5756
J. Phys. Chem. C 2007, 111, 5756-5759
Oxygen Incorporation Mechanism during Atomic Layer Deposition of Al2O3 onto H-Passivated Si(100)-2×1 Zheng Hu and C. Heath Turner* Department of Chemical and Biological Engineering, UniVersity of Alabama, Tuscaloosa, Alabama 35487-0203 ReceiVed: December 18, 2006; In Final Form: February 5, 2007
Using electronic structure calculations, a new oxygen incorporation mechanism is predicted for the atomic layer deposition of Al2O3 from trimethylaluminum (TMA) and H2O. During the TMA exposure, trace amounts of H2O can react with the H/Si(100) surface to form a surface hydroxyl group through a mechanism involving the coadsorbed reactants (TMA and H2O). As compared to the isolated H2O + H/Si(100) reaction, the H2O + TMA + H/Si(100) reaction is strongly preferred both kinetically and thermodynamically. This reaction is relevant to the formation of the SiO2 interfacial oxide layer, which is an unwanted feature produced during the deposition process.
Introduction It is important to develop high κ gate dielectric replacement materials for SiO2, to overcome current leakage problems and maintain the capacitance on smaller devices. Al2O3 is a promising candidate, exhibiting high thermal stability on Si, a large band gap and band offsets, and an abrupt Al2O3/Si interface,1 and hence, has recently received much attention.2-4 An ideal technology for Al2O3 deposition is atomic layer deposition (ALD). In ALD, precursor vapors are alternately pulsed onto the surface of a substrate, with a purge cycle of an inert gas between the precursor pulses. The surface reactions in ALD are complementary and self-limiting, which enables highly uniform deposition and conformal growth, with thickness control at the atomic level. The most common precursors used in ALD of Al2O3 are TMA and H2O. These precursors have high volatility, high reactivity toward each other, high stability against self-decomposition, and chemically inert reaction byproducts (CH4). The ALD reactions occurring during the growth of Al2O3 havebeenexaminedbothexperimentally5,6 andtheoretically.7-10 The ideal growth mechanism involves two alternating exchange reactions between the precursors (TMA and H2O) and the surface functional groups (where surface species are denoted by asterisks)
Al2O3 - OH* + Al(CH3)3 (g) f Al2O3 - O - Al (CH3)2* + CH4 (g) (1) Al2O3 - CH3* + H2O (g) f Al2O3 - OH* + CH4 (g)
(2)
A critical problem during the ALD of Al2O3 onto Si surfaces is the formation of interfacial SiO2, which increases the equivalent oxide thickness (EOT) of the gate stack due to the low permittivity of SiO2. To reduce the thickness of the interfacial SiO2 layer, the H-passivated Si surface is typically used. Unfortunately, it has been found9,11 that the surface hydrogen is inert toward H2O and significantly inhibits the oxidization of the Si substrate. The in situ study conducted by Frank et al.11 found that ALD of Al2O3 onto an H-passivated Si surface was initiated by the reaction of the metal precursor (TMA) with the surface. More importantly, they found unex-
pected oxygen incorporation during the TMA exposure, and this is attributed to trace amounts of H2O left from the previous H2O pulse. Accordingly, Halls and co-workers12 have proposed a possible mechanism of the oxygen incorporation using electronic structure calculations. They concluded that TMA can react with trace amounts of H2O in the gas phase to produce Al(CH3)2OH, which can then react with the H-passivated Si surface, leaving a hydroxyl group on the surface. However, according to a Langmuir adsorption model, the fraction of the surface sites covered with Al(CH3)2OH is proportional to the Al(CH3)2OH partial pressure. Because of the trace amounts of Al(CH3)2OH in the gas phase, these species are unlikely the primary source of the oxygen incorporation. While the previous calculations have predicted one of the possible mechanisms, we suggest that there is another (more dominant) mechanism that may lead to oxygen incorporation onto the Si surface. This new mechanism provides some needed insight into the formation of the interfacial SiO2 layer, which may ultimately lead to improved thin film deposition techniques. Computational Details The single dimer cluster model (Si9H14) shown in Figure 1 was used to represent the reaction site on the H-passivated Si(100)-2×1 surface. The cluster model was generated by truncating a local structure of the surface and terminating the boundary Si atoms with hydrogen. This preserves the electronic structure of the boundary Si atoms and avoids the unphysical electronic effect resulting from the excision of the Si-Si bonds. To avoid over-relaxation of the cluster model and to mimic the constraints imposed by an extended surface, the Si atoms in the two bottom layers were fixed at their bulk positions. The H atoms that represent the third and lower layer bulk Si atoms were also fixed in tetrahedral directions, with the Si-H distance of 1.48 Å. This single dimer cluster model has been successfully used to represent the Si(100)-2×1 surface in previous studies of ALD growth mechanisms and was found to render consistent results with larger cluster models.9,13,14 The potential energy surfaces (PESs) for the reacting systems were constructed with electronic structure calculations using Gaussian03.15 The equilibrium structures and transition state
10.1021/jp068696r CCC: $37.00 © 2007 American Chemical Society Published on Web 03/24/2007
Deposition of Al2O3 onto H-Passivated Si(100)-2×1
Figure 1. Stationary points for the TMA + H2O + H/Si(100) reaction calculated using B3LYP/BSI: (a) co-adsorption of TMA with H2O on the H-terminated Si(100)-2×1 surface and (b) transition state. Distances are shown in angstroms. Terminated H atoms representing the bulk Si atoms are not shown explicitly for clarity in the figure.
