chemical shift calculations could reproduce the experimental (averaged) value and predict the individual Se chemical shifts of the static C2" structure. In the optimization and chemical shift calculations, only 641(d) and 641+(d) basis sets,55 respectively, could be employed. Structural and NMR data are summarized in Figure 3a. Again, optimized bond lengths appear to be overestimated at the MP2 level, compared t o the crystal structure (AsF6-
J. Phys. Chem., Vol. 99, No. 12, 1995 4005
Computation of 77Se NMR Chemical Shifts
TABLE 4: IGLO-SCF LMO Contributions to the Se Chemical Shieldinp compd Lb Mb SexC SeH lpd Re MeZSe HzSe MeSeH EtSeH(anti) EtSeH(gauche) Se(SiH3)z SeF4
1.769
S (average)
N
2.392 2.33413) Geometry: SCF/661(d) M W 6 4 I (d)
14 C2”
X-Ray
we): p
!?auhwl
seF6 SeOF2
I I
Se-
HzC=Sd MezC-Sd SeCO CSe2
I
Se-
Se
’=% Lm
15 D4h
Figure 3. Geometries of cyclic dications 14 and 15 including computed and experimental 77Sechemical shifts.
a
832 879 1031 990
290 311 384 375
-88 -89 -95 -122 -28 -254(eq) - 198 (ax) -166 -225(SeF) -667 (SeO) -869 -370 -1349 -16oh
958 321 -263
c s e
Se
I
2.274 2.360 2.28
320 333 332 330 327 362 293
996 232 939 265
b)
1 2+
975 998 976 974 971 1008 982
-47 -54 -65 -55
-213 -160 -191 -191 -225 6 -462
-15 -21 -31 -31 -129
1913 2164 2016 1986 1887 2542 1027
-631
-93 -62
1387 576
-1131 -98 -788 - 1 1 1 -539 -37 -44V -209
-871 11 2332 1475
-273
-28
u
-84
1370
Contributions in ppm of the localized MO’s to the total shielding
u;basis II and experimental geometries have been employed except where otherwise noted. Sum of the MO contributions of the L- and the M-shell, respectively;the contribution of the 1s orbital is constant (1247 ppm). Contribution of each S e x bond; for doubly bonded Se sum of “d’and “n” MO’s. Contribution of each lone pair. e Other. fMP2/641(d) geometry employed. 8 The localization yielded one SeC bond and three lone pairs. The localization yielded three SeC “bonds” and one lone pair.
counter ion^).^^ The correspondingly averaged GIAO-SCF Se chemical shift, 2657 ppm, is considerably downfield from experiment, but the GIAO-MP2 value, 1894 ppm, deviates by more than 500 ppm. For one set of Se nuclei, GIAO-SCF and 20001 10 GIAO-MP2 chemical shifts differ by as much as 1000 ppm. Hence, this may be a case where perturbation theory cannot be applied. 1500. *8 The same is apparent, even more drastically, for Se2+, another cyclic dication (15, Figure 3b). As far as the Se *6 1000. chemical shift is concerned, both SCF and MP2 methods fail completely, with deviations from experiment of roughly f 2 0 0 0 13 500 ppm (Figure 3b). According to a preliminary CI(SD) calcula12 tion,56 15 is most probably a case where multireference methods, 4 * *1 0 : Le., MC-IGLOlI or MC-GIA012 would be needed to compute *3 magnetic pr0perties.5~ The ground-state determinant of 15 has -500 2’*1* a weight of only ca. 91% in the CI(SD) wave function. It is p5 noteworthy that the DFPT13 approach performs very well in . NPACharge this case: the DFPT-11 computed &Se) value, 2030 ~ p m is, ~ ~ 0 1 2 3 in good agreement with experiment, 1958 ppm. Figure 4. Plot of experimental 6(77Se)chemical shifts vs. computed Interpretation of Se Chemical Shifts. In the IGLO apselenium atomic charges from natural population analysis (NPA). proach, the total shielding o is obtained as a sum of contributions of localized MO’s. LMO contributions have been discussed, selenides, the corresponding MO contributions are quite similar. e.g., for the interpretation of 13Cchemical shifts.3c It was found Nevertheless, the differences in the total shieldings cannot be that the contributions of specific bond types are transferable explained in terms of transferable MO contributions. For doubly only between closely related molecules (e.g., saturated hydrobonded Se in HZC=Se (9) and MezC=Se (lo), the S e x and carbons). lone pair contributions are strongly dependent on the substituents in p-position. The larger the paramagnetic contributions are, The IGLO Se total shieldings of compounds 1-13 are the larger is the sensitivity to small perturbations of the decomposed