Ga Cleaning of Al2O3 Substrate: Low Coverage Adsorption of Ga on a

The low coverage adsorption of gallium on a hydrogen-contaminated Al-terminated α-Al2O3(0001) surface is studied theoretically using first-principles...
0 downloads 0 Views 891KB Size
Download PDF
3384

J. Phys. Chem. C 2007, 111, 3384-3392

Ga Cleaning of Al2O3 Substrate: Low Coverage Adsorption of Ga on a Hydrogen-Contaminated r-Al2O3(0001) Surface Rui Yang* and Alistair P. Rendell† Department of Computer Science, ANU College of Engineering and Computer Science, The Australian National UniVersity, Canberra ACT 0200, Australia ReceiVed: June 23, 2006; In Final Form: October 13, 2006

The low coverage adsorption of gallium on a hydrogen-contaminated Al-terminated R-Al2O3(0001) surface is studied theoretically using first-principles density functional calculations with periodic boundary conditions. Initial computations to form the hydrogen-contaminated surface identified two stable hydrogen adsorption sites, one where the hydrogen is chemisorbed and one where it is physisorbed. Although interaction between the gallium atom and the chemisorbed hydrogen leads to a weakening of the hydrogen to Al2O3 surface adhesion, this weakening is insufficient to result in a loss of hydrogen from the Al2O3 surface. For the physisorbed hydrogen, addition of gallium can lead to a strongly bound gallium/hydrogen pair that shows much less adhesion to the Al2O3 surface. This provides direct theoretical support for a gallium cleaning effect.

1. Introduction Because of structural and economic considerations, R-Al2O3 (sapphire) is commonly chosen as a substrate for the growth of GaN semiconductors. To offset the still large lattice mismatch and thermal expansion coefficient difference between Al2O3 and GaN, the relaxed rhombohedra surface of R-Al2O3 is transformed to a polar wurtzite structure by forming several layers of AlN prior to the heteroepitaxial growth of GaN.1 This is accomplished using a number of initial growth steps such as substrate surface cleaning, nitridation, and growth of a nucleation layer.2 X-ray photoelectron spectroscopy studies have shown that an epitaxial-polished sapphire substrate is usually contaminated with hydrocarbon and fluorine atoms.3,4 For example, hydrogen was found to be stable on the Al2O3 surface even after annealing at 1100 °C under ultrahigh vacuum conditions,3 while carbon was found to be stable on an Al2O3 surface after annealing at 1100 K in an O2 partial pressure of 5 × 106 Torr.4 Firstprinciples calculations of the hydrogen and carbon adsorption on Al2O3 surface phases and metal deposition have also found stable surface phases that have bound C and H atoms.5 Although C and H impurities can normally be removed by using an O2 or H2/Ar plasma,6 recent work by Davidsson et al. has suggested an alternative process based on the deposition and evaporation of a layer of Ga atoms.7 In their work, Davidsson et al. used reflection high-energy electron diffraction (RHEED) to investigate surface atomic periodicity and verified that Ga can remove impurities such as hydrocarbons and oxides from the Al2O3 surface. This offers a significant advantage in the preparation of GaN devices because it removes the need for an extra plasma source. In addition to the Al2O3 substrate, Ga cleaning has also been used with SiC substrates,8 AlN buffer layers,9 and GaN surfaces.10-12 In spite of this widespread use, there appears to be little understanding of the physical processes involved, and to the best of our knowledge, there are no theoretical studies of * Corresponding author. E-mail: [email protected]. † E-mail: [email protected].

the Ga cleaning effect. To address this deficiency, we present in this paper a detailed theoretical investigation of the adsorption of Ga on a hydrogen-contaminated Al-terminated R-Al2O3(0001) surface. The work presented here builds on our previous study of Ga adsorption on the same Al-terminated R-Al2O3 surface.19 In line with this previous work, we also employ a periodic supercell approach13-15 with localized basis functions. We note that, in comparison to plane wave based techniques, use of localized basis functions facilitates analysis of the local chemical bonding properties. Initially, the hydrogen adsorptive surface is explored starting from several different H positions. Then, after determining the possible hydrogen adsorption sites, the Ga cleaning effect is studied by placing one gallium atom at various positions above each unit cell of the H-contaminated Al2O3 surface. The unit cell size used in this work is in line with our previous study,19 having one gallium atom per unit cell. This corresponds to 1/3 monolayer (ML) gallium coverage and, as a consequence, we term this low coverage gallium adsorption. In all cases, the studied Al2O3 surface was fully relaxed. 2. Models and Computational Details Density functional theory (DFT) calculations were performed using the Gaussian 0316 quantum chemistry package. The Becke-Lee-Yang-Parr (BLYP) GGA functional17,18 was used in conjunction with a 3-21g* or 6-31g* basis set. This level of theory was chosen, as it was found in our previous work19 to give good agreement between the experimental and theoretical results for the bulk R-Al2O3 crystal lattice constants, R-Al2O3 (0001) surface energy, and surface structure. We note also that work by J. R. B. Gomes et al.20-22 on relaxation of the Al2O3 surface has suggested that Hartree-Fock (HF), hybrid HF/DFT, and pure DFT approaches all give the same qualitative description of the surface structure. To make our computations tractable, the Coulomb energy has been evaluated using the Gaussian 03 automatically generated fitting basis sets. The basic model for the Al2O3 surface was also taken from our previous study of Ga adsorption. It comprises a 1 × 1 Al-

10.1021/jp063931r CCC: $37.00 © 2007 American Chemical Society Published on Web 02/02/2007

Ga Cleaning of Al2O3 Substrate

J. Phys. Chem. C, Vol. 111, No. 8, 2007 3385

Figure 1. Al2O3 surface slab (18 layers) with larger circles used to indicate Al. (a) On-top view showing the four supercells and eight initial adsorption positions within one of the supercells. (b) Surface view of a single supercell showing the six triangles defined by the locations of the O(2) O(2) O(1) atoms and distinguished by the central O(2), A1(1), A1(2), or A1(3) atom. Positions B1, B2, and B3 correspond to the BO(2) Al(1), BAl(2), and BAl(3) bridge positions, respectively. (c) Outmost five surface layers of the model slab showing the Al(1), O(2), Al(2), Al(3), and O(2) layers and using gray scale to characterize distance from the surface.

terminated R-Al2O3(0001) surface containing 18 atomic layers. Periodicity was imposed in two dimensions using the periodic boundary condition (PBC) method in Gaussian 03. This offers advantages compared to methods that require periodicity in three dimensions because it eliminates possible effects that result from interactions between the supercell surface and image surfaces. In our previous work, a model with 18 layers was found to be large enough to obtain a converged surface energy and a reasonable free surface structure. An illustration of the Al-terminated R-Al2O3 surface model with its 18 atomic layers is shown in Figure 1. The top five surface layers dominate the chemical bonding that occurs on adsorption. These five layers are plotted in progressively lighter shades of gray the further they are from the surface and are explicitly labeled as Al(1), O(1), Al(2), Al(3), and O(2). For the 1/3 ML surface coverage considered in this work, there is one hydrogen atom per supercell. Eight possible positions were considered as starting points for both the initial hydrogen adsorption and the subsequent Ga adsorption. These starting sites are shown in Figure 1a and are similar to those

used in other studies of hydrogen and copper adsorption on an Al2O3 surface,23,24 except that some extra bridge positions are also considered. The initial positions are classified according to the type of atom under them. For example, positions Al1, Al2, and Al3 lie above atoms in the Al(1), Al(2), and Al(3) layers, respectively. Similarly, the O1 and O2 positions lie above one of oxygen atoms in the O(1) and O(2) layers, respectively. Both the Al1 and O1 positions should be regarded as on-top positions, owing to the well-known large inward relaxation of the Al(1) layer,19,25-29 whereas Al2, Al3, and O2 are hollow positions. It is convenient to divide the stoichiometric Al2O3(0001) surface into triangles with vertices defined by the location of O(1) atoms. In which case, there are four different types of triangles that are distinguished by the atom (from a lower layer) that resides at the center of each triangle. These triangles are shown in Figure 1b and will be referred to as the OAl(1), OAl(2), OAl(3), and OO(2) triangles accordingly. Between pairs of triangles, there are three possible bridge positions, which we

3386 J. Phys. Chem. C, Vol. 111, No. 8, 2007

Yang and Rendell work concerned with Ga adsorption on the same Al2O3 surface.19 3. Hydrogen Adsorption on r-Al2O3(0001) Surface For all but one of the eight possible H starting positions (see Figure 1a), structural optimization resulted in the hydrogen atom moving to sit directly on top of the surface Al(1) atom. The only exception was for the Al3 position, in which case, the H stayed just above the Al(3) atom. Considering first H adsorbed at the Al1 position, the value of the H adsorption energy (Etotal H ads) at this location is -0.59 eV. This value, which agrees well with a value of -0.5 eV computed using the norm-conserving pseudopotential method,18 suggests significant interaction between the hydrogen and the Al2O3 surface. To analyze this binding further, it is convenient to use our previously proposed scheme32 that partitions ETotal H ads Stru into chemical (EChem H ads ) and structural (EH ads) contributions: Chem Stru ETotal H ads ) EH ads + EH ads

Figure 2. Charge redistribution evaluated on a diagonal plane containing Al(1), Al(2), and Al(3) atoms for H atom sitting at the Al1 positions of the Al2O3 surface. Positive contours (solid lines) denote charge accumulation due to H adsorption and negative contours (dotted lines) denote charge depletion. O(2) O(2) O(2) label as BAl(1) , BAl(2) , and BAl(3) , depending on which two O(1) triangles they connect. Hydrogen and gallium atoms were deposited on only one side of the Al2O3 slab without imposing any symmetry constraints. The location of the adsorbed atom was optimized using the Berny algorithm developed by Schlegel et al.30,31 and as implemented in Gaussian 03. This algorithm uses the forces acting on the atoms together with an approximate second derivative or Hessian matrix to predict energetically more favorable structures, stopping when a minimum is located. Initial structural optimizations were performed using BLYP/321g*, which were then refined using BLYP/6-31g*; such an optimization process was found to be valid in our previous

The chemical contribution is defined as the energy difference between the fully optimized H-adsorbed Al2O3 system and an isolated Al2O3 system with the same structural configuration as for the H-adsorbed system plus an isolated hydrogen. The structural contribution is defined as the energy difference between the fully optimized isolated Al2O3 system and an isolated Al2O3 system with the same structural configuration as that found in the optimized H-adsorbed Al2O3 system. At the Al1 site, EChem H ads is -1.48 eV and serves to bind the hydrogen atom to the surface. To determine the origin of this binding, we plot, in Figure 2, the charge redistribution (∆F) that occurs between the H and the Al2O3 surface. This has been calculated as ∆F ) F(H/Al2O3) - F(Al2O3 isolated surface) F(H atom), and is plotted along the diagonal plane of the supercell containing the Al(1), Al(2), and Al(3) atoms. A clear gain of electron density appears in the region between the H and Al(1) atoms, indicating that Al(1)-H covalent bonding dominates EChem H ads . This bonding is further evident from Figure 3, which gives density of state (DOS) curves for the isolated

Figure 3. (a) DOS curves for H atom, isolated Al2O3 surface slab, and the H adsorbed Al2O3 surface with H sitting at the Al1 position. The occupied states are filled with gray background. The vertical downward arrow indicates the energy level of the LUMO of the isolated Al2O3 surface slab. (b) Real space distribution of the LUMO of the isolated Al2O3 surface slab.

Ga Cleaning of Al2O3 Substrate hydrogen atoms, the isolated Al2O3 surface, and the H-adsorbed Al2O3 system with hydrogen sitting at the Al1 position, i.e., the Al1/H-adsorbed Al2O3 system. Specifically, Figure 3a shows that there is an unoccupied state that lies within the band gap of the isolated Al2O3 surface. A plot of the corresponding real space molecular orbital, given in Figure 3b, indicates that this is a surface state arising mainly from an Al(1) pz dangling orbital. Such a surface state has also been found in other studies of the Al2O3 surface that used periodic HF theory33 and in our study of the Cu/Al2O3 interface that used a plane-wave pseudopotential method.34 In our previous study of Ga adsorption on the Al2O3 surface, this surface state was shown to play a key role in promoting charge transfer between the adsorbed Ga and the Al2O3 surface.19 For the H-absorbed Al2O3 system considered here, the disappearance of this surface state when hydrogen is adsorbed is evidence of similar strong chemical bonding involving the H s and Al(1) pz orbitals. The structural contribution to the H adsorption energy (EStru H ads) at the Al1 position is 0.89 eV. This reduces the binding between H and the Al2O3 surface and results in a substantially Chem smaller value for ETotal ads compared to Eads . Adsorption of H also causes a marked change in the Al2O3 surface, shifting the Al(1) atom away from the rest of the slab by a distance of 0.52 Å. The resulting interlayer distances for Al(1)-O(1) and Al(1)-H are 0.65 and 1.62 Å, respectively, and are in good agreement with the values of 0.64 and 1.63 Å obtained from first-principles pseudopotential calculations.24 Also, the calculated Al(1)-O(1) interplanar spacing is consistent with the results of ion-scattering experiments,3 where large Al2O3 surface relaxation was observed for a system that was believed to be contaminated with surface hydrogen. We now consider hydrogen adsorption at the Al3 site; although the hydrogen atom remains at this site, the overall binding energy is just -0.0147 eV, with values for EChem H ads and EStru of -0.015 and 0.0003 eV, respectively. Thus there is H ads only very weak chemical bonding between the hydrogen atom and Al2O3 slab and very small structural changes occur on adsorption. Indeed, for adsorption at this position, there is no obvious charge redistribution, with the overlap charge between H and any one of the Al2O3 surface atoms being less than 0.05 electrons. Also, in contrast to adsorption at the Al1 position, the DOS for the Al3/H-adsorbed Al2O3 systems, shown in Figure 4, is just a superposition of the DOS for the H atom and the Al2O3 surface. In summary, it appears that there are two different adsorption sites for a H atom sitting on the Al2O3(0001) surface: chemisorption with the H atom sitting at the Al1 position and physisorption with the hydrogen atom sitting at the Al3 position. The fact that hydrogen optimizes to the Al1 position from all but one of the eight possible starting positions and that the binding energy is so small at the Al3 site suggests that interaction between hydrogen and the Al2O3 surface is rather weak at all locations except near the Al(1) atoms. In the following two sections, we consider in detail the affect of Ga adsorption on these two types of adsorbed H atoms. 4. Ga Adsorption on H-Chemisorbed r-Al2O3(0001) Surface The optimized R-Al2O3(0001) surface with one hydrogen atom located at the Al1 position can be considered as a model for a hydrogen-contaminated R-Al2O3 surface where the H atom is chemisorbed. One gallium atom was added to this system located at one of the eight positions given in Figure 1. The whole system was then optimized. This gave rise to only two stable

J. Phys. Chem. C, Vol. 111, No. 8, 2007 3387

Figure 4. DOS curves for H atom, isolated Al2O3 surface slab, and the H adsorbed Al2O3 surface with H sitting at the Al3 position. The occupied states are filled with gray background. The downward arrow indicates the DOS superposition peak from H and Al2O3 surface.

configurations: one with the Ga atom sitting at the Al3 position and another energetically less favorable structure with the Ga atom positioned at Al2. In both cases, the hydrogen remains at Al1. For most starting configurations, the Ga atom moves to the Al3 position; the exceptions are for the Al2 and B2 starting positions, where it moves to the Al2 position. Interestingly, the two stable Ga adsorption sites are identical to those observed for Ga interacting with a clean Al2O3 surface.19 This suggests that similar interactions are involved, specifically, that the highest occupied molecular orbital (HOMO) of the combined Ga/H-chemisorbed Al2O3 system is formed through interaction of the Ga HOMO with the lowest unoccupied molecular orbital (LUMO) of the H-chemisorbed Al2O3 surface19 and that at the start of the adsorption process, this interaction gives rise to a net charge transfer from the Ga atom to the Al2O3 slab, leaving the Ga atom slightly positively charged and repelled away from the positively charged Al(1) atoms. The final Al3 and Al2 adsorption sites are then locations where the Ga-Al(1) repulsive interactions are balanced, i.e., a Ga atom located at any other position is repulsed by the nearest Al(1) ions until it inevitably moves into one of these two positions. Values for the Ga adsorption energy (ETotal Ga ads) at the Al2 and Al3 sites and for both the H-chemisorbed Al2O3 and clean Al2O3 surface are given in Table 1. Similarity between H-chemisorbed and clean Al2O3 systems is further supported by the fact that the difference between the adsorption energies at the Al2 and Al3 positions is identical at 0.33 eV for both systems. Differences, however, are also evident because the absolute values of the adsorption energies (ETotal Ga ads) are quite different, with the values being much larger for the H-chemisorbed Al2O3 surface compared to the clean Al2O3 surface (-4.53 eV compared to

3388 J. Phys. Chem. C, Vol. 111, No. 8, 2007

Yang and Rendell

TABLE 1: Adsorption Energies for Ga on the Chemisorbed Total Chem H/Al2O3 Surface (EGa ads), Its Chemical (EGa ads) and Stru Structural (EGa ads) Components, and Changes in the Al(1) to O(1) interlayer Distance (∆dAl(1)-O(1)) Resulting from Ga Adsorptiona Ga on H/Al2O3 surface

Total E Ga ads(eV) Chem E Ga ads(eV) Stru E Ga ads(eV) ∆dAl(1)-O(1) (Å)

Ga on clean Al2O3 surface19

Al2 position

Al3 position

Al2 position

Al3 position

-4.53 -4.80 0.27 -0.07

-4.86 -5.07 0.21 -0.06

-1.46 -2.18 0.72 0.43

-1.79 -2.74 0.95 0.49

a Values for Ga adsorption on the same Al2O3 surface without hydrogen are also listed for comparison.

TABLE 2: Chemical Contributions to the Adsorption Chem Chem Energies for H (EHChem ads ), Ga (EGa ads), and Ga-H (EGa-H ads) Ga-H Adsorption Plus Derived Binding Energies for H-Ga (Ebinding ), Ga-Al2O3 H-Al2O3 Ga-Al2O3 (Ebinding ), and H-Al2O3 (Ebinding ) Interactions on the Chemisorbed H/Al2O3 Surface with and without Ga Atoma H/Al2O3 surface EChem H ads (eV) Chem EGa ads (eV) Chem EGa-H ads(eV) EGa-H binding(eV) 2O3 EGa-Al (eV) binding H-Al2O3 Ebinding (eV) a

-1.46 eV at the Al2 site, and -4.86 eV compared to -1.79 eV at the Al3 site). To determine the origin of this difference, we have again calculated chemical and structural contributions to the adsorption energy of Ga on the H-chemisorbed Al2O3 surface. These values are also given in Table 1. They indicate that, on the H-contaminated Al2O3 surface, the chemical contribution to the adsorption energy is significantly larger than the structural contribution, accounting for the majority of the total binding energy. And, in absolute terms, the chemical contribution for Ga interacting with the H-contaminated Al2O3 surface is more than twice that observed for Ga interacting with a clean Al2O3 surface. These observations imply that there are significant additional interactions occurring between Ga and the adsorbed H atom. For the clean Al2O3, surface changes in the Al(1)-O(1) interlayer distance of 0.43 and 0.49 Å were observed for Ga adsorption at the Al2 and Al3 sites, respectively.19 These large movements give rise to large positive values for the structural contribution to the adsorption energy. In Section 3, it was noted that a similar large movement occurs when hydrogen is adsorbed at the Al1 site. As a consequence, it is not surprising that when Ga is adsorbed on the H-contaminated Al2O3 system, further changes in the Al(1)-O(1) interlayer distance are only slight (shortening by 0.07 and 0.06 Å for Al2 and Al3 sites, respectively) and therefore that the values for EStru Ga ads are substantially smaller compared to those for clean Al2O3. To determine whether the presence of Ga leads to a weakening of the H-Al2O3 binding energy, we have evaluated this binding energy in two ways. The first approach approximates the interaction of Ga and H with the Al2O3 surface as a sum of pairwise interactions involving Ga-Al2O3, Ga-H, and H-Al2O3. Within this model, EChem Ga ads involves the loss of two interactions, namely Ga-Al2O3 and Ga-H. If the pairwise Ga-H H-Al2O3 2O3 binding energies are written as EGa-Al binding , Ebinding, and Ebinding , Chem Ga-Al2O3 Ga-H then EGa ads ) Ebinding + Ebinding. By computing analogous Chem quantities to EChem Ga ads through the removal of either H (EH ads ) or Chem both the H and Ga (EGa-H ads), we obtain two further equations: H-Al2O3 Chem H-Al2O3 + EGa-H + EChem H ads ) Ebinding binding and EGa-H ads ) Ebinding Ga-Al2O3 Ebinding . With three equations and three unknowns, values for all three binding energies can be derived. These, together with Chem values for EChem H ads and EGa-H ads, are shown in Table 2. This gives derived binding energies for the H-Al2O3 interaction ( 2O3 EH-Al binding ) when Ga is at the Al2 and Al3 sites as -1.319 and -1.361 eV, respectively. 2O3 involves reThe second approach to determine EH-Al binding moving Ga from the Ga adsorbed H/Al2O3 system and then performing calculations with and without the H atom (while at all times keeping the structure of the Al2O3 system fixed at its

