«Mechanism-Based Design of Precursors for MOCVD Lisa McElwee-White,a,* Jürgen Koller,a Dojun Kim,b and Timothy J. Andersonb a Department of ...»
ECS Transactions, 25 (8) 161-171 (2009)
10.1149/1.3207587 © The Electrochemical Society
Mechanism-Based Design of Precursors for MOCVD
Lisa McElwee-White,a,* Jürgen Koller,a Dojun Kim,b and Timothy J. Andersonb
Department of Chemistry, University of Florida, Gainesville, Florida 32611 USA
Department of Chemical Engineering, University of Florida, Gainesville, Florida
A chemistry-based approach to designing precursors for the
deposition of inorganic films requires consideration of the physical properties of the precursor compound (e.g., volatility for transport in the reactor) and its probable decomposition pathways, both in the gas phase and on the surface during growth. We have been using Aerosol-Assisted Chemical Vapor Deposition of tungsten carbonitride (WNxCy) films from tungsten imido complexes and tungsten hydrazido complexes as a model system to investigate the relationship between data obtained from conventional organometallic mechanistic study of the precursors and the resulting film deposition kinetics and film properties. Among the typical techniques for the elucidation of decomposition pathways in organometallic compounds that will be discussed in this context are NMR kinetics, mass spectrometry, DFT calculations, and single crystal X-ray diffractometry.
Introduction Deposition of thin films of inorganic materials has been of great recent interest, in part due to technological applications in the electronics industry. The surface reactions that result in film formation during chemical vapor deposition (CVD) and the related technique atomic layer deposition (ALD) provide opportunities for growth of highly conformal thin layers of material on patterned substrates, an advantage of chemical deposition methods as compared to physical deposition methods such as sputtering.
Careful choice of reaction chemistry could also be used to control composition and properties of the resulting films (1). These considerations have led to recent efforts to design precursors using chemical strategies (2-8).
Our approach to precursor design is a mechanism-based strategy in which known pathways for thermal decomposition of organometallic complexes have been used to predict the reactivity of precursors during CVD. The challenge is the difference between reaction conditions used in the majority of mechanistic studies (complexes in solution at or below room temperature) and the conditions used in CVD (gas/surface reaction at temperatures 300 °C). Under CVD conditions, not only are the typical solution-phase organometallic reaction pathways faster but new reaction manifolds can become accessible (9). An additional challenge in precursor design is the need to transport the precursor to the substrate surface in the vapor phase. In the case of conventional CVD reactors that utilize bubblers or sublimators for volatilization of the precursor, attention to the relationship between molecular structure and the volatility of the complexes is required.
Downloaded 26 Oct 2009 to 188.8.131.52. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp ECS Transactions, 25 (8) 161-171 (2009) Because of the importance of transition metal nitrides as diffusion barrier materials to prevent migration of Al or Cu metallization into the Si device layer or the dielectric layers of integrated circuits (7), we have chosen deposition of metal nitride and the related carbonitride films for testing of our design strategies. We have recently been using Aerosol-Assisted Chemical Vapor Deposition (AACVD) of tungsten carbonitride (WNxCy) films from the tungsten imido complexes WCl4(NR)(CH3CN) [1: R = Ph; 2: R = iPr; 3: R = allyl] (10-12) and the tungsten hydrazido complexes WCl4(NNR2)(CH3CN) [4: R2 = Me2; 5: Ph2; 6: R = -(CH2)5-] (13, 14) as a model system to investigate the relationship between data obtained from conventional organometallic mechanistic study and the resulting film deposition kinetics and film properties.
Among the standard techniques for the elucidation of decomposition pathways in organometallic compounds that we have applied in this context are NMR kinetics, mass spectrometry, DFT calculations, and small molecule structure determination by single crystal X-ray diffractometry. Information on the precursor chemistry found through these means can be correlated to film composition (AES, XRD), film growth kinetics (X-SEM), bonding in the films (XPS), and electrical resistivity (four-point probe). Selected examples will be discussed below.