structures along the reaction pathways were computed using gradient-corrected density functional theory with the Becke three-parameter exchange functional16 and the Lee-Yang-Parr correlation functional (B3LYP).17 The optimized geometries were predicted using a hybrid basis set scheme. The atoms in the top layers of the H/Si(100) substrate (which were fully relaxed) and the precursor molecules are the crucial region for the surface reactions and, hence, were treated with a polarized 6-31+G(d,p) double-ζ basis set with a diffuse function added. The remaining frozen atoms were approximated at the less expensive 6-31G basis set. This hybrid basis set scheme is denoted as BSI. The nature of the stationary points on the PES was identified by subsequent frequency calculations, which also yielded zero point energy (ZPE) corrections. Intrinsic reaction coordinate (IRC) calculations were performed to ensure that the transition states connected reactants to products. To obtain an accurate prediction of the energetics of the reactions, single point energy calculations were carried out using high level model chemistry with the geometries predicted from the B3LYP/BSI calculations. The atoms treated with the 6-31G basis set during geometry optimization were expanded using the 6-311++G(2df,2p) basis set during the single point energy calculations, whereas the frozen atoms were still approximated with the 6-31G basis set. This hybrid basis set scheme is denoted as BSII. Heyman and Musgrave have shown that B3LYP is very accurate for predicting reaction enthalpies during the ALD of Al2O3.7 Also, for a similar ALD reaction,18 it was found that, based on the agreement with high level CCSD(T) results, second-order Møller-Plesset perturbation theory (MP2) performs quite well for predicting the heats of adsorption and the activation barriers and that B3LYP can give an accurate prediction of the reaction enthalpies. Therefore, in this work, the heats of adsorption and the activation barriers were calculated with MP2/BSII, and the reaction enthalpies were calculated with B3LYP/BSII. Results and Discussion A previous in situ infrared spectroscopic study has shown unexpected O incorporation during the exposure of an Hterminated Si surface to TMA.11 The oxygen must come from trace amounts of H2O (left from the previous H2O exposure) since this is the only O source in the ALD process. However, both experimental and theoretical studies9,11 have demonstrated that the H-passivated Si surface is very inert toward H2O. The H-passivated Si surface remains nearly unchanged after H2O
J. Phys. Chem. C, Vol. 111, No. 15, 2007 5757 exposure, and therefore, there is not an intuitive explanation for the observed O incorporation. Although a N2 purge cycle is performed between precursor pulses, trace amounts of H2O may remain adsorbed on the surface. Subsequently, when a vapor dose of TMA precursor is then injected, TMA molecules can adsorb on the surface alone or with H2O coadsorbed. Therefore, we have explored an oxygen incorporation mechanism, which is initiated with the coadsorption of TMA and H2O on the H-terminated Si surface. The calculated co-adsorption geometry of this complex is illustrated in Figure 1a, with a 0 K heat of adsorption calculated to be -11.0 kcal/mol. With the coadsorbed TMA and H2O, the aqueous O can attack the Si to form a surface hydroxyl group with the assistance of TMA. The transition state structure with interatomic distances is shown in Figure 1b. In the transition state, the O-SiI, Al-HI, and C-HII bonds are partially formed with the partial breakage of the SiI-HI, Al-C, and O-HII bonds. After passing through the transition state, the reacting system generates a surface hydroxyl group, accompanied by the release of HAl(CH3)2 and CH4. This reaction features simultaneous movement of three species, which leads to only minor structural distortion in the transition state and, hence, a relatively low activation barrier (27.0 kcal/mol). The reaction is exothermic by 17.4 kcal/mol at 0 K. To evaluate the relative probability of this oxygen incorporation mechanism, we compared the TMA + H2O + H/Si reaction with the H2O + H/Si and TMA + H/Si reactions, based on the same computation scheme. The energy profiles of these three reactions are compared in Figure 2. Our predictions of the adsorption energies and activation barriers