Condensed Matter Physics, 2011, Vol. 14, No 2, 23701: 1–13 DOI: 10.5488/CMP.14.23701 http://www.icmp.lviv.ua/journal

A comparative study for structural and electronic properties of single-crystal ScN

arXiv:1107.0590v1 [cond-mat.mtrl-sci] 4 Jul 2011

˘ 2 R. Mohammad1,2 , S. ¸ Katırcıoglu 1 Palestine Technical University, Applied Science College, WestBank, Palestine 2 Middle East Technical University, Physics Department, 06530 Ankara, Turkey

Received November 4, 2010, in final form February 3, 2011 A comparative study by FP-LAPW calculations based on DFT within LDA, PBE-GGA, EVex -PWco -GGA, and EVex -GGA-LDAco schemes is introduced for the structural and electronic properties of ScN in RS, ZB, WZ, and CsCl phases. According to all approximations used in this work, the RS phase is the stable ground state structure and makes a transition to CsCl phase at high transition pressure. While PBE-GGA and EVex -PWco GGA’s have provided better structural features such as equilibrium lattice constant and bulk modulus, only EVex -PWco -GGA and EVex -GGA-LDAco ’s have given the non zero, positive indirect energy gap for RS-ScN, comparable with the experimental ones. The indirect band gap of ScN in RS phase is enlarged to the corresponding measured value by EVex -PWco -GGA+USIC calculations in which the Coulomb self and exchangecorrelation interactions of the localized d-orbitals of Sc have been corrected by the potential parameter of U. The EVex -PWco -GGA calculations have also provided good results for the structural and electronic features of ScN in ZB, WZ, and CsCl phases comparable with the theoretical data available in the literature. EVex -PWco GGA and EVex -PWco -GGA+USIC schemes are considered to be the best ones among the others when the structural and electronic features of ScN are aimed to be calculated by the same exchange-correlation energy approximations. Key words: ScN, FP-LAPW, DFT, structural properties, electronic properties PACS: 71.15.Mb, 71.15.Nc, 71.20.Nr, 71.18.+y

1. Introduction ScN (Scandium nitride) has many potential applications due to its high mechanical strength [1], good thermal stability with a melting temperature above 2000o C [2] and high hardness of 21 GPa with respect to load deformation [1]. After the epitaxial growth of smooth and singly oriented ScN films [1–8], ScN was also considered to be a potential semiconductor in electronic device applications. X-Ray diffraction analysis indicated that, unlike the other nitride semiconductors, all ScN films crystallize into a single rock-salt (RS) phase (B1) with a lattice constant of ∼4.5 Å [1, 3–8]. The RS structure was also determined to be the most stable structure for ScN by the comparative theoretical works [9–11]. The lattice constant of ScN in RS phase has been calculated to be in the range of 4.42–4.651 Å [9, 11–15] by the first principles calculations within local density approximation (LDA) [16, 17] and generalized gradient approximation (GGA) [18, 19] of exchange and correlation energies. In these works [9–14], the bulk modulus of ScN has been reported in the range of 196–235 GPa with respect to the experimental value of 182∓40 [1]. In the literature the most stable RS phase with the cohesive energy of –13.69 [9] and –13.428 eV [10] was found to be followed by the wurtzite (WZ or B4), zinc-blende (ZB or B3), and CsCl (B2) type meta-stable structures of ScN with the ordered cohesive energies of –13.35 [9], –13.03 [9], and –11.58 eV [9] (–11.34 eV [10]), respectively. These cohesive energies correspond to the lattice constants of 3.49 (c/a = 1.6, u = 0.38) [9], 4.88 [9] and 2.81 Å [9] (2.79 Å [10]) for WZ, ZB, and CsCl phases of ScN, respectively. In a recent work [11], the lattice constants have been calculated to be 3.45, 4.939, and 2.926 Å for ScN in WZ, ZB and CsCl phases. The bulk modulus of WZ phase of ScN has been reported to be 156 [9] and 132.91 GPa [11] by GGA and LDA of exchange and correlation energies, respectively. In the same works, the bulk modulus of CsCl phase was calculated to be 170 [9] and c R. Mohammad, S. ˘ 2011

