31 March 1995
CHEMICAL PHYSICS LETTERS Chemical Physics Letters 235 (1995) 528-534
Electronic structure of trimethylamine alane in the solid state C. Fauquet a, p. Dannetun a, C. Fredriksson a, S. Stafstr/~m a, J.E. Bour6e P. Brillaud c, N. Bouanah c, W.R. Salaneck a
" Department of Physics, IFM, LinklJping Unit,ersity, S-581 83 Link#ping, Sweden h CNRS, Laboratoirc de Meudon, Meudon, France ¢ SMi Organometailics Dia,ision, 13 000 Marseille, France
Received 12 November 1994: in final form 18 January 1995
Abstract The chemical and electronic structure of ultrathin molecular films of trimethylamine alane (TMAA), condensed in UHV at - t00°C, have been studied in the solid state, using both X-ray and ultraviolet photoelectron spectroscopy. The results are analyzed with the help of quantum chemical calculations at the ab initio Hartree-Fock 6-310" level. Based upon the good agreement between theory and experiment, it is determined that clean, oxygen-free, condensed molecular solid films consist of the 2 : 1 adduct of TMAA, which was previously uncertain, in addition, based upon the electronic structure results, it is clear that the mechanism of the photodecomposition of TMAA can be explained in terms of the wavefunction of electrons photoexcited into the first unoccupied molecular orbital.
Recently, light emitting diodes (LEDs) have been fabricated from organic conjugated polymer films [1-4]. Studies of the electronic structure of conjugated polymer-metal interfaces involved in these devices have shown that changes occur in the "rrelectronic structure at the surfaces of the conjugated polymer films during physical vapour deposition (PVD) of metals [5-8]. These modifications are, in general, related to chemical reactions of individual metal atoms with the organic surface. In certain instances, however, surface modification may be induced by the PVD process itself, for example, from the high thermal energy of the individual metal atoms impinging on the surface. The use of a laserassisted chemical vapor deposition (LACVD) technique for the metallization of polymer surfaces could
suppress the kinetic energetic factor involved in the PVD technique. The goal of this work is to provide a spectroscopic basis for the interpretation of spectra acquired in the process of the application of metal contacts on conjugated polymer surfaces via LACVD of trimethylamine alane, or TMAA. LACVD can be carried out in different ways . The method of interest here would involve ultrahigh vacuum (in the 10 -I° Tort range), since TMAA is unstable in air. The organometallic molecules would be adsorbed in the form of an ultrathin molecular film on the surface to be metallized, followed by the removal of the ligands with light or heat (laser beam). Recently, it has been shown that trimethylamine alane (TMAA) is an interesting candidate for the organometailic precursor for aluminum deposits, since aluminum films made from TMAA contain essentially no carbon, as opposed to those
0009-2614/95/$09.50 © 1995 Elsevier Science B,V. All rights reserved SSD! 0009.26 t 4(95 )00133 -6
C. Fauquet et al. / Chemical Physics Letters 235 (1995) 528-534
cu3 H H3C-~ N-~,i/ I CH 3 H
S/- CH3 \ CH 3
H3C-~N-AI/-H / \
Fig. I, (A) 2: I adduct of Irimethyh|mine to ahtrle (TMAA); (B) I : 1 adduct.
obtained with other precursors such as trimethyl aluminum or dimethylaluminum hydride [10-13]. In the context of a study of LACVD of aluminum on the surfaces of conjugated polymer films, the first stage is to determine the electronic structure of condensed ultrathin molecular films of the organometallie compound ia the solid state, since the ~xa,~i nature of the chemical structure of the molecule itself is not known. There is disagreement as to whether the molecule exists as the 1:1 or the 2:1 adduct of trimethylamine to alane in different phases, since in the gas phase there is an equilibrium between the 2:1 adduct to the 1 : 1 adduct [14,15]. The chemical structure of the two adducts is shown in Fig. 1. Thus, the question of the structure of the TMAA in solid thin-film form following vapor deposition in UHV is a basic point to be studied. A second step is to characterize the chemical and electronic structure of a monomolecular layer on the surface of a conjugated polymer of interest. Finally, a third step is to follow the chemical and electronic structure modifications of the early stages of interface formation with a polymer surface, during the removal of the organic ligands with light. These results could then be compared with those from studies of the early stages of Al-interface formation by thermal PVD on the surface of the conjugated polymer of interest. In this Letter, the results of the first stage studies are reported. TMAA multilayer films were condensed on gold substrates at low
temperatures under UHV conditions. The electronic and chemical structure has been studied by X-ray and ultraviolet photoelectron spectroscopy (XPS and UPS). The results have been interpreted with the help of ab initio Hartree-Fock/6-31G* calculations.
