Halogen-Induced Reconstruction of the c-BN(100) Surface - American

ARTICLE pubs.acs.org/JPCC

Halogen-Induced Reconstruction of the c-BN(100) Surface Johan Karlsson* and Karin Larsson Department of Materials Chemistry, Uppsala University, Box 538, Uppsala 751 21, Sweden ABSTRACT: The cubic phase of boron nitride (c-BN) is an extremely promising multifunctional material. However, to exploit all possible applications, large area chemical vapor deposition (CVD) of c-BN films is required. To be successful in the CVD growth of high-quality c-BN films, one must be able to stabilize the sp3 hybridization of the surface atoms; and in the present study, the surface stabilizing effect of F and Cl on the B- and N-terminated c-BN(100)-(1  1) surfaces has been investigated using density functional theory (DFT) calculations. It was found that Cl, most probably, will induce large sterical hindrance on both the B- and N-terminated c-BN(100) surface. F, on the other hand, was found to be a promising surface stabilizing agent for the B- and N-terminated c-BN(100) surface. However, the F atoms must be abstracted with H atoms. It can therefore be concluded that the optimal gas-phase composition for growth of c-BN consists of a mixture of H and F.

1. INTRODUCTION Boron nitride (BN) exists in two main crystalline phases: the diamond-like sp3 hybridized cubic phase (c-BN) and the graphitelike sp2 hybridized hexagonal phase (h-BN; Figure 1).1,2 The hexagonal phase is the thermodynamically stable phase under normal laboratory conditions. However, at 6 GPa and 2000 °C, it changes into the cubic phase. This cubic phase has several outstanding physical and chemical properties, e.g., extreme hardness (second only to diamond), low density (3.48 g/cm3), high thermal conductivity (13 W/(cm K) at T = 300 K), high electrical resistivity (1016 Ω cm), wide band gap (6
6.4 eV, the largest among all IV and III
V materials), high chemical stability, and transparency from near-ultraviolet to infrared (η = 2.1 for λ = 600 nm).3,4 Furthermore, it possesses a high thermal stability, both in oxidizing environments (up to 1300 °C) and in contact with Fe, Co, and Ni. It is therefore promising as a tool coating for machining of steel, cast iron, and ferrous alloys. It can also be made as both p- and n-type semiconductors, suitable for p
n junction diodes. These properties make c-BN an extremely promising multifunctional material, which could be tailored for a very large range of advanced mechanical, tribological, thermal, electronic, and optical applications. However, to exploit these applications, large area chemical vapor deposition (CVD) of c-BN films is required. To be successful in the CVD growth of high-quality c-BN films one must be able to stabilize the sp3 hybridization of the surface atoms, and thereby hindering the surface from a structural collapse to the hexagonal phase.5 This is done by covering the surface with surface stabilizing species. However, the surface stabilizing species must also be able to undergo abstraction reactions with gaseous species and, hence, leave room for an incoming B- or N-containing growth species. Hydrogen, H, has earlier been found to be a good surface stabilizing species for the dominant c-BN growth surfaces under CVD conditions, i.e., the (100), (110), and (111) surfaces.6
11 However, both experimental and theoretical studies indicate that halogen atoms also might be r 2011 American Chemical Society

Figure 1. The two main phases of BN: (A) c-BN (νsp3 = 109.5°) and (B) h-BN (νsp2 = 120°).

efficient as surface stabilizing species.9
12 The theoretical studies, however, only treated the (110) and the B- and N-terminated (111) surfaces of c-BN. The purpose with the present study has been to investigate the surface stabilizing effect of F and Cl on the B- and N-terminated c-BN(100)-(1  1) surfaces using density functional theory (DFT) calculations (Figure 2). For this surface type, the cleavage plane cuts two bonds per surface atom. A fully halogen-covered B- or N-terminated c-BN(100)-(1  1) surface would therefore contain two halogen atoms per surface atom. However, since the dihydride surfaces are unstable, due to strong repulsion between the H atoms, the dihalogenated surfaces are also expected to be unstable. Hence, only monohalogenated surfaces have been calