Nitrogen adsorption phases on the 4HâSiC(0001) Si-face

Viewer
Transcript

Surface Science 602 (2008) 3617–3622

Contents lists available at ScienceDirect

Surface Science journal homepage: www.elsevier.com/locate/susc

Nitrogen adsorption phases on the 4H–SiC(0 0 0 1) Si-face Christopher R. Ashman a,*, Gary Pennington b a b

High Performance Technologies Inc., 11955 Freedom Drive, Suite 1100, Reston, VA 20190-5673, United States US Army Research Laboratory, 2800 Powder Mill Road, Adelphi, MD 20783, United States

a r t i c l e

i n f o

Article history: Received 24 May 2008 Accepted for publication 1 October 2008 Available online 18 October 2008 Keywords: Density functional calculations Chemisorption Growth Interface states Surface structure SiC Nitrogen

a b s t r a c t Using the density functional theory this paper identiﬁes two phases of nitrogen which form on the 4H– SiC(0 0 0 1) Si-face. At 13 ML, N-adatoms occupy the sites between three surface silicon atoms bonding to each of the three available half-ﬁlled silicon dangling bonds. This passivates the surface dangling bonds and removes states from the upper half of the band gap. Above this coverage nitrogen atoms pair on the surface to form dimers with a corresponding change in the chemical potential. The nitrogen dimers reintroduce states into the SiC band gap. Between 13 ML and 1 ML coverage, the nitrogen redistributes in patches corresponding to regions of 13 ML coverage and 1 ML coverage. At 1 ML the nitrogen dimers populate all the silicon dangling bonds, thus forming a new surface phase. Above 1 ML a third bonding conﬁguration appears in which the nitrogen dimers are only a singly bonded to the surface. This conﬁguration saturates at 2 ML. Ó 2008 Elsevier B.V. All rights reserved.

1. Introduction The 4H–SiC polytype of SiC is a promising candidate for use in high power, high temperature semiconductor devices due to its wide range of physical properties [1,2]. Progress with this material has been hindered due to poor electron ﬁeld-effect mobility thought to arise from a high density of interface traps (Dit) [3–6]. The electron traps which are most detrimental are located just below the conduction band edge in 4H–SiC [7–9]. Possible sources for the traps have been assigned to interface defects due to carbon clusters and silicon suboxide bonds [8,10]. Evidence of graphitic regions at or near the SiC/SiO2 interface has not been consistently found, however the traps seem to exist independent of interface preparation, bringing this model into question [11–14]. Experiments have shown that the introduction of N or NO into the interface reduces Dit and improves electron mobility [1,11,15– 18]. Measurements on nitrogen treated gate oxides indicate that the nitrogen coverage at the 4H–SiC/SiO2 interface corresponds to 0.35 ± 0.13 ML and 1.02 ± 0.02 ML for the silicon and carbon faces, respectively [19]. Current explanations propose that nitrogen breaks up the interface carbon clusters or passivates suboxide bonds [20,21]. One problem associated with nitrogenation of the interface is that it has been found that, upon reoxidation, the nitrogen is depleted from the interface region and may be completely removed [21–23]. Thus the question of nitrogen stability at the surface in the presence of oxygen is an open question. * Corresponding author. Tel.: +1 202 767 3886. E-mail address: [email protected] (C.R. Ashman). 0039-6028/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.susc.2008.10.013

In terms of theoretical work, Olander et al. have done an extensive study of the adsorption of nitrogen containing species onto one and two fold sites of the Si surface [24]. In order to understand the role of nitrogen in passivating interface states, we recently presented work on nitrogen adsorption onto the SiC–Si surface [25]. That work showed that the interface states near the conduction band edge of the 4H–SiC silicon surface may be accounted for by the silicon surface dangling bonds. It explained the density of nitrogen on the silicon surface. It also offered an explanation for how nitrogen is able to passivate the near interface traps (NITs). This paper is a follow up to our previous work. In particular we examine the effects of nitrogen adsorption above 13 ML in greater detail. We ﬁnd that, although the 13 ML results in a stable low energy phase, a second stable phase appears at a coverage of 1 ML which reintroduces states into the band gap. These two phases are described by two different bonding conﬁgurations for the nitrogen. Up to 13 ML each nitrogen bonds to three surface silicon atoms at which coverage the silicon dangling bonds are saturated. Above 13 ML, additional nitrogen atoms combine to form N2 molecules. Each nitrogen in an N2 bonds with a single silicon atom until the surface dangling bonds become saturated at 1 ML. Above 1 ML, a third bonding conﬁguration takes place in which each N2 is bonded by one of the nitrogens to a single silicon atom. This bonding conﬁguration saturates at 2 ML. 2. Computational details The calculations are performed within the density functional theory (DFT), using the projector augmented wave method to solve

3618

C.R. Ashman, G. Pennington / Surface Science 602 (2008) 3617–3622

the electronic structure problem [26,27] and using periodic boundary conditions to consider the extended surface. A Car–Parrinello type ab initio molecular dynamics scheme was used for relaxation of the atomic positions [28]. A fundamental (1 1) SiC surface slab was constructed by ﬁrst optimizing the bulk SiC unit cell for lattice parameters and atomic positions. For the study of the silicon face, 15 Å of vacuum was added along the c-axis direction between repeated unit cells. The carbon surface dangling bond was passivated with an hydrogen atom to simulate the effects of bulk termination. This was achieved in two steps. First, both the silicon surface and the carbon surface were terminated with hydrogen. The SiC atomic positions were held at the bulk determined values while the hydrogens were allowed to relax to their optimum positions. Second, the silicon surface hydrogen atoms were removed while the carbon surface bilayer and hydrogen atom were ﬁxed at those values for the remainder of the study. The remaining, unfrozen atoms were optimized for atomic positions within the unit cell prior to the adsorption study. Three different supercells were used depending on the coverage being investigated. The principle supercell was composed of 9 (1 1) unit surface cells arranged in a (3 3) conﬁguration. In some cases we have also used a supercell composed of 4 (1 1) unit surface cells arranged in a (2 2) conﬁguration and a supercell composed of 3 unit cells. We used a k-point mesh consisting of 6 k-points along each of the surface reciprocal space directions of the (1 1) cell for the (3 3) and (2 2) conﬁgurations. For the supercell composed of 3 unit cells we used 3 k-points in each surface reciprocal space direction. The total error introduced by the k-point mesh mismatch is less than 0.007 eV per (1 1) unit cell which we consider negligible.

3. Results and discussion This paper uses the same nomenclature for identifying surface bonding sites as in [25]. To summarize, there are two types of hollow sites, h1, and h2. These sites are centered between three surface silicon atoms arranged in an equilateral triangle. The h1 sites do not contain a carbon atom in the surface bilayer while the h2 sites do. Throughout this paper we will refer to the triangular region including the three surface silicon atoms as an h1 region or h2 region where appropriate. We label the bonding sites on top of the silicon atoms with t and the bridge sites between two surface silicon atoms with b. Within this scheme there are four distinct types of sites on the surface for an adatom to bond to, namely h1, h2, t and b. To determine the energetics of a particular site, an adatom is deposited at a best guess of the optimum position for that site. Then the atomic positions of the slab are allowed to relax to lower the total energy via a Car–Parrinello molecular dynamics simulation at 0 K. We consider the adsorption of nitrogen onto the surface up to 2 ML coverage. A number of conﬁgurations have been sampled at each coverage to take into consideration the most likely surface bonding conﬁgurations. In addition to considering the range of bonding sites mentioned above, we have considered stacking of two nitrogen atoms particularly for coverages above 1 ML. From the starting conﬁgurations, the adsorbed surfaces are allowed to relax until the total energy has converged to at least 5 104 eV. Fig. 1 is a schematic of the phases identiﬁed in this work at 13 ML and 1 ML and the low energy conﬁgurations at 23 ML and 43 ML which illustrate intermediates between phases. Nitrogen phases are identiﬁed by an increase in the nitrogen chemical potential above the coverage corresponding to the phase combined with a unique, lowest energy pattern which appears at that coverage. We previously found that the h1 regions are the stable adsorption sites atoms up to 13 ML (see Fig. 1A). This coverage pﬃﬃﬃ forpnitrogen ﬃﬃﬃ has a ð 3 3ÞR30 surface pattern. At this site the three nitrogen

A

B

C

D

Surface Si

Surface C

Nitrogen

2nd bilayer C

Fig. 1. A top view schematic of the nitrogen coverage on the Si surface of 4H– SiC(0 0 0 1) at (A) 13 ML; (B) 23 ML; (C) 1 ML; and (D) 43 ML. (A), and (C) have been determined to be phases of on the SiC surface however (B) and (D) are not. The cell depicts a (3 3) surface including atoms which lie on the boundary of the region. Thick black lines connecting nitrogens indicate bonds. Thin black triangles are meant to aid the eye and do not represent bonds. The parallelograms (red) superimposed on the diagrams indicate the fundamental unit cells. (For interpretation of the references to colour in this ﬁgure legend, the reader is referred to the web version of this article.)

