8604
J. Phys. Chem. 1995, 99, 8604-8607
Equilibrium Structure of SiH3NCO: Comparison of Theory and Experiments Mikl6s FehCr*3? and Michael J. Smit Institute of Physical Chemistry, University of Basel, Klingelbergstrasse 80, CH-4056 Basel, Switzerland
Tibor Pasinszki and Tamas Veszpremi Inorganic Chemistry Department, Technical University of Budapest, H-1.521 Budapest, Hungary Received: December 20, 1994; In Final Form: March 6, 1995@
The bending potential function of SiHsNCO, including relaxation effects, was calculated at the DZP/QCISD and TZV2P/QCISD levels of theory to establish reasons for the disagreement between the ab initio and experimental equilibrium structures. The microwave, infrared, and Raman spectra were reinvestigated using this bending potential. It was shown that the asymmetric top geometry with a double-minimum bending potential is not a unique interpretation. Two alternative assignments for the microwave spectra were put forward, based on a symmetric top structure and a strongly anharmonic bending potential. It was shown that including triple excitations in the ab initio calculations at the TZV2P/QCISD(T) level generates a small barrier in the potential. Also, the importance of allowing for relaxation in the analysis of the microwave spectra was demonstrated, as this substantially decreases the barrier size. These results bring the experimental and ab initio structures closer to each other. The effects of small-amplitude motions and the existence of a small internal rotation barrier are also discussed.
Introduction
In our previous publication,' we calculated the bending potential function of SiH3NCO at the ab initio 6-31G**/MP2 level by optimizing the geometry of the molecule at different values of the SiNC angle (e). We argued that this treatment allows for the interactions of all vibrations with the largeamplitude bending mode (Le. relaxation). The resulting potential was found to be described well by a polynomial function, which had no local maximum at e = 180". Although this result is in agreement with structures derived from early microwave2 and infrared3 spectra, it is incompatible with the potentials from microwave,"-' far-infrared,s and Raman9 studies that inferred a bent SiNC frame and a barrier to linearity of -30 cm-'. On the other hand, an important common feature of these later experimental potentials and the ab initio one is that they are strongly anharmonic. Both kinds of potentials are compatible with the results of electron diffraction.'O As was shown in ref 1, ab initio calculations can satisfactorily represent the structures of most pseudohalides with a quasilinear structure, except for SiH3NCO. In our previous work we made a number of suggestions for the possible causes of this discrepancy. We recommended that it may be necessary to include the effects of relaxation and the torsion of the SiH3 group in the analysis of the microwave spectra, as well as to allow fully for the effects of small-amplitude motions. It was even suggested that alternative assignments may also be possible and should be tested. The aim of this work was to investigate these issues in a more quantitative manner. Calculations
The ab initio calculations in this work were performed using the CADPAC" and Gaussian-9212quantum chemistry packages. Bending potential energy surfaces were obtained by varying the
* To whom correspondence should be addressed.
