8930
J. Phys. Chem. 1994, 98, 8930-8936
Aluminum-Nitrogen Multiple Bonds in Small AINH Molecules: Structures and Vibrational Frequencies of AlNH2, AlNHs, and AlNH4 Randall D. Davy' and Kent L. Jaffrey Department of Chemistry, Liberty University, Lynchburg, Virginia 24506 Received: April 13, 1994; In Final Form: May 31, 1994"
Ab initio molecular electronic structure methods have been used to study aluminum-nitrogen analogues of the simplest alkyne and alkene molecules and their isomers and the experimentally studied AlNH3. Geometries, relative energies, and harmonic vibrational frequencies of isomers of formulas AlNH2, AlNH3, and AlNHI were obtained a t the self-consistent field (SCF), single and double excitation configuration interaction (CISD), and coupled cluster including all single and double substitutions (CCSD) levels of theory. The lowest energy isomers were AI-NH2, HA1-NH2, and H2Al-NH2. The HA1-NH isomer is quasi-linear and has a theoretical bond length indicative of a triple bond. It, however, is very high energy, as evidenced by the very large negative heat of hydrogenation, -62 kcal/mol, and the lower energy of the Al-NH2 isomer. The double bond ( u bond plus N to A1 a donation) is much more favorable. The strength of the A1-N bond depends strongly on the number of hydrogens on the aluminum atom because the hydrogens increase the Lewis acid strength of the aluminum atom. The addition and insertion products of aluminum atoms and ammonia are fully characterized and compared to previous experimental and theoretical studies.
Introduction The chemistry of 111-V compounds has been developed mainly along two lines: toward the development of solid semiconductors, for obvious practical reasons, and for comparison with the isoelectronic organic compounds as in the case of borazine and benzene. These two lines converge in the study of multiple bonds between group I11 and group V atoms. Multiply bonded compounds have been pursued for use in organometallic chemical vapor deposition, as recently reviewed by Cowley.' The key is to make the 111-V bond the most stable in the organometallic precursor, so that the deposition process will result in the loss of the ligands at low temperature, but retain the 1:l ratio of the desired product. A 111-V multiple bond will of course favor this process. Although much practical work has been done to find the conditions that favor the stability of multiple bonds,2 few theoretical studies have been conducted to examine the nature of multiple bonds in various 111-V compounds. Notable exceptions are B-N compounds, especially borazine3 and recently phosphaborazine and a l ~ m a z i n e . ~ Study of A1-N multiple bonds in particular is pertinent for two reasons. First is the synthesis of the A1-N analogue of borazine by Waggoner, Hope, and Power5 and the theoretical study by Finkand Richards.4 These studies show that even though both B-P and AI-N will form planar six-membered rings (thus superficially resembling benzene) the bonding in the two compounds is quite different and in fact presents a puzzle. The theoretical study indicates a greater "aromatic" character for the B-P analogue of benzene but also a greater tendency for it to be nonplanar. The B-P ?r system delocalizes more than the A1-N ?r system, but, according to theory, the B-P structure must be forced planar by bulky substituents, while the AI-N system need not be. Second, there has been continued interest in formation of solid aluminum nitride from organometallic precursors." These processes usually involve chemical vapor deposition, and to understand them one would like to delve into the chemistry of small A1-N molecules. Although there have been many studies of the reactions of A1 atoms with small organic molecules, there are few papers that report studies of A1 atoms with ammonia. A recent paper by Howard, Joly, Edwards, Singer, and Logan reported EPR studies of the reaction of A1 atoms with NH3 Abstract published in Advance ACS Absrracrs, July 15, 1994.
molecules9 and assigns a structure for the insertion product. A theoretical study of the insertion of A1 atoms into X-H bonds, including the N-H bond of ammonia, is reported by Sakai in his 1992 paper.10 The theoretical study of Sakai gives an optimized structure for the insertion product quite different from that postulated by Howard et al. and predicts a substantial (25.8kcal/ mol) energy barrier for insertion. Vibrational frequencies for NH3 coordinated to A1 are reported by Wang, Hipps, and Mazurl * in their experimental study of hydrogenated AI-N thin films. The molecule ammonia alane, NH3AlH3, has been the subject of several theoretical studies, as reviewed in the most recent study of Marsh, Hamilton, Xie, and Schaefer.12 The motivation for the study of ammonia alane is interest in its derivative trimethylamine alane, (CH3)pNA1H3,l3 as a source of aluminum in applications such as chemical vapor deposition.l3 Our paper reports work on the AI-N analogues of the simplest organic compounds that contain multiple bonds. We report theoretical geometries, energies, and vibrational frequencies for compounds of formulas, AlNH2, AlNH3, and AlNH4. Theoretical results for AlH3NH3 were also obtained for the sake of comparisons of bond lengths and frequencies and energies of hydrogenation. However, the geometries and frequencies were essentially identical with those of Marsh et al.; therefore, only theoretical energies and certain vibrational frequencies are included from the present study for AlH3NH3. We will compare AI-N bond lengths and charge distributions among compounds that contain (formally) single, double, and triple bonds. At the outset it was not certain whether HAl-NH would remain linear or H2Al-NH2 would remain planar to preserve the possible triple and double bond, respectively, or, as Allen, Scheiner and SchaferI4 found for the case of HzP-BHz, theory would predict a nonplanar structure. The energy of rotation about the AI-N axis in H2Al-NH2 will be determined to give another measure of the ?r bond strength. We will also examine the insertion of A1 into NH3 at a higher theoretical level than Sakai and make further comparison with the experimental results of Howard et al. and Wang et al. The ultimate goal is to further the understanding of the fundamental natureof the bonding between A1 andN and tolay thegroundwork for studies of larger AI-N clusters.