p electrons bond to the three nearest neighbor silicon dangling bonds, thus removing the silicon states from the upper half of the band gap while contributing states just above the valence band edge. Up to 13 ML coverage the nitrogen atoms populate h1 sites randomly on the surface. At 13 ML, all the h1 sites are occupied. This also means that there are no unbonded silicon surface dangling bonds. As a result the nitrogens occupy all the periodically arranged sites on the surface and there are no states in the upper half of the band gap with respect to bulk SiC. Fig. 2 shows the density of states (DOS) for the 13 ML coverage. It can be seen that the band gap between the bulk valence band and conduction band is state free. Also included in the top panel of the ﬁgure for comparison purposes is the total DOS for the clean 4H–SiC(0 0 0 1) surface. We ﬁnd this surface to be insulating with a ﬁlled state lying in the gap, above the bulk valence band edge and an empty state below the bulk conduction band edge, in agreement with calculations done at a higher level of theory and with recent experiments [29,30]. In Fig. 3 we plot the relative binding energies per (1 1) unit cell of the various conﬁgurations studied in this work. This is equivalent to a plot of the Gibbs free energy at zero temperature and pressure [31]. The rectangles in the ﬁgure represent the lowest energy conﬁgurations at a given coverage. The triangles represent higher energy conﬁgurations and are included to illustrate the completeness of our search for stable phases. This is necessary since there is no current method which guarantees that one can ﬁnd the lowest energy conﬁgurations. All of the high energy conﬁgurations (triangles) are at local minima which may have been reached either by a fortuitous starting conﬁguration or relaxed to such a conﬁguration during the simulation. The solid line connecting the squares represents the low energy boundary. The slope of this line gives the chemical potential relative to the chosen reference, namely gas phase N2 [31]. No structures exist below this line. Conﬁgurations which lie on this line have been determined in this work to be the lowest energy conﬁgurations at a given coverage. Conﬁgurations which lie above this line but below the zero of energy are bound to the surface with respect to gas phase N2 at zero

C.R. Ashman, G. Pennington / Surface Science 602 (2008) 3617–3622

3619

Fig. 3. Plot of the nitrogen adsorption energy per (1 1) surface cell. The squares indicate the lowest energy conﬁgurations as determined in this study and in [25]. The triangles indicate higher energy conﬁgurations at a given coverage. The solid line connecting the squares represents the low energy boundary. The adsorption energies are taken relative to 12 the DFT total energy of N2.

Fig. 2. The layer resolved density of states for the 13 ML N coverage. The solid black line in the uppermost panels shows the total density of states of all layers combined. The next panel below shows the density of states for the surface nitrogen only. Below that is the surface SiC layer with subsequent panels representing deeper layers. The black line in the lowest layer is the DOS for the hydrogen terminated carbon surface. The grey (red) line on the graphs is the DOS for the clean SiC silicon surface (top panel), and bulk SiC (bottom panel). The vertical hatched lines represent the valence band (VB), fermi level (FL) and conduction band (CB) edges, respectively, of the bulk SiC. The fermi levels were aligned with 0 eV. (For interpretation of the references to colour in this ﬁgure legend, the reader is referred to the web version of this article.)

temperature and pressure. However, those that lie above the low energy boundary are considered to be only metastable. Stable phases occur at coverages for which there is a conﬁguration lying on the low energy boundary line and at which there is an increase in the chemical potential as indicated by a change in the slope of the line. Thus we identify phases at 13 ML and 1 ML. 3.1. Nitrogen surface migration barriers Regarding the higher energy conﬁgurations indicated by the triangles in Fig. 3, it can be considered that there are two types of metastable conﬁgurations, namely those for which there exists a more stable conﬁguration at the same coverage and those for which no such lower energy conﬁguration exists but which do not lie on the low energy boundary. For this latter case, at coverages above and below such a conﬁguration there exist conﬁgurations which have lower chemical potentials. One can therefore create a lower energy surface by redistributing the adsorbed atoms in such a way that the total number of atoms does not change, but the surface is now constituted of two lower energy surfaces. In order to quantify the energy barriers separating various conﬁgurations at a given coverage we plot the total energy as the