of the H2O + H/Si and TMA + H/Si reactions differ somewhat from the previous calculations of Halls and Raghavachari,9 due to the different computational methods used. In Halls and Raghavachari’s work, DFT with the B3LYP functional is used for all of the calculations. Just for comparison, we also used DFT with the B3LYP functional to study the TMA + H/Si reaction, and we obtained identical results (not shown). As shown in Figure 2, the TMA + H2O + H/Si and TMA + H/Si reactions have similar reaction barriers, indicating similar intrinsic kinetics. However, the TMA + H2O + H/Si reaction is thermodynamically more favored, with a reaction enthalpy of -17.4 kcal/ mol. This is 10.1 kcal/mol more exothermic than the TMA + H/Si reactions. Both TMA + H2O + H/Si and H2O + H/Si reactions generate a hydroxyl group on the surface. However, as shown in Figure 2, the activation barrier of the H2O + H/Si reaction (with respect to the adsorption complex) is 42.1 kcal/mol. This high activation barrier prohibits direct reaction of the H2O with H/Si and, consequently, leads to a relatively unchanged H/Si surface after H2O exposure.11 However, the additional presence of TMA changes the hydroxylation reaction mechanism and reduces the activation barrier to 27.0 kcal/mol. Therefore, coexistence of TMA and H2O greatly enhances the surface hydroxylation rate. The resulting surface hydroxyl group (Si-OH), which is very reactive toward the HAl(CH3)2 product and TMA in the gas phase,8 can undergo further reaction to form a Si-O-Al bond, by releasing a CH4 molecule. This result is consistent with the in situ infrared spectroscopic measurement, which is observed in the Al-O and Si-O-Al bands during TMA exposure.11 This mechanism is expected to contribute significantly to the formation of the interfacial oxide layer. To compare another possible mechanism of the oxygen incorporation proposed by Halls and co-workers,12 we performed calculations for the Al(CH3)2OH + H/Si f Al(CH3)2H + Si-
5758 J. Phys. Chem. C, Vol. 111, No. 15, 2007
Hu and Turner
Figure 2. ZPE-corrected PES for the Al(CH3)2OH + H/Si (I), TMA + H/Si (II), H2O + H/Si (III), and TMA + H2O + H/Si reactions (IV). IR ) Isolated reactants; RC ) reactant-like adsorption complex; TS ) transition state; PC ) product-like adsorption complex; and IP ) isolated products.
OH reaction using our current computational methodology. While the methods are slightly different, our calculations and the DFT calculations of Halls and co-workers are generally consistent. Our calculations and those of Halls et al. (shown in parenthesis) are -8.1 (-2.8) kcal/mol for adsorption, an activation barrier of 17.2 (13.8) kcal/mol, a product complex of -3.6 (3.0) kcal/mol, and the separated products at 18.4 (17.9) kcal/mol. The main source of discrepancy among the values is due to the treatment of the adsorption energy. In our calculations, we use the MP2 method to evaluate the heats of adsorption and the activation barrier. This method incorporates dispersion interactions, which typically lead to stronger adsorption and a more reliable estimate for evaluating interactions with a surface. Since we used DFT to calculate the overall reaction enthalpy, our value is very close to the previous estimate, as expected. In our calculations, we found that the Al(CH3)2OH + H/Si reaction is more kinetically favored with an activation barrier 9.8 kcal/mol smaller (17.2 vs 27.0 kcal/mol) than that of the TMA + H2O + H/Si reaction. However, although the Al(CH3)2OH + H/Si reaction is kinetically favored, we still believe that the TMA + H2O + H/Si reaction is an important mechanism of oxygen incorporation based on the following two reasons: (1) the overall ALD dynamics cannot be determined solely by the rate of the surface reaction. For a multistep LangmuirHinshelwood reaction, the rate-determining step (RDS) can be any one of the events (adsorption, desorption, surface diffusion, or surface reaction) depending upon reaction conditions. If the partial pressure of Al(CH3)2OH (the gas phase partial hydroxylation product formed from trace amounts of H2O in the gas phase) is low, according to a Langmuir adsorption model, the adsorption rate of Al(CH3)2OH is expected to be very small and hence to be the RDS. As a result, the probability of oxygen incorporation through the Al(CH3)2OH + H/Si reaction is greatly reduced by the very low probability of adsorption (since the gas phase is purged between each ALD cycle), although the surface reaction step is relatively fast. On the contrary, during a vapor dose of the TMA precursor, with trace amounts of H2O remaining adsorbed on the surface, the co-adsorption of TMA