' Present address: Chinoin Pharmaceutical and Chemical Works Ltd., Research Analytical Laboratories, T6 u. 1-5., H-1045 Budapest, Hungary. Abstract published in Advance ACS Abstracts, May 1, 1995. @
0022-365419512099-8604$09.00/0
SiNC angle and optimizing all other geometrical parameters at each point. This procedure ensures that the interaction of all vibrations is being explicitly taken care of. In addition to our previously obtained potential surface at the 6-3 lG**/MET! level,' the potential energy curves were recalculated using full double-9 (D95*)I3 and TZV2PI4-l5basis sets. The polarization functions were taken from ref 16. Gradient geometry optimizations were performed at the QCISD level of theory (with only the valence orbitals correlated). These optimizations, however, failed at 180" for the basis sets D95* and TZV2P: the values of the gradients and forces showed oscillatory behavior. For this reason, optimizations were carried out at 179" instead. Additionally, single-point calculations were performed at each point of the potential to assess the contribution from triple and quadruple excitation. Unfortunately, due to the size of the problem, the quadruple contribution could only be estimated from MP4 calculations. A full optimization was also performed at SiNC angles of 165", 175", and 179" including the triple contribution (Le. TZVZP/QCISD(T)) using the Fletcher-Powell method. The entire potential, including triple substitutions, could not be obtained due to the time-consuming nature of these calculations. For the calculation of rovibrational energy levels, the program RCFITI7 was employed. The rotation-bending internal rotation Hamiltonian used in the program7,l8 is directly applicable for symmetric, asymmetric, and quasisymmetric top molecule^.^,' An important feature of the program is that it directly includes the change of the reduced mass as a function of the bending angle. The energy and the rotational constants in different rovibrational states are calculated, and if required, the parameters of the potential function can be fitted to achieve minimum standard deviation between the observed and calculated constants. The original program was modified by us to include an internal rotation barrier and to calculate rotational energy levels and transitions for J > 0 and K > 0 levels. 0 1995 American Chemical Society
J. Phys. Chem., Vol. 99, No. 21, 1995 8605
Equilibrium Structure of SiH3NCO
TABLE 1: Calculated Potentials at Different Levels of Theory for the SiNC Bend in SiH3NCO" Q (deg)
6-31G**/MP2 [1Ib
DZP/QCISD'
QCISD(T)d
115.0 125.0 135.0 145.0 155.0 165.0 175.0 179.0
-458.45547 1 -458.458734 -458.460771 -458.461955 -458.462599 -458.46294 1 -458.463099 -458.4631 199
-458.459032 -458.462246 -458.464015 -458.464908 -458.465359 -458.465637 -458.465 804 -458.465 804
-458.47909 1 -458.482115 -458.483722 -458.484462 -458.484773 -458.484940 -458.485045 -458.485059
(1
DZP/QCISD with torsion
TZV2P/QCISDe
-458.458998 -458.462293 -458.464096 -458.464971 -458.465388 -458.465644 -458.465804
TZV2P/QCISD(T)'
-458.554508 -458.557822 -458.559697 -458.560683 -458.561136 -458.561210 -458.561339 -458.561342
-458.585161 -458.585 131 -458.585125
Energies were calculated with full optimization (unless otherwise indicated) for each angle of the SiNC bond. Best fitting quartic potential:
+
+
e
+
e
V(Q) = 3 9 9 . 4 ~ ~6 9 5 . 7 ~ (in~ cm-I, in radians).' Best fitting hextic potential: V ( e )= 4 4 4 . 2 4 ~ ~1 2 4 . 7 6 ~ ~3 3 3 . 0 4 ~(in~ cm-', in radians). ~ cm-I, in Single-point calculations at the QCISD optima (Le. DZP/QCISD(T)//DZP/QCISD). e Best quartic potential: V ( g ) = 9 0 1 . 8 7 ~(in
e
radians). The hextic component is negligibly small. f Full optimizations (TZV2P/QCISD(T)//TZV2P/QCISD(T))using the Fletcher-Powell method. 8 Obtained at 180".