adsorbate species migrates along the surface. We do this for only a limited number of cases because of the large computational resources required however we believe it is sufﬁcient to conﬁrm that the high energy conﬁgurations presented in Fig. 3 are at local minima. Likewise, the barrier to the formation of a surface bound N2 molecule from a single nitrogen atom migrating along the surface into the bonding vicinity of a second may also be determined. We determine the barriers as follows. A line segment connecting the starting point and the desired end point on the surface is determined. This line segment is divided into approximately 4000 parts representing steps along the path for the adatom to migrate. At each of these steps, the adatom is incremented along the line segment but allowed to move freely on the surface of a plane orthogonal to the line segment. Thus the line segment gives one component of the adatom displacement vector while the plane gives the other two. All other atoms on the surface are given completed freedom to relax. The total energy is calculated at each step and a plot of the total energy versus displacement along the line segment is made. Because the step size is kept small the species stays close to the Born Oppenheimer surface. This allows the species to follow a nearly minimum energy path. Note, however, that it does not take into consideration the effects of temperature or vibrational modes and therefore only presents an upperbound to the migration barriers. Fig. 4 shows such a plot for one of the adsorbed nitrogens to mi grate from the lowest energy h1 location to a second h1 location at 1 a surface coverage of 3 ML. The hatched line in the inset shows the path that the nitrogen follows. The h1 differs from the starting point in that it shares nearest neighbor surface Si atoms with other nitrogens. From this plot we see that the bridge sites are not local minima. There is a barrier of about 1.7 eV for a nitrogen to migrate from the h1 to h2 site. There is a much smaller barrier of about 0.2 eV for the nitrogen to migrate from the h2 to the h1 site which lies 0.5 eV higher in energy than the h1 site. We were unable to determine an energy barrier to the formation of an N2 dimer at coverages up to 13 ML. As a nitrogen atom is incrementally moved into an occupied h1 region from a neighboring h2 region the occupying nitrogen atom migrates out of the h1 region indicating that there is a strong repulsion between the nitrogens. We were, however, able to identify an energy barrier

3620

C.R. Ashman, G. Pennington / Surface Science 602 (2008) 3617–3622

bonds to a neighboring silicon atom in the plane. The top of the barrier lies about 0.6 eV above the starting conﬁguration. 3.2. Coverage above

Fig. 4. Plot of the variation of the total energy as a single nitrogen atom migrates along the surface of a 13 ML coverage (3 3) surface cell. The path is indicated in the inset by a hatched line. The vertical hatched lines identify points at which the nitrogen atom lies on the h1, b, h2 and h1 locations. The curve is shifted so that the lowest energy point occurs at 0 eV.

1 3

ML

At coverages above 13 ML there are no surface dangling bonds available for the nitrogen to bond to. The only choice is for the nitrogen to disrupt Si–N bonds. There are two possibilities for the nitrogen to adsorb onto this surface. The ﬁrst is that one nitrogen atom bonds to two surface sites and the other atom bonds to a single surface site. This leaves one of the nitrogen p electrons unpaired on one of the atoms and two unpaired on the second, thus it is unstable. Alternatively, the nitrogen form double bonded dimers which bond to two surface sites along a bridge between adjacent Si atoms in an h1 region. In this conﬁguration, as an additional nitrogen is adsorbed onto an h1 region, a nitrogen dimer is formed, bridging two surface sites and leaving one of the surface silicon dangling bonds unpassivated (see Fig. 1B). We ﬁnd the N–N bond length to be 1.22 Å which is marginally shorter than a N2 double bond length of 1.25 Å. This is in good agreement with the work of Olander et al. [24]. At 23 ML each of the h1 regions is populated by a single nitrogen dimer and has a single half-ﬁlled surface silicon dangling bond. Each such silicon dangling bond contributes a half-ﬁlled state centered below the middle of the bulk silicon band gap and a halfﬁlled empty state below the conduction band edge. These states

of about 3.7 eV to form a single N2 molecule in the unit cell at a coverage of 49 ML. The resulting plot is presented in Fig. 5. A similar energy was found for the formation of a single N2 on the 1 ML surface. We have also looked at the barrier for an N2 to walk along the surface. In this migration, one of the nitrogen atoms travels tangentially to the nitrogen–nitrogen bond from the top of one silicon atom to the top of a neighboring silicon atom, while the other nitrogen acts as a pivot point, remaining essentially stationary. We ﬁnd the barrier to be a bit less than 0.5 eV at a coverage of 23 ML. A second migration path occurs when one of the nitrogen atoms in the pair is incremented along a path parallel to the nitrogen–nitrogen bond. The incremented nitrogen travels parallel to the bridge between two surface silicon atoms. The second nitrogen rotates in a vertical plane around the incremented nitrogen and

Fig. 5. Plot of the variation of the total energy as a single nitrogen atom migrates from an h2 site to a neighboring, occupied h1 site. The conﬁguration is a 49 ML coverage (3 3) surface cell. The vertical hatched line identiﬁes a local minimum in which one nitrogen is at a bridge site N(b) and the second is at a top site N(t). The curve is shifted so that the lowest energy point occurs at 0 eV.