-1.480

H/Al2O3 +Ga (site Al2) surface

H/Al2O3 +Ga (site Al3) surface

-3.656 -4.801 -3.783 -2.337 -2.451 -1.319

-3.662 -5.073 -4.133 -2.301 -2.772 -1.361

See text for further details.

original Ga adsorbed H/Al2O3 optimized geometry). Performing these calculations gives hydrogen binding energies of -1.027 and -1.091 eV for Ga at the Al2 and Al3 sites, respectively. These values are somewhat less that those derived above, but differences are to be expected given the lack of many-body effects in the derived model and errors due to basis set superposition effects in the computed values. The consistent observation, however, is that in both cases the values obtained 2O3 for EH-Al binding in the presence of Ga are less than the value of -1.480 eV computed for H interacting with a clean Al2O3 surface. Thus it does appear that Ga adsorption leads to a modest weakening in the adhesion of chemisorbed hydrogen to the Al2O3 surface. The values derived above for EGa-H binding and given in Table 2 indicate strong interaction between the Ga and H atoms. This is supported by the density of states (DOS) plots given in Figure 5a for Ga at the Al2 position and Figure 5b for Ga at the Al3 position. These show that the HOMO of the (Ga + H)/Al2O3 surface is indeed formed as a result of interactions between the Ga HOMO and the LUMO of the H/Al2O3 surface and from the real space distributions of two respective HOMOs (also shown in Figure 5) that this is an orbital that is distributed around both the H and Ga atoms. In comparison to the experimental bond enthalpy of -2.84 eV35 for diatomic GaH, the values for EGa-H binding are somewhat less, at around -2.3 eV for both adsorption sites. This, however, is to be expected given that the interatomic distance between Ga and H is 2.841 Å for Ga at the Al2 site and 2.848 Å for Ga at the Al3 site, while the equilibrium bond length of Ga-H is 1.7 Å.36,37 2O3 The value derived for EGa-Al (Table 2) at the Al3 position binding (-2.772 eV) is somewhat larger than at the Al2 position (-2.451 eV). This reflects the relative stability of these two positions and is in agreement with the similar values computed previously for Ga adsorption on a clean Al2O3 surface.19 In summary, analysis of the binding energies for the interaction between the Ga and H adsorbates and the Al2O3 surface gives results that are consistent with analysis of the charge distributions. Our results support the view that Ga adsorption may have a cleaning effect on a hydrogen-contaminated Al2O3 surface, but the weakening of the H-Al2O3 binding computed here is insufficient to lead to a loss of hydrogen from the Al2O3 surface. It should be remembered, however, that our calculations are performed at 1/3 monolayer gallium coverage, and it is possible that higher coverage levels may enhance this bond weakening.

Ga Cleaning of Al2O3 Substrate

J. Phys. Chem. C, Vol. 111, No. 8, 2007 3389

Figure 5. DOS curves for Ga atom, isolated chemisorbed H/Al2O3 surface slab, and the chemisorbed H/Al2O3 surface with Ga adsorbed at the (a) Al2 position and (b) Al3 position. The occupied states are filled with gray background. The vertical downward arrow indicates the energy level of the HOMO of the combined Ga/H/Al2O3 surface, while the right-hand side of the figure shows the real space distribution of this HOMO.

Figure 6. (a) Top view of the starting and the final relaxed configuration for physisorbed H/Al2O3 surface and Ga starting from the Al3 position. (b) Charge density plot for final relaxed system.

5. Ga Adsorption on H-Physisorbed r-Al2O3(0001) Surface To study the effect of the Ga adsorption on physisorbed hydrogen, structural optimizations were performed on the Al3/ H-adsorbed R-Al2O3(0001) system augmented with a single Ga atom positioned at one of the eight possible starting positions given in Figure 1. For positions Al2, B2, and O1, the Ga moved to the Al3 position, while the H atom was pushed into the Al1 position. This gave rise to final configuration that was identical to the Al3 chemisorbed structure identified in Section 4. For

the remaining five starting positions, essentially two different final structures were obtained. Both of these were characterized by a Ga-H diatom pair located above the Al2O3 surface. Because formation of this Ga-H pair is likely to play an important role in the Ga cleaning process, we now consider these different stable structures in detail. If the Ga atom starts from the Al3 position, located immediately above the hydrogen atom, then the main structural changes that occur are vertical height adjustments for the Ga atom and slight relaxation of the Al2O3 surface. This results in

3390 J. Phys. Chem. C, Vol. 111, No. 8, 2007

Yang and Rendell

TABLE 3: Adsorption Energies for Ga on the Physisorbed Total Chem H/Al2O3 Surface (EGa ads), Its Chemical (EGa ads) and Stru Structural (EGa ads) Components, Equivalent Values for Ga-H Adsorption, the Chemical Contribution for H Adsorption (EHChem ads ), and the Bond Length of the Ga-H Pair (dGa-H)a Ga starting position Total EGa ads (eV) Chem EGa ads (eV) Stru EGa ads(eV) Total EGa-H ads(eV) Chem EGa-H ads (eV) Stru E Ga-H ads(eV) EChem H ads (eV) dGa-H (Å) Ga-H orientation

Al3

Al1

BO(2) Al(1)

O2

BO(2) Al(3)

-2.584 -2.581 0.003 -0.190 -0.193 0.003

-4.320 -5.288 0.968 -1.879 -2.900 1.021 -2.944 1.591 in-plane

-4.327 -5.290 0.963 -1.916 -2.933 1.017 -2.953 1.590

-4.325 -5.287 0.962 -1.887 -2.902 1.015 -2.990 1.590

-4.325 -5.292 0.967 -1.885 -2.907 1.022 -2.943 1.590

1.689 vertical

a Data is reported for the five Ga starting positions that lead to formation of a Ga-H pair. The Ga-H pair is characterized as either vertical or in-plane as discussed in the text.

the “vertical” Ga-H diatom pair shown in Figure 6a. For this case, the Ga adsorption energy is 2.581 eV and comes almost exclusively from chemical interactions (see Table 3). Clearly, however, a Ga-H pair located perpendicular to the Al2O3 surface is not going to be the energetically most favorable structure. So it is not surprising that, for the other Ga starting positions, Ga-H pairs are obtained lying almost in the plane of the Al2O3 surface. For example, when starting with the Ga atom located at Al1, the final structure has a Ga-H pair lying on the Al2O3 surface with its bond midpoint located approximately above the Al(3) atom. This is shown in Figure 7a. The Ga-H pair is tilted slightly, with the Ga atom having a slightly shorter vertical distance from the Al2O3 surface than the H atom. The energies for three other “in-plane” systems are given in Table 3, with their geometries illustrated in Figure 8. For all of these configurations, the Ga adsorption energies are roughly equivalent at -4.3 eV, a value that is significantly larger than that for the vertical Ga-H diatom. For these systems, there is also greater structural reorganization of the Al2O3 lattice, as evident from larger structural components to the Ga adsorption energies; values that are comparable with EStru Ga ads computed for Ga adsorption on a clean Al2O3 surface. In the context of Ga cleaning, the adsorption energies for the Ga-H diatom in these systems are of particular interest. Values for these have been computed by subtracting the energies of fully optimized Ga-H and clean Al2O3 from the total energy of the combined system. These values are shown in Table 3, together with their breakdown into chemical and structural

Figure 8. Top view of the starting and the final relaxed configuration for physisorbed H/Al2O3 surface with Ga starting from positions (a) O(2) BO(2) Al(3), (b) BAl(1), and (c) O2.

components. Clearly, the vertical Ga-H is physisorbed, requiring just 0.190 eV to be removed in a process that has minimal effect on the underlying Al2O3 lattice. This result is to be expected given the charge density plot shown in Figure 6b, where strong interaction between Ga and H are evident, and there appears to be very little interaction between the Ga-H pair and the Al2O3 surface. By comparison, for the in-plane adsorbed system, the charge density plots, shown in Figure 7b for the Al1 system, indicate strong interaction between the Ga-H pair and the Al2O3 surface, explaining the substantial chemical and structural contributions to the Ga-H adsorption energy. In Section 4, values for the Ga-H binding energy of approximately -2.3 eV were derived. These values were contrasted with the experimental bond energy for Ga-H of -2.84 eV. For the vertical Ga-H pair, there is only very slight interaction with the Al2O3 surface, hence it is not surprising that the adsorption energy for Ga at this position (-2.58 eV) is roughly in line with the binding energy of Ga-H. It is interesting to note, however, that regardless of exactly what value for the Total Ga-H binding energy is used, all values for EGa+H ads are significantly less that for the Ga-H binding energy. This indicates that Ga-H could be removed from the Al2O3 surface as a single entity by high-temperature treatments. In contrast to the vertical Ga-H system, all in-plane Ga-H pairs exhibit much larger adhesion to the Al2O3 surface. For

Figure 7. (a) Top view of the starting and the final relaxed configuration for physisorbed H/Al2O3 surface and Ga starting from the Al1 position. (b) Charge density plot for the final relaxed system.

Ga Cleaning of Al2O3 Substrate

J. Phys. Chem. C, Vol. 111, No. 8, 2007 3391

Figure 9. (a) Charge redistribution of the isolated Ga-H pair. Positive contours (solid lines) denote charge accumulation due to H adsorption, while negative contours (dotted lines) denote charge depletion. (b) DOS curves of H atom, Ga atom, and the isolated Ga-H pair, together with the crystal orbital overlap population (COOP). The occupied states are filled with gray background and the unoccupied ones keep blank. See text for further details.

Figure 10. Charge redistribution evaluated on a diagonal plane containing Al(1), Al(2), and Al(3) atoms for (a) vertical and (b) in-plane Ga-H pairs. Positive contours (solid lines) denote charge accumulation due to H adsorption and negative contours (dotted lines) denote charge depletion.

these configurations, the charge density plots, illustrated in Figure 7b for the in-plane Ga-H pair at the Al1 position, show relatively little interaction between the hydrogen and Al2O3 surface. From this observation, it might be expected that the energy associated with removing hydrogen from these in-plane configurations would also be roughly equivalent to the Ga-H bond energy. This value has been computed by removing hydrogen from the Ga/H adsorbed system while keeping all other parameters fixed and is therefore the chemical contribution to the H adsorption energy, or EChem H ads . Values for this quantity are also given in Table 3, and at ∼2.9 eV, they are indeed in line with the Ga-H bond energy. Also, as the values for EChem H ads Chem are somewhat larger than EGa-H ads, this implies that the Ga-H pair would be removed intact before the Ga-H bond breaks. Such strong binding within the Ga-H pair provides further support for Ga-H playing a key role in Ga cleaning. To explore the bonding between the Ga-H pair and the Al2O3 surface more fully, the charge redistribution of the separate Ga-H pair is plotted in Figure 9a (using the same Ga-H orientation as found in Figure 7b). From this, it is apparent that

there is a large charge polarization between Ga and H atoms and that the Ga-H pair has substantial ionic bonding character. In Figure 9b, the atomic orbitals involved in the charge polarization are identified by evaluating the crystal orbital overlap population (COOP) for the Ga-H pair. This shows that the two highest occupied orbitals have bonding characteristics and result from interactions between the occupied H s orbital and the Ga s and pz orbitals. Meanwhile, the antibonding peaks consist of the virtual H s orbital and unoccupied Ga s and p orbitals, while between the bonding and antibonding orbitals there are empty Ga px and py orbitals that are essentially nonbonding in character. The charge polarization that occurs for the vertical and Al1 in-plane adsorbed Ga-H systems is apparent from the charge difference plot for Ga-H on the Al2O3 surface shown in Figure 10. For the vertical system, the charge polarization is very similar to that observed for the isolated Ga-H, but for the inplane system, substantial additional charge polarization occurs between the Ga and H atoms owing to the ionization of the Ga atom by the Al2O3 surface O atoms. As a consequence, it is

3392 J. Phys. Chem. C, Vol. 111, No. 8, 2007 not surprising that binding to the Al2O3 surface is very weak for the vertical Ga-H pair, while for the in-plane Ga-H pair, it is relatively strong. In summary, for certain starting geometries, it appears that the low coverage addition of Ga to an H-contaminated Al2O3 surface can lead to the formation of a Ga-H diatom. When this occurs, binding of the Ga-H diatom pair to the Al2O3 surface is always weaker than the Ga-H bond strength. This suggests that removal of H from an H-contaminated Al2O3 surface could proceed by Ga adsorption followed by a hightemperature treatment to remove Ga-H. The fact that for some other starting geometries, addition of Ga leads to isolated chemisorbed Ga and H atoms implies, however, that low coverage adsorption of Ga on an H-contaminated Al2O3 surface is insufficient to clean an entire Al2O3 surface of all adsorbed H atoms. 6. Conclusions In the present work, the adsorption of hydrogen on an Al-terminated R-Al2O3(0001) surface together with the cleaning effect of Ga at 1/3 ML coverage have been studied using a firstprinciples density functional approach with a 2D periodic supercell. Two hydrogen adsorption sites were identified: one corresponds to chemisorption, while the other corresponds to physisorption. Addition of Ga to the chemisorbed system leads to a weakening of the hydrogen to Al2O3 surface adhesion, but this is insufficient to lead to a loss of hydrogen from the surface. For the physisorbed system, addition of Ga can lead to the formation of a Ga-H diatom that in one case is only very weakly bound to the Al2O3 surface. This, plus the fact that the Ga-H bond energy was found to be stronger than the Ga-H/ Al2O3 surface binding, provides direct theoretical evidence in support of a Ga cleaning effect. The results reported here are for 1/3 ML Ga coverage, and it is possible that higher-level Ga coverage would give rise to a larger cleaning effect. Work is currently underway to investigate this. Acknowledgment. This work is funded by Australian Research Council Linkage Grant LP0347178 and is in association with Gaussian Inc. and Sun Microsystems. Provision of computer time from the Australian Partnership in Advanced Computing is gratefully acknowledged, as is the donation of computing resources from Alexander Technology. References and Notes (1) Jain, S. C.; Willander, M.; Narayan, J.; Overstraeten, R. V. J. Appl. Phys. 2000, 87, 965. (2) Seelmann-Eggebert, M.; Zimmermann, H.; Obloh, H.; Niebuhr, R.; Wachtendorf, B. J. Vac. Sci. Technol., A 1998, 16, 2008. (3) Ahn, J.; Rabalais, J. W. Surf. Sci. 1997, 388, 121. (4) Niu, C.; Shepherd, K.; Martini, D.; Tong, J.; Kelber, J. A.; Jennison, D. R.; Bogicevic A. Surf. Sci. 2000, 465, 163. (5) Wang, X. G.; Smith, J. R. Phys. ReV. B 2004, 70, 081401(R). (6) Namkoong, G.; Doolittle, W. A.; Brown, A. S.; Losurdo, M.; Capezzuto, P.; Bruno, G. J. Appl. Phys. 2002, 91, 2499. (7) Davidsson, S. K.; Andersson, T. G.; Zirath, H. Appl. Phys. Lett. 2002, 81, 664.

Yang and Rendell (8) Ploog, K.; Brandt, O.; Muralidharan, R.; Thamm, A.; Waltereit, P. J. Vac. Sci. Technol., B 2000, 18, 2290. (9) Khan, M. A.; Kuznia, J. N.; Olson, D. T. J. Appl. Phys. 1993, 73, 3108. (10) Bermudez, V. M.; Kaplan, R.; Khan, M. A.; Kuznia, J. N. Phys. ReV. B 1993, 48, 2436. (11) Kampen, T. U.; Mo¨nch, W. Appl. Surf. Sci. 1997, 117/118, 388. (12) Maffeis, T.; Clark, D.; Dunstan, P.; Wilks, S.; Evas, D.; Peiro, F.; Riechert, H.; Parbrook, P. Phys. Status Solidi A 1999, 176, 751. (13) Kudin, K. N.; Scuseria, G. E. Chem. Phys. Lett. 1998, 289, 611. (14) Kudin, K. N.; Scuseria, G. E. Chem. Phys. Lett. 1998, 283, 61. (15) Kudin, K. N.; Scuseria, G. E. Phys. ReV. B 2000, 61, 16440. (16) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03, revision C.02; Gaussian, Inc.: Wallingford, CT, 2004. (17) Becke, A. D. Phys. ReV. A 1988, 38, 3098. (18) Lee, C.; Yang, W.; Parr, R. Phys ReV. B 1988, 37, 785. (19) Yang, R.; Rendell, A. P. J. Phys. Chem. B 2006, 110, 9608. (20) Gomes, J. R. B.; Moreira, I. de P. R.; Reinhardt, P.; Wander, A.; Searle, B. G.; Harrison, N. M.; Illas, F. Chem. Phys. Lett. 2001, 341, 412. (21) Carrasco, J.; Gomes, J. R. B.; Illas, F. Phys. ReV. B 2004, 69, 064116. (22) Wander, A.; Searle, B. G.; Harrison, N. M. Surf. Sci. 2000, 458, 25. (23) Hernandez, N. C.; Sanz, J. F. J. Phys. Chem. B 2002, 106, 11495. (24) Tepesch, P. D.; Quong, A. A. Phys. Status Solidi B 2000, 217, 377. (25) Yang, R.; Tanaka, S.; Kohyama, M. Philos. Mag. 2005, 85, 2961. (26) Wang, X. G.; Chaka, A.; Scheffler, M. Phys. ReV. Lett. 2000, 84, 3650. (27) Siegel, D. J.; Hector, L. G., Jr.; Adams, J. B. Phys. ReV. B 2002, 65, 085415. (28) Verdozzi, C; Jennison, D. R.; Schultz, P. A.; Sears, M. P. Phys. ReV. Lett. 1999, 82, 799. (29) Felice, R. D.; Northrup, J. E. Phys. ReV. B 1999, 60, R16278. (30) Peng, C.; Ayala, P. Y.; Schlegel, H. B.; Frisch, M. J. J. Comput. Chem. 1996, 17, 49. (31) Peng, C.; Schlegel, H. B. Isr. J. Chem. 1994, 33, 449. (32) Yang, R.; Wang, Y. M.; Ye H. Q.; Wang, C. Y. J. Phys.: Condens. Matter 2001, 13, 4485. (33) Nagy, L. T.; Micov, M.; Benco, L.; Liska, M.; Mach, P.; Tunega, D. Int. J. Quant. Chem. 1998, 70, 341. (34) Yang, R.; Tanaka, S.; Kohyama, M. Philos. Mag. Lett. 2004, 84, 425. (35) Kerr, J. A. In CRC Handbook of Chemistry and Physics 19992000: A Ready-Reference Book of Chemical and Physical Data (CRC Handbook of Chemistry and Physics, 81st ed.; D.R. Lide, Ed.; CRC Press: Boca Raton, FL, 2000. (36) Pulham, C. R.; Downs, A. J.; Goode, M. J.; Rankin, D. W. H.; Robertson, H. E. J. Am. Chem. Soc. 1991, 113, 5149; Baxter, P. L.; Downs, A. J.; Rankin, D. W. H.; Robertson, H. E. J. Chem. Soc., Dalton Trans. 1985, 807; Downs, A. J.; Goode, M. J.; Pulham, C. R. J. Am. Chem. Soc. 1989, 111, 1936; Baxter, P. L.; Downs, A. J.; Goode, M. J.; Rankin, D. W. H.; Robertson, H. E. J. Chem. Soc., Dalton Trans. 1990, 2873. (37) Wang, Z. P.; Ley, L.; Cardona, M. Phys. ReV. B 1982, 26, 3249.