Chemical Physics Letters 457 (2008) 69–73

Contents lists available at ScienceDirect

Chemical Physics Letters journal homepage: www.elsevier.com/locate/cplett

Initial adsorption mechanisms of TiCl4 on OH/Si(1 0 0)-2  1 Manik Kumer Ghosh, Cheol Ho Choi * Department of Chemistry, College of Natural Sciences, Kyungpook National University, Taegu 702-701, South Korea

a r t i c l e

i n f o

Article history: Received 9 January 2008 In final form 21 March 2008 Available online 29 March 2008

a b s t r a c t The initial adsorption mechanisms of TiCl4 on OH/Si(1 0 0)-2  1 surface were theoretically investigated with the help of ab initio theories. Four reaction channels were identified. The penta-coordinated Ti, which effectively blocks the surface adsorption sites, plays a significant role in all processes. Ó 2008 Elsevier B.V. All rights reserved.

1. Introduction Continued miniaturization in according with ‘‘Moore’s Law” will render SiO2 unusable as a gate oxide layer with in a decade due to the quantum tunneling effect [1]. For oxide thickness below 20 Å, tunneling current through the SiO2 gate dielectric becomes substantial [2,3]. Therefore, the research for alternative materials to replace SiO2 has been critical in recent years [3]. The most promising alternative dielectric materials are the high-j binary metal oxides. If a higher dielectric constant material, for example Al2O3, TiO2, HfO2, or ZrO2, is used as the gate dielectric, the tunneling problem can be avoided [4]. Atomic layer deposition (ALD) is a promising deposition technique for gate oxides because exceptionally uniform and conformal ultra-thin films with precisely controlled thickness can be obtained [5]. Various titanium chloride and alkoxide precursors, such as, TiCl4 [6], TiI4 [7], Ti(OMe)4 [8], Ti(OEt)4 [9,10], Ti(OiPr)4 [11–13] have been employed with oxygen sources, O2, O3, H2O, H2O2 etc. for the growth of TiO2 thin films of high quality in both ALD and CVD (chemical vapour deposition) processes [14]. Theoretical growth rate is one monolayer per deposition cycle in ALD process. It is generally significantly lower then the expectation especially when the ligands of precursor are bulky. However, in the case of TiCl4 ALD process, typical growth rates have ranged from 0.04 to 0.07 nm per cycle, which is smaller than those with bulky ligand containing titanium. The low growth rate of TiCl4 precursor may be due to the direct chlorination [15–18]. Haukka et al. [16–18] also found that the reaction of TiCl4 with silica resulted in a Cl/Ti ratio above 3 and concluded that direct chlorination occurred to account for the high Cl/Ti ratio. They further reported that TiCl4 molecule reacts with one to two surface hydroxyl groups and as a result up to two chlorine ligands are released from an adsorbed TiCl4. The readsorption possibility of the HCl released during chemisorptions of a metal chloride was also discussed [18,19], which * Corresponding author. Fax: +82 53 950 6330. E-mail address: [email protected] (C.H. Choi). 0009-2614/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.cplett.2008.03.053