Fig. 6. The layer resolved density of states for the 23 ML N coverage. The construction of the graph is similar to Fig. 2 with the exception that the top (surface) layer of the SiC includes the DOS for the nitrogen bonded component of the surface bilayer in black and the unpassivated bilayer component in grey(red). (For interpretation of the references to colour in this ﬁgure legend, the reader is referred to the web version of this article.)

C.R. Ashman, G. Pennington / Surface Science 602 (2008) 3617–3622

3621

Fig. 7. The layer resolved DOS for the 1 ML N coverage (A) and 1 ML NH coverage (B). The construction of the graph is similar to Fig. 2.

have been observed in experiments for the clean surface [30]. Interestingly they are similar in energy to those contributed by the highest occupied and lowest unoccupied states (HOMO and LUMO, respectively) of the N2 dimer as shown in Fig. 6. This means that, for coverages of nitrogen above 13 ML, it is not possible to remove the silicon surface states from the gap by incorporating nitrogen into the surface alone. Below 23 ML coverage of nitrogen there are still some h1 regions which are fully passivated by the 13 ML phase. Above this coverage N2 populate the remaining dangling bonds until at 1 ML the surface is completely populated so that there is one N2 per two surface silicon sites. This represents the second phase above which there is an increase in the chemical potential as may be seen in Fig. 3. Similar to the 23 ML case, a plot of the layer resolved density of states (DOS) shows that the region just above the valence band edge is populated by the HOMO states from the nitrogen dimers and just below the conduction band edge are the LUMO states (Fig. 7A). This is very much analogous to the clean SiC silicon surface which we have superimposed (red) on the total DOS for the nitrogenated surface. Our study indicates that these states can be pulled down in energy, out of the gap, by adsorption of one hydrogen per atom (Fig. 7B). The 1 ML pﬃﬃﬃ nitrogen pﬃﬃﬃ surface no longer has the ð 3 3ÞR30 surface symmetry of the 13 ML phase. Instead, it is characterized by a (2 1) surface imposed by the nitrogen dimers, overlaying the original 4H SiC surface symmetry as can be seen in Fig. 1C. Above 1 ML we ﬁnd the onset of a new surface pattern formation. There is no way for the nitrogens to rearrange their bonding on the surface to allow for additional surface bonding sites. Instead, the nitrogens stack, breaking the bonds between surface nitrogens and forming N2 molecules. In this conﬁguration, only one of the nitrogen atoms in the molecule bonds to the surface while the second is displaced away from the surface. This bonding conﬁguration continues to replace that of the 1 ML coverage until each silicon surface atom is bonded by one N2. The 43 ML coverage pictured in Fig. 1D serves to illustrate an intermediate coverage containing both the 1 ML coverage and the 2 ML coverage. We have not investigated the coverage above 2 ML and cannot make any deﬁnite statements with regards to changes in the chemical potential for nitrogen adsorption above 2 ML therefore we cannot conﬁrm this coverage to be associated with an additional phase.