0022-3654/94/2098-893Q~Q4.5Q/Q0 1994 American Chemical Society
The Journal of Physical Chemistry, Vol. 98, No. 36, 1994 8931
Aluminum-Nitrogen Multiple Bonds
TABLE 1: H-AI-N-H Theoretical Geometries and Harmonic Vibrational Freauencies' ~~
AI-N A1-H N-H LH-A1-N LA1-N-H N-H str AI-H str AI-N str linear bend linear bend torsion 0
TABLE 2 AI-NHz Theoretical Geometries and Harmonic Vibrational Freauencies'
~
SCF
CISD
CCSD
1.594/1.594 1.54611.550 0.99010.982 1801180 1801180 405214022 214812123 124411229 5831554 336/294
1.60811.608 1.540/1.546 0.99910.988 1801176.3 1801173.2 391613907 214912102 119411177 5491527 120/88 -1527
1.63011.633 1.547/1.556 1.00610.997 166.11165.2 158.41154.5 381213788 209812042 114411104 6581498 2181262 5061497
Notation is DZPITZZP.
Theoretical Methods Because these are little studied molecules several theoretical models were used to assess the level of theory necessary for consistent results. The PSI programs written by the Schaefer group were used for all methods.15 The lowest level approximation was the Hartree-Fock method with a double-{plus polarization (DZP) basis set. Electron correlation was then taken via the methods of configuration interaction including all single and double excitations (CISD) and coupled cluster expansion including all singles and doubles (CCSD). This trio of methods was repeated by using a larger basis set, a triple-{plus two polarization functions (TZZP) expansion for the AlNH2 isomers, which are the smallest and highest energy molecules. For the AlNH3 and AlNH4isomers geometry optimizations were performed at the SCF/TZ2P and CISD/TZ2P level. All geometry optimizations were done with analytic gradient methods. Geometries were optimized at all levels except for coupled cluster on open shell molecules due to unavailability of programs to calculate gradients for the open shell coupled cluster method. For open shell molecules and coupled cluster energy was evaluated at the CISD optimal geometry. Vibrational frequencies were evaluated by using analytic second derivatives for the SCF method and by finite difference for the CISD and CCSD (closed shell) methods. The DZP basis set is a DZ contracted Gaussian basis set consisting of the Huzinaga and Dunning16J7(9s5p/4s2p) basis on nitrogen, the (4s/2s) set on hydrogen, and the HuzinagaI8 and Dunning19 (1 ls7p/6s4p) set on aluminum, with an added d function for A1 and N (exponent = 0.40 for Al, 0.80 for N) and an added p function for H (exponent = 0.75). The TZ2P basis is the Huzinaga16 and Dunning20 (lOs6p/5s3p) set on nitrogen and (5s/3s) set on hydrogen and the Mclean-Chandler1s~21 (12s9p/ 6s5p) set on aluminum, with two d functions added onto A1 and N and two p functions on H (exponents = 0.80,0.20 for Al; 1.60, 0.40forN;and 1.5,0.375forH). Forcorrelatedmethods(CISD, CCSD) two core orbitals (corresponding to the A1 and N 1s orbitals) were frozen and two virtual orbitals deleted.
Results AINHz. Several structures including H2Al-N, HAl-NH, and AI-NH2 and bridging structures were initially studied. No stable structures with bridging hydrogens were found, and H2Al-N proved to be very high energy-56.2 kcal/mol above AI-NH2 at theSCF/DZPleveland 75.0 kcal/mol at the CI/DZPlevel-and was not studied further. The results for HAl-NH and AI-NH2 and the transition state that connects them are summarized in Tables 1-111 and Figure 1. HAl-NH is one of the unusual cases that has a qualitatively incorrect geometry at the CISD/DZP level. The linear geometry becomes a saddle-point when either the size of the basis set or the level of correlation correaction is increased. Both the geometry of Al-NH2 and the relative energy show good convergence as the level of approximation is increased. HAl-NH is substantially higher energy than AI-NH2 at all levels of theory, with the best estimate of their energy difference being 41.4 kcal/mol from the CCSD/TZ2P method, including zero
AI-N N-H LH-N-H N-H asym str N-H sym str H-N-H bend AI-N str NH2 rock out-of-plane
SCF
CISD
CCSD
1.79111.788 1.003/0.997 108.41108.9 384313831 375613750 169111692 7901780 5391532 4981445
1.79511.797 1.013/ 1.004 108.01108.6 374813732 365813645 161711613 7781762 4991488 4521428
1.801/1.803 1.018/1.009 107.91108.5 368213660 358913571 159311585 7691750 4911478 4361413
Notation is DZPITZZP.
