Interstellar Silicon–Nitrogen Chemistry. III. the Spectral Signatures of The

Home , Silicon

Interstellar silicon–nitrogen chemistry. III. The spectral signatures ؉ of the H2SiN molecular ion O. Parisel,a) M. Hanus, and Y. Ellinger Laboratoire de Radioastronomie Millime´trique, Equipe d’Astrochimie Quantique, E.N.S. et Observatoire de Paris, 24, rue Lhomond, F. 75231 Paris Cedex 05, France ͑Received 23 May 1995; accepted 16 October 1995͒ The recent detection of SiN in the outer envelope of the IRCϩ10216 carbon star has renewed the interest for the gas phase interstellar silicon chemistry. In this contribution, we present a theoretical ϩ study of the H2SiN molecular ion, the silicon hydrogenated counterpart of the previously studied ϩ ϩ SiNH2 . On many points, the differences relative to the SiNH2 isomer have been found to be dramatic. As an example, the dipole moment is computed to be 3.8 D while being only 0.5 D in ϩ SiNH2 . The radio, infrared and electronic signatures have been evaluated at a quantitative level. The rotational constants and vibrational frequencies have been determined using Mo¨ller–Plesset MPn ͑nϭ2,3,4͒, coupled cluster ͑CCSDT͒ and complete active space self-consistent field ϩ ͑CASSCF͒ methods for H2SiN and some of its isotopomers. These quantities have been corrected using a scaling procedure derived from previous studies on the HNSi, HSiN, HSiNH2 ,H2SiNH, and ϩ SiNH2 species in order to provide quantitative results. The failure of single-reference perturbation theories to predict a relevant infrared spectrum is discussed. Intense bands around 550, 950, and 2300 cmϪ1 are predicted. The electronic spectrum has been obtained using a coupled multiconfiguration SCF–perturbation treatment ͑MC/P͒: It is characterized by a large number of excited states, none of them having a strong transition moment. The lowest excited state is predicted to lie 0.54 eV above the ground state, but the first allowed transition having a nonnegligible oscillator strength has to be searched at 6.44 eV. © 1996 American Institute of Physics. ͓S0021-9606͑96͒00404-3͔

ϩ I. INTRODUCTION with those of HSiNH2 and H2SINH but, contrary to SiNH2 whose elusive existence has been confirmed in some mass- Due to the recent detection of the SiN free radical in spectroscopy experiments,12 no report on the production of space,1,2 a renewed attention has been given to the gas phase H SiNϩ has been published. chemistry of silicon–nitrogen compounds.3–7 By analogy 2 In this connection, we have initiated an accurate study of with the carbon chemistry,8 HSiN and HNSi are supposed to the ͓H,H,Si,N͔ϩ potential energy surface to get a better un- be formed by a dissociative recombination reaction on a ϩ derstanding of the chemistry and spectroscopy of the related ͓Si,N,H,H͔ system which could be produced either by the 30 molecules. In a previous study ͑Part II in this series͒, it was action of the Siϩ cation onto NH or by the combination of 3 shown that the 1A ground state of SiNHϩ is the absolute SiH with Nϩ.2,3,7 The route via SiH seems, however, to be 1 2 3 3 minimum of this potential surface, confirming earlier rather unrealistic in interstellar conditions regarding the pos- calculations:12,19,29 Its spectral signatures ͑radio, ir, and elec- tulated abundance for this species, although it is a possible tronic͒ were determined in order to contribute to its identifi- mechanism of formation in recent laboratory experiments cation either in the interstellar medium or in laboratory ex- that involve electric discharges in a silane/dinitrogen periments. In this paper, we present a theoretical plasma.9–12 determination of the rotational, vibrational, and electronic Following the interstellar chemistry of the carbon– spectroscopic signatures of the H SiNϩ molecule which nitrogen system which led to the observation of the CN, 2 might be anticipated as a possible key intermediate on the HCN, and HNC species, a number of theoretical, experimen- way to HSiN. At each step of the computation, comparison tal and observational studies have been devoted to the linear ϩ will be done with the SiNH2 isomer. HNSi and HSiN and to related structures ͑see Refs. 9–30, 31 32 33 The GAUSSIAN92, HONDO8.5, ALCHEMY II codes and and references therein . Although chemical models assume ͒ the MC/P modules34 were used. these two species to be abundant,3,7 none of them has yet been detected in space to our knowledge, in spite of the ؉ support of recent laboratory experiments which provide ac- II. THE ELECTRONIC STRUCTURE OF H2SiN curate rotational constants for the HNSi isomer ͑while failing ϩ The structure of H2SiN has been determined in C2v to produce HSiN using the same experimental conditions͒. symmetry. Levels of theory cover from Hartree–Fock wave The infrared spectra of both species HSiN and HNSi trapped 27,28 functions in a restricted formalism ͑ROHF͒ to correlated in a solid matrix have been obtained recently together wave functions obtained by Mo¨ller–Plesset perturbation theory ͑MP2, MP3, and MP4͒, by the coupled-cluster meth- a͒Author to whom correspondence should be addressed. odology including triple excitations ͑Refs. 35 and 36 and

J. Chem. Phys. 104 (5), 1 February 1996 0021-9606/96/104(5)/1979/10/$10.00 © 1996 American Institute of Physics 1979 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 129.194.51.204 On: Fri, 13 Dec 2013 11:10:31 1980 Parisel, Hanus, and Ellinger: Interstellar silicon–nitrogen chemistry. III

ϩ 3 TABLE I. Optimized structures of H2SiN (X A2).

Method SiN ͑Å͒ SiH ͑Å͒ NSiH ͑°͒ Energy ͑a.u.͒⌬E͑kcal/mol͒b

ROHF/3-21G 1.856 1.463 118.46 Ϫ342.313 583 95.6 ROHF/6-31G* 1.763 1.454 118.26 Ϫ344.132 666 81.1 ROHF/6-311ϩϩG** 1.757 1.457 118.17 Ϫ344.168 503 85.3 MP2/6-31G* 1.781 1.467 118.16 Ϫ344.308 446 120.8 MP2/6-311ϩϩG** 1.772 1.459 118.89 Ϫ344.497 320 123.7 MP3/6-311ϩϩG** 1.776 1.460 118.30 Ϫ344.524 288 116.2 MP4/6-311ϩϩG** 1.782 1.462 118.37 Ϫ344.534 903 119.0 CCSD͑T͒/6-311ϩϩG** 1.785 1.464 118.37 Ϫ344.538 411 117.5 CASSCF/6-311ϩϩG** 1.794 1.485 118.60 Ϫ344.242 081 123.6 FOCI/6-311ϩϩG** a 1.794 1.485 118.60 Ϫ344.317 341 158.2 SOCI/6-311ϩϩG** a 1.794 1.485 118.60 Ϫ344.385 545 115.2

aCASSCF geometry. b ϩ 1 Energy difference relative to SiNH2 (X A1).