¸ Katırcıoglu,

23701-1

˘ R. Mohammad, S. ¸ Katırcıoglu

159.759 GPa [11]. In another work [10], the CsCl meta-stable phase of ScN has the bulk modulus of 178.60 GPa by GGA calculations. The first principles electronic band structure calculations within LDA [9, 11, 12, 14] and GGA [9, 10, 13] schemes indicated that RS-ScN was a semi-metal with an almost zero indirect band gap. However, ScN in RS phase was a semiconductor with the indirect gap of 2.4 [8], 1.3∓0.3 [20], and 0.9∓0.1 eV [7] measured by optical transmission, reflection and absorption experiments, respectively. The recent ab-initio calculations [12, 14, 15, 20, 21] using pseudopotential (PP) method within Hedin’s [22] Green’s functions quasiparticle corrections on LDA (LDA-Go Wo ), exact exchange of LDA [23] [OEPx(cLDA)], and Hedin’s Green’s functions quasiparticle corrections on exact exchange of LDA [OEPx(cLDA)-Go Wo ], Linear Muffin Tin Orbital (LMTO) method within Bechstedt’s [24] Green’s functions quasiparticle corrections on LDA (LDA-GW) and full-potentiallinearized augmented plane waves (FP-LAPW) method within GGA [19] and screened exchange of LDA (sx-LDA) [25] have all supported the semiconductor nature of the stable ScN by giving the indirect gap in the range of 0.54–1.70 eV at X symmetry point (EgΓ−X ). In [12, 14, 20, 21], the optical direct gap of RS structure at X point (EX−X ) was calculated in the range of 1.98–2.90 eV g with respect to the experimental direct absorptions in the range of 1.8–2.4 eV [3, 4, 7, 8, 20]. The recent DFT calculations within GGA scheme [9] have indicated that the WZ and ZB metastable structures of ScN were non-metallic with large indirect gaps of ∼3 eV along M-Γ symmetry line (EM−Σ ) and 2.3 eV at W symmetry point (EX−W ), respectively. The nonmetallic nature of g g ScN in ZB phase was also obtained by an indirect gap of 2.36 eV at W point (EX−W ) by LDA g scheme [11]. These GGA and LDA calculations of ScN in ZB phase [9, 11], have given the direct gap of 2.4 and 2.42 eV at X symmetry point, respectively. In the literature, ScN in B2 phase was reported to be metallic by DFT calculations within GGA [9, 10] and LDA [11] schemes. In the literature, although ScN has not been worked so far, hybrid FP-LAPW calculations within the framework of DFT and different exchange-correlation functionals have given accurate electronic features for nitride compounds and alloys [26, 27] comparable with the corresponding measured ones due to the possible strong hybridization between the 2p orbitals of N and the corresponding cationic states. In addition, the orthogonalized norm-conserving pseudopotential (NCPP) method in which the plane wave basis functions are orthogonalized to core-like orbitals has been reported to be a very promising method in electronic band structure calculations of nitride alloys [28] for describing the experimental optical data, together with the FP-LAPW method within virtual crystal approximation [28]. In the present work, we have examined the structural and electronic properties of ScN in stable (RS) and meta-stable phases (CsCl, ZB, WZ) by DFT calculations mainly within two GGA schemes which have not been used before for ScN. We have aimed to introduce a comparative study for the structural and electronic features of ScN such as the lattice constant, bulk modulus, cohesive energy, energy gaps and the effective masses of electrons and holes. The present work has also comprised the electronic band structure of stable ScN corrected by an on-site Coulomb self- and exchange-correlation potential approximation (USIC ) [29].