occupies adsorption sites preventing the growth of metal oxide because of self–site–blocking effects in adsorption on a semiconductor. Experimental mechanism study [20] of TiCl4 on Si(1 0 0) and on amorphous silica showed that the growth rate depended on the temperature and film structure. The low growth rate was found to be a result of significant chlorine adsorbed during TiCl4 pulse. The desorption and decomposition of the initial surface intermediates also influenced the deposition rate. Recently, Hu and Turner studied the initial surface reactions involved in the atomic layer deposition (ALD) of TiO2 from TiCl4 [21] and TiI4 [22] using electronic structure calculations based on cluster models. They showed that the effects of tunneling were found to be negligible for all reactions. However, the rotational contributions to the rate constants must be taken into account in certain cases. Lee et al. [23] devised a novel procedure that can create hydroxyl-only terminated Si surface, which would provide a well-defined model environment for the study of TiCl4 chemisorptions. In this Letter, the reaction mechanism of TiCl4 on OH/Si(1 0 0)2  1 surface were studied with the help of quantum mechanical methods to elucidate the repeatedly suggested secondary surface reactions and their consequences. 2. Computational details The second order Moller–Plesset Perturbation theory (MP2) were adopted in combination with cluster models of the Si(1 0 0)2  1 surface. All calculations reported here were performed with the GAMESS [24] (general atomic and molecular electronic structure system) electronic structure program. All stationary points were obtained with 6-31G(d) basis sets. In addition, single point energy calculations were performed with the 6-311G(2df,2p) [25] basis sets at the stationary points optimized with 6-31G(d) basis set. Minimum energy reaction paths were determined by first optimizing the geometries of the minimum and the transition states. To follow the minimum energy path (MEP), also called intrinsic reaction coordinate (IRC), the Gonzalez-Schlegel second-order method [26] was used. To characterize each stationary point, the

70

M.K. Ghosh, C.H. Choi / Chemical Physics Letters 457 (2008) 69–73

Chart 1. Stationary points along the titanium chloride adsorption on OH/Si(1 0 0)-2  1 surface as calculated with SIMOMM:MP2/6-311G(2df,2p). The corresponding values in parentheses and in braket are obtained with SIMOMM:MP2/6-31G(d) with and without zero point energy corrections, respectively. Geometric data are from the MP2/631G(d). The bond distances are in Angstrom and the energies are in kcal/mol.

Chart 2. Stationary points along the released HCl re-adsorption to the mixed Si(1 0 0)-2  1 surface as calculated with SIMOMM:MP2/6-311G(2df,2p). The corresponding values in parentheses and in braket are obtained with SIMOMM:MP2/6-31G(d) with and without zero point energy corrections, respectively. Geometric data are from the MP2/6-31G(d). The bond distances are in Angstrom and the energies are in kcal/mol.

M.K. Ghosh, C.H. Choi / Chemical Physics Letters 457 (2008) 69–73

Hessian (matrix of energy second derivatives) was calculated and diagonalized at each stationary point which also yielded zero point energy corrections. In order to study surface size effects, a hybrid quantum mechanics/molecular mechanics (QM/MM) method called SIMOMM (surface integrated molecular orbital molecular mechanics) was adopted [27]. In this work, the QMMM models were designed such that TiCl4O2Si9H14 and TiCl3O2Si9H13 quantum regions are embedded in TiCl4O6Si48H42 and TiCl3O6Si48H41 clusters for titanium chloride studies, respectively. These models have one and three surface Si dimer(s) in the QM and MM regions, respectively. MM3 [28–30] parameters were used for the molecular mechanics part of the computations. 3. Results and discussion Four reaction channels that are relevant to the initial reactions of TiCl4 were identified. They shall be discussed in order. The very first reaction starts with the physisorption of TiCl4 as described in the next paragraph. Note that the energy values in the Charts were obtained by the single point energy calculations using MP2/6311G(2df,2p) basis sets. The corresponding numbers in parenthesis and in brakets are MP2/6-31G(d) with and without zero point energy corrections, respectively. 3.1. H–Cl desorption path Upon approaching the OH/Si(1 0 0)-2  1 surface, titanium chloride initially forms a penta-coordinated intermediate I1 without a reaction barrier (see Chart 1). The initial intermediate is exothermic by 17.1 kcal/mol, which is similar with the corresponding species between trimethylaluminum(TMA) and OH/Si(1 0 0) [31]. The lone pair electrons of O9 interact with empty d orbitals of Ti1 resulting a bonding interaction of Ti1–O9 (2.202 Å). The transition state

