Effect of the adsorption of ethylene carbonate on Si surfaces on the Li insertion behavior

Chemical Physics Letters 585 (2013) 157–161

Contents lists available at ScienceDirect

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

Effect of the adsorption of ethylene carbonate on Si surfaces on the Li insertion behavior Alexandra Carvalho a,b, Mark J. Rayson c, Patrick R. Briddon d, Sergei Manzhos e,⇑ a

Department of Physics, I3N, University of Aveiro, 6 Science Drive 2, 3810-193 Aveiro, Portugal Graphene Research Centre, National University of Singapore, Campus Santiago, 117576, Singapore c Department of Engineering Sciences and Mathematics, Luleå University of Technology, Luleå S-97187, Sweden d Electrical, Electronic and Computer Engineering, University of Newcastle upon Tyne, Newcastle upon Tyne NE1 7RU, United Kingdom e Department of Mechanical Engineering, Faculty of Engineering, National University of Singapore, Block EA #07-08, 9 Engineering Drive 1, Singapore 117576, Singapore b

a r t i c l e

i n f o

Article history: Received 18 May 2013 In final form 2 September 2013 Available online 7 September 2013

a b s t r a c t Effects of the ethylene carbonate (EC) solvent on Li insertion and diffusion in Si anodes are studied using density functional theory. On both (1 0 0) and (1 1 1) reconstructed surfaces of Si, a semi-dissociated (SD) configuration of EC is stable and most favorable for Li insertion, lowering its barrier by up to 0.2 eV vs a clean surface. The less stable molecular adsorption has little effect on Li insertion and diffusion, while the surface ketone formed by dissociating the SD configuration at a cost of 0.6 eV has a strong detrimental effect on Li insertion, increasing its barrier by up to 0.4 eV. Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction Metal ion batteries are a key technology for powering portable electronics and for achieving sustainability. Li ion batteries provide today the highest energy density at about 160–180 Wh/kg, which is still only about one tenth of that of gasoline [1]. Further increases in capacity, energy density, and rate capability are required to enable a wide use of all-electric vehicles [1] and bulk electrochemical storage [2]. This can be achieved by designing new electrode materials and nanostructures. For the longest time, carbon-based anodes have been used in commercial batteries, with the capacity of about 370 mAh/g, still exceeding that of common cathodes (about 150 mAh/g for LiCoO2 [3]). However, as the capacity of cathodes continues to increase, reaching about 500 mAh/g in experimental cathodes [4,5], more benefit can be obtained by increasing the specific capacity of the anode. One of the most promising anode materials is Si, with a theoretical capacity of about 4200 mAh/g. It has been suggested that the issues of cyclability of Si anodes due to a large volume expansion (more than 300%) upon lithiation and pulverization with a loss of connectivity can be solved via nanostructuring [6–10]. Si and other group IV elements can today be considered as practical anode materials for Li ion batteries and are also actively studied as potential anode materials for other metal ion batteries [11–13]. It should be noted, however, that good cyclability has been achieved at specific capacities significantly lower than the theoretical capacity [6–10]. The rate capability,

⇑ Corresponding author. Fax: +65 6779 1459. E-mail address: [email protected] (S. Manzhos). 0009-2614/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.cplett.2013.09.006

which is determined by insertion barriers and diffusivity of Li, is also lower than that of carbon-based anodes. This is confirmed by multiple experiments and by theoretical calculations of the barriers in bulk c-Si [14–18] and at different surface facets of Si [19,20]. Barriers to insertion into the (1 0 0)/(1 1 1) surfaces and for subsurface diffusion were computed to be about 0.84–0.88/ 0.51 eV and 0.48–0.63/0.53–0.74 eV respectively, and the barrier for single-atom diffusion in the bulk was found to be 0.55– 0.60 eV. These numbers compare very unfavorably with barriers to insertion and diffusion in graphite, as low as 0.2 eV [21,22] as well as with barriers in new high-capacity cathode materials [23]. The barriers computed so far are also very approximate in the sense that they ignore the presence of other atomic and molecular species. In real batteries, the interaction with the electrolyte and with Li leads to surface modification, eventually leading to the formation of an oxide or silicate layer [24]. These are expected to change drastically the barrier to insertion and diffusion. For Si specifically, the influence of Li–Li interactions has already been confirmed by density functional theory and MD studies [14,25,26]. Adsorbed species have also been shown to modify the insertion dynamics of other dopants [27]. The presence of molecules at the anode surface is also expected to influence significantly insertion dynamics, yet is much less studied. The surface of the anode is in contact with electrolyte species, and specifically with solvent molecules. Further, common electrolytes are reduced at low voltages, and the reduction products depend on the kind of lithium salt used. Although surface complexes might be formed, these are often ignored in computational models. Continuum solvent models would not be appropriate to analyze the effects due to the solvent, especially, as shown in this letter, if solvent molecules form strong