2. Method of calculations The present DFT calculations on the structural and electronic properties of ScN compound have been performed using FP-LAPW method implemented in WIEN2k code [30]. In the literature, the exchange-correlation energy of DFT has been defined by local density approximation (LDA) [16] for the systems having uniform electron charge density. But for the systems of non-uniform charge density, the exchange-correlation energy of LDA has been corrected by gradient of the charge density within different generalized gradient approximations (GGA). In the present total energy and electronic band structure calculations of ScN in B1-B4 phases, four different approximations of exchange-correlation energies have been considered. In one of the approximations, exchange and correlation energies have been defined simply by LDA [16], without regarding the homogeneity of the real charge density. In the second approximation, exchange and correlation energies of LDA have been corrected by GGA of Perdew-Burke-Ernzerhof (PBE) [19]. The generalized gradient functional 23701-2

A comparative study for structural and electronic properties

of Perdew-Burke-Ernzerhof [19] has retained correct features of LDA [16] and satisfied only those which are energetically significant. In the third approximation, GGA of Engel-Vosko (EV) [31] and GGA of Perdew and Wang (PW) [32] have been used to correct the exchange and correlation energies, respectively. Since the generalized gradient functional of Engel-Vosko [31] was designed to give a better exchange potential (Vx ) only, the standard correlation potential of LDA [16] in the third approximation has been corrected by another functional, namely, GGA of Perdew and Wang [32]. The functional of Perdew and Wang [32] has incorporated some inhomogeneity effects while retaining many of the best features of the local density approximation. In the last approximation, the exchange energy of LDA [16] was corrected by GGA of Engel-Vosko [31], but the correlation energy was defined directly by LDA [16]. The exchange-correlation energy approaches considered in this work have been labeled as LDA, PBE-GGA, EVex -PWco -GGA and EVex -GGALDAco . The acronyms have been produced either based on the key word of the approach (LDA) or on the name of the authors (PBE-GGA, EVex -PWco -GGA, EVex -GGA-LDAco ) who developed the corresponding exchange and correlation functionals. Here, the subscripts of exchange (ex) and correlation (co) functionals are exclusively used for the cases where the exchange and correlation functionals are different. It has been considered that EVex -PWco -GGA and EVex -GGA-LDAco schemes can provide significant improvement for the structural and electronic properties of ScN, respectively. In the literature, it was reported that LDA+U and GGA+U schemes can also improve the band gap energies of transition metal compounds and alloys by reproducing quite well the localized nature of the d-electrons (or f -electrons) [33–35]. Since the ab-initio calculations are difficult to perform, the strong correlations like in transition metal compounds and alloys are often based on a model Hamiltonian approach in which the important parameter of U improves the effective Coulomb interactions between the localized d-electrons. In the present work, the electronic band structure of RS-ScN calculated by EVex -PWco -GGA has been improved by USIC method [29, 36] introduced in WIEN2k code [30]. The present EVex -PWco -GGA+USIC scheme has rectified on-site Coulomb self- and exchange-correlation interactions of the localized d-orbitals of Sc by the potential parameter of U [29, 36]. In Anisimov et al’s paper [29], the meaning of the U parameter were defined as the cost Coulomb energy for thePplacement of two d-electrons on the same site. The Coulomb interaction was defined to be (1/2)U i6=j ni nj [37] for d-orbitals. Here, ni are d-orbital occupancies. The Coulomb interaction term included into the total energy functional of EVex -PWco -GGA has given the orbital energies of ScN as ξi =ξ (EVex −PWco −GGA) +U(1/2 − ni) [37]. The shifting of the corresponding orbital energies in EVex -PWco -GGA+USIC calculations gives a qualitative improvement for the energy gap of RS-ScN. The present Coulomb interaction parameter of U is calculated to be 4.08 eV in WIEN2k [30] to have the maximum approach to the measured indirect band gap of RS-ScN. In the present work, ScN has been studied in RS, CsCl, ZB, and WZ structures. The unit cells of RS, CsCl, and ZB consist of two basis atoms; Sc at (0, 0, 0) and N at (0.5a, 0.5a, 0.5a) in fcc structure, Sc at (0, 0, 0) and N at (0.5a, 0.5a, 0.5a) in bcc structure, and Sc at (0, 0, 0) and N at (0.25a, 0.25a, 0.25a) in fcc structure, respectively, where a is the lattice constant parameter. The WZ structure with space group of F 63 mc has four atoms in the unit cell; Sc atoms at (a/3, 2a/3, 0) and (2a/3, a/3, 0.5c), N atoms at (a/3, 2a/3, u ∗ c) and (2a/3, a/3, (0.5 + u) ∗ c), where a and c are the periods in x-y plane and along z direction, respectively. The z directional distance, u, is defined between the layers of Sc and N atoms. In FP-LAPW calculations, each unit cell is partitioned into non-overlapping muffin-tin spheres around the atomic sites. Basis functions are expanded in combinations of spherical harmonic functions inside the non-overlapping spheres. In the interstitial region, a plane wave basis is used and expansion is limited with a cutoff parameter, RMT KMAX =7. Here, RMT is the smallest radius of the sphere in the unit cell, KMAX is the magnitude of the largest K vector used in the plane wave expansion. The muffin-tin radius is adopted to be 1.8 and 1.67 a.u. for Sc and N atoms, respectively. In the calculations, the electrons of Sc and N atoms in 3s2 3p6 4s2 3d1 and 2s2 2p5 shells respectively, are treated as valence electrons by choosing a cutoff energy of –6.0 Ry. The core states are treated within the spherical part of the potential only and are assumed to have a spherically symmetric charge density totally confined inside the muffin-tin spheres. The expansion of spherical harmonic functions inside the muffin-tin spheres is truncated 23701-3

˘ R. Mohammad, S. ¸ Katırcıoglu

at l =10. The cutoff for Fourier√expansion of the charge density and potential in the interstitial region is fixed to be GMAX = 16 Ry. The FP-LAPW parameters presented in this work have been obtained after a few trials around their fixed values. The ScN in RS, CsCl and ZB structures have been optimized with respect to the volume of the unit cells by minimizing the total energy. The equilibrium lattice constants of ScN in RS, CsCl, and ZB phases are determined by fitting the total energies to the Murnaghan’s equation of state [38]. The equilibrium structure of ScN in WZ phase that corresponds to the minimum total energy has been obtained by the application of both volume and geometry optimizations. The volume optimization used for all structures is provided with the energy criterion of 0.01 mRy. The optimum volume for WZ phase corresponds to the optimum c/a ratio and a value has been found by fitting the total energies to a quadratic function in a least square fitting method. The z directional distance u between the Sc and N layers in WZ phase has been obtained by geometry optimization at the optimum volume of the unit cell. The geometry optimization forces the atoms in the unit cell to move towards their equilibrium positions. In the geometry optimization, all forces on the atoms are converged to less than 1 mRy/a.u. The variation of total energy with respect to the volume and c/a ratio of WZ structure is plotted in figure 1. The present structural and electronic band calculations have been performed using 21x21x21 grids and correspond to 1000 k points sufficiently defined in the irreducible wedge of the Brillouin zone for ScN in B1-B4 phases.

Figure 1. (Color on-line) The relative total energy (per unit cell) versus volume and c/a within EVex -PWco -GGA for WZ-ScN.

In the present work, the cohesive energies (energy/atom-pair) of ScN structures (B1-B4) have been calculated by

Ecoh. = EScN − M EN − M ESc .

(2.1)

Where, EN and ESc are the values of the self-atomic energies of N and Sc atoms. EScN is the minimum of the total energy per unit cell that corresponds to the equilibrium structures of the ScN phases. Here, M defines the number of N and Sc atoms in the unit cell of the phases. The selfatomic energies of either atoms are calculated accurately in a fcc super cell with a lattice constant of 25 a.u. 23701-4

A comparative study for structural and electronic properties

3. Results and discussion 3.1. Structural properties The total energies (per unit cell) of RS, CsCl, WZ and ZB structures of ScN calculated within EVex -PWco -GGA scheme are plotted as a function of the volume of the structures in figure 2. The curvature of variations is made clear by plotting the total energies (per unit cell) of the structures relative to the minimum total energy of the RS-ScN. The minimum of the total energies of the phases with E(RS-ScN)

A comparative study for structural and electronic properties of single-crystal ScN

arXiv:1107.0590v1 [cond-mat.mtrl-sci] 4 Jul 2011

˘ 2 R. Mohammad1,2 , S. ¸ Katırcıoglu 1 Palestine Technical University, Applied Science College, WestBank, Palestine 2 Middle East Technical University, Physics Department, 06530 Ankara, Turkey

Received November 4, 2010, in final form February 3, 2011 A comparative study by FP-LAPW calculations based on DFT within LDA, PBE-GGA, EVex -PWco -GGA, and EVex -GGA-LDAco schemes is introduced for the structural and electronic properties of ScN in RS, ZB, WZ, and CsCl phases. According to all approximations used in this work, the RS phase is the stable ground state structure and makes a transition to CsCl phase at high transition pressure. While PBE-GGA and EVex -PWco GGA’s have provided better structural features such as equilibrium lattice constant and bulk modulus, only EVex -PWco -GGA and EVex -GGA-LDAco ’s have given the non zero, positive indirect energy gap for RS-ScN, comparable with the experimental ones. The indirect band gap of ScN in RS phase is enlarged to the corresponding measured value by EVex -PWco -GGA+USIC calculations in which the Coulomb self and exchangecorrelation interactions of the localized d-orbitals of Sc have been corrected by the potential parameter of U. The EVex -PWco -GGA calculations have also provided good results for the structural and electronic features of ScN in ZB, WZ, and CsCl phases comparable with the theoretical data available in the literature. EVex -PWco GGA and EVex -PWco -GGA+USIC schemes are considered to be the best ones among the others when the structural and electronic features of ScN are aimed to be calculated by the same exchange-correlation energy approximations. Key words: ScN, FP-LAPW, DFT, structural properties, electronic properties PACS: 71.15.Mb, 71.15.Nc, 71.20.Nr, 71.18.+y

1. Introduction ScN (Scandium nitride) has many potential applications due to its high mechanical strength [1], good thermal stability with a melting temperature above 2000o C [2] and high hardness of 21 GPa with respect to load deformation [1]. After the epitaxial growth of smooth and singly oriented ScN films [1–8], ScN was also considered to be a potential semiconductor in electronic device applications. X-Ray diffraction analysis indicated that, unlike the other nitride semiconductors, all ScN films crystallize into a single rock-salt (RS) phase (B1) with a lattice constant of ∼4.5 Å [1, 3–8]. The RS structure was also determined to be the most stable structure for ScN by the comparative theoretical works [9–11]. The lattice constant of ScN in RS phase has been calculated to be in the range of 4.42–4.651 Å [9, 11–15] by the first principles calculations within local density approximation (LDA) [16, 17] and generalized gradient approximation (GGA) [18, 19] of exchange and correlation energies. In these works [9–14], the bulk modulus of ScN has been reported in the range of 196–235 GPa with respect to the experimental value of 182∓40 [1]. In the literature the most stable RS phase with the cohesive energy of –13.69 [9] and –13.428 eV [10] was found to be followed by the wurtzite (WZ or B4), zinc-blende (ZB or B3), and CsCl (B2) type meta-stable structures of ScN with the ordered cohesive energies of –13.35 [9], –13.03 [9], and –11.58 eV [9] (–11.34 eV [10]), respectively. These cohesive energies correspond to the lattice constants of 3.49 (c/a = 1.6, u = 0.38) [9], 4.88 [9] and 2.81 Å [9] (2.79 Å [10]) for WZ, ZB, and CsCl phases of ScN, respectively. In a recent work [11], the lattice constants have been calculated to be 3.45, 4.939, and 2.926 Å for ScN in WZ, ZB and CsCl phases. The bulk modulus of WZ phase of ScN has been reported to be 156 [9] and 132.91 GPa [11] by GGA and LDA of exchange and correlation energies, respectively. In the same works, the bulk modulus of CsCl phase was calculated to be 170 [9] and c R. Mohammad, S. ˘ 2011

¸ Katırcıoglu,

23701-1

˘ R. Mohammad, S. ¸ Katırcıoglu

159.759 GPa [11]. In another work [10], the CsCl meta-stable phase of ScN has the bulk modulus of 178.60 GPa by GGA calculations. The first principles electronic band structure calculations within LDA [9, 11, 12, 14] and GGA [9, 10, 13] schemes indicated that RS-ScN was a semi-metal with an almost zero indirect band gap. However, ScN in RS phase was a semiconductor with the indirect gap of 2.4 [8], 1.3∓0.3 [20], and 0.9∓0.1 eV [7] measured by optical transmission, reflection and absorption experiments, respectively. The recent ab-initio calculations [12, 14, 15, 20, 21] using pseudopotential (PP) method within Hedin’s [22] Green’s functions quasiparticle corrections on LDA (LDA-Go Wo ), exact exchange of LDA [23] [OEPx(cLDA)], and Hedin’s Green’s functions quasiparticle corrections on exact exchange of LDA [OEPx(cLDA)-Go Wo ], Linear Muffin Tin Orbital (LMTO) method within Bechstedt’s [24] Green’s functions quasiparticle corrections on LDA (LDA-GW) and full-potentiallinearized augmented plane waves (FP-LAPW) method within GGA [19] and screened exchange of LDA (sx-LDA) [25] have all supported the semiconductor nature of the stable ScN by giving the indirect gap in the range of 0.54–1.70 eV at X symmetry point (EgΓ−X ). In [12, 14, 20, 21], the optical direct gap of RS structure at X point (EX−X ) was calculated in the range of 1.98–2.90 eV g with respect to the experimental direct absorptions in the range of 1.8–2.4 eV [3, 4, 7, 8, 20]. The recent DFT calculations within GGA scheme [9] have indicated that the WZ and ZB metastable structures of ScN were non-metallic with large indirect gaps of ∼3 eV along M-Γ symmetry line (EM−Σ ) and 2.3 eV at W symmetry point (EX−W ), respectively. The nonmetallic nature of g g ScN in ZB phase was also obtained by an indirect gap of 2.36 eV at W point (EX−W ) by LDA g scheme [11]. These GGA and LDA calculations of ScN in ZB phase [9, 11], have given the direct gap of 2.4 and 2.42 eV at X symmetry point, respectively. In the literature, ScN in B2 phase was reported to be metallic by DFT calculations within GGA [9, 10] and LDA [11] schemes. In the literature, although ScN has not been worked so far, hybrid FP-LAPW calculations within the framework of DFT and different exchange-correlation functionals have given accurate electronic features for nitride compounds and alloys [26, 27] comparable with the corresponding measured ones due to the possible strong hybridization between the 2p orbitals of N and the corresponding cationic states. In addition, the orthogonalized norm-conserving pseudopotential (NCPP) method in which the plane wave basis functions are orthogonalized to core-like orbitals has been reported to be a very promising method in electronic band structure calculations of nitride alloys [28] for describing the experimental optical data, together with the FP-LAPW method within virtual crystal approximation [28]. In the present work, we have examined the structural and electronic properties of ScN in stable (RS) and meta-stable phases (CsCl, ZB, WZ) by DFT calculations mainly within two GGA schemes which have not been used before for ScN. We have aimed to introduce a comparative study for the structural and electronic features of ScN such as the lattice constant, bulk modulus, cohesive energy, energy gaps and the effective masses of electrons and holes. The present work has also comprised the electronic band structure of stable ScN corrected by an on-site Coulomb self- and exchange-correlation potential approximation (USIC ) [29].