references therein͒ and variational calculations in the com- part of the electronic correlation as will be emphasized in plete active space self-consistent field ͑CASSCF͒ formalism Sec. V: Using ␲ and ␲* orbitals will simplify the description ͑Ref. 37 and references therein͒. In correlated calculations, of the variational spaces to be used. all electrons were active except when specifically indicated. A bond orbital analysis shows that the ␴SiN orbital is In order to allow for a large flexibility in the one-particle mostly built on the 2sN and 3sSi atomic orbitals with no space, a triple-zeta basis set extended by polarization and contribution from the 2pzN or 3pzSi functions. The spN lone 38,39 diffuse functions on each atom was used, which is pair involves an adapted combination of both the 2sN and 40 known as 6-311ϩϩG** for the hydrogen and nitrogen at- 2pzN orbitals. As a consequence of the interaction between oms, while the expanded basis set by McLean and the ␴SiN and spN orbitals, silicon contributions are present in Chandler41 was used for the silicon atom. Smaller 3-21G42 the description of the lone pair for orthogonality reasons. The 43 and 6-31G* basis sets were also used for preliminary in- very weak singly occupied ␲SiN orbital is in fact the qua- vestigations. When not specified, the Si and N labeling refers sipure 2p␲ orbital of the nitrogen atom which acts as a very furthermore to the 28Si and 14N isotopes, respectively. The weak ␲–donor.14,20 It is almost degenerated ͑within 7 kcal/ optimized geometries are collected in Table I together with mol at the ROHF/6-311ϩϩG** level͒ with the in-plane re- their corresponding absolute energies. maining 2pN orbital that presents small, but non-negligible, At the ROHF level, the electronic ground state configu- contributions from the H atoms. As a consequence of the ϩ 3 ration of H2SiN is A2 , corresponding to the following or- weakness of the ␲SiN bond, the positive charge appears to be bital occupancy: essentially located at the silicon atom whatever the type of calculation performed. It was, however, checked that this ͑1a ͒2͑2a ͒2͑3a ͒2͑1b ͒2͑4a ͒2͑5a ͒2͑6a ͒2͑2b ͒2 1 1 1 2 1 1 1 2 bond remains strong enough to ensure that the ground state 1 2 1 ϫ͑2b1͒ ͑7a1͒ ͑3b2͒ . Although the ROHF calculations show a strong coupling between the orbitals describing the lone pair located on the terminal nitrogen atom and the ␴SiN bond, the electronic structure of the ground state can be described in terms of Lewis-localized orbitals as ͓Fig. 1͑a͔͒ 2 2 2 6 ϩ 2 Ϫ 2 2 ͑1sSi͒ ͑1sN͒ ͑2sSi͒ ͑2pSi͒ ͑␴SiH͒ ͑␴SiH͒ ͑␴SiN͒ 1 2 1 ϫ͑␲SiN͒ ͑spN͒ ͑2pN͒ , ͑1͒ ϩ Ϫ where ͑␴SiH͒ and ͑␴SiH͒ stand for the symmetry-adapted combinations of the SiH bonds, respectively, of a1 and b2 symmetries. The spN orbital is the ␴ lone pair orbital localized at the terminal nitrogen atom, and the 2pN orbital rep- resents the remaining in-plane nitrogen atomic orbital. Such a representation that uses a ␲ bond between Si and N should only be seen as a possible mesomeric formula: As will be discussed below, the ␲ bond appears to be essentially located on the nitrogen atom. However, the use of such a representation has the advantage of suggesting a planar molecule and ϩ FIG. 1. Lewis structures of H2SiN . ͑Dots stand for in-plane ␴ electrons will make it possible to design specific treatments to recover while crosses stand for out-of-plane ␲ electrons.͒

J. Chem. Phys., Vol. 104, No. 5, 1 February 1996 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 129.194.51.204 On: Fri, 13 Dec 2013 11:10:31 Parisel, Hanus, and Ellinger: Interstellar silicon–nitrogen chemistry. III 1981

does not pyramidalize and retains an sp2 hybridization at the scription are represented by the following electronic distri- silicon atom. butions with weights of 5.2% and 3.2%, respectively ͑CSF2 Why the open-shell triplet state is lower in energy than and CSF3͒ the closed-shell singlet state, contrary to the SiNHϩ isomer, 2 ...͑␴ϩ ͒2͑␴Ϫ ͒1͑␴ ͒2͑␲ ͒1͑sp ͒2͑2p ͒2͑␲* ͒0, ͑2͒ can be nicely explained when considering the dramatic SiH SiH SiN SiN N N SiN weakness of the ␲ bond. As a consequence, and due to the ϩ 2 Ϫ 2 2 0 2 1 1 SiN ...͑␴SiH͒ ͑␴SiH͒ ͑␴SiN͒ ͑␲SiN͒ ͑spN͒ ͑2pN͒ ͑␲SiN* ͒ . ͑3͒ high electronegativity of the nitrogen atom relative to silicon, The first CSF accounts for a strong coupling between the there is an electronic transfer from silicon towards nitrogen Ϫ ␴SiH and the in-plane 2pN terminal orbital. Such a coupling which results in having two unpaired electrons on the termi- ϩ nal nitrogen atom. According to Hund’s rule, these electrons was not observed for SiNH2 . As a consequence of both the ϩ weight and the nature of this CSF, it may be expected that all couple to give a triplet. In the case of SiNH2 , a similar electronic transfer occurs but the two electrons localized on properties involving a variation of the SiH distance will be nitrogen can only couple in singlet since only one empty poorly described by single-reference post-SCF treatments orbital is available on this atom, the others being involved in based on the HF determinant. Among these properties are of the NH bonds. special interest the SiH bond length, the dipole moment and On a more quantitative point of view, a detailed analysis its derivatives, the low-lying SiH2 wagging and out-of-plane shows that, at the MP2/6-31G* level of theory for example, deformation frequencies as well as their intensities. Mulliken charges are ϩ0.99 ͑Si͒, Ϫ0.11 ͑N͒, and ϩ0.06 ͑H͒; The second important CSF describes the backdonation summing the hydrogen charges on silicon gives ϩ1.11 ͑Si͒, from the nitrogen atom towards the silicon atom and contrib- ϩ utes to slightly reinforces the weak ␲SiN bond. It is, however, which is slightly larger than the same quantity in SiNH2 ͑ϩ0.96͒, and Ϫ0.11 ͑N͒. Such distributions are confirmed by worth noting that these three CSFs account only for 76% of a natural bond orbital analysis ͑NBO͒͑Refs. 44, 45, and the complete CASSCF wave function which means that references therein͒ that gives natural charges as: ϩ1.57 ͑Si͒, about 25% of this wave function arise from CSFs having Ϫ0.30 ͑N͒ and Ϫ0.14 ͑H͒. Such a NBO analysis confirms each very small contributions. The last column of Table I reports the electronic energy the previous orbital analysis that localized the positive ϩ differences between the ground state of SiNH2 and that of charge on the silicon and reveals that about 90% of the elec- ϩ H2SiN . For the purpose of comparison between the two tronic density of the ␲SiN bond resides on nitrogen. The hy- 2.2 isomers, the level of the calculations was increased up to a bridization at the silicon atom is roughly sp which is co- 46 herent with a planar molecule and a very weak ␲ bond. complete FOCI ͑first-order configuration interaction ͒ which involved 599 768 CSFs, and to a truncated SOCI Moreover, the weight of the silicon atomic orbitals in the 46 ␴ bond is as low as 26%, among which 67% come from ͑second-order configuration interaction ͒ that involved only, SiN as references for the double excitations to the external space, pzSi atomic orbitals. On the contrary, 86% of the electronic density of the sp axial lone pair on the terminal nitrogen the CSFs generated by single excitations from the leading N determinant: 1 168 408 CSFs were thus considered in the atom come from sN atomic orbitals. In order to account for the nondynamic correlation, final variational space. If we except the FOCI treatment, which is expected to be of essential importance because of which is known to be inadapted for the evaluation of energy differences involving states of different multiplicities, all two valence holes ͑one in the terminal in-plane 2pN orbital and the other one on the 2p␲ orbital on the nitrogen atom͒, correlated treatments converge to a value of about 115 kcal/ a multiconfigurational description was generated within the mol, which is 40 kcal/mol higher than at the ROHF level. CASSCF framework. The wave function was expanded in This large difference can be entirely explained by nondy- the configuration space spanned by all possible distributions namic correlation effects: The correlation recovered when going from ROHF/6-311ϩϩG** to CASSCF/6-311ϩϩG** of the 10 valence electrons among the 10 corresponding ϩ is 85 kcal/mol for SiNH2 while being only 46 kcal/mol for bonding, antibonding and nonbonding orbitals ϩ H2SiN . It must, however, be kept in mind that ROHF ap- ϩ Ϫ ϩ Ϫ 10 ͕␴SiH␴SiH␴SiN␲SiNspN2pN␴SiH*␴SiH*␴SiN* ␲SiN* ͖ proaches intrinsically include some part of correlation effects as a consequence of nonvanishing exchange interaction provided the spin and spatial symmetries of the generated terms arising from the unpaired electrons; the definition of configuration state functions ͑CSF͒ were 3A . This led to 2 the nondynamic correlation energy as the difference between 7400 CSFs in the calculation. the ROHF and CASSCF energies is somewhat biaised in this The analysis of the CASSCF results shows that the case. The valence–virtuals dynamic correlation energy, weight of the leading CSF ͑similar to the ROHF determi- which we defined as the difference between the CASSCF nant͒ is about 67% in the multireference wave function while and SOCI energies in the present treatment, is 90 kcal/mol, this weight is more than 90% for the CASSCF wave function ϩ ϩ slightly larger than in SiNH2 ͑82 kcal/mol͒. describing SiNH2 . Such a small weight of the reference determinant makes it a priori questionable to use MPn meth- III. THE RADIO SIGNATURE OF H SiN؉ ods without careful attention to the properties being investi- 2 gated. It should also be mentioned that no significant spin The rotational spectrum is an unambiguous fingerprint of contamination was observed in our MPn treatments. The a molecule; the corresponding transition frequencies are di- next important configurations involved in the CASSCF de- rectly linked to the geometry through rotational constants