71

TS1 connects the initial intermediate I1 and another intermediate I2. TS1 is more stable than the reactants by 2.9 kcal/mol. The net forward reaction barrier of TS1 is 14.2 kcal/mol, which is 2.2 kca/mol lower than the backward barrier. As compared to the surface Si–O–Al bond formation using TMA on OH/Si(1 0 0) surface, the Si–O–Ti bond forming barrier is 5.5 kcal/mol higher showing that TiO2 film growth requires a little more reaction energies. In TS1, Ti1 is making a bond with O9 while H7 migrates from O9 to Cl2. The generated HCl is trapped as shown in intermediate I2, where the H7 is hydrogen bonded to surface O8 and the leaving Cl2 is still interacting with Ti1, making it quite stable intermediate. Endothermic energy of 9.1 kcal/mol is necessary for the trapped HCl to be released as shown in P1. It is interesting that the extra interaction between O8–Ti1 exists in P1. Both the trapped HCl in I2 and the extra O8-Ti1 interaction in P1 occupy adsorption site that can reduce the growth rate. Overall, initial TiCl4 deposition on OH/ Si(1 0 0) are both kinetically and thermodynamically favorable. 3.2. H–Cl re-adsorption path Experimentally suggested re-adsorption path were studied (see Chart 2). It starts from the intermediate I2 of the previous channel. By Cl2 of HCl nuclophilically attacking surface Si9 while forming a O8–H7 bond, the transition state TS2 connects the intermediate I2 and the H–Cl re-adsorbed intermediate I3. It is also noted that the Cl2 is also making a weak interaction with the already adsorbed Ti1 atom making a penta-coordinated Ti in TS2. Therefore, it cooperatively stabilizes this transition state assisting H–Cl re-adsorption channel. The net forward reaction barrier of TS2 is 16.9 kcal/mol, which is 2.7 kcal/mol higher than that of TS1. But, the overall relative energy of TS2 is 2.4 kcal/mol which is only 0.5 kcal/mol higher than the initial transition state TS1. The resulted intermediate I3 is more stable then I2 by 4.5 kcal/mol. And, the resulted product, P2 is more stable than the product P1 of earlier channel by 6.2 kcal/

Chart 3. Stationary points along the direct chlorination reaction of titanium chloride on OH/Si(1 0 0)-2  1 surface as calculated with SIMOMM: MP2/6-311 G(2df,2p). The corresponding values in parentheses and in braket are obtained with SIMOMM:MP2/6-31 G(d) with and without zero point energy corrections, respectively. Geometric data are from the MP2/6-31 G(d). The bond distances are in Angstrom and the energies are in kcal/mol.

72

M.K. Ghosh, C.H. Choi / Chemical Physics Letters 457 (2008) 69–73

Chart 4. Stationary points along the ring closing reaction of titanium chloride on mixed Si(1 0 0)-2  1 surface as calculated with SIMOMM: MP2/6-311 G(2df,2p). The corresponding values in parentheses and in braket are obtained with SIMOMM:MP2/6-31G(d) with and without zero point energy corrections, respectively. Geometric data are from the MP2/6-31G(d). The bond distances are in Angstrom and the energies are in kcal/mol.

mol. Therefore, the re-adsorption channel has a similar reaction barrier and it produces much more stable surface product, making the reaction thermodynamically and kinetically favored and justifying the possibilities of HCl re-adsorption channel. In this reaction, H2O forms as a byproduct. Previously, Haukka et al. [18] reported that water must be formed during the reaction and it is due either to the HCl re-adsorption or direct chlorination. According to our calculation, we propose that water is formed during HCl re-adsorption reaction. 3.3. Direct Cl adsorption path The chlorine can also directly adsorb to the surface as shown in Chart 3. I1 is the starting point of this reaction. In stead of the Cl2 abstracting H7 as in TS1, it can interact with surface Si11 directly as in TS3 producing a new intermediate I4. The net forward reaction barrier of TS3 is calculated to be 19.0 kcal/mol, which is 4.8 kcal/mol higher than that of TS1. However, the overall relative energy of TS3 is calculated to be 1.9 kcal/mol, which is slightly higher than the other previous channels. The resulted intermediate I4 is exothermic by 11.3 kcal/mol. By releasing TiCl3OH, the final product P3 is formed. Although, the overall reaction enthalpy is higher than the previous two channels by 9.9 and 16.1 kcal/mol, the overall reaction barrier is almost the same allowing this channel to be accessible. 3.4. The second H–Cl desorption path A bidentate Ti species was suggested by Finnie et al. [32]. The possibility of bidentate Ti was studied and the results are presented in Chart 4. In order to study this part of potential energy surface, a new reactant R2 was introduced where the first desorbed