2. Method of calculations The present DFT calculations on the structural and electronic properties of ScN compound have been performed using FP-LAPW method implemented in WIEN2k code [30]. In the literature, the exchange-correlation energy of DFT has been defined by local density approximation (LDA) [16] for the systems having uniform electron charge density. But for the systems of non-uniform charge density, the exchange-correlation energy of LDA has been corrected by gradient of the charge density within different generalized gradient approximations (GGA). In the present total energy and electronic band structure calculations of ScN in B1-B4 phases, four different approximations of exchange-correlation energies have been considered. In one of the approximations, exchange and correlation energies have been defined simply by LDA [16], without regarding the homogeneity of the real charge density. In the second approximation, exchange and correlation energies of LDA have been corrected by GGA of Perdew-Burke-Ernzerhof (PBE) [19]. The generalized gradient functional 23701-2

A comparative study for structural and electronic properties

of Perdew-Burke-Ernzerhof [19] has retained correct features of LDA [16] and satisfied only those which are energetically significant. In the third approximation, GGA of Engel-Vosko (EV) [31] and GGA of Perdew and Wang (PW) [32] have been used to correct the exchange and correlation energies, respectively. Since the generalized gradient functional of Engel-Vosko [31] was designed to give a better exchange potential (Vx ) only, the standard correlation potential of LDA [16] in the third approximation has been corrected by another functional, namely, GGA of Perdew and Wang [32]. The functional of Perdew and Wang [32] has incorporated some inhomogeneity effects while retaining many of the best features of the local density approximation. In the last approximation, the exchange energy of LDA [16] was corrected by GGA of Engel-Vosko [31], but the correlation energy was defined directly by LDA [16]. The exchange-correlation energy approaches considered in this work have been labeled as LDA, PBE-GGA, EVex -PWco -GGA and EVex -GGALDAco . The acronyms have been produced either based on the key word of the approach (LDA) or on the name of the authors (PBE-GGA, EVex -PWco -GGA, EVex -GGA-LDAco ) who developed the corresponding exchange and correlation functionals. Here, the subscripts of exchange (ex) and correlation (co) functionals are exclusively used for the cases where the exchange and correlation functionals are different. It has been considered that EVex -PWco -GGA and EVex -GGA-LDAco schemes can provide significant improvement for the structural and electronic properties of ScN, respectively. In the literature, it was reported that LDA+U and GGA+U schemes can also improve the band gap energies of transition metal compounds and alloys by reproducing quite well the localized nature of the d-electrons (or f -electrons) [33–35]. Since the ab-initio calculations are difficult to perform, the strong correlations like in transition metal compounds and alloys are often based on a model Hamiltonian approach in which the important parameter of U improves the effective Coulomb interactions between the localized d-electrons. In the present work, the electronic band structure of RS-ScN calculated by EVex -PWco -GGA has been improved by USIC method [29, 36] introduced in WIEN2k code [30]. The present EVex -PWco -GGA+USIC scheme has rectified on-site Coulomb self- and exchange-correlation interactions of the localized d-orbitals of Sc by the potential parameter of U [29, 36]. In Anisimov et al’s paper [29], the meaning of the U parameter were defined as the cost Coulomb energy for thePplacement of two d-electrons on the same site. The Coulomb interaction was defined to be (1/2)U i6=j ni nj [37] for d-orbitals. Here, ni are d-orbital occupancies. The Coulomb interaction term included into the total energy functional of EVex -PWco -GGA has given the orbital energies of ScN as ξi =ξ (EVex −PWco −GGA) +U(1/2 − ni) [37]. The shifting of the corresponding orbital energies in EVex -PWco -GGA+USIC calculations gives a qualitative improvement for the energy gap of RS-ScN. The present Coulomb interaction parameter of U is calculated to be 4.08 eV in WIEN2k [30] to have the maximum approach to the measured indirect band gap of RS-ScN. In the present work, ScN has been studied in RS, CsCl, ZB, and WZ structures. The unit cells of RS, CsCl, and ZB consist of two basis atoms; Sc at (0, 0, 0) and N at (0.5a, 0.5a, 0.5a) in fcc structure, Sc at (0, 0, 0) and N at (0.5a, 0.5a, 0.5a) in bcc structure, and Sc at (0, 0, 0) and N at (0.25a, 0.25a, 0.25a) in fcc structure, respectively, where a is the lattice constant parameter. The WZ structure with space group of F 63 mc has four atoms in the unit cell; Sc atoms at (a/3, 2a/3, 0) and (2a/3, a/3, 0.5c), N atoms at (a/3, 2a/3, u ∗ c) and (2a/3, a/3, (0.5 + u) ∗ c), where a and c are the periods in x-y plane and along z direction, respectively. The z directional distance, u, is defined between the layers of Sc and N atoms. In FP-LAPW calculations, each unit cell is partitioned into non-overlapping muffin-tin spheres around the atomic sites. Basis functions are expanded in combinations of spherical harmonic functions inside the non-overlapping spheres. In the interstitial region, a plane wave basis is used and expansion is limited with a cutoff parameter, RMT KMAX =7. Here, RMT is the smallest radius of the sphere in the unit cell, KMAX is the magnitude of the largest K vector used in the plane wave expansion. The muffin-tin radius is adopted to be 1.8 and 1.67 a.u. for Sc and N atoms, respectively. In the calculations, the electrons of Sc and N atoms in 3s2 3p6 4s2 3d1 and 2s2 2p5 shells respectively, are treated as valence electrons by choosing a cutoff energy of –6.0 Ry. The core states are treated within the spherical part of the potential only and are assumed to have a spherically symmetric charge density totally confined inside the muffin-tin spheres. The expansion of spherical harmonic functions inside the muffin-tin spheres is truncated 23701-3