Results and Discussions The calculated bending potentials of SiH3NCO at different levels of theory are given in Table 1. It can be seen that although an increased basis set (TZV2P instead of 6-31G**) and a higher level of theory (QCISD instead of MP2) were used in the optimizations, the bending potential still has no maximum at the linear configuration. The effect of including triple substitutions in the CI matrix on the total energy or conformation of the molecule was first checked by single-point calculations at the DZP/QCISD-optimized points of the potential curve. The fact that the QCISD and QCISD(T) curves are parallel gives an indication that the contribution from triple substitutions has little influence on the geometry and only lowers the total energy. Nevertheless, the geometries at 165", 175", and 179" were optimized with allowance for triple substitutions (TZV2P/ QCISD(T)), and at this level, a small barrier to linearity appears. On fitting a parabola to these points, the barrier size was estimated to be -8 cm-'. However, this barrier must be treated with caution. It is known that QCISD(T) calculations overestimate the contribution from triple e x c i t a t i o n ~ . l ~It*has ~ ~also been shown that calculations at the QCISD(T) level may predict erroneous dissociation energies (Le. an erroneous potential);*' thus, the barrier may be exaggerated by the calculations and the actual height may be smaller than 8 cm-'. The effect of quadruple excitations could not be calculated directly for such large basis sets and was estimated from the difference between MP4D and MP4DQ energies. This interaction appears to make only a negligible contribution to the total energy of SiH3NCO. Firstly B-values for different rovibrational states were calculated directly using the three ab initio potentials (MP2/631G**, DZP/QCISD, and TZV2P/QCISD) and compared with those from microwave spectroscopy (from Table 3 of ref 6). The standard deviation between the experimental and calculated B-values was below 30 MHz for all three potentials, which represents an absolute error of -1%. The level of theory of the ab initio calculations did not have an appreciable effect on the quality of the fits. This demontrates that the inadequacy of the ab initio calculations is not to blame for this discrepancy and another explanation must be sought. Relaxation (Le. the change of bond lengths and angles during the bending motion) is usually neglected in microwave spectroscopy in contrast to the calculated ab initio potential that includes this effect. From the TZV2P/QCISD calculations the change of the three most sensitive bond lengths and bond angles as a function of the SiNC bending angle, e, was fitted to the following expressions (bond lengths in angstroms, bond angles in degrees):
rNc(Q)= 1.1975
+ 0 . 0 2 6 3 ~-~
+ rsiN(Q)= 1.7030 + 0 . 1 4 1 1 ~-~0 . 1 6 9 3 ~+~0 . 0 7 6 4 ~ ~ a,,&) = 180.00 - 1 8 . 0 4 0 4 ~+~2 4 . 5 2 3 6 ~-~9 . 8 5 1 7 ~ ~
0 . 0 0 4 0 8 8 ~ ~0 . 0 0 0 0 2 7 8 3 ~ ~
Of these, relaxation of the NCO angle was found to affect the calculated levels most. Using these relaxation parameters, the potential from microwave spectroscopy was refitted with the semirigid-bender model. It was found that the barrier to linearity decreased (from 33.5 to 22.7 cm-I). The equilibrium angle did not change substantially, which is not surprising, since the potential is rather flat. This shows that for molecules with small barriers to linearity allowance for relaxation is important. Another factor that might be responsible for the difference between the experimental and theoretical values is the proper consideration of the internal SiH3 rotation. In the microwave spectrum analysis this group was taken as a free rotor.' Ab initio calculations in this work indicate a small barrier of rotation (-10 cm-I at 155'). This bamer, however, is insignificant, and the silyl group can practically be considered as freely rotating. It must be noted that, in contrast to what had been expected, the rotational barrier is at the sterically more favorable staggered configuration. We earlier recommended' that the difference between the experimental and theoretical bending potential curves may have to be found in the neglect of small-amplitude motions. As has been shown for HCN0,22the effective bending potential as a function of the bond angle, e, can be expressed as
which is in effect the pure ab initio bending potential modified by the zero-point energy. This effective potential has been used in an attempt to reconcile a similar disagreement between ab initio and experimental potentials in HCNO 23 and trimethylene oxide.24 In both cases it was found that although the inclusion of this effect does not entirely resolve the discrepancy, the effective potential lies between the experimental and the equilibrium potentials. In the present work, the zero-point energy contribution was calculated from the sum of zero-point energies of the highfrequency vibrations as a function of the bending angle (e) at the 6-31G**/MP2 level. The following polynomial was fitted to the results: Veff= 8725.4904 - 0 . 9 7 8 3 ~ ~0 . 4 0 5 6 ~(energy ~ in cm-', e in radians). This function, however, is very shallow and produces no difference in the calculated energy levels (to
+
8606 J. Phys. Chem., Vol. 99, No. 21, 1995
FehCr et al.