158

A. Carvalho et al. / Chemical Physics Letters 585 (2013) 157–161

bonds with or dissociate at the surface. The interaction of solvent molecules with the anode has to be studied explicitly at the atomic level, which is done here. It has been computed that ethylene carbonate (EC) – a common solvent species used in batteries [28,29] – decomposes on the surface of some cathode materials, with implications for cathode cycling properties [30]. Chemical reactions at the graphite anode involving EC have been the subject of several density functional theory and molecular dynamics studies [31–33]. Recently, the interaction of (molecularly adsorbed) and fluorinated EC with graphene – a common component of experimental anodes [34– 36] – was considered at the atomistic level [37]. However, atomistic studies of the formation of adsorbate complexes at the Si anode surface, possibly following solvent dissociation, and their effect on the rate, are still lacking. In the present work, we consider the EC–Si interface and address the following questions: (i) what surface complexes are formed by EC adsorption at the Si(1 0 0) and (1 1 1) surfaces? (ii) how do these surface complexes affect the insertion of Li? and (iii) is it in principle possible to use adsorbed molecules to lower the insertion and/or diffusion barrier. We use density functional theory, which has been successfully used to model diffusion in Li batteries [14,25,26] and other electrochemical devices (see for example Ref. [38]) The article is organized as follows. In Section 2, we describe the details of the modeling method. In Section 3, we consider the results for the Si(1 0 0) surface, followed by those for the Si(1 1 1) surface. Section 4 concludes.

2. Method The simulations were carried out within the framework of density functional theory, as implemented in the AIMPRO code [39,40]. The core electrons were accounted for by using the dual space separable pseudopotentials of Hartwigsen et al. [41]. All three electrons of Li were explicitly included in the calculation. The basis set consists of Cartesian s; p and d-type Gaussian functions centered on each atom with four different exponents optimized for each species [42]. The basis sets used on Si, Li, O, C sites were, respectively, of type ddpp (28 functions per atom), pppp (16 functions per atom), dddd (40 functions per atom), and pddd (34 functions per atom). For hydrogen, a contracted basis set (C44G*) with 13 functions per atom, including a polarization function, was used. The exchange and correlation energy functional was approximated by the gradient approximation of Perdew, Burke and Ernzerhof (PBE) [43]. The Si(1 0 0) and Si(1 1 1) surfaces were represented by slabs, with periodic boundary conditions along the three dimensions. The slabs consisted of six or ten layers of silicon atoms. Unless otherwise stated, the results of the six-layer calculation will be given. The slabs were separated by at least 1.7 nm from the image slabs, and identical points in the slabs were separated by more than 1.3 nm from their images in the neighboring supercells. The reciprocal space was sampled at 2  2  1 special k-points [44]. Each slab had one reconstructed surface to be used for modeling. We considered the ð4  2Þ reconstruction for the Si(1 0 0) surface (Figure 1a)), which is its lowest energy configuration [45]. For the Si(1 1 1), we considered a ð4  2Þ reconstruction, since it is locally very similar to the most stable ð7  7Þ surface reconstruction[46], but can be simulated at a lower computational cost. A similar strategy has been employed in previous studies of molecular adsorption and Li in-diffusion [47,20]. The back surface of the slabs was passivated with hydrogen, and both the passivizing hydrogen atoms and the closest layers of Si atoms (two in the case of the 6-layer slab, or four in the case of the 10-layer slab) were kept fixed during the structural optimiza-

Fig. 1. Si(1 0 0) (4  2) surface: (a) clean surface, (b) surface with an adsorbed EC molecule in the OM configuration (c) or in the SD configuration, and (d) with an adsorbed ketone group. Si, C, O and H atoms are represented by light gray, dark gray, black and white spheres, respectively. Points represent the positions considered for the Li atoms. The orientation of figures (b) and (e) is the same as (a). Higher and lower-lying atoms of the reconstructed dimer rows are indicated by ‘h’ and ‘l’.