into the LMO contributions in Table 4. As has electronic structure (in 9 and 10, the induced interactions of been found for other nuclei,3c the K-shell contribution is transferable, whereas the other core contributions are generally the ocx bonds with the low-lying z*c-se gives rise to large deshielding contributions). Except for Se(SiH3)z, the lone pair(s) not. The variations of the L-shell contributions are smaller than on Se provide the largest single LMO contribution from the those of the M-shell, but are nevertheless significant: 980 f 20 ppm for most compounds, with larger variations in systems valence shell. This may rationalize the necessity of employing sufficiently diffuse basis functions in chemical shift calculations with strong shielding or deshielding MO contributions, cf. in order to describe the selenium lone pairs properly. Se(SiH3)z and HZC=Se. AI1 contributions of valence shell MO’s As is apparent from Figure 4, no general correlation exists (comprising 4s and 4p orbitals on Se) are paramagnetic, i.e., between d(Se) for 1-13 and atomic Se charges, computed with deshielding. Shielding or deshielding of a(Se) with respect to natural population analysis The values of the singly 1 cannot be traced back to specific types of LMO’s. Rather, all contributions change simultaneously. For H2Se and the alkyl bonded (1-5) and the stabilized multiply bonded systems (11-
2500E
-
1
4006 J. Phys. Chem., Vol. 99, No. 12, 1995 13) might be interpreted in terms of such a correlation: the 6(77Se)values between ca. -650 and f 6 0 0 ppm correspond to computed charges from ca. -0.5 to +0.5e. However, even though the NPA charge in the selenoketone 10, ca. O.Oe, is within this range, the selenium is strongly deshielded (6 > 2000 ppm). The formally “hypervalent” compounds 6, 7,and 8 are characterized by highly positive central Se atoms. The NPA charge in SeF6 (8), +3.3e, is comparable to that of sulfur in SF6 (+2.9e).60 In both cases, the highly positive charges on the central atoms are not reflected in exceptional downfield shifts in the corresponding NMR spectra. The selenium NPA charge in SeF4 (6), +2.5e, is less positive than in 8; 6 is deshielded with respect to 8, however, due to the large paramagnetic contribution of its lone pair (-450 ppm, Table 4) which 8 is lacking. Conclusions The IGLO and GIAO methods have been employed to compute relative Se NMR chemical shifts to an accuracy of typically better than f l O O ppm, Le., a few percent of the selenium chemical shift range (ca. 3000 ppm). At the IGLOSCF/II//Expt., GIAO-SCF, and GIAO-MP2/962+(d)//MP2/ 962(d) levels, all methods perform comparably well in terms of mean absolute deviations from experiment. In some cases, substantial electron correlation effects on 6(Se) are computed when going from SCF to MP2 (e.g., more than 300 ppm for CSe2, cf. Table 2). As this correlation effect appears to be overestimated by MP2, accurate computations of Se NMR chemical shifts may have to go beyond this level. In addition, the comparison of theoretical and experimental 6 values is hampered by possible medium effects on 6(77Se)and by the lack of gas-phase NMR data. For HZSe, where 6(77Se)is known in the gas phase (and where a considerable gas-to-liquid shift of ca. 120 ppm is apparent),24 the best agreement with experiment is found at GIAO-MP2 levels. Chemical shift calculations need to employ sufficiently diffuse basis functions on Se for a proper description of its lone pairs, as rationalized by an analysis of the contributions of the localized MO’s to the total Se shielding: the lone pair(s) on Se provide the largest single LMO contribution from the valence shell. The computed Se chemical shifts are affected by the specific molecular conformation (cf. the 80 ppm difference between the trans and the cis rotamers of EtSeH) and by substituents one bond away (cf. the ca. 700 ppm difference between H2C=Se and Me2C=Se). The cyclic dication Se42f turns out to be a multiconfiguration case. Accordingly, both SCF and MP2 methods, which are single-reference based, fail to reproduce its experimental Se chemical shift.
Acknowledgment. This work was supported by the AlfriedKrupp-Stiftung. We thank Dr. J. Gauss for a copy of the GIAOMP2 program as well as Dr. U. Meier, Dr. C. van Wullen, and Dr. M. Schindler for the direct IGLO version. Discussions with Dr. V. I. Malkin were very helpful, and we thank him for supplying us with unpublished data and with a preprint of ref 13b. Calculations were performed on IBM-RS6000 workstations of the Rechenzentrum and the Organisch-Chemisches Institut of the Universitat Zurich, and on a Cyber 205 of the Rechenzentrum der Rub-Universitat Bochum. We also thank a referee for helpful comments. References and Notes (1) Reviews: (a) Webb, G. A. In Nuclear Magnetic Shieldings and Molecular Structure; NATO AS1 Series; Tossell, J. A,, Ed.; Kluwer Academics: Amsterdam, 1993; p 1. (b) Chesnut, D. B. Annu. Rep. NMR Spectrosc. 1989, 21, 51.
Buhl et al. (2) (a) Ditchfield, R. Mol. Phys. 1974, 27, 789. (b) Wolinski, K.; Hinton, J. F.; Pulay, P. J. Am. Chem. Soc. 1990, 112, 8251. (3) (a) Kutzelnigg, W. Zsr. J. Chem. 1980, 19, 193. (b) Schindler, M.; Kutzelnigg, W. J. Chem. Phys. 1982, 76, 1919. (c) Review: Kutzelnigg, W.; Fleischer, U.; Schindler, M. NMR Basic Principles and Progress; Springer: Berlin, 1990; Vol. 23, p 165. (4) Hansen, Aa.E.; Bouman, T. D. J. Chem. Phys. 1985, 82, 5035; 1989, 91, 3552. (5) Meier, U.; van Wullen, C.; Schindler, M. J. Comput. Chem. 1992, 13, 551. (6) Haser, M.; Ahlrichs, R.; Baron, H. P.; Weiss, P.; Hom, H. Theor. Chim. Acta 1992, 83, 455. (7) Bouman, T. D.; Hansen, Aa. E. Chem. Phys. Lett. 1990, 175,292. (8) (a) Gauss, J. Chem. Phys. Lett. 1992, 191, 614. (b) Gauss, J. J. Chem. Phys. 1993, 99, 3629. (9) Fukui, H.; Baba, T.; Matsuda, H.; Miura, K. J. Chem. Phys. 1994, 100, 6608. (10) Gauss, J. Chem. Phys. Lett. 1994, 229, 198. (11) (a) Kutzelnigg, W., van Wullen, C.; Fleischer, U.; Franke, R.; v. Mourik, T. In Nuclear Magnetic Shieldings and Molecular Structure; NATO AS1 Series; Tossell: J. A., Ed.; Kluwer Academics: Amsterdam, 1993; p 141. (b) van Wiillen, C.; Kutzelnigg, W. Chem. Phys. Lett. 1993,205, 563. (12) (a) Jaszunski, M.; Helgaker, T.; Ruud, K.; Bak, K. L.; Jfirgensen, P. Chem. Phys. Lett. 1994,220, 154. (b) Ruud, K.; Helgaker, T.; Kobayashi, R.; Jorgensen, P.; Bak, K. L.; Jensen, H.-J. Aa., J . Chem. Phys. 1994, 100, 8178. (13) (a) Malkin, V. G.; Malkina, 0. L.; Salahub, D. Chem. Phys. Lett. 1993,204, 80, 87. (b) Malkin, V. G.; Malkina, 0. L.; Casida, M. E.; Salahub, D. J. Am. Chem. Soc. 1994, 116, 5898. (14) E.g: Chesnut, D. B.; Rusiloski, B. E.; Moore, K. D.; Egolf, D. A. J. Comput. Chem. 1993, 11, 1364. (15) See e.g.: (a) Schindler, M. J. Am. Chem. SOC. 1987, 109, 1020. (b) Schleyer, P. v. R., Carneiro, J. W. de M. J. Am. Chem. Soc. 1990, 112, 4046. (c) Koch, W.; Schleyer, P. v. R.; Buzek, P.; Liu, B. Croat. Chim. Acta 1992, 65, 655, and references cited therein. (16) See e.g.: (a) Buhl, M.; Schleyer, P. v. R. J. Am. Chem. Soc. 1992, 114, 