2. Experimental The UHV system used consists of three interconnected vacuum cilambers (introduction, preparation and analysis). Vacuum is obtained through pumping by a system consisting of turbo and ion pumps, in combination with a cryogenic shielding/pumping system at 20 K. In the preparation and analysis chambers, the residual vacuum is lower than 10-Io Torr, even after the vapor deposition of TMAA. XPS is carried out using unmonochromatized Mg Kot (1253.6 eV) radiation, while UPS is performed using monochromatized He I (21.2 eV) or He II (40.8 eV) radiation. The analyzer is run with a net resolution of 0.2 eV for XPS and 0.1 eV for UPS. Under these conditions, &e Au(4f7/2) line would be recorded with a full-width at half-maximum of about 0.9 eV. The XPS peak energies are reproducible to within about 0.2 eV, and the UPS to within about 0.1 eV. For TMAA, the molecules were introduced into the preparation chamber through a capillary close to the sample, in order to obtain a nigh local vapor pressure. Although the 2:1 adduct is known to dissociate into the 1:1 adduct, recombination leads to the 2:1 adduct in equilibrium at -196°C . Therefore, the TMAA source was cooled in a liquid nitrogen bath at the start of any experiment. The remaining gas in the TMAA source was mainly hydrogen, as indicated by mass spectroscopy. This hydrogen over pressure was then pumped down to about 10 -5 Torr, followed by subsequent evacuation of the stainless steel gas handling manifold down in to the 10 -9 range. The TMAA source was then heated to about 35°C, and maintained at this temperature during subsequent deposition. At this temperature, the vapor pressure of the product is sufficiently high to fill the gas manifold. In order to prepare an ultrathin film of TMAA in UHV, a gold substrate was cooled to 200 K, cleaned by ion etching to remove hydrocarbons, and then
C, Fauquet et al. / Chemical Physics Letters 235 (1995) 528-534
exposed to TMAA introduced from the gas handling system through a leak valve. Films were formed by exposure of the cold substrate for 2 to 3 min. During deposition, the partial pressure of the mass m/e = 58 peak in the cracking pattern of TMAA, monitored by a VG Quadrupole SX-2-743 residual gas analyzer, was kept at about 2 × 10 -9 Tort. The rate of disappearance of the Au(4fT/2) line in the XPS spectra was used to estimate the rate of growth of the TMAA films on the substrate, using the mean free path value of A = 22 ,~, for electrons at a kinetic energy of 1.17 keV . Tri~nethylamine (TMA) was condensed under essentially the same conditions as TMAA, in order to keep all parameters as identical as possible. The only difference was that put down at -150°C, because of a somewhat higher vapor pressure for TMA than for TMAA.
BINDING ENERGY (eV) !
290 288 286 284 282 280 BINDING ENERGY (eV)
3. Results and discussion In Fig. 2 the XPS spectra for a typical condensed film of TMAA are shown. At the top of the figure the core-electron lines corresponding to aluminum in TMAA and the gold substrate are shown. Note that at 1254 eV, the photoionization cross section for gold is approximately 23 times higher than that for aluminum . The Au(4fT/2) and Au(4fs/:) lines are at the same binding energy as before the deposition of TMAA. From the attenuation of the Au(4f,/, ) line, the thickness is estimated to be about 100 A. The binding energy of the Al(2p) line is 73.7 eV. The binding energies of the C(ls) and N(ls) electrons are found to be 286,4 and 400.7 eV, as shown in the middle and bottom of Fig. 2. No oxygen was detected. The intensity ratios of the lines, corrected for X-ray cross sections, yield a ratio for the A! : N : C content close to the theoretical value of 1:2:6. Thus, the data indicate that the condensed molecular solid films, at 200 K, consist of the 2:1 adduct, at least at the surface. The core-electron binding energies in the XPS spectra can be compared with those in standard reference compounds. The binding energy of the Al(2p) line is higher (by about I eV) than that in pure aluminum, indicating that the aluminum atom in the TMAA molecule is positively charged, as expected from the known ability of the Al atom to
406 404 402 400 398 396 BINDING ENERGY (eV) Fig. 2. XPS spectra of a thin film of (A) TMAA and (B) TMA condensed on gold, Top: Au(4f~/,), Au(4fv/:) and Al(2p) lines; middle: C(Is) spectrum; bottom: N(ls) spectrum.