4. Conclusion In conclusion, in addition to the previously identiﬁed phase at 13 ML we ﬁnd a phase at 1 ML. The 13 ML phase coincides with the disappearance of silicon surface states from the upper half of the SiC band gap while ﬁlled states appear just above the valence band edge. This phase is stable relative to higher coverages of nitrogen and can be selected by controlling the partial pressure of nitrogen at the surface. Above 13 ML ﬁlled states shift upwards, into the band gap while empty states appear below the conduction band edge due to the reappearance of silicon surface states and the onset of the 1 ML phase. The 1 ML phase exhibits gap states similar to those of the clean SiC silicon surface which can be passivated with hydrogen. Acknowledgements Simulations were performed on computers at the Army Research Laboratory with the assistance of the PET program of the Department of Defense High performance Computing Modernization Ofﬁce. We thank those institutions for their valuable support. References [1] G.Y. Chung, J.R. Williams, K. McDonald, L.C. Feldman, J. Phys. C 16 (2004) S1857. [2] M. Losurdo, M.M. Giangregorio, P. Capezzuto, G. Bruno, A.S. Brown, T.-H. Kim, C. Yi, JEM 34 (2005) 457. [3] R. Schörner, P. Friedrichs, D. Peters, IEEE Trans. Electron Devices 46 (1999) 533. [4] N.S. Saks, S.S. Mani, A.K. Agarwal, APL 76 (2000) 2250. [5] G.Y. Chung, C.C. Tin, J.H. Won, J.R. Williams, Mat. Sci. For. 338–342 (2000) 1069. [6] M.K. Das, B.S. Um, J.A. Cooper Jr., Mat. Sci. For. 338–342 (2000) 1069. [7] V.V. Afanas’ev, A. Stesmans, Phys. Rev. Lett. 78 (1997) 2437. [8] V.V. Afanasev, M. Bassler, G. Pensl, M. Schultz, Phys. Stat. Sol. A 162 (1997) 321. [9] J.M. Knaup, P. Deák, T. Frauenheim, A. Gali, Z. Hajnal, W.J. Choyke, Phys. Rev. B 72 (2005) 115323. [10] B. Hornetz, H.-J. Michel, J. Halbritter, J. Mater. Res. 9 (1994) 3088. [11] C. Virojanadara, L.I. Johansson, Surf. Sci. 530 (2003) 17. [12] E. Pippel, J. Woltersdorf, H.Ö. Olafsson, E.Ö. Sveinbjörnsson, J. Appl. Phys. 97 (2005) 034302. [13] J.W. Chai, J.S. Pan, Z. Zhang, S.J. Wang, Q. Chen, C.H.A. Huan, Appl. Phys. Lett. 92 (2008) 092119. [14] V. van Elsbergen, M. Rohleder, W. Moench, Appl. Surf. Sci. 134 (1998) 197. [15] H. Li, S. Dimitrijev, H.B. Harrison, D. Sweatman, APL 70 (1997) 2028.

3622

C.R. Ashman, G. Pennington / Surface Science 602 (2008) 3617–3622

[16] G.Y. Chung, J.R. Williams, T. Isaacs-Smith, F. Ren, K. McDonald, L.C. Feldman, APL 81 (2002) 4266. [17] V.V. Afanas’ev, A. Stesmans, F. Ciobanu, G. Pensl, K.Y. Cheong, S.D. Dimitrijev, Appl. Phys. Lett. 82 (2003) 568. [18] S. Dhar, Y.W. Song, L.C. Feldman, T. Isaacs-Smith, C.C. Tin, J.R. Williams, G. Chung, T. Nishimura, D. Starodub, T. Gustafsson, et al., APL 84 (2004) 1498. [19] K.-C. Chang, Y. Cao, L.M. Porter, J. Bentley, S. Dhar, L.C. Feldman, J.R. Williams, J. Appl. Phys. 97 (2005) 104920. [20] V.V. Afanas’ev, A. Stesmans, M. Bassler, G. Pensl, M.J. Schulz, C.I. Harris, J. Appl. Phys. 85 (1999) 8292. [21] K. McDonald, R.A. Weller, S.T. Pantelides, L.C. Feldman, G.Y. Chung, C.C. Tin, J.R. Williams, J. Appl. Phys. 93 (1999) 2719.

[22] [23] [24] [25] [26] [27] [28] [29]

K. Chatty, V. Khemka, T.P. Chow, R.J. Gutmann, JEM 28 (1999) 161. P. Hoffman, A. Goryachko, D. Schmeisser, Mater. Sci. Eng. B 118 (2005) 270. J. Olander, K.M.E. Larsson, Phys. Rev. B 67 (2003) 115306. G. Pennington, C.R. Ashman, App. Phys. Lett. 91 (2007) 072106. W. Kohn, L.J. Sham, Phys. Rev. 140 (1965) A1133. P.E. Blöchl, Phys. Rev. B 50 (1994) 17953. R. Car, M. Parrinello, Phys. Rev. Lett. 55 (1985) 2471. K.V. Emtsev, T. Seyller, L. Ley, A. Tadich, L. Broekman, J.D. Riley, R.C.G. Leckey, M. Preuss, Surf. Sci. 600 (2006) 3845. [30] V.I. Anisimov, A.E. Bedin, M.A. Korotin, G. Santoro, S. Scandolo, E. Tosatti, Phys. Rev. B 61 (2000) 1752. [31] C.R. Ashman, C.J. Först, K. Schwarz, P.E. Blöchl, Phys. Rev. B 69 (2004) 075309.

First-principles study of oxygen adsorption on the ...