Precursor design. Our choice of the tungsten imido complexes WCl4(NR)(CH3CN) (1-3) and the tungsten hydrazido complexes WCl4(NNR2)(CH3CN) (4-6) as precursors
for the deposition of WNxCy films shares some common features with the selection of the dimeric complexes [MCl2(NtBu)(NHtBu)(NH2tBu)]2 (M = Nb, Ta) (15) as CVD precursors for films of NbN and Ta3N5, respectively, and the use of TiCl2(NtBu)(py)3 to deposit TiN (16, 17). All of these precursors contain combinations of an imido group, chlorides and other N-bound ligands (amines, amides and nitriles). In each case, the imido ligand provides a strong M-N multiple bond, which is likely to survive decomposition of the complex during deposition and provide a building block for the Downloaded 26 Oct 2009 to 184.108.40.206. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp ECS Transactions, 25 (8) 161-171 (2009) assembly of metal nitrides on the substrate surface. In our complexes 1-6, the strong trans effect of the imido or hydrazido ligand is expected to facilitate dissociation of the nitrile ligand, which is trans to the imido or hydrazido moiety in all of these compounds.
Our synthetic routes to complexes 1-6 allow for considerable flexibility in the imido and hydrazido substituents, which in turn allows manipulation of bond dissociation energies within those moieties. During film deposition, the chloride ligands are removed from precursors 1-6 as HCl by addition of H2 or NH3 to the carrier gas (9, 18-20). Although the imido and hydrazido complexes 1-6 are non-volatile solids, they can be volatilized for AACVD by nebulizing a solution of the precursor in benzonitrile (11).
Mass Spectrometry. In evaluating our precursor designs and screening precursors before deposition experiments, we utilize the correlation between mass spectrometric fragmentation patterns and likely decomposition pathways during deposition (21, 22).
Since mass spectrometry involves gas phase ions while CVD is a heterogeneous thermal process, over-interpretation should be avoided but we have found that fragmentation patterns can be useful qualitative predictors of CVD behavior. As an example, the mass spectra of the series WCl4(NR)(CH3CN) (1-3) exhibit significant differences that are reflected in the behavior of the complexes during CVD.
One critical piece of information involves the facility of N-C bond cleavage in the imido moiety. The tendency for cleavage of the N(imido)-C bond can be discerned in the negative ion chemical ionization mass spectrometry (NCI-MS) data for 1-3 (Table 1) through the abundance of the [WCl4N]- ion, in which the alkyl substituent has already dissociated to leave a tungsten nitrido fragment. For complexes 2 (R = iPr) and 3 (R = allyl) this nitrido ion gives rise to the base peak of the NCI mass spectrum. The behavior of the phenylimido complex 1 is very different in that the [WCl4N]- fragment exhibits only 4% of the integrated intensity of the mass envelope for its base peak [WCl4(NPh)]-.
These data suggest difficulty in cleavage of the N-C(imido) bond of phenylimido complex 1, which will result in corresponding problems with growth of WNxCy. Further insight can be gained from examination of the positive ion EI spectrum of 1, in which observation of [WCl4]+ and [WCl3]+ fragments implies preferential cleavage of the W-N bond instead of cleavage of the N-C(imido) bond (11). W-N cleavage results in loss of the intact NPh ligand during film growth and thus low N levels in the resulting films.
This qualitative information on trends for preferred bond cleavage in the imido complexes 1-3 is consistent with the nitrogen content of films grown from the three precursors. For low temperature film growth, precursors which exhibit facile N-C bond dissociation in the NCI-MS (2 and 3) afford films with higher nitrogen levels than those grown from 1, in which W-N dissociation competes well with N-C dissociation during precursor decomposition (Figure 2).
Figure 2. Comparison of nitrogen content in the films grown from 1-3.
Figure reproduced with permission from reference (9). Copyright 2006, American Chemical Society.