477. (b) Buhl, M.; Gauss, J.; Hofmann, M.; Schleyer, P. v. R. J. Am. Chem. Soc. 1993, 115, 12385, and references cited therein. (17) See extensive bibliography in ref 16a; more recent applications include e.g.: (a) McKee, M. L.; Buhl, M.; Schleyer, P. v. R. Inorg. Chem. 1993, 32, 1712. (b) Onak, T.; Tran, D.; Tseng, J.; Diaz, M.; Arias, J.; Herrera, S. J. Am. Chem. Soc. 1993, 115, 9210. (18) (a) Pyykko, P.; Gorling, A.; Rosch, N. Mol. Phys. 1987, 61, 195. (b) Kolb, D.; Johnson, W. R.; Shorer, P. Phys. Rev. 1982, A16, 19. (c) Malli, G.; Froese, C. Int. J. Quantum Chem. 1967, 15, 95. (19) Relativistic effects may also influence the chemical shifts of neighboring nuclei considerably, cf. refs l l a and 18. (20) Loos, D.; Schnockel, H.; Gauss, J.; Schneider, U. Angew. Chem. 1992, 104, 1376; Angew. Chem., In?. Ed. Engl. 1992, 31, 1362. (21) E.g.: (a) Sugimoto, M.; Kanayama, M.; Nakatsuji, H. J. Phys. Chem. 1993,97,5868. (b) Fleischer, U. Ph.D. Thesis, Bochum (Germany), 1992. (22) Tossel, J. A.; Lazzeretti, P. J. Magn. Reson. 1988, 80, 39. (23) Jokisaari, J.; Lazzeretti, P.; Pyykko, P. Chem. Phys. 1988, 123, 339. (24) Ellis, P. D.; Odom, J. D.; Lipton, A. S. Chen, Q.; Gulick, J. M. In Nuclear Magnetic Shieldings and Molecular Structure; NATO AS1 Series; Tossell, J. A,. Ed.; Kluwer Academics: Amsterdam, 1993; p 539. (25) (a) Mason, J., Ed. Multinuclear NMR: Plenum Press: New York, 1987. (b) Brevard, C.; Granger, P. Handbook of High Resolution Multinuclear NMR; Wiley: New York, 1981. (26) E.g.: (a) Krief, A.; Hevesi, L. Organoselenium Chemistv: Springer Verlag: Berlin, 1988. (b) Paulmier, C. Seleenium Reagents and Intermediates in Organic Synthesis; Pergamon Press: Oxford, U.K., 1986. (27) Aces 11: Stanton, J, F.; Gauss, J.; Watts, J. D.; Lauderdale, W. J.; Bartlett, R. J., Quantum Theory Project, University of Florida, Gainesville, FL, 1991. (28) Binning, R. C.; Curtiss, L. A. J. Comput. Chem. 1990. 11, 1206. (29) Dunning, T. H. J. Chem. Phys. 1977, 66, 1382. (30) (a) Hehre, W. J.; Ditchfield, R.; Pople, J. A. J. Chem. Phys. 1972, 56, 2257. (b) Hariharan, P. C.; Pople, J. A. Theor. Chim. Acta 1973, 28. 213. (c) Gordon, M. S. Chem. Phys. Lett. 1980, 76, 163. (31) (a) Huzinaga, S. J. Chem. Phys. 1965, 42, 1293. (b) Dunning, T. J. Chem. Phys. 1970, 53, 2823. (32) (a) Dunning, T. H. J. Chem. Phys. 1970, 53, 2823; 1971, 55, 716. (b) McLean, A. D.; Chandler, G. S. J. Chem. Phys. 1980, 72, 5639. (33) Huzinaga, S. Approximate Atomic Wave Functions; University of Alberta: Edmonton, Alberta, 1971. (34) Basis I1 has been generated from the primitive set of Dunning (ref 29) by keeping the steepest functions (7s and 5p) and choosing the other exponents in an “even-tempered‘’way @ = 2.5,48.8-0.08, and 7.2-0.074). (35) Fleischer, U.; Buhl, M., unpublished results; the differences of calculations using Gaussian lobes instead of Cartesian Gaussians are even smaller (provided the same localization scheme is used in both cases).