donate electrons to organic ligands. The N(ls) and C(ls) core levels of TMA are compared with those of TMAA in the middle and lower panels of Fig. 2. The binding energies are 400.2 and 286.1 eV, respectively. The differences in the binding energies (between TMAA and TMA) although small, are consistent with the relative binding energies obtained in ab initio Hartree-Fock calculations, indicating that the bond between the alane moiety and the nitrogen in TMAA does not extensively change the charge density on the trimethylamine ligand. In other words, the charge transfer between the alane group
C. Fauquet et aL /Chemical Physics Letters 235 (! 995) 528-534
and the ligands is small. This points to an explanation of why essentially pure (carbon-free) aluminum films can be prepared, by pyrolytic or photolytic decomposition, from this precursor molecule. Note, however, that the electron transfer from the AI atom to ~he N atom within the alane group, as well as from the C H 3 group to the N atom, within the TMA group, is quite large. This electron transfer results in Mulliken charges of 0.641 qe I and - 0.631 qe I (on the Al and N atoms, respectively). Thus, even though there is no charge transfer between the AI and N atoms, this bond is highly polar in the ground state. In contrast, in the excited state, the net charges on the AI and N atoms are much smaller, which reduces the ionic contribution to the Al-N bond and weakens the bond considerably. The results of ab initio quantum chemical calculations lead to an estimate of the ground state geometric configuration, through the minimization of the total ground state energy. Without any symmetry restriction in the geometry optimization procedure, the ground state geometries of the 2:1 and 1:1 adducts relax to Cah and C3v, respectively, as shown in Fig. 1, as determined in previous studies [15,18]. The bond lengths and bond angles are compared with those of related molecules in Table 1. Although the A l - N bond in these compounds is longer than in other molecules containing a four-coordinated AI atom [19,20], the Al-N distance in the mono- and di-amine compounds is found to be in agreement with previously published results on the molecules [19,21,22]. Also, the calculated N-C and C-H and A l - H bond lengths are found to be about the same as those reported previously for the 2:1 and 1:1
adducts as well as those of related compounds (see "Fable 1). Previous suggestions to explain the high polarization of the diamine molecule, and its dipole moment in solution, in terms of the flexing of the highly polar AI-N bonds [18,21,22] or of the AI-H bonds  have not been examined in this work. The results of this present work, however, indicate that the ground state value of the N - A l - N angle is about 180°, in agreement with previously reported data from X-ray diffraction studies in the solid state . The N - C - H angle is in good agreement with that in the 1:1 adduct and that found in (CH03AI-N(CH3)3 . The calculated A l - N - C angle is found to be the same for both compounds and is intermediate to that found experimentally for the 2:1 adduct in the gas phase  and in the solid single crystalline phase . Furthermore, it is in good agreement with that found tbr trimethylaluminum-trimethylamine . Note that the structure of the molecule is almost the same in the gas phase as in the solid state and in solution. The structural differences between the 1:1 and 2:1 adducts as calculated in this work are small, consistent with the discrepancy found in the literature between these two compounds under the generic name of TMAA. In the geometry optimization, the main difference between the mono- and the di-amine consists in the angle between the main axis of the molecule and the hydrogen atoms on the aluminum, as can be understood intuitively. The N - A l - H angle changes from 90 ° to 109° upon removal of a trimethylamine group (see "Fable 1), due to the geometrical rearrangement of the hydrogen atoms bound to the
Table 1 A comparison of bond lengths and bond angles Angle (deg)
Bond lengths (,~)
TMAA 2:1 (this work) Ref.  Ref.  TMAA i : l (this work) Ref.  (CH3) 3 AI-N(CHa) 3 
2.26 2.19 2.18
1.46 1.48 1.48
1.08 l.l I
108.8 106.5 I I 0.0
2.26 2.06 2.1
1.46 !.48 1.47
1.08 1.11 1.11
1.62 1.56 -
108.8 109,0 209.3
110. I 107.6 112. I
109.0 104.3 -
C. Fauquel et al, / Chemical Physics Letters 235 (1995) 528-534
BINDING ENERGY eV Fig, 3. UPS valcncc band spectrum of a thin film of T M t . • condensed on gold, Top: UPS (He l) spectrum; middle: 2:1 adduct calculated DOVS; bottom: 1 : I adduct calculated DOVS.