H–Cl which was generated during the initial reaction is removed. All the relative energies are calculated with respect to R2. The ring closing transition state TS4 connects the reactant R2 and an intermediate I5. During the reaction, Ti1 attacks O8 while Cl5 abstracts H6 from the surface with an energy barrier of 16.8 kcal/mol. The H6 of leaving H–Cl is making a hydrogen bonding with Cl3 in I5. With 6.1 kcal/mol endothermic energy, the weakly bound H–Cl desorbs yielding the final product P4 by forming bidentate Ti species, (–O–)2–Ti–Cl2. The net forward reaction barriers of these four channels are 14.2, 14.7, 19.0 and 16.8 kcal/mol suggests that all reaction have similar possibility of being formed. In all of the transition states, the penta-coordination of Ti plays a major role. 4. Conclusion We have theoretically investigated the initial adsorption mechanisms of TiCl4 on OH/Si(1 0 0)-2  1 surface in order to elucidate the atomistic details of surface reactions and to explain the slow growth rate. The potential energy surfaces of four reaction paths were explored. According to our calculations, the TiCl4 initially interact with surface oxygen forming the penta-coordinate intermediate I1 without a reaction barrier. The initially formed I1 is more stable than the reactants by 17.1 kcal/mol, indicating that the penta-coordination yields quite stable intermediate. Two different reaction channels are possible from the intermediate I1. The hydrogen abstraction by Cl via TS1 yields HCl trapped intermediate I2, while the exchange reaction by Cl via TS3 yields direct Cl adsorbed species I4. The reaction barrier of TS3 is higher than TS1 by 4.8 kcal/mol. And the HCl trapped intermediate I2 is more stable than I4 by 8.0 kcal/mol. As a result, the HCl desorption path is both kinetically and thermodynamically favorable. However, the difference in reaction barrier is

M.K. Ghosh, C.H. Choi / Chemical Physics Letters 457 (2008) 69–73

minor. Therefore, the direct Cl adsorption channel should also be accessible. The HCl as trapped in I2 via penta-coordination to Ti can either desorb as P1 or readsorb to the surface via TS2. The relative energies of P1 and TS2 are calculated to be 10.2 and 2.4 kcal/ mol with respect to the reactant R1 indicating that the former is more favorable. However, the HCl readsorption is also possible due to its low reaction barriers. The second HCl can desorb via TS4. The reaction is initiated by the hydrogen abstract of Cl. Since the reaction barrier of this channel as measured from R2 is comparable with those of the other channels, the bidentate Ti shown in P4 can be formed. In summary, the penta-coordination ability of Ti plays a big role in all process of surface reactions. One of the consequences is that the surface adsorption sites are effectively blocked reducing the film growth rate. Therefore we speculate that, if the bulky ligands such as titanium ethoxide and titanium isopropoxide are used, the penta-coordination to Ti would be sterically hindered. Therefore, its ability to block the adsorption site would be reduced, explaining why the relatively small ligand Cl has smaller growth rate than those with larger ligands. More studies are expected to clarify our propositions. Acknowledgement This work was supported by Korea Research Foundation Grant (KRF-2005-070-C00065). References [1] D.A. Muller, T. Sorsch, S. Moccio, F.H. Baumann, K. Evans-Lutterodt, G. Timp, Nature 399 (1999) 758.