and H2O on the H-terminated Si surface can be easily obtained and hence initiate the TMA + H2O + H/Si reaction. (2) Although the Al(CH3)2OH + H/Si reaction is kinetically favored, it is strongly endothermic by 18.4 kcal/mol. On the contrary, the TMA + H2O + H/Si reaction is the most thermodynamically favored reaction with a reaction enthalpy of -17.4 kcal/mol, which is 35.8 kcal/mol more exothermic than the Al(CH3)2OH + H/Si reaction. The large exothermicity suggests that oxygen incorporation through the TMA + H2O + H/Si reaction can easily proceed toward completion. Conclusion In conclusion, a new oxygen incorporation mechanism during the exposure of the H/Si(100) surface to TMA during ALD processing has been predicted using electronic structure calculations. With the assistance of TMA, H2O can react with H/Si(100) to form a hydroxyl group on the surface (Si-OH). Although the direct reaction of H2O and H/Si is difficult at typical ALD temperatures, the presence of TMA lowers the activation barrier significantly and enhances surface hydroxylation. In addition, among the four reactions considered in this work, the TMA + H2O + H/Si reaction is the most thermodynamically favored reaction and is kinetically competitive. The large thermodynamic driving force suggests that the surface hydroxylation reactions can easily proceed toward completion. Since the surface hydroxyl group is very reactive, it can quickly react further with HAl(CH3)2 or TMA to form a Si-O-Al bond. As a result, our work shows that during TMA exposure, trace amounts of adsorbed H2O can lead to oxygen incorporation onto the Si surface and that this mechanism may contribute to the formation of the interfacial oxide layer. Acknowledgment. The authors thank Prof. Tonya Klein for helpful discussions of the experimental systems and the University of Alabama Graduate Council Fellowship for financial support. The computing resources were partially supported by the NSF Teragrid (Project TG-CTS060011T) and the Alabama Supercomputer Center.
Deposition of Al2O3 onto H-Passivated Si(100)-2×1 References and Notes (1) Gusev, E. P.; Copel, M.; Cartier, E.; Baumvol, I. J. R.; Krug, C.; Gribelyuk, M. A. Appl. Phys. Lett. 2000, 76, 176. (2) Manchanda, L.; Morris, M. D.; Green, M. L.; van Dover, R. B.; Klemens, F.; Sorsch, T. W.; Silverman, P. J.; Wilk, G.; Busch, B.; Aravamudhan, S. Microelectron. Eng. 2001, 59, 351. (3) Ritala, M.; Kukli, K.; Rahtu, A.; Raisanen, P. I.; Leskela, M.; Sajavaara, T.; Keinonen, J. Science 2000, 288, 319. (4) Higashi, G. S.; Fleming, C. G. Appl. Phys. Lett. 1989, 55, 1963. (5) Rahtu, A.; Alaranta, T.; Ritala, M. Langmuir 2001, 17, 6506. (6) Yates, D. J. C.; Dembinski, G. W.; Kroll, W. R.; Elliott, J. J. J. Phys. Chem. 1969, 73, 911. (7) Heyman, A.; Musgrave, C. B. J. Phys. Chem. B 2004, 108, 5718. (8) Xu, Y.; Musgrave, C. B. Chem. Mater. 2004, 16, 646. (9) Halls, M. D.; Raghavachari, K. J. Chem. Phys. 2003, 118, 10221. (10) Halls, M. D.; Raghavachari, K. J. Phys. Chem. A 2004, 108, 2982. (11) Frank, M. M.; Chabal, Y. J.; Wilk, G. D. Appl. Phys. Lett. 2003, 82, 4758. (12) Halls, M. D.; Raghavachari, K.; Frank, M. M.; Chabal, Y. J. Phys. ReV. B 2003, 68, 161302.
J. Phys. Chem. C, Vol. 111, No. 15, 2007 5759 (13) Widjaja, Y.; Han, J. H.; Musgrave, C. B. J. Phys. Chem. B 2003, 107, 9319. (14) Haran, M.; Engstrom, J. R.; Clancy, P. J. Am. Chem. Soc. 2005, 128, 836. (15) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A.; Stratmann, R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, M.; Cammi, R.; Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.; Ochterski, J.; Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz, J. V.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Gonzalez, C.; Challacomba, M.; Gill, P. M. W.; Johnson, B. G.; Chen, W.; Wong, M. W.; Andres, J. L.; Head-Gordon, M.; Replogle, E. S.; Pople, J. A. Gaussian 03; Gaussian, Inc.: Pittsburgh, PA, 2003. (16) Becke, A. D. J. Chem. Phys. 1993, 98, 5648. (17) Lee, C. T.; Yang, W. T.; Parr, R. G. Phys. ReV. B 1988, 37, 785. (18) Hu, Z.; Turner, C. H. J. Am. Chem. Soc. 2007, accepted.