J. Chem. Phys., Vol. 104, No. 5, 1 February 1996 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 129.194.51.204 On: Fri, 13 Dec 2013 11:10:31 1982 Parisel, Hanus, and Ellinger: Interstellar silicon–nitrogen chemistry. III

ϩ 3 ϩ 3 TABLE II. Unscaled rotational constants for H2SiN (X A2). TABLE III. Scaled rotational constants for H2SiN (X A2).

Method Ae ͑GHz͒ Be ͑GHz͒ Ce ͑GHz͒ Method A0 ͑GHz͒ B0 ͑GHz͒ C0 ͑GHz͒ MP2/6-31G* 149.878 15.422 13.983 MP2/6-31G* 153.322 15.776 14.304 MP2/6-311ϩϩG** 151.620 15.543 14.098 MP2/6-311ϩϩG** 154.652 15.854 14.380 MP3/6-311ϩϩG** 151.060 15.502 14.066 MP3/6-311ϩϩG** 150.154 15.409 13.982 MP4/6-311ϩϩG** 151.449 15.393 13.973 MP4/6-311ϩϩG** 155.993 15.855 14.392 CCSDT͑T͒/6-311ϩϩG** 151.087 15.335 13.922 CCSDT͑T͒/6-311ϩϩG** 153.958 15.627 14.187 CASSCF/6-311ϩϩG** 147.452 15.162 13.749 CASSCF/6-311ϩϩG** 149.811 15.405 13.969

and intensities scale with the square of the electric dipole. level of calculation is the one that is the closest to unity, the Inspection of Table I reveals that the computed geometry of scaled MP3 values should be of higher quality, which gives the molecule is only marginally sensitive to the level of A ϭ150.154 GHz, B ϭ15.409 GHz, theory or to the basis set used, apart from the poor ROHF/ 0 0

3-21G level of calculation. The SiN bond length remains C0ϭ13.982 GHz. close to 1.79 Å, the SiH bond close to 1.47 Å and the Even after the scaling procedure, signiﬁcant differences ЄSiNH bond angle close to 118.5°. The length of the SiN remain between the various methods. Due to the high sensi- bond in the ground state of H SiNϩ is larger than observed 2 tivity of the rotational constants ͑especially B and C͒ to the for SiN ͑X 2⌺: rϭ1.572 Å47͒ and HNSi ͑1⌺ϩ: rϭ1.551 Å24͒ SiH bond length, one may wonder whether the CASSCF or computed for related species:48 HSiN ͑X 1⌺ϩ: rϭ1.59 Å͒, geometry would not be of higher quality to reproduce experi- HSiNHϩ ͑X 1AЈ: rϭ1.52 Å, a 3AЈ: rϭ1.65 Å͒, HSiNH 2 mental values since the MP3 approach does not properly ͑X 1AЈ: rϭ1.72 Å͒,HSiNH ͑X 1AЈ: rϭ1.62 Å͒. It is also 2 describe the interaction between the ␴Ϫ and the in-plane larger than the SiN bond length determined in the ground SiH 2p terminal orbitals. It is unfortunate that experimental ro- state of SiNHϩ ͑X 1A : rϭ1.66 Å͒ as a consequence of a N 2 1 tational constants for the HSiN species are not available. much weaker ␲ bond. As anticipated in the previous sec- SiN Turning to theoretical approaches, it has been found that a tion, the SiH bond is lengthened by 0.02 Å when increasing multireference description is necessary to properly describe the level of calculation from MPn to CASSCF. It is a con- the infrared spectrum of this species and that a multirefer- sequence of the weight of CSF2 in the CASSCF wave func- ence methodology is also essential to get a consistent infra- tion that weakens the SiH bond. The CASSCF value of 1.485 red spectrum of H SiNϩ as will be seen in the next section. Å is consistent with theoretical12,15,48 and experimental 2 As a consequence, we expect the scaled CASSCF rotational data49–51 available for the SiH bond lengths in a large variety constants to be closer to the experimental values than those of compounds. It is possible that such a long bond is respon- obtained with any single reference approach and in the end sible for the fact that multireference treatments that include will recommend the use of CSF2 are necessary to properly account for the spectral sig-

nature of H–Si compounds: this fact was pointed out in our A0ϭ149.811 GHz, previous studies on HSiN and will be emphasized here in the next sections. Surprisingly, the SiN length also increases by B0ϭ15.405 GHz, 0.01 Å at the CASSCF level of calculation contrary to what C0ϭ13.969 GHz. would have been expected from the contribution of CSF3. However, such an increase, although smaller, was also ob- Such constants are gathered in Table IV for the most abun- ϩ dant H28Si14Nϩ isotopomer together with those expected for served in the SiNH2 isomer and in HSiNH2 as well. 2 The calculated geometries lead to the rotational con- other isotopomers. Analysis of the data in Table IV allows to stants presented in Table II. Following the remark by De- get the shifts induced by isotopic substitution on these con- 52 28 14 ϩ Frees et al. that MP3 optimized geometries give rotational stants relative to the H2 Si N . Even if the absolute values constants that match more closely the experimental A0 , B0 , and C0 than those obtained by computing MP2 or MP4 structures, the rotational constants of H SiNϩ can be ﬁrst approxi- 2 TABLE IV. Recommended rotational A0 , B0 , and C0 constants ͑scaled mated as CASSCF/6-311ϩϩG** Ae , Be , and Ce rotational constants͒ for some iso- ϩ 3 topomers of H2SiN (X A2). A0ϭ151.060 GHz, B0ϭ15.502 GHz, Isotopomer A0 ͑GHz͒ B0 ͑GHz͒ C0 ͑GHz͒

C0ϭ14.066 GHz. 28 14 ϩ H2 Si N 149.811 15.405 13.969 A further improvement consists in the correction of the pure HD28Si14Nϩ 102.432 14.690 12.847 28 14 ϩ D2 Si N 74.982 14.131 11.890 Ae , Be , and Ce constants with the scaling factors obtained, 29 14 ϩ 48 H2 Si N 149.811 15.275 13.862 at each level of calculation, in our previous study on HNSi: 30 14 ϩ H2 Si N 149.811 15.153 13.761 This leads to the A , B , and C constants reported in Table 28 15 ϩ 0 0 0 H2 Si N 149.811 14.726 13.408 III. Since the scaling factor obtained for HNSi at the MP3

a Ϫ1 ϩ 3 TABLE V. Scaled frequencies ͑wave numbers in cm ͒ and relative intensities normalized to the total intensity ͑KM/mole͒ for H2SiN (X A2).

ROHF MP2 MP2 MP3 MP4 CCSD͑T͒ CASSCF 6-31G* 6-31G* 6-311ϩϩG** 6-311ϩϩG** 6-311ϩϩG** 6-311ϩϩG** 6-311ϩϩG**

Total intensity 285 252 415 142 Vibrations

b2 SiH2 wagging 543 20.9% 524 21.0% 520 20.9% 506 500 496 542 19.3% b1 SiH2 out of plane 650 37.0% 621 36.0% 627 37.6% 612 604 580 411 9.1% b a1 SiN stretch 731 0.8% 750 0.3% 754 0.2% 725 739 718 719 3.0% c a1 SiH2 bending 920 39.4% 933 41.5% 923 41.0% 895 913 903 955 38.9% a1 SiH2 stretch ͑s͒ 2255 0.7% 2307 0.3% 2287 0.1% 2229 2280 2254 2315 11.4% a b2 SiH2 stretch 2306 1.1% 2369 0.3% 2346 0.3% 2287 2346 2316 2330 18.4%

aScaling factors are ͑see the text for details͒: ROHF 0.806, MP2 0.920, MP3 0.882, MP4 0.949, CCSD͑T͒ 0.933, CASSCF 0.941. b Strongly coupled with the SiH2 bending mode ͑80%/20%͒. cCoupled with the SiN stretching mode ͑99%/1%͒.

of the constants suffer from a few percent uncertainty, it is and for which experimental infrared spectra are available: As expected that these shifts, that could be observed experimen- a consequence, we use here scaling factors deduced from tally, are of higher accuracy. comparisons with experiments devoted to the parent HNSi, The determination of the dipole moment of charged spe- HSiN species and to the close HSiNH2 and HNSiH2 cies is always delicate since that observable depends on the molecules.48 This leads to use the following values ͑that refer choice of the origin retained for its evaluation. Assuming that to scales applied on frequencies͒: RHF/6-31G*: 0.898, this quantity has to be evaluated at the center of mass of the MP2/6-31G* or MP2/6-311ϩϩG**: 0.959, molecule in order to get a relevant value for the determina- MP3/6-311ϩϩG** 0.939, MP4/6-311ϩϩG** 0.974, tion of the rotational intensities,53 all the calculations re- CCSDT/6-311ϩϩG** 0.966, and CASSCF/6-311ϩG**: ported in this paper have been performed accordingly, using 0.970. These values have been discussed elsewhere: The the convention that a positive value indicates that the dipole MP4 and CASSCF approaches seem to be both of better is directed along the NSi direction. The inherent difﬁculties quality than the CCSDT method since their corresponding in the determination of the dipole moment for such species scaling factor is slightly larger. According to our experience ϩ have been discussed and developed in the case of SiNH2 .It for other compounds of this kind, we expect the deviation of was concluded that the CASSCF dipole moment computed our scaled predicted infrared frequencies from the experi- ϩ with a 6-311ϩϩg** basis set was relevant. For H2SiN , this mental values to be smaller than 5% on the whole spectrum. value is 3.40 D. The FOCI dipole value is increased to 4.03 The scaled vibrational frequencies and normalized inten- D. At the SOCI level, the dipole is found to be 3.8 D; we sities are presented in Table V and the corresponding spectra expect this last value to be accurate within 5%. in Figs. 2 and 3 where they have been obtained by ﬁtting each vibrational transition with a Lorentzian havinga4cmϪ1 ؉ IV. THE ir SIGNATURE OF H2SiN FWHM. The vibrational spectrum calculated at the ROHF, MP2 The large amount of data available on ir spectra has now ͑Fig. 2͒, MP3, MP4, and CCSDT levels of theory are con- well established the fact that calculated frequencies should 54 sistent with each other, both in frequencies and intensities. be scaled to reproduce experimental values whatever the The analysis of the normal coordinates shows that the vibra- level of calculation used. This is done here at the same time as the normal coordinate analysis. The appropriate B matrix55 is evaluated and then used to transform the force constant matrix from a Cartesian to an internal coordinate representation which leads to annihilate the contaminations of the molecular motions arising from translation and rota- tion. In order to correct the general overestimation of the diagonal force constants, it is an usual procedure to scale these terms by factors of 0.792 ͑ROHF͒, 0.884 ͑MP2͒, and 0.893 ͑MP4͒. In those cases where couplings between normal modes are negligible, the force constant matrix is diagonal in the internal representation, so that the scaling procedure is rigorously equivalent to the application on the initial frequencies of the usual scaling factors 0.89 ͑ROHF͒, 0.94 ͑MP2͒, and 0.945 ͑MP4͒. These scaling factors however de- pend on both the method of calculation and the basis set. It was found more attractive here to use scaling factors ob- FIG. 2. Simulated infrared spectrum of H SiNϩ (X 3A ) from scaled ϩ 2 2 tained by preliminaries studies on species related to H2SiN MP2/6-311ϩϩG** calculations.

J. Chem. Phys., Vol. 104, No. 5, 1 February 1996 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 129.194.51.204 On: Fri, 13 Dec 2013 11:10:31 1984 Parisel, Hanus, and Ellinger: Interstellar silicon–nitrogen chemistry. III

lane derivatives.56 While having their frequencies almost un- changed, the SiH2 stretching modes have their relative intensities increased which is also observed for the SiN stretching vibration. The SiH2 bending mode remains almost unaf- fected; the wagging mode retains its intensity. It is thus clear that the consideration of properly described nondynamic cor- ϩ relation effects is essential in the case of the H2SiN molecule. The CASSCF infrared spectrum is presented in Fig. 3 and is expected to be, by far, of higher accuracy, than the MP2 presented in Fig. 2. In the CASSCF calculation, the SiH stretching modes are located at 2315 and 2330 cmϪ1 after scaling, which appears to be in between the experimental values of H2SiI2 and H3SiCCH cited in Ref. 12. They also FIG. 3. Simulated infrared spectrum of H SiNϩ (X 3A ) from scaled compare very favorably with the spectra of silane 2 2 56 CASSCF/6-311ϩϩG** calculations. derivatives, for example, and with the experimental spectra Ϫ1 28 of the more exotic H2SiNH ͑2253, 2175 cm ͒ and Ϫ1 21 HSiNH2 ͑1975 cm ͒ as well. The CASSCF level of cal- tion modes are almost pure, except the SiN stretching mode culations was then used to obtain the infrared frequencies that strongly couples to the SiH2 bending. Three strong and the corresponding force constants for selected isoto- bands are easily identiﬁed: Two bands at about 500 and 600 pomers of H SiNϩ as reported in Table VI. Ϫ1 2 cm which, respectively, correspond to the SiH2 wagging The dramatic failures of the MPn approaches to cor- and to the SiH out-of-plane deformation that tends to the ϩ 2 rectly predict a relevant ir spectrum of the H2SiN species is pyramidalization of the molecule. The third strong band at a perfect illustration of the too-often forgotten basic principle Ϫ1 about 900 cm is the SiH2 symmetric in-plane bending of perturbation theories that the zeroth-order wave function mode. The three remaining bands are of very low intensity: has already to be of high enough quality to properly describe Ϫ1 One at about 720 cm for the SiN stretching mode ͑to be the property under investigation. This was already empha- Ϫ1 22 Ϫ1 13 compared with 1151 cm for SiN, 1198 cm for HNSi, sized for the description of excited states57,58 but this non- Ϫ1 21 Ϫ1 28 866 cm in HSiNH2 , 1097 cm for H2SiNH, 1162 academic example provides the interesting evidence that this Ϫ1 28 Ϫ1 ϩ cm for HSiN and 920 cm for SiNH2 ͒, the next ones rule also applies to ground states. Ϫ1 around 2250 and 2320 cm correspond to the SiH2 symmetric and asymmetric stretchings. It is worth noting that such ؉ calculations predict a quasicomplete extinction of the SiH V. THE ELECTRONIC SIGNATURE OF H2SiN stretching modes, which is in complete disagreement with A. Methodology and computational details what is known on silicated compounds,49–51,56 and also in contradiction with the nice experimental studies by Maier In order to provide a quantitative description of the ver- 21,28 ϩ et al. on the more closely related HSiN, H2SiNH, or tical electronic absorption spectrum of the H2SiN molecular HSiNH2 species. ion, we used the MC/P approach developed in our laboratory. The CASSCF spectrum is, however, somewhat different. It consists in a coupled variation-perturbation treatment that Ϫ As a consequence of the interaction between the ␴SiH and the aims to treat the nondynamic correlation effects involved in in-plane 2pN terminal orbitals, the lowest two frequencies the zeroth-order description of a particular state at a MCSCF corresponding to the SiH2 wagging and the out-of-plane de- level, followed by the recovering of the dynamic correlation formation modes that involve modiﬁcation of the SiH dis- through a large scale perturbation treatment on this varia- tances are strongly shifted: The out-of-plane vibration fre- tional wave function. This methodology whose details have quency is dramatically reduced from about 600 to 411 cmϪ1 been presented recently34 has been successfully applied to and has its intensity reduced by a factor 4. Such vibrations the description of the vertical spectra of the ethylene, form- about 400 cmϪ1 have been observed experimentally for si- aldehyde and cyclopropene molecules taken as

TABLE VI. Scaled CASSCFG/6-311ϩϩG** computed vibrational frequencies ͑wave numbers in cmϪ1͒ for ϩ 3 some isotopomers of H2SiN (X A2).

28 14 ϩ 28 14 ϩ 28 14 ϩ 28 15 ϩ 29 14 ϩ 30 14 ϩ Isotopomer H2 Si N HD Si N D2 Si N H2 Si N H2 Si N H2 Si N Vibrations

b2 542 453 414 541 541 540 b1 411 364 310 410 410 409 a1 719 700 604 703 716 713 a1 955 872 814 954 953 951 a1 2315 1688 1653 2316 2315 2314 b2 2330 2302 1692 2330 2328 2326

ϩ ϩ TABLE VII. Conﬁguration spaces for MCSCF1 calculations of H2SiN . TABLE VIII. Conﬁguration space for MCSCF2 calculations of H2SiN .

Symmetry Set 1 ͑frozen͒ Set 2 Set 3 Set 4 Set 5 Symmetry Set 1 ͑frozen͒ Set 2 Set 3 Set 4

a 1s 1s 2s 2p ␴ sp 8 R a 1s 1s 2s 2p ␴ ϩ ␴ sp * 1 Si N Si Si SiN N 1 Si N Si Si SiH SiN N ␴SiHϩ ␴SiN* ϩ ␴SiH ␴SiN* spN* b1 2pSi ␲␲* b 2p ␲␲* 4R b 2p ␴ Ϫ 2p * 1 Si 2 Si SiH N ␴SiHϪ b Ϫ 2p 4 R 2 2pSi ␴SiH N a2 a2 2 R Electronic Electronic distributions distributions 12 10 0 0 162220 12 9 0 1 161300 12 8 2 0 162040 12 8 0 2 162121 12 7 2 1 162031 12 7 0 3 162211 12 6 2 2 162130 12 5 2 3 161230 161221

space to account for the unusual bonding capabilities of the silicon atom which may lead to peculiar bridged structures benchmarks.58,59 As shown in these previous calculations, the when engaged in multiple bonds.61–63 The variational zeroth- use of a Mo¨ller–Plesset partition of the electronic Hamil- order wave function is developed in the MCSCF space built tonian combined with a well-designed variational zeroth- on all the spin and spatial symmetry adapted CSFs arising order space built on averaged MCSCF orbitals provides from the electronic distributions presented in Table VIII. All quantitative results. For example, the root-mean-square de- valence electrons are correlated. These calculations will viation from relevant experimental data for test molecules hereafter be referred to as MCSCF2 and MC/P2. such as formaldehyde and ethylene does not exceed a few tenths of an electron volt. B. Results and discussion All calculations were done at the fully optimized MP4/6-311ϩϩG** geometry of the ground state. Variational The MCSCF1 calculations lead to the unambiguous con- zeroth-order spaces were constructed with the orbitals ob- clusion that no diffuse states have to be expected in the spec- tained by averaging the MCSCF calculations, for each spin trum up to 10 eV above the ground state: It is consistent with and spatial symmetry, on those states lying up to 10 eV the electronic spectrum of a positively charged molecule above the ground state. The perturbation step of the MC/P whose core may suffer a Coulombian explosion upon such a method was done with all valence electrons active and with Rydbergization. However, the MCSCF1 space, which could ϩ no virtual orbitals frozen. Singlet, triplet, and quintet states be used from preliminary investigations on the SiNH2 spec- ϩ were studied. trum fails to describe H2SiN , even at a qualitative level. ϩ Two series of calculations were performed using the If NH bonds were not involved in excitations of SiNH2 6-311ϩϩG** basis expanded by a set of diffuse s, p, and d and could be frozen, it is not possible here to freeze the SiH functions on both heavy atoms. The exponents for these dif- electrons at the zeroth-order level of calculations, as illus- fuse orbitals have been taken as 0.017, 0.014, and 0.015 for trated in Tables IX, X, and XI that collect all the excited the 4s,4p, and 3d functions on the silicon atom. The values states up to 10 eV from the ground state. As a consequence, of 0.028, 0.025, and 0.015 have been used for the corre- the forthcoming discussion will focus on MCSCF2 and sponding 3s,3p, and 3d orbitals on the nitrogen atom. MC/P2 calculations. Before going to the results in more de- The first series of calculation aimed at checking for the tail, it is worth noting that the ␴SiN bonding orbital and the ϩ existence of diffuse states. It involved the MCSCF space ␴SiH orbital remain doubly occupied, and that at least one spanned by all CSFs arising from the electronic distributions electron remains in the spN orbital. It was also observed that presented in Table VII. The diffuse orbitals ͑denoted as ‘‘R’’͒ putting the ␴SiN and spN electrons in different sets led to are explicitly taken into account in the variational treatment numerical troubles in the convergence of the MCSCF pro- by allowing monoexcitations from the valence occupied or- cess due to the strong coupling between these orbitals. 58,59 bitals to the diffuse orbital set. In addition, a spN corre- The first important remark to emphasize concerns the lating function ͑spN͒* has been explicitly included in the large number of excited states obtained within 10 eV from 58–60 ϩ zeroth-order configuration space to allow for the spatial the ground state of H2SiN : 37 excited states are obtained ϩ relaxation of this important orbital that would be poorly ac- which is three times as much as for the SiNH2 ion within the counted for using a variational space restricted to valence same energy range. Such a variation in the number of excited orbitals. In this calculation, which will be referred to as states is usually observed when comparing a closed-shell MCSCF1, the electrons in the SiH bonds are frozen. system to similar open-shell ones and is coherent with their The second series of calculations focuses on the valence multiplicity. If, however, we omit those excited states that spectrum. The SiH bonds are included in the variational involve the SiH orbitals, 22 states still remain which is larger

J. Chem. Phys., Vol. 104, No. 5, 1 February 1996 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 129.194.51.204 On: Fri, 13 Dec 2013 11:10:31 1986 Parisel, Hanus, and Ellinger: Interstellar silicon–nitrogen chemistry. III

ϩ TABLE IX. Vertical electronic absorption spectrum of H2SiN : Triplet states.