A! atom. This leads to the appearance of a dipole mom¢nt of 5.7 D in the molecule, in good agreement with that assumed in previous work [15,18]. In Fig. 3 the experimental UPS valence band spectrum of the condensed film of TMAA is shown. From the energy of the cut-off in the secondary electron spectrum (not shown in the figure), the work function of the molecular solid is found to be about 3.5 eV. The density-of-valence-states (DOVS) is obtained from the calculated electronic structure in the standard way, by broadening the energy eigenvalue spectrum by a Gaussian function, in order to simulate the experimental UPS spectrum. A comparison between the theoretical DOVS for 2:1 and 1:1 compounds and the UPS spectrum of a condensed molecular solid film of TMAA also is seen in Fig. 3. A standard contraction and an energy shift (by 0.8 × and 2 eV, respectively) are applied to the calculated DOVS in order to correct for electron correlation and solid-state effects which are not included in the
present calculation. We have no reason to believe that the 1:1 or 2:1 adducts differ in this respect. The comparison between the theoretical and experimental data is, therefore, of the same level of accuracy for both adducts. Also, photoionization cross sections are not included in the calculated DOVS. In addition, it should be noted that by using the DOVS to simulate UPS spectra, final state effects are inherently neglected. The overall agreement between the DOVS and the UPS spectra, however, is good, as shown in Fig. 3, which shows that such final state effects are unimportant for the type of systems studied here. The calculated DOVS for both the 1 : 1 and 2:1 adducts appear to be in overall good agreement with the UPS spectrum. Note, however, that slight differences between the two compounds appear in the calculated data. The calculated DOVS for the 2:1 adduct are more consistent with the experimental UPS spectrum than the DOVS for the 1 : 1 adduct. A shoulder is observed on the experimentai spectrum, where, instead of a peak, a valley is observed. A shoulder observed on the high binding energy side of peak A in the experimental UPS spectrum is in better agreement with the energy splitting of the two corresponding peaks in the calculated DOVS for the 2 : 1 adduct than for the 1 : 1 adduct. Moreover, the shoulder at 8.9 eV in the calculated 1 : 1 DOVS does not appear in the experimental spectrum, nor in the calculated DOVS for the 2:1 adduct. These indications themselves are not conclusive enough to differentiate between the two different adducts using only the UPS data. The conclusion that there is better agreement with the 2:1 adduct, however, is consistent with the indications from the relative intensities of the XPS core electron lines, which provide strong evidence for the existence of the 2 : 1 adduct in the condensed molecular solid films. Given the agreement above, the theoretical DOVS results can be discussed further. Peak A contains contributions from three orbitals, two of which are degenerate and are localized on the alane group, while one extends over the whole molecule, but with small contributions from either AI or C atoms. Peaks B and C in Fig. 3 arise mostly from the hybridization of atomic orbitals of the carbons and the nitrogen in both compounds. The LUMO in both adducts corresponds mainly to an antibonding interaction between
C. Fauquet et al. /Chemical Physics Letters 235 (1995) 528-534
the alane and TMA groups. In this simple picture, pimtodecomposition of TMAA is consistent with electron transfer from the HOMO to the LUMO. The antibonding nature of the AI-N bond leads to a destabilization of the molecule, inducing the trimethylamine group(s) to separate from the alane group. Detailed local spin density calculations have shown that an energy of 1.5 eV is required to break two A1-N bonds . These results togethe," indicate that electron-induced bond breaking may occur. Photoexcitation, for example, of an electron to the first unoccupied molecular orbital should preferentially weaken the bonding of the AI atom, which could then be liberated. This electronic structural feature most likely indicates why it is possible to form relatively pure aluminum films from the TMAA precursor using low energy transfer techniques.
4. Summary and conclusions
The chemical structure of clean, oxygen-free thin films of trimethylamine alane (TMAA) condensed at - 100°C consist of the diamine compound, i.e. 2: 1 adduct. XPS line intensities indicate that the films are in the form of the 2 : 1 adduct. Quantum chemical calculations at the ab initio Hartree-Fock level have been used to study the structure of the 2:1 and of the 1:1 adducts, and the results are in agreement with previous results on these compounds and on comparable molecules. The DOVS calculated from the Hartree-Fock results, when compared with the UPS valence band spectra, also indicate that the electronic structure of the condensed films correspond to the 2:1 adduct. Subsequently, various features on the UPS spectrum can be discussed in detail. The nature of the HOMO and that of the LUMO in TMAA are consistent with simple bond breaking in connection with electron transfer, for example, from optical absorption across the HOMO-LUMO gap.