TABLE 3: Theoretical Geometries and Harmonic Vibrational Frequencies for the H-AI-N-H to AI-NHz Transition State AI-N AI-H N-H N-H (bridge) LH-A1-N LA1-N-H N-H str AI-H str AI-N str AI-N-H bend torsion H-A1-N bend 0
SCF
CISD
CCSD
1,68411.679 1.61911.622 1.002/0.995 1.99311.986 74.2174.0 170.91171.1 387513865 188111863 9761973 9281928 3461270 1095i/llOSi
1.68111.677 1.61311.618 1.01211.001 1.93311.938 7 1.9172.0 170.81170.7
1.704/1.700 1.62811.634 1.018/1.007 1.92411.903 70.4171.0 172.61172.3
Notation is DZPITZZP. 1.803
H
1'700
/1
71.0
AI k N ' H
97.7
Figure 1. Best values (CCSDITZZP) for geometries of AlNHz isomers. point vibrational energy (ZPVE) corrections. The CCSD/TZ2P method should be taken as the best estimates of the geometry of the isomers. AINHJ. The isomers of AI-NH3, HAl-NH2, and H2Al-NH were initially investigated. For all AlNH3 isomers CCSD results are energies calculated at CISD optimal geometries because CCSD geometry optimizations by gradient methods are not available. A search was made for bridging isomers, but no stable structures (minima) were found. Extensive studies were therefore only done on Al-NH3, HAI-NH2, HzAINH, and the transition state connecting AI-NH3 and HAl-NH2. To help understand experimental studies of A1 reactions with ammonia, the energy of separated A1 and NH3 was also estimated. For the SCF and CCSD methods, which are size consistent, the energy of the two separate species can simply be summed; for CISD, which is not size consistent, the energy of Al..NHp was calculated at a large (100 A) fixed AI-N distance, with the NH3 geometry fixed at
8932 The Journal of Physical Chemistry, Vol. 98, No. 36, 1994
TABLE 4: AI-NH3 Addition Product Geometries and Harmonic Vibrational Frequencies. SCF CISD 2.45512.403 2.37612.350 AI-N l.006/1.001 1.017/1.008 N-H’ 1.015l1.006 N-H l.005/1.000 107.31108.3 LH-N-H’ 108.21108.6 106.31107.2 LH-N-H 107.41107.6 112.2/112.0 113.711 12.7 AI-N-H N-H str A’ 382113994 3710 3708 N-H str A” 381613783 N-H str A‘ 369213680 3580 1714 HNH bend A’ 178811795 1531 HNH bend A’‘ 149611494 1258 HNH bend A’ 127611272 4281440 426 NH3 wag AI-N str 1681177 200 tors 1331161 182 a Notation is DZPJTZZP. Where only one number is given, it is DZP. b H’ is the symmetry unique H on NH3. See Figure 2.
Davy and Jaffrey
TABLE 7: Hfl-NH Theoretical Geometries and Harmonic Vibrational Frequencies. SCF CISD A1-H 1.56911.573 1.56111.567 AI-N 1.75711.776 1.769J1.778 N-H 0.99710.994 1.010/ 1.001 LH- AI-H 123.71123.4 123.9/124.2 LAI-N-H 159.91141.7 151.41143.6 N-AlH2 out-of-plane 3.414.8 4.9J5.9 N-H stretch 391813849 3775 AI-H sym str 205212024 2062 AI-H asym str 204812017 2057 AI-N stretch 8931868 873 H-AI-H bend 8281809 807 N-AlH2 out-of-plane 6681658 645 H-AI-H rock 5771545 549 torsion 3991350 377 H-N-AI bend 1771317 255 a Notation is DZPJTZZP. Where only one number is given, it is DZP. 1.008
+
TABLE 5 HAl-NHp AI NH3 Insertion Product Theoretical Geometries and Harmonic Vibrational hequenies’ SCF
CISD
AI-H 1.585/1.587 1.578/1.584 AI-N 1.76911.766 1.77311.772 N-H (cis) 1.001/0.993 1.011/0.999 N-H (trans) 1.000/0.994 1.010/1.000 AI-N-H (cis) 124.71124.4 12431124.4 125.6/125.5 125.7/125.5 AI-N-H (trans) 116.61116.5 1 16.111 16.2 LH-AI-N N-H str 3891 3788 3799 3693 N-H str 1970 1972 AI-H str H-N-H bend 1705 1627 AI-N str 848 836 NH2 rock 798 770 torsion 534 519 H-AI-N bend 520 510 out-of-plane+ torsion 498 447 Notation is DZPJTZZP. Where only one number is given, it is DZP.
1.008
H 1.822 ,
SCF
42.9
5.9 out of plane 143.8
HH l.584\
TABLE 6 Transition State for A1 Insertion into NH3. Theoretical Geometries and Harmonic Vibrational Frequencies.
H
116.2
124.4
u
1.587
H
‘3(--AAI __ N