TABLE 2: Calculated B-Values Using a Symmetric Top Model and Two Alternative Assignments for the Microwave SDectrum of the SiHlNCO Molecule B (exp.)6 vlb B (calc) Ac vld B (calc) A' 2517.9 2542.3 2543.1 2562.8 2565.2 2581.1 2577.1 2584.7 2598.1 2593.7 2602.6 2614.1 2607.0 2610.5 2619.3 2629.4 2621.4 2626.7
On 1' 2' 22 3' 3' 4' 42 44 5' 55 6' 6' 64 66 7' 73
2514.3 2538.5 2552.5 2559.0 2569.5 2577.3 2582.9 2586.1 2594.2 2597.3 2602.1 2610.1 2609.8 2611.8 2617.4 2625.2 2622.8 2626.1
3.6 3.8 -9.4 3.8 -4.3 3.8
-5.8 -1.4 3.9 -3.6 0.5 4.0 -2.8 -1.3 1.9 4.2 -1.4 0.6
0' 1' e 22 2' 3' 3' 4' 44 42 5'
53 6' 62 66 64 7'
2520.3 2544.2
-2.4 -1.9
2564.3 2558.0 2582.3 2574.7 2587.8 2599.0 2591.0 2602.0 2614.6 2606.7 2614.3 2616.3 2629.4 2621.7 2627.1
-1.5 7.2 -1.2 2.4 -3.1 -0.9 2.7 0.6 -0.5 0.3 -3.8 3.0
0.0 -0.3 -0.4
a In MHz. Parameters of the potential were fitted to get the best standard deviation on the B-values. The fit was started using the TZV2P/QCISD potential, and the resulting fitted potential was V ( e )= 1792.4~ (in~ cm-', e in radians). The u of the fit was 4.165. Observed minus calculated in MHz. Parameters of the potential were fitted to get the best standard deviation on the B-values after reassignment. The fit was started using the TZV2P/QCISD potential, and the resulting fitted potential was V(Q)= 1 9 0 0 . 7 ~(in~ cm-I, e in radians). The u of the fit was 2.672. e This transition was not fitted.
within 0.05 MHz). Thus unlike in refs 23 and 24, the inclusion of the zero-point energy has a negligible effect on the calculated energy levels. The TZV2P/QCISD bending potential could be fitted well by the quartic function, V(Q)= 9 0 1 . 8 7 ~(in~ cm-I, e in radians), and this was then fitted using the experimental B-values.6 It was found that, within the symmetric top model, best fit can be achieved with a similar quartic potential, V(Q)= 1 7 9 2 . 3 9 (in ~~ cm-I, in radians). As shown by Table 2, the standard deviation of the rotational constants is acceptable and similar to that with a double-minimum potentiaL6 In the first microwave2 and infrared3 studies on SiH3NC0, only the v = 0 levels were considered and the molecule was thought to be a symmetric top. A similar conclusion could be drawn on the basis of the microwave spectrum of the analogous trimethylsilyl i s ~ c y a n a t e . ~In~the , ~ ~present work it was found that B-values from ref 6 in the known N , = 0 states (Nv = 2v 111) were already predicted rather well with the TZV2P/ QCISD quartic ab initio potential, given above (with u = 5.9). Also, the potential could be fitted with the assumption of a symmetric top structure (with o = 0.06). Using this fit, the rotational constants in the N , > 0 states were predicted and reassigned, and the potential was refitted with the new assignment. The resulting best-fitting quartic potential is again similar, V(g)= 1 9 0 0 . 7 ~ (in~ cm-I, g in radians). The standard deviation was u = 2.7, even better than with the original assignment (see Table 2). From the fitted bending potentials, vibrational energy levels were also calculated and are given in Table 3. It can be seen immediately that the vibrational spacing is strongly dependent on the level of theory and decreases on improving the ab initio calculations. In fact, using the ab initio potential obtained at the TZV2P/QCISD level of theory, most transitions observed between 90 and 140 cm-' in the experimental low-resolution far-infrared8 and Raman9 spectra can be qualitatively reproduced. From this data, the first (unobserved) band (2O-Oo) in the Raman spectrum is at 81.3 cm-l, whereas the second one
+
TABLE 3: Lowest Energy Vibrational Levels (in cm-') for SiHJNCO MP2 QCISD QCISD V' 6-31G** DZP TZV2P 00 1' 20 22 3' 33 40 42 44
ZE" a
0.0
0.0
0.0
55.5 110.2 119.1 172.0 190.4 23 1.4 240.8 269.1
52.9 101.8 112.3 157.5 178.4 209.6 220.0 25 1.O
37.4 81.3 85.9 131.7 144.2 183.0 189.9 211.4
49.8
49.4
26.4
Zero-point energy.