States MCSCF ͑eV͒ MC/P ͑eV͒ f

3 2 2 2 2 1 1 A1 : SiHϩ␴ sp ␲ SiHϪ 2pN 5.56 4.79 0.0 2 2 2 1 1 2 SiHϩ ␴ sp ␲ ␲* SiHϪ 6.22 5.34 0.0 2 2 2 1 1 1 1 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 8.72 8.03 0.0 2 2 2 1 1 1 1 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 9.98 8.79 0.0 3 2 2 2 1 2 1 A2 : SiHϩ ␴ sp ␲ SiHϪ 2pN 0.0 0.0 2 2 2 1 2 1 SiHϩ ␴ sp ␲* SiHϪ 2pN 7.23 6.44 0.073 (z) 2 2 2 2 1 1 SiHϩ ␴ sp ␲ ␲* SiHϪ 9.26 8.12 0.032 (z) 2 2 2 1 2 1 SiHϩ ␴ ␲ ␲* SiHϪ 2pN 9.51 8.74 0.011 (z) 3 2 2 1 1 2 2 B1 : SiHϩ ␴ sp ␲ SiHϪ 2pN 1.81 1.65 0.001 (y) 2 2 1 2 1 0 2 Ϫ3 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 7.49 6.83 Ͻ10 (y) 2 2 1 1 2 2 SiHϩ ␴ sp ␲ SiHϪ 2pN 9.40 8.09 0.011 (y) 2 2 1 2 1 1 1 Ϫ3 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 9.92 8.70 Ͻ10 (y) 3 2 2 1 2 2 1 Ϫ3 B2 : SiHϩ ␴ sp ␲ SiHϪ 2pN 0.80 0.54 Ͻ10 (x) 2 2 1 1 1 2 1 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 6.45 5.59 0.001 (x) 2 2 1 1 1 2 1 Ϫ3 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 7.46 6.92 Ͻ10 (x) 2 2 1 1 1 2 1 Ϫ3 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 7.73 6.64 Ͻ10 (x) 2 2 1 2 1 2 SiHϩ ␴ sp ␲ SiHϪ 2pN 8.90 5.77 0.002 (x)

ϩ than in SiNH2 by a factor of 2. This larger density of states higher than the corresponding deviation between MCSCF originates from numerous excited states described by four and MC/P treatments observed in the cases of the ethylene, ϩ 30,59 open-shell configurations: Each of them generates two sin- formaldehyde and SiNH2 molecules. glet, three triplet, and one quintet states. The coupling of It is known that the oscillator strengths f are usually such configurations to the excitations from the SiH orbitals is little affected by dynamic correlation:64 Consequently, they responsible for the large increase in the number of states. have been determined using the MCSCF transition moments, One may, however, wonder why such states are so low-lying but taking the MC/P corresponding transition energies. As ϩ in energy contrary to SiNH2 . Possible qualitative explana- for the dipole moment, nonvanishing transition moments de- ϩ tions would invoke the larger size of H2SiN ͑SiNϭ1.78 Å, pend on the choice of the origin of coordinates: according to ϩ SiHϭ1.46 Å͒ compared to SiNH2 ͑SiNϭ1.66 Å, NHϭ1.02 Sec. III, it was located at the center of mass. These energies Å͒; the larger spatial extension which is possible in the and oscillator strengths were used to product the pictorial former ion allows a better accommodation of the constraints representation of the vertical electronic absorption spectrum between electronic distributions. In addition, we have a presented in Fig. 4 where the electronic transition have been smaller energy gap between occupied and virtual orbitals in fitted with a Gaussian having a 900 cmϪ1 FWHM. ϩ ϩ H2SiN than in SiNH2 . Analysis of the electronic spectrum. The lowest excited The position of the MCSCF states is corrected when state is obtained by the promotion of a nonbonding spN elec- dynamic correlation effects are taken into consideration at tron to the hole in the ␲SiN orbital ͓Fig. 1͑c͔͒. The resulting 3 the perturbation level. The quadratic deviation from MCSCF 1 B2 state lies at 0.54 eV above the ground state and has the to MC/P transition energies is about 1.1 eV which is slightly following electronic configuration: 2 2 2 6 ϩ 2 Ϫ 2 ͑1sSi͒ ͑1sN͒ ͑2sSi͒ ͑2pSi͒ ͑␴SiH͒ ͑␴SiH͒ ϩ TABLE X. Vertical spectrum for the singlet states of H2SiN . 2 2 1 1 0 ϫ͑␴SiN͒ ͑␲SiN͒ ͑spN͒ ͑2pN͒ ͑␲SiN* ͒ . States MCSCF ͑eV͒ MC/P ͑eV͒ Although the transition is symmetry-allowed, the oscillator 1 2 2 2 2 2 A1 : SiHϩ ␴ sp ␲ SiHϪ 1.08 0.86 strength is so weak that its detection using absorption tech- 2 2 2 2 2 SiHϩ ␴ sp SiHϪ 2pN 2.51 2.55 niques might be difficult. The orbital occupancy of this first 2 2 2 2 1 1 SiHϩ ␴ sp ␲ SiHϪ 2pN 5.62 4.99 2 2 2 2 2 excited state is completely analogous to the first excited state SiHϩ ␴ ␲ SiHϪ 2pN 5.92 5.20 3 ϩ 2 2 2 1 1 1 1 ͑1 B2͒ state of SiNH2 which, at the same level of calcula- SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 9.48 8.34 1 2 2 2 1 2 1 A2 : SiHϩ ␴ sp ␲ SiHϪ 2pN 1.61 1.33 2 2 2 1 2 1 SiHϩ ␴ sp ␲* SiHϪ 2pN 7.64 6.59 2 2 2 1 1 2 SiHϩ ␴ sp ␲ SiHϪ 2pN 9.43 7.98 ϩ 1 2 2 1 1 2 2 TABLE XI. Vertical spectrum for the quintet states of H2SiN . B1 : SiHϩ ␴ sp ␲ SiHϪ 2pN 3.87 3.30 2 2 1 2 1 0 2 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 7.79 7.23 States MCSCF ͑eV͒ MC/P ͑eV͒ 2 2 1 1 2 2 SiHϩ ␴ sp ␲* SiHϪ 2pN 10.19 8.76 2 2 1 2 1 1 1 5 2 2 2 1 1 1 1 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 10.87 9.49 A2 : SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 6.83 7.03 1 2 2 1 2 2 1 5 1 2 1 2 1 2 1 B2 : SiHϩ ␴ sp ␲ SiHϪ 2pN 3.15 2.61 A2 : SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 11.20 9.30 2 2 1 2 1 2 5 2 2 1 1 1 2 1 SiHϩ ␴ sp ␲ SiHϪ 2pN 7.47 6.03 B1 : SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 8.60 7.93 2 2 1 1 1 2 1 5 2 2 1 1 1 2 1 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 8.28 7.03 B2 : SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 4.87 4.20 2 2 1 2 2 1 2 2 1 1 1 1 2 SiHϩ ␴ sp ␲ SiHϪ 2pN 9.12 7.84 SiHϩ ␴ sp ␲ ␲* SiHϪ 2pN 9.84 7.11