˘ R. Mohammad, S. ¸ Katırcıoglu

at l =10. The cutoff for Fourier√expansion of the charge density and potential in the interstitial region is fixed to be GMAX = 16 Ry. The FP-LAPW parameters presented in this work have been obtained after a few trials around their fixed values. The ScN in RS, CsCl and ZB structures have been optimized with respect to the volume of the unit cells by minimizing the total energy. The equilibrium lattice constants of ScN in RS, CsCl, and ZB phases are determined by fitting the total energies to the Murnaghan’s equation of state [38]. The equilibrium structure of ScN in WZ phase that corresponds to the minimum total energy has been obtained by the application of both volume and geometry optimizations. The volume optimization used for all structures is provided with the energy criterion of 0.01 mRy. The optimum volume for WZ phase corresponds to the optimum c/a ratio and a value has been found by fitting the total energies to a quadratic function in a least square fitting method. The z directional distance u between the Sc and N layers in WZ phase has been obtained by geometry optimization at the optimum volume of the unit cell. The geometry optimization forces the atoms in the unit cell to move towards their equilibrium positions. In the geometry optimization, all forces on the atoms are converged to less than 1 mRy/a.u. The variation of total energy with respect to the volume and c/a ratio of WZ structure is plotted in figure 1. The present structural and electronic band calculations have been performed using 21x21x21 grids and correspond to 1000 k points sufficiently defined in the irreducible wedge of the Brillouin zone for ScN in B1-B4 phases.

Figure 1. (Color on-line) The relative total energy (per unit cell) versus volume and c/a within EVex -PWco -GGA for WZ-ScN.

In the present work, the cohesive energies (energy/atom-pair) of ScN structures (B1-B4) have been calculated by

Ecoh. = EScN − M EN − M ESc .

(2.1)

Where, EN and ESc are the values of the self-atomic energies of N and Sc atoms. EScN is the minimum of the total energy per unit cell that corresponds to the equilibrium structures of the ScN phases. Here, M defines the number of N and Sc atoms in the unit cell of the phases. The selfatomic energies of either atoms are calculated accurately in a fcc super cell with a lattice constant of 25 a.u. 23701-4

A comparative study for structural and electronic properties

3. Results and discussion 3.1. Structural properties The total energies (per unit cell) of RS, CsCl, WZ and ZB structures of ScN calculated within EVex -PWco -GGA scheme are plotted as a function of the volume of the structures in figure 2. The curvature of variations is made clear by plotting the total energies (per unit cell) of the structures relative to the minimum total energy of the RS-ScN. The minimum of the total energies of the phases with E(RS-ScN)