(3I-ll) is at 94.3 cm-' (to be compared with the first observed experimental band at 99.9 cm-' 9). This assignment is in agreement with that performed for the Raman spectrum using the bending potential from microwave spectroscopy.6 This demonstrates again that an anhannonic potential with a single minimum is a possible altemative to the one with a double minimum in the interpretation of experimental spectra. Despite the fact that the ab initio potential qualitatively reproduces parts of the low-resolution far-infrared8and Raman9 spectra, no fitting was performed for two reasons. Firstly, both experimental spectra have poor signal to noise ratios. More importantly, however, both spectra contain strong contributions from contaminants. In the Raman spectrum? the existence of "strong lines of an unidentified impurity" has been noted. In the same work it was also noted that the presence of the decomposition products (H3Si)zO and HNCO was identified from the infrared spectrum. Indeed, disyloxane has a broad unstructured feature at 68 cm-' 27 and a weaker structured band at -120 cm-',28 just as the Raman spectrum of SiH3NCO. In the same way it seems likely that the band at -70 cm-' in the far-infrared spectrum, which was attributed to the Av = 1 transition of SiH3NCO but was not assigned to individual vibrational transitions, is also largely due to the strong absorption of the decomposition product, (H3Si)20.27 The other band should similarly contain some weak contribution from this molecule. Thus, it is difficult to establish which peaks genuinely belong to the SiH3NCO molecule. Conclusions
Our previous ab inito structure of SiH3NCO' and the YIO bending potential in particular have been reinvestigated at a higher level of theory (TZV2P/QCISD). The present calculations reaffirm a symmetric top structure. On the basis of the ab initio potential, the rotational constants from microwave spectroscopy could be fitted using a symmetric top model. Moreover, an altemative interpretation for the microwave spectra could also be provided. The prediction of a symmetric top geometry is in agreement with early microwave and infrared but the proposed bending potential strongly differs from these in that it is strongly anhannonic. The anhannonicity of the potential, however, is in line with the conclusions of later investigations of the microwave ~ p e c t r u m ~and - ~ electron diffractionlo of SiH3NCO. Improvement in the level of theory of the ab initio calculations does not improve the quality of the prediction of rotational constants, which indicates that the key to the discrepancy lies elsewhere. Nonetheless, further improvements of the ab initio calculation may lead to a small barrier, although its size is not expected to reach the -30 cm-I originally put forward by microwave spectroscopy. On the other hand, including relaxation in the interpretation of the microwave
Equilibrium Structure of SiHsNCO spectra is important: on allowing for this effect the potential barrier is substantially shallower. Although a choice among the three assignments could not be made, it could be demonstrated that the asymmetric top structure with a double-minimum bending potential is not a unique interpretation. From the previous arguments it follows that the “true” structure of silyl isocyanate is somewhere between the previous microwave structure and the symmetric top form. This result also demonstrates that the determination of the equilibrium structure of pseudolinear molecules may still hold many surprises. Acknowledgment. Dr. Jan Makarewicz (A. Mickiewicz University, Posnan, Poland) is thanked for providing his vibrational energy level program and for useful discussions. Thomas Speck (University of Basel) is acknowledged for his contributions at the initial stages of this work. Financial assistance from the Swiss National Science Foundation and the Hungarian Scientific and Research Fund (OTKA TO07424 and T014339)is also acknowledged. References and Notes (1) Fehtr, M.; Pasinszki, T.; Veszprtmi, T. J. Phys. Chem. 1993, 93, 1538. (2) Gerry, M. C. L.; Thompson, J. C.; Sudgen, T. M. Narure 1966, 211, 846. (3) Ebsworth, E. A. V.; May, M. J. J. Chem. Soc. 1962, 4844. (4) Duckett, J. A,; Robiette, A. G.; Mills, I. M. J. Mol. Spectrosc. 1976, 62, 34. ( 5 ) Duckett, J. A.; Robiette, A. G.; Gerry, M. C. L. J. Mol. Spectrosc. 1981, 90, 374. (6) Kreglewski, M.; Jensen, P. J. Mol. Spectrosc. 1984, 103, 312. (7) Kreglewski, M. J. Mol. Spectrosc. 1984, 105, 8. (8) Cradock, S.; Skea, D. C. J. J. Chem. Soc., Faraday Trans. 2 1980, 76, 860. (9) Durig, J. R.; Kalasinsky, K. S.; Kalasinsky, V. F. J. Chem. Phys. 1978, 69, 918.