The triplet ␲–␲* transition is found to occur at 6.44 eV 3 ͑2 A2͒, with an oscillator strength of 0.073 which is the largest of all the oscillator strengths reported for the transitions studied in this paper: This transition is certainly the first transition to be observed in an absorption spectrum due to the weakness of the lower lying ones. The ␲–␲* transitions are located at 6.04 eV in the case of 1 1 SiNH2ϩ(X A1 –2 A1) with the largest oscillator strength 1 3 and at 4.82 eV for the forbidden X A1 –1 A1 transition. The lowest-lying state involving an excitation from the 1 SiHϪ orbital appears at 4.99 eV ͑3 A1͒, but the first one of that type having a non-zero transition strength lies 3 eV higher at 8.12 eV ͑3 3A ͒: it is important to point out that it ϩ 3 2 FIG. 4. Simulated electronic absorption spectrum of H2SiN (X A2). presents a relatively large f value ͑f ϭ0.032͒ among the symmetry-allowed states and that it would probably provide the second transition observed in absorption since the lower- 1 lying transitions have oscillator strengths smaller by a factor tion, lies 2.99 eV above the X A1 ground state and at 1.13 3 of 5, at least. The fact that the dominant configuration, as eV below the 1 A2 excited state if we report to the triplet equivalent distribution. reported in Table VI, corresponds to a double excitation from The second excited state is obtained by promoting the the ground state which is normally forbidden in one-photon absorption processes, should not be misleading since its nonbonding 2pN electron to the hole in the ␲SiN orbital ͓Fig. 1 weight in the CI wave function is barely over 50%; single 1͑d͔͒. The corresponding 1 A1 state has the following electronic configuration: excitations from the ground state have enough weight in the MC/P wave function to increase the oscillator strength of the 2 2 2 6 ϩ 2 Ϫ 2 ͑1sSi͒ ͑1sN͒ ͑2sSi͒ ͑2pSi͒ ͑␴SiH͒ ͑␴SiH͒ transition to a value which is of the same order of magnitude 3 2 2 2 0 0 as that computed for the singly excited 2 A2 state. Close to ϫ͑␴SiN͒ ͑␲SiN͒ ͑spN͒ ͑2pN͒ ͑␲SiN* ͒ . 3 this state ͑8.74 eV͒, the 4 A2 state is also described by a The orbital occupancy for this second excited state is then double excitation from the terminal spN orbital to the ␲SiN ϩ and ␲* orbitals; for the same reason, it has f ϭ0.011. Both completely analogous to the ground state of SiNH2 . The SiN 3 1 ϩ the previous transitions are polarized along the C axis of the X A2 –1 A1 singlet–triplet gap in H2SiN is 0.86 eV, far 2 1 3 molecule. from the 4.14 eV corresponding X A1 –1 A2 separation in SiNHϩ obtained at the same level of calculation. As a conclusion, it is essential to emphasize that the 2 ϩ As a consequence, the ordering of the first three states of electronic spectrum of H2SiN is completely different from ϩ 3 3 1 ϩ that of its SiNHϩ isomer on both qualitative and quantitative H2SiN (X A2,1 B2,1 A1) is reversed relative to SiNH2 2 1 3 3 points of view: Within 10 eV from the ground state, the (X A1,1 B2,1 A2). It has been previously hypothesized 3 density of the excited states is larger in the latter by a factor that using highly correlated methods should switch the 1 B2 1 ϩ 16 greater than 2. In this range of energy, 12 symmetry-allowed and 1 A1 states in H2SiN ; our calculations, however, that include most part of the nondynamic correlation and a large states can be reached by absorption techniques, but none of Ϫ2 part of the dynamic correlation, do not confirm this point of them presents an oscillator strength greater than 7.3ϫ10 , 3 1 in opposition to SiNHϩ that exhibits only four allowed view, maintaining the 1 B2 state 0.32 eV below the 1 A1 2 state. states, three of them having an oscillator strength greater 1 than 0.120. The next excited state ͑1 A2͒ lies 1.33 eV above the ground state and has the same orbital occupancies but exhibits a singlet coupling between the two unpaired electrons 3 VI. CONCLUSIONS ͓Fig. 1͑b͔͒. The following state ͓1 B1 , Fig. 1͑e͔͒ at 1.65 eV is the second symmetry-allowed state, but once again with a In this paper, we have reported a theoretical investiga- very small oscillator strength. Three singlet states appear at tion of the spectroscopic signatures of the ground state of ϩ 3 2.55, 2.61, and 3.30 eV and involve excitations among the H2SiN ( A2 – C2v) and have compared its properties to ϩ ͑spN,2pN,␲͒valence orbitals. The spectrum then presents a those of the SiNH2 isomer. 5 large gap: The next state is located at 4.20 eV and is a 1 B2 —The rotational constants and dipole moment have been corresponding to the following orbital occupancy: determined in order to help radio identification. The large value of the dipole moment: 3.8Ϯ0.2 D makes detection in 2 2 2 6 ϩ 2 Ϫ 2 ͑1sSi͒ ͑1sN͒ ͑2sSi͒ ͑2pSi͒ ͑␴SiH͒ ͑␴SiH͒ laboratory experiments possible, provided a relevant tech- ϩ ϫ͑␴ ͒2͑␲ ͒1͑sp ͒1͑2p ͒1͑␲* ͒1. nique is used to produce H2SiN since it lies about 115 SiN SiN N N SiN kcal/mol above the absolute minimum of the H, H, Si, Nϩ 1 ϩ The other states arising from the same electron distribution potential energy surface ͑X A1SiNH2 ͒. 3 3 3 lie at 5.59, 6.92, and 6.64 eV (2 B2,3 B2,4 B2), and 7.03 —At the reliable CASSCF level of theory, the infrared spec- 1 eV ͑3 B2͒. trum is characterized by intense bands around 550, 950, and

J. Chem. Phys., Vol. 104, No. 5, 1 February 1996 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 129.194.51.204 On: Fri, 13 Dec 2013 11:10:31 1988 Parisel, Hanus, and Ellinger: Interstellar silicon–nitrogen chemistry. III