Information from the mass spectra can also be correlated to changes in the apparent activation energies for film growth from precursors 1-3. The integrated intensities of the peaks for the [WCl4N]- ion rise as the imido substituent changes from phenyl to isopropyl to allyl (1 3). There is a parallel decrease in the bond dissociation energy (BDE) of the N C bond in the imido moiety, as modeled by N-C BDEs of the corresponding primary amines RNH2. From the amine data, the N C bond of allylimido complex 3 should be approximately 11 kcal/mol weaker than that of isopropylimido complex 2 and 32 kcal/mol weaker than that of phenylimido complex 1 (23). The relative magnitudes of the apparent activation energy Ea for film deposition from 1-3 follow a trend that is consistent with both the intensity of the [WCl4N]- ion in the mass spectra and with the strength of the imido N C bond, as predicted from the organic model compounds. The linear relationship between the estimated N-C BDE and Ea (Figure 3) is consistent with cleavage of the N C imido bond before or during the rate-determining step for film
growth. Thus, the strength of the N-C bond in the imido moiety is critical in two regards:
1) partitioning between W-N and N-C cleavage, which affects the film composition and
2) determination of the apparent activation energy, which determines the rate of film growth. This type of information can then be used in design of additional precursors.
Downloaded 26 Oct 2009 to 220.127.116.11. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp ECS Transactions, 25 (8) 161-171 (2009) Figure 3. Variation of apparent activation energy (Ea) for film growth from 1-3 with the N C bond energies of the corresponding amines R-NH2 as models for the imido N-C bonds. Figure reproduced with permission from reference (12). Copyright 2005, American Chemical Society.
NMR kinetics. One of the design features of precursors 1-6 was a facile initial dissociation of the CH3CN ligand during film growth, which was postulated on the basis of the strong trans effect of the imido/hydrazido ligand in the dissociative reactions of imido complexes (24). This reaction pathway is consistent with the mass spectra of 1-3, in which molecular ions cannot be observed and the highest m/z values in both positive and negative modes correspond to ions from which CH3CN has been lost (10-12).
To obtain experimental values for the activation energy of CH3CN dissociation, complex 2 was chosen as a representative case. 1H NMR kinetics were used to study the exchange of the acetonitrile ligand of 2 with free CH3CN in CDCl3 solution. Both bound and free acetonitrile could be detected in the 1H NMR spectra obtained at -20˚C and the two signals coalesced with increasing temperature. The exchange rate k for the exchange of acetonitrile was determined by line-shape analysis in the temperature range -6 to 34 °C.
The activation energy of 18.52 ± 0.14 kcal/mol and an entropy of activation of 15.8 ± 0.5 cal/mol•K were obtained from a plot of ln(k/T) vs. 1/T (Figure 4) and correspond to H‡ and S‡ for dissociative loss of CH3CN from isopropylimido complex 2. Since this process can be easily followed at 34 °C, loss of acetonitrile from 1-6 must be kinetically facile under CVD conditions, which involve temperatures above 300 °C.
-4 0.0032 0.0033 0.0034 0.0035 0.0036 0.0037 0.0038
Figure 4. Plot of ln(k/T) vs.
1/T(K) for acetonitrile exchange in complex 2. Figure reproduced with permission from reference (9). Copyright 2006, American Chemical Society.
X-Ray Crystallography. The tungsten hydrazido complexes 4-6 were prepared to provide an alternative to addition of hydrazine derivatives into the carrier gas during deposition of metal nitride films, a process that has been reported to lower the deposition temperature of TiN films (25-27). Further support for this strategy came from the report that hydrazido complexes were viable single-source precursors for the deposition of TiN thin films while maintaining the positive effects of hydrazine as a co-reactant (28). The observation that the apparent activation energy for deposition of WNxCy depended on the N(imido)-C bond dissociation energy raised an interesting question about the hydrazido complexes, for which there are two limiting resonance structures (Figure 5). The bond lengths obtained for 4-6 by single crystal X-ray crystallography were consistent with structure A being the major resonance contributor (13, 29). Although the solid state Figure 5. Limiting resonance structures A and B for hydrazido(2-) complexes of W(VI) and structure of the analogous imido complexes.
Downloaded 26 Oct 2009 to 18.104.22.168. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp ECS Transactions, 25 (8) 161-171 (2009) structures of 4-6 suggest partial double bonding character between the two nitrogens, 4-6 are viable precursors for the deposition of WNxCy, with deposition possible at temperatures as low as 300 °C (14, 30).