The authors would like to thank M. Fahlman, M. Boman and J. Rasmusson for helpful discussions arid comments. They would also like to acknowledge
particularly A. Evaldsson and A. Robertson for elaborate technical support. This work was performed within, and supported by, the SCIENCE program (project 0661 POLYSURF). In general, research on organic molecules and polymers in Link6ping is supported by grants from the Swedish Natural Sciences Research Council (NFR), the Swedish National Technical Research Board (TFR), the Swedish National Board for Industrial and Technical Development (NUTEK), the Neste Corp., Finland, Philips Research, NL, the Commission of the European Community, the ESPRIT program (project number 8013 LEDFOS) and the ESPRIT network-of-excellence NEOME.
References  J.H. Burroughes, D.D.C. Bradley, A.R. Brown, R.N. Marks, K. Mackay, R.H. Friend, P.L. Burns and A.B. Hohnes, Nature 347 (1990) 539.  D. Braun and A.J. Heegcr, Appl. Phys. Letters 58 (1991) 1982.  G. Grem, G. Leditzky, B. Ullrich and G. Leising, Advan. Mater. 4 (1992) 36.  G. Gustafsson, Y. Cao, G.M. Treacy, F. Klavetter, N. Colaneri and A.]. Heeger, Nature 357 (1992) 477.  P. Dannetun, M. Boman, S. StafatrSm, W.R. Salaneck, R. Lazzaroni, C. Fredriksson, J.-L. Br,~das, R. Zamboni and C. Taliani, J. Chem. Phys. 99 (1992) 664.  M. 133gdlund, P. Dannetun, S. Stafstfiim, W.R. Salaneek, M.G. Ramsey, C.W. Spangler, C. Fredriksson and J.-L Br~das, Phys. Rcv. Letters 70 (1993)970.  P. Dannetun, M. L6gdlund, C. Fredriksson, M. Boman, S. Stafstr/~m, W.R. Salaneck, B.E. Kohler and C. Spangler, Polymer-solid interfaces, eds. J.J. Pireaux, P. Berlrand and J.-L. Br6das (lOP Publishing, Bristol, 1992) p. 201.  P. Dannetun, M. l_.6gdlund, W.R. Salaneck, C. Fredriksson, S. StafslrSm, A.B. Holmes, A. Brown, S. Graham, R.H. Friend and O. Lhost, Mol. Cryst. Liquid Cryst. 228 (1993) 43.  J. Flicstein and J.E. Bour~e, Photochemical processing of electronic materials (Academic Press, New York, 1991) p. 105.  Th. Bcucrmann and M. Stuke, Appl. Phys. B 49 (1989) 145.  J. Flicstein and J.-E. Bour~e, Appl. Surface Sci. 36 (1989) 443.  T. Cacouris, G. Scelsi, P. Shaw, R. Searmozzino, R.M. Osgood and R.R. Krchnavck, Appl. Phys. Letters 52 (1988) 1865.  D. Labben, T. Motooka, J.E. Greene, J.F. Wendelken, J.-E. Sundgren and W.R. Salancck, MRS 1987, Procced!ngs on Laser and Particle-Beam Chemical Processing for Microelee-
    
C. Fauquet et al. /ChesJzical Physics Letters 235 (1995) 528-534 tronics, eds. D.J. Ehflich, G.S. Higashi and M.M. Oprysko (Elsevier, Amsterdam, 1988) p. 15. C.W. Heitsch, Nature 195 (1962) 995. G.W. Fraser, N.N. Greenwood and B.P. Straughan, .I. Chem. Soc. (1963) 3742. D.T. Clark and H.R. Thomas, J. Polym. Sci. Polym. Chem. Ed. 15 (1977) 2843. J.H. Scofield, J. Electron Spectry Relat. Phenom. 8 (1976) 129. D.G. Hendricker and C.W. Heitsch, J. Phys. Chem. 71 (1967) 2683.
 A. Almenningen, G. Gundersen, T. Haugen and A. Haaland, Acta Chem. Scaad. 26 (1972) 3928.  G.A. Anderson, F.R. Forgaard and A. Haaland, Acta Chem. Scand. 26 (1972) 1947.  V.S. Mastryukov, A.V. Golubinskii and L.V. Vilkov, J. Struct. Chem. 20 (1979) 788.  C.W. Heitsch, C.E. Nordman and R.W. Parry, inorg. Chem. 2 (1963) 508. 1947.  J. Andzelm, Density functional methods in chemistry, eds., J. Labanowski and J. Andzelm (Springer, Berlin, 1991).