J. Phys. Chem., Vol. 99, No, 21, 1995 8607 (10) Glidewell, C.; Robiette, A. G.; Sheldrick, M. Chem. Phys. Lett. 1972, 16, 526.
(11) Amos. R. D.; Rice, J. E. CADPAC: The Cambridge Analytical Package, issue 5.1; Cambridge, 1992. (12) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Gill, P. M. W.; Johnson, B. G.; Wong, M. W.; Foresman, J. B.; Robb, M. A,; Head-Gordon, M.; Replogle, E. S.; Gomperts, R.; Andres, J. L.; Raghavachari, K.; Binkley, J. S.; Gonzalez, C.; Martin, R. L.; Fox, D. J.; Defrees, D. J.; Baker, J.; Stewart, J. J. P.; Pople, J. A. Gaussian 92/DFT, Revision G.2; Gaussian, Inc.: Pittsburgh, PA, 1993. (13) Dunning, T. H.; Hay, P. J. Modern Theoretical Chemistry; Plenum: New York, 1976; p 1. (14) Krishnan, R.; Binkley, J. S.; Seeger, R.; Pople, J. A. J. Chem. Phys. 1980, 72, 650. (15) McLean, A. D.; Chandler, G. S. J. Chem. Phys. 1980, 72, 5639. (16) Frisch, M. J.; Pople, J. A,; Binkley, J. S. J. Chem. Phys. 1984, 80, 3265. (17) Makarewicz, J.; Mickiewicz, A. University, Posnan, Poland, 1993, private communication. (18) Wierzbicki, A,; Koput, J.; Kreglewski, M. J. Mol. Spectrosc. 1983, 99, 102. (19) He, Z.; Cremer, D. Int. J. Quantum Chem. Symp. 1991, 25, 43. (20) He, Z.; Cremer, D. Theor. Chim. Acta 1993, 85, 305. (21) Bohme, M.; Frenking, G. Chem. Phys. Lett. 1994, 224, 195. (22) Bunker, P. R.; Landsberg, B. M.; Winnewisser, B. P. J. Mol. Spectrosc. 1979, 74, 9. (23) Famell, L.; Nobes, R. H.; Radom, L. J. Mol. Spectrosc. 1982, 93, 271. (24) Szalay, V.; BBnhegyi, G.; Fogarasi, G. J. Mol. Spectrosc. 1987, 126, 1. (25) Kroto, H. W. Molecular Rotation Spectra; Wiley: London, 1975. (26) Careless, A. J.; Green, M. C.; Kroto, H. W. Chem. Phys. Lett. 1972, 16, 414. (27) Robinson, D. W.; Lafferty, W. J.; Aronson, J. R.; Durig, J. R.; Lord, R. C. J. Chem. Phys. 1961, 35, 2245. (28) Durig, J. R.; Flanagan, M. J.; Kalasinsky, V. F. J. Chem. Phys. 1977, 66, 2775.
JP943353R