Ϫ 2300 cm 1. The evaluation of the corresponding intensities 26 N. Goldberg, M. Iraqi, J. Hrusak, and H. Schwarz, Int. J. Mass Spectrom. was shown not to be relevant if single-reference approaches Ion Processes 125, 267 ͑1993͒. 27 ͑ROHF, MP2, MP3, MP4, and CCSDT͒ are used. R. Damrauer, M. Krempp, and R. A. J. O’Hair, J. Am. Chem. Soc. 115, 1998 ͑1993͒. —The electronic spectrum, obtained by our MC/P treatment 28 G. Maier and J. Glatthaar, Angewandte Chemie ͑Int. Ed. Engl.͒ 33, 473 presents only 2 significant features that should be observed ͑1994͒. in the 6.4 and 8.1 eV regions: Although some lower-lying 29 O. Parisel, M. Hanus, and Y. Ellinger, Molecules and Grains in Space, symmetry-allowed excited states have been found, their tran- edited by I. Nenner ͑AIP Press, New York, 1994͒, p. 515. 30 O. Parisel, M. Hanus, and Y. Ellinger, J. Phys. Chem. ͑in press͒. 31 sition moments seem to be too weak to allow their detection GAUSSIAN 92 ͑Revision B͒, M. J. Frisch, G. W. Trucks, M. Head-Gordon, P. using absorption techniques. M. W. Gill, M. W. Wong, J. B. Foresman, B. G. Johnson, H. B. Schlegel, We hope that the present results will stimulate experi- M. A. Robb, E. S. Replogle, R. Gomperts, J. L. Andres, K. Raghavachari, mental work to get more information on a molecular ion J. S. Binkley, C. Gonzalez, R. L. Martin, D. J. Fox, D. J. DeFrees, J. Baker, J. J. P. Stewart, and J. A. Pople ͑Gaussian Inc., Pittsburgh, 1992͒. 32 which may be a leading intermediate in the silicon–nitrogen HONDO 8.5, M. Dupuis, F. Johnston, and A. Marquez ͑IBM Corp., Kingston, chemistry. 1994͒. 33 ALCHEMY II, A. D. McLean, M. Yoshimine, B. H. Lengsfield, P. S. Bagus, ACKNOWLEDGMENTS and B. Liu, MOTECC-90 1990. 34 O. Parisel and Y. Ellinger, Chem. Phys. 189,1͑1994͒. We wish to thank N. Talbi for developing graphic inter- 35 B. O. Roos, Lecture Notes in Quantum Chemistry II ͑Springer, Berlin, faces to our codes. Part of the calculations presented in this 1994͒, Vol. 64. 36 R. J. Bartlett, J. Phys. Chem. 93, 1697 ͑1989͒. report were supported by the CNRS ‘‘Institut du De´veloppe- 37 B. O. Roos, Lecture Notes in Quantum Chemistry I ͑Springer, Berlin, ment et des Ressources en Informatique Scientifique’’ IDRIS 1992͒, Vol. 58. supercomputing center. This research was also partly con- 38 T. Clark, J. Chandrasekhar, G. W. Spitznagel, and P. v. R. Schleyer, J. ducted with supercomputer resources at the Jet Propulsion Comput. Chem. 4, 294 ͑1983͒. 39 M. J. Frisch, J. A. Pople, and J. S. Binkley, J. Chem. Phys. 80, 3265 Laboratory, California Institute of Technology, under con- ͑1984͒. tract to the National Aeronautics and Space Administration. 40 R. Krishnan, J. S. Binkley, R. Seeger, and J. A. Pople, J. Chem. Phys. 72, 650 ͑1980͒. 41 A. D. McLean and G. S. Chandler, J. Chem. Phys. 72, 5639 ͑1980͒. 1 B. E. Turner, BAAS 23, 933 ͑1991͒. 42 W. J. Pietro, M. M. Francl, W. J. Hehre, J. S. Binkley, M. S. Gordon, D. J. 2 B. E. Turner, Astrophys. J. 388, L35 ͑1992͒. 3 DeFrees, and J. A. Pople, J. Am. Chem. Soc. 104, 5039 ͑1982͒. E. Herbst, T. J. Millar, S. Wlodek and D. K. Bohme, Astron. Astrophys. 43 222, 205 ͑1989͒. M. M. Francl, W. J. Pietro, W. J. Hehre, J. S. Binkley, M. S. Gordon, D. J. 4 DeFrees, and J. A. Pople, J. Chem. Phys. 77, 3654 ͑1982͒. W. D. Langer and G. A. E., Astrophys. J. 352, 123 ͑1990͒. 44 5 A. E. Glassgold, Astrochemistry of Cosmic Phenomena, edited by P. D. NBO Version 3.1 by E. D. Glendening, A. E. Reed, J. E. Carpenter, and F. Weinhold. Singh ͑Kluwer, Dordrecht, 1992͒, p. 379. 45 6 B. E. Turner, Astrophys. J. 376, 573 ͑1991͒. F. Weinhold and J. E. Carpenter, The Structure of Small Molecules and 7 B. E. Turner, in: Astrochemistry of Cosmic Phenomena, edited by P. D. Ions ͑Plenum, New York, 1988͒. 46 A. D. McLean and B. Liu, J. Chem. Phys. 58, 1066 ͑1973͒. Singh ͑Kluwer, Dordrecht, 1992͒, p. 181. 47 8 T. L. Allen, J. D. Goddard, and H. F. Schaeffer III, J. Chem. Phys. 73, S. Saito, Y. Endo, and E. Hirota, J. Chem. Phys. 78, 6447 ͑1983͒. 48 3255 1980 . O. Parisel, M. Hanus, and Y. Ellinger, Chem. Phys. ͑submitted͒. ͑ ͒ 49 9 M. Elhanine, R. Farrenq and G. Guelachvili, J. Chem. Phys. 94, 2529 E. A. V. Ebsworth, Volatile Silicon Compounds ͑Pergamon, London, ͑1991͒. 1963͒. 50 10 M. Bogey, C. Demuynck, J.L. Destombes, and A. Walters, Astron. Astro- U. Wannagat, The Chemistry of silicon-nitrogen compounds, edited by H. phys. 244, L47 ͑1991͒. J. Emeleus and A. G. Sharpe ͑Academic, New York, 1964͒, p. 225. 51 11 M. Elhanine, B. Hanoune, and G. Guelachvili, J. Chem. Phys. 99, 4970 H. Bu¨rger and R. Eujen, Low-Valent Silicon, Vol. 50 ͑Springer, Berlin, ͑1993͒. 1974͒. 52 12 N. Goldberg, J. Hrusak, M. Iraqi, and H. Schwarz, J. Phys. Chem. 97, D. J. DeFrees, K. Raghavachari, H. B. Schlegel, and J. A. Pople, J. Am. 10687 ͑1993͒. Chem. Soc. 104, 5576 ͑1982͒. 53 13 J. F. Ogilvie and S. Cradock, Chem. Commun. 1966, 364. P. Cassam-Chenaı¨, F. Pauzat, and Y. Ellinger, J. Mol. Struct. ͑Theochem͒ 14 R. Preuss, R. J. Buenker and S. D. Peyerimhoff, J. Mol. Struct. 330, 167 ͑1995͒. 54 ͑Theochem͒ 49, 171 ͑1978͒. D. J. DeFrees and A. D. McLean, J. Chem. Phys. 82, 333 ͑1985͒. 55 15 B. T. Luke, J. A. Pople, M. B. Krogh-Jespersen, Y. Apeloig, M. Karni, J. E. Wilson, J. C. Decius, and P. C. Cross, Molecular Vibrations ͑McGraw Chandrasekhar, and P. von Rague Schleyer, J. Am. Chem. Soc. 108, 270 Hill, 1955͒. 56 ͑1986͒. E. A. V. Ebsworth, M. Onyszchuk, and N. Sheppard, J. Chem. Soc. 1958, 16 J. R. Flores and J. Largo-Cabrerizo, Chem. Phys. Lett. 142, 159 ͑1987͒. 1453. 17 J. R. Flores, F. Gomez Crespo and J. Largo-Cabrerizo, Chem. Phys. Lett. 57 B. O. Roos, K. Andersson and M. P. Fu¨lscher, Chem. Phys. Lett. 192,5 147,84͑1988͒. ͑1992͒. 18 S. Wlodek, C. F. Rodriquez, M. H. Lien, A. C. Hopkinson, and D. K. 58 O. Parisel, and Y. Ellinger, in Strategies and Applications in Quantum Bohme, Chem. Phys. Lett. 143, 385 ͑1988͒. Chemistry, edited by Y. Ellinger and M. Defranceschi ͑Kluwer Academic, 19 J. R. Flores and J. Largo-Cabrezio, J. Mol. Struct. ͑Theochem͒ 183,17 Dordrecht, 1996͒. ͑1989͒. 59 O. Parisel and Y. Ellinger ͑submitted͒. 20 A. C. Hopkinson and M. H. Lien, Can. J. Chem. 67, 991 ͑1989͒. 60 A. D. McLean, B. H. Lengsfield, J. Pacansky, and Y. Ellinger, J. Chem. 21 G. Maier, J. Glatthaar and H. P. Reisenauer, Chem. Ber. 122, 2403 ͑1989͒. Phys. 83, 3567 ͑1985͒. 22 S. C. Foster, J. Mol. Spectrosc. 137, 430 ͑1989͒. 61 J. S. Binkley, J. Am. Chem. Soc. 106, 603 ͑1984͒. 23 S. Ye and S. Dai, J. Mol. Struct. ͑Theochem͒ 236, 259 ͑1991͒. 62 M. Cordonnier, M. Bogey, C. Demuynck, and J. L. Destombes, J. Chem. 24 P. Botschwina, M. Tommek, P. Sebald, M. Bogey, C. Demuynck, J. L. Phys. 97, 7984 ͑1992͒. Destombes, and A. Walters, J. Chem. Phys. 95, 7769 ͑1991͒. 63 R. S. Grev and H. F. Schaefer III, J. Chem. Phys. 97, 7990 ͑1992͒. 25 D. P. Chong, D. Papousek, Y. T. Chen, and P. Jensen, J. Chem. Phys. 98, 64 L. Serrano-Andre´s, M. Merchan, I. Nebot-Gil, R. Lindh, and B. O. Roos, 1352 ͑1993͒. J. Chem. Phys. 98, 3151 ͑1993͒.