Table 1 Relative energies of molecularly adsorbed (OM), semidissociated chemisorbed (SD) and dissociated (K) EC on (0 0 1) and (1 1 1)-surfaces. Ead is the enthalpy of adsorption, whereas Eb is the energy barrier. All values are in eV. Adsorption mode

Ead (1 0 0)

OM SD K

1.1 3.2 2.5

Eb (1 0 0)

Ead (1 1 1)

Eb (1 1 1)

1.63 1.88

0.4 1.9 1.7

3.20 4.34

tion. The lattice parameters were also kept constant, and identical to the bulk value (5.50 Å) along the directions parallel to the surface. The minimum energy paths for diffusion and the respective saddle points were determined using the climbing image nudged elastic band method (NEB) [48] with five images. The highest energy image was allowed to move along the direction of the band (climb) after five iterations of the regular NEB method. 3. Results 3.1. Si(1 0 0) surface We first consider the adsorption modes of the EC molecule on the Si(1 0 0) surface. The oxygen forming the C@O bond in the isolated molecule can form an additional bond to a surface Si atom, as

159

A. Carvalho et al. / Chemical Physics Letters 585 (2013) 157–161

Table 2 Energetics of Li for the surface and subsurface layers of the Si(1 0 0) surface, with and without absorbed EC or ketone. Numbers inside brackets were obtained using the 10-layer slab. All values are in eV. surface

Clean surf.

Er ðT4Þ Er ðT40 Þ Er ðS1Þ Er ðS10 Þ Er ðS2Þ Em ðT4 ! S1Þ Em ðT40 ! S10 Þ Em ðS1 ! S2Þ

0.00 0.16 0.51 0.54 0.62 1.00 0.79 0.62

EC-OM (0.00) (0.16) (0.51) (0.54) (0.62) (1.04) (0.78) (0.62)

0.00 0.18 0.53 0.51 0.60 1.07 0.70 0.65

EC-SD (0.00) (0.17) (0.49) (0.71) (0.50) (1.07) (0.70) (0.65)

0.00 0.00 0.39 0.44 0.34 0.87 0.77 0.52

K (0.00) (0.01) (0.37) (0.43) (0.31) (0.84) (0.84) (0.52)

0.00 0.13 0.68 0.54 0.58 1.38 0.84 0.59

(0.00) (0.13) (0.75) (0.53) (0.58) (1.36) (0.72) (0.58)

Table 3 Energetics of Li for the surface and subsurface layers of the Si(1 1 1) surface, with and without absorbed EC or ketone. Numbers inside brackets were obtained using the 10-layer slab. All values are in eV. Surface Er ðH3Þ

Er ðS1Þ

Er ðH30 Þ Er ðS10 Þ Er ðS2Þ Em ðH3 ! S1Þ

Em ðH0 3 ! S10 Þ Em ðS1 ! S2Þ

Clean surf. (i) (ii) (iii) (iv) (i) (ii) (iii) (iv)

(i) (ii) (iii) (iv)

EC-OM

EC-SD

0.00

(0.00)

0.00

(0.00)

0.56

(0.56)

0.41

(0.40)

0.38 0.46 0.58 1.10

(0.42) (0.49) (0.58) (1.10)

0.28 0.29 0.50 0.92

(0.28) (0.30) (0.52) (0.98)

0.34 0.60

(0.34) (0.69)

0.28 0.70

(0.27) (0.69)

shown in Figure 1b. This configuration will be noted OM. The reaction is exothermic with an adsorption energy of 1.1 eV (Table 1). A more stable configuration results from breaking one of the C–O bonds to form another Si–O bond and a Si–C bond (Figure 1c). In this semi-dissociated (SD) configuration EC has a binding energy of 3.2 eV. The dissociation of a free EC molecule into a surface ketone (K) group (Figure 1 and a free ethylene molecule is also energetically favorable, with an enthalpy change of 2.6 eV. However, the energy barrier to semidissociate EC is lower than for complete dissociation into K and an ethylene molecule (Table 1). Thus, SD is more likely to be found on the Si(1 0 0) surface than K. The insertion of Li into (clean) Si(1 0 0) and Si(1 1 1) surfaces has been analyzed in detail in previous studies [49,20]. We follow the same site labeling convention as used in Ref. [20] for the clean surface. The most stable sites for Li adsorption are on the valleys in between the dimer rows, either between the higher-lying atoms (T4) or between the lower-lying atoms (T40 ). These are nearly degenerate in energy for the clean surface (Table 2). The positions in the sub-surface layer (S1 and S10 ) are higher in energy by about 0.5 eV. The lowest energy path for Li insertion is T40 ! S10 , which requires an activation energy Em ðT40 ! S10 Þ ¼ 0:8 eV. The T4 ! S1 path requires an activation energy of 1.00 eV. These values are in agreement with previous calculations, which obtained Em ðT4 ! S1Þ ¼ 0:88-1:126 eV and Em ðT40 ! S10 Þ ¼ 0:837 eV [49,20]. It is also converged within 0.07 eV with the thickness of the slab used in the simulation. The direct interaction between adsorbed EC molecules and Li in some cases stabilizes Li at the surface positions. However, EC adsorption on the Si surface also weakens the Si–Si bonds, potentially decreasing the Li migration barrier. Hence, the insertion barriers are different in the proximity of adsorbed EC molecules or