[2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32]

73

G.D. Wilk, R.M. Wallace, Appl. Phys. Lett. 74 (1999) 2854. P.A. Packan, Science 285 (1999) 2079. G.D. Wilk, R.M. Wallace, J.M. Anthony, J Appl. Phys. 89 (2001) 5243. S.M. Goerge, A.W. Ott, J.W. Claus, J. Phys. Chem. 100 (1996) 13121. W. Gu, C.P. Tripp, Langmuir 21 (2005) 211. J. Arik, A. Aidla, T. Uustare, K. Kukli, V. Sammelselg, M. Ritala, M. Leskelä, Appl. Surf. Sci. 193 (2002) 277. V. Pore, A. Rahtu, M. Leskelä, M. Ritala, T. Sajavaara, J. Keinonen, Chem. Vap. Deposition 10 (2004) 143. A. Rahtu, K. Kukli, M. Ritala, Chem. Mater. 13 (2001) 817. I.-D. Kim, H.L. Tuller, H.-S. Kim, J.-S. Park, Appl. Phys. Lett. 85 (2004) 4705. D.-S. Kil, J.-M. Lee, J.-S. Roh, Chem. Vap. Deposition 8 (2002) 185. D. Jeong, J. Lee, H. Shin, J. Lee, J. Kim, M. Sung, J. Korean Phys. Soc. 45 (2004) 1249. S.K. Kim, W.-D. Kim, K.-M. Kim, C.S. Hwang, J. Jeong, Appl. Phys. Lett. 85 (2004) 4112. R.L. Puurunen, J. Appl. Phys. 97 (2005) 121301. M.L. Hair, W. Hertl, J. Phys. Chem. 77 (1973) 2070. S. Hukka, E.-L. Lakomma, T. Suntola, Thin Solid Films 225 (1993) 280. S. Hukka, E.-L. Lakomma, O. Jylhä, J. Hornytskyj, S. Vilhunen, Langmuir 9 (1993) 3497. S. Haukka, E.-L. Lakomma, A. Root, J. Phys. Chem. 97 (1993) 5085. H. Siimon, J. Aarik, J. Phys. D: Appl. Phys. 30 (1997) 1725. J. Aarik, A. Aidla, H. Mändar, T. Uustare, App. Surf. Sci. 172 (2001). Z. Hu, C.H. Turner, J. Phys. Chem. B 110 (2006) 8337. Z. Hu, C.H. Turner, J. Am. Chem. Soc. 129 (2007) 3863. S.S. Lee, J.Y. Baik, K.S. An, Y.D. Suh, J.H. Oh, Y. Kim, J. Phys. Chem. B. 108 (2004) 15128. M.W. Schmidt et al., J. Comp. Chem. 14 (1993) 1347. W.J. Herhe, R. Ditchfield, J.A. Pople, J. Chem. Phys. 56 (1972) 2257. C. Gonzalez, H.B. Schlegel, J. Phys. Chem. 94 (1990) 5523. Jr. Shoemaker, L.W. Burgruff, M.S. Gordon, J. phys. Chem. A 103 (1999) 3245. N.L. Allinger, Y.H. Yuh, J.H. Lii, J. Am. Chem. Soc. 111 (1989) 8551. J.H. Lii, N.L. Allinger, J. Am. Chem. Soc. 111 (1989) 8566. J.H. Lii, N.L. Allinger, J. Am. Chem. Soc. 111 (1989) 8576. M.K. Ghosh, C.H. Choi, Chem. Phys. Lett. 426 (2006) 365. K.S. Finnie, G. Triani, K.T. Short, D.R.G. Mirchell, D.J. Attard, J.R. Bartlett, C.J. Barbe, Thin Solid Films 440 (2003) 109.