J. Phys. Chem., Vol. 99, No. 12, I995 4007
Computation of 77Se NMR Chemical Shifts (36) Standard methods have been employed, cf.: Hehre, W.; Radom, L.; Schleyer, P. v. R.; Pople, J. A. A b Initio Molecular Orbital Theory; Wiley: New York, 1986. (37) Gaussian 92, Revision B. Frisch, M. J.; Trucks, G. W.; HeadGordon, M.; Gill, P. M. W.; Wong, M. W.; Foresman, J. B.; Johnson, B. G.; Schlegel, H. B.; Robb, M. A.; Replogle, E. S.; Gomperts, R.; Andres, L.; Raghavachari, K.; Binkley, J. S.; Gonzales, C.; Martin, R. L.; Fox, D. J.; DeFrees, D. J.; Baker, J.; Stewart, J. J. P.; Pople, J. A. Gaussian, Inc.: Pittsburgh, PA, 1992. (38) Hellwege, K. H., Hellwege, A. M., Eds. Landoldt-Biimstein, New Series 1117, Structure Data of Free Polyatomic Molecules; Springer Verlag: Berlin, 1987. (39) Chesnut, D. B.; Phung, C. G. J. Chem. Phys. 1989, 91, 6238. (40) For MezSe, the principal values ull are more strongly affected by electron correlation, but these changes are largely canceled in the isotropic average, u (cf. Table 3). (41) Fieischer, U.; Schindler, M. Chem. Phys. 1988, 120, 103. (42) The geometry effect for 6 and 8 is somewhat attenuated at GIAO-MP2/962+(d): the corresponding differences between MP2/962(d) and experimental geometries are 57 and 40 ppm, respectively, resulting in &Se) chemical shifts of 1118 and 1325 ppm, respectively, at GIAO-MP2/ 962+(d)//Expt. (43) (a) Nagakagawa, J.; Okutani, H.; Hayashi, M. J. Mol. Spectrosc. 1982, 94, 410. (b) Durig, J. R.; Bucy, W. E. J. Mol. Spectrosc. 1982, 94, 410. (44)Almenningen, A,; Femholt, L.; Seip, H. M. Acta Chem. Scand. 1986, 22, 51. (45) Klobukowski, M. K. J. Comput. Chem. 1993, 14, 1234. (46) Dobbs, K. D.; Hehre, W. J. J. Comput. Chem. 1986, 7, 359. (47) Jameson, C. A.; Jameson, A. K. Chem. Phys. Lett. 1987,135,254. (48) Tossell and Lazzeretti obtained a value of 1595 ppm employing CHF, cf. ref 22. (49) In contrast to HzSe and MeSeH, the gas-to-liquid shift of 6(77!k) in SeF6 is very small, ca. 1 ppm, cf.: (a) Jameson, C. I., Jameson, A. K.;
Oppusunggu, D. J. Chem.Phys. 1986,85,5480. (b) Jameson, C. J.; Jameson, A. K. J. Chem. Phys. 1986, 85,4584. (50) Cullen, E. R.; Guziec, F. S.; Murphy, C. J.; Wong, T. C.; Andersen, K. A. J. Am. Chem. SOC. 1981, 103, 7055. (51) Hutchinson, M.; Kroto, H. W. J. Mol. Spectrosc. 1978, 70, 347. (52) (a) Overview: Sebald, A. In NMR, Basic Principles and Progress; Springer Verlag: Berlin, 1994; Vol. 31, p 91. (b) Collins, M. J.; Ratcliffe, C. I.; Ripmeester, J. A. J. Magn. Reson. 1986, 68, 172. (c) Facey, G.; Wasylishen, R. E.; Collins, M. J.; Ratcliffe, C. I.; Ripmeester, J. A. J. Chem. Phys. 1986, 90, 2074. (53) Tossell, J. A., private communication to U. Fleischer. (54) Awere, E. G.; Passmore, J.; White, P. S. J. Chem. SOC.,Dalton Trans. 1993, 299. (55) Basis 641+(d): same as 641(d), augmented with the same diffuse functions as 962+(d). (56) Configuration interaction with single and double excitations; the calculation employed the Gaussian 92 program (ref 37), 641(d) basis set, and the MP2/641(d) geometry. (57) Unfortunately, 15 is presently too big for a proper treatment with MC-IGLO. (58) Malkin, V. G., private communication. (59) Reed, A. E.; Weinstock, R. B.; Weinhold, F. J. Chem. Phys. 1985, 83, 735. (60) Reed, A. E.; Weinhold, F. J. Am. Chem. SOC. 1986,108,3586; the question of hypervdency is addressed there, see also: Reed, A. E.; Schleyer, P. v. R. J. Am. Chem. SOC.1990, 112, 1434. (61) Magyarfalvi, G.; Pulay, P. Chem. Phys. Lett. 1994, 225, 280. (62) Nakatsuji, H.; Higashioji, T.; Sugimoto, M. Bull. Chem. SOC. Jpn. 1993, 66, 3235. JP942121P