1.28 0.73 0.85 0.00 0.77 1.27 0.62 0.80 0.51 0.73 0.69 0.65 1.13 0.98 1.36 0.50 0.62

K (1.30) (0.77) (0.91) (0.00) (0.78) (1.28) (0.84) (0.79) (0.51) (0.73) (0.70) (0.67) (0.99) (0.83) (1.34) (0.44) (0.63)

0.00

(0.00)

1.05

(1.03)

0.62 1.52 1.05 1.65

(0.69) (1.59) (1.07) (1.67)

1.11 0.58

(1.12) (0.60)

ketone groups. Whereas OM or SD EC adsorption changes little or even decreases the insertion barriers, a ketone group resulting from dissociation of EC increases the activation barriers for both insertion channels by up to 0.4 eV (Table 3). The barrier for the next insertion step S1 ! S2 is already little affected. The lowest energy paths for Li insertion on the clean Si(1 0 0) surface and in the neighborhood of the SD EC molecule or the ketone are shown in Figure 3. 3.2. Si(1 1 1) surface The modes of adsorption on the Si(1 1 1) surface are very similar to those considered for the Si(1 0 0) surface. Non-dissociative adsorption in the OM configuration is most stable if the oxygen forms a bond with the Si ad-atom (Figure 2b). The adsorption energy in this case is only Ead ¼ 0:4 eV. In the SD configuration, the molecule forms only two bonds with the surface, the carbon binding to the Si ad-atom and the oxygen to the Si rest-atom (Figure 2c), in contrast with the Si(1 0 0) surface where three bonds were formed. The adsorption energy is Ead ¼ 1:9 eV. Dissociation into a surface-anchored ketone (Figure 2) and a free ethylene molecule has an enthalpy gain of Ead ¼ 1:7 eV, so that, as in the case of the (1 0 0) surface, the SD configuration was also found to be the most stable. The most stable site for Li adsorption on the Si(1 1 1) surface is the H3 site (Figure 2a). The H30 site, closer to the Si rest-atom than to the Si ad-atom, is higher in energy by about 0.4 eV. Nevertheless, it is energetically favorable to use H30 as an intermediate position for insertion into the subsurface layer, since the activation energy for the H30 ! S10 step is only 0.3 eV, in contrast with Em ðH3 ! S1Þ ¼ 1:1 eV.

160

A. Carvalho et al. / Chemical Physics Letters 585 (2013) 157–161

Fig. 3. Configuration-coordinate diagram for the T4 ! S1 ! S2 path at the clean Si(1 0 0) surface and in the neighborhood of SD EC molecule or a ketone.

Fig. 2. Si(1 1 1) (4  2) surface: (a) clean surface, (b) surface with an adsorbed EC molecule in the OM configuration (c) or in the SD configuration, and (d) with a surface ketone group. Si, C, O and H atoms are represented by light gray, dark gray, black and white spheres, respectively. Points represent the positions considered for the Li atoms. The orientation of figures (c) and (d) is the same as (a).

Similar to the Si(1 0 0) surface, the presence of OM-adsorbed EC decreases very slightly the insertion barriers. As the energy step between the outer surface and the first subsurface layer is smaller, the activation energy for migrating from H3 to S10 via H30 is reduced from 0.7 eV to 0.6 eV. The presence of SD EC affects several near H3 sites. The most important ones are identified in Figure 2c. Of those, the lowest energy is H3ðiv Þ. The H3ðiÞ site, which is between the Si rest-atom and the Si ad-atom the semidissociated EC molecule binds to, is higher in energy by 1.30 eV. Once Li is trapped on the H3ðiv Þ site, it is difficult to diffuse in, as the required activation energy is close to 1.4 eV. However, other H3 ! S1 channels, namely (i) and (ii), have lower activation energies between 0.6 and 1.0 eV. The barrier for H30 ! S10 increases. The ketone is even more harmful, as it interacts directly with the Li atom increasing the H30 ! S10 barrier to 1.1 eV, and may react with Li close to the nearby H3 sites, releasing CO2. The lowest energy paths for Li insertion on the clean Si(1 1 1) surface and in the neighborhood of the SD EC molecule or the ketone are shown in Figure 4.

4. Conclusions We have used density functional theory to study the effects of the interaction between electrolyte species and the surface of a silicon anode on Li insertion properties. It was shown that adsorption

Fig. 4. Configuration-coordinate diagram for the T4 ! S1 ! S2 path at the clean Si(1 1 1) surface and in the neighborhood of SD EC molecule or a ketone.

of ethylene carbonate in different configurations on (1 0 0) and (1 1 1) reconstructed surfaces can have a significant effect on the barrier to insertion, promoting or hindering it, depending on the surface complex formed. Specifically, the presence on the surface of the semi-dissociated ethylene carbonate is (i) likely, as it is energetically favored both on the (1 0 0) and (1 1 1) surfaces, and (ii) favorable for Li insertion into the (100) surface, reducing the T4 ! S1 insertion barrier by up to 0.2 eV. The effect on the T40 ! S10 barrier is small, and an overall speedup of insertion kinetics is expected. The ketone configuration, on the contrary, leads to a significant increase of the barrier for both surfaces, by up to 0.38 eV. Dissociation of EC leaving a surface ketone group is energetically favorable and should be avoided in real batteries. In contrast, the SD configuration is lower in energy, and therefore its dissociation into ethylene and a ketone group is not likely in the system considered here due a large enthalpy cost. The lowering of the barriers by the SD configuration is only relatively small (0.2 eV out of 1.0 eV for insertion and 0.1 out of 0.6 eV for diffusion) but is much larger than kT at room temperature

A. Carvalho et al. / Chemical Physics Letters 585 (2013) 157–161

(about 0.025 eV) and is therefore expected to lead to improved rate capability of the anode. More generally, our results show that it is possible to facilitate Li insertion by functionalizing a crystalline anode with simple molecules. In the present case, this is achieved by using none other than a solvent species, but other molecules deserve to be tried. It is important to do so, as the insertion and diffusion barriers achieved in high-capacity anode materials are still much higher than in graphite [21]. Acknowledgement A.C. was supported by the Marie Curie Program PEOPLE (Ref. REG/REA.P1(2010)D/22847) FCT (Pest) and COST NanoTP (MP0901). S.M. acknowledges the support from the AcRF Tier 1 Grant R-265-000-430-133 from the Ministry of Education of Singapore. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]

M.M. Thackeray, C. Wolverton, E.D. Isaacs, Energy Environ. Sci. 5 (2012) 7854. J. Rugolo, M.J. Aziz, Energy Environ. Sci. 5 (2012) 7151. T. Ohzuku, A. Ueda, J. Electrochem. Soc. 141 (1994) 2972. L. Li, F. Meng, S. Jin, Nano Lett. 12 (2012) 6030. W. Bensch et al., Phys. Chem. Chem. Phys. 14 (2012) 7509. H. Wu, Y. Cui, Nano Today 7 (2012) 414. H. Wu, G. Zheng, N. Liu, T.J. Carney, Y. Yang, Y. Cui, Nano Lett. 12 (2012) 904. M. Ge, J. Rong, X. Fang, C. Zhou, Nano Lett. 12 (2012) 2318. H. Zhang, P.V. Braun, Nano Lett. 12 (2012) (2012) 2778. N. Liu, H. Wu, M.T. McDowell, Y. Yao, C. Wang, Y. Cui, Nano Lett. 12 (2012) 3315. M. Mortazavi, J. Deng, V.B. Shenoy, N.V. Medhekar, J. Power Sources 225 (2013) 207. L.D. Ellis, T.D. Hatchard, M.N. Obrovac, J. Electrochem. Soc. 159 (2012) A1801. N. Singh, T.S. Arthur, C. Ling, M. Matsui, F. Mizuno, Chem. Commun. 49 (2013) 149. O.I. Malyi, T.L. Tan, S. Manzhos, Appl. Phys. Express 6 (2013) 027301. V. Milman, M.C. Payne, V. Heine, R.J. Needs, J.S. Lin, M.H. Lee, Phys. Rev. Lett. 70 (1993) 2928.

161

[16] H. Kim, K.E. Kweon, C.-Y. Chou, J.G. Ekerdt, G.S. Hwang, J. Phys. Chem. C 114 (2010) (2010) 17942. [17] W. Wan, Q. Zhang, Y. Cui, E. Wang, J. Phys.: Condens. Matter 22 (2010) 415501. [18] K. Zhao, W.L. Wang, J. Gregoire, M. Pharr, Z. Suo, J.J. Vlassak, E. Kaxiras, Nano Lett. 11 (2011) 2962. [19] B. Peng, F. Cheng, Z. Tao, J. Chen, J. Chem. Phys. 133 (2010) 034701. [20] S.C. Jung, Y.-K. Han, Phys. Chem. Chem. Phys. 13 (2011) 21282. [21] L. Mandeltort, J.T. Yates, J. Phys. Chem. C 116 (2012) 24962. [22] K. Persson et al., J. Phys. Chem. Lett. 1 (2010) 1176. [23] M. Nakayama, M. Kaneko, M. Wakihara, Phys. Chem. Chem. Phys. 14 (2012) 13963. [24] B. Philippe et al., Chem. Mater. 24 (2012) 1107–1115. [25] Z. Cui, F. Gao, Z. Cui, J. Qu, J. Power Sources 207 (2012) 150. [26] P. Johari, Y. Qi, V.B. Shenoy, Nano Lett. 11 (2011) 5494. [27] A. Carvalho, M.J. Rayson, P.R. Briddon, J. Phys. Chem. C 116 (2012) 8243. [28] W.A. van Schalkwijk, B. Scrosati, Advances in Lithium-ion Batteries, Kluwer, New York, 2002. [29] K. Xu, Chem. Rev. 104 (2004) 4303. [30] K. Leung, J. Phys. Chem. C 116 (2012) 9852. [31] L. Leung, Y. Qi, K.R. Zavadil, Y.S. Jung, A.C. Dillon, A.S. Cavanagh, S.-H. Lee, S.M. George, J. Am. Chem. Soc. 133 (2011) 14471. [32] J. Vatamanu, O. Borodin, G.D. Smith, J. Phys. Chem. C 116 (2012) 1114. [33] S.-P. Kima, A.C.T. van Duinb, V.B. Shenoy, J. Power Sources 196 (2011) 8590. [34] B. Luo, Y. Fang, B. Wang, J. Zhou, H. Song, L. Zhi, Energy Environ. Sci. 5 (2012) 5226. [35] L.-J. ZHou, Z.F. Hou, L.-M. Wu, J. Phys. Chem. C 116 (2012) 21780. [36] Y. Sun, X. Hu, W. Luo, Y. Huang, J. Phys. Chem. C 116 (2012) 20794. [37] S. Abe, F. Watari, H. Tachikawa, Jpn. J. Appl. Phys. 51 (2012) 10NE34. [38] A. Chroneos, B. Yildiz, A. Tarancón, D. Parfitt, J.A. Kilner, Energy Environ. Sci. 4 (2011) 2774. [39] M.J. Rayson, P.R. Briddon, Phys. Rev. B 80 (2009) 205104. [40] M.J. Rayson, P.R. Briddon, Comput. Phys. Commun. 178 (2008) 128. [41] C. Hartwigsen, S. Goedecker, J. Hutter, Phys. Rev. B 58 (1998) 3641. [42] J.P. Goss, M.J. Shaw, P.R. Briddon, Topics Appl. Phys. 104 (2007) 69. [43] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865. [44] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188. [45] A. Ramstad, G. Brocks, P.J. Kelly, Phys. Rev. B 51 (1995) 14504. [46] A.A. Baski, S.C. Erwin, L.J. Whitman, Surf. Sci. 392 (1997) 69. [47] S.-H. Lee, M.-H. Kang, Phys. Rev. Lett. 82 (1999) 968. [48] G. Henkelman, B.P. Uberuaga, H. Jonsson, J. Chem. Phys. 113 (2000) 9901. [49] B. Peng, F. Cheng, Z. Tao, J. Chen, J. Chem. Phys. 133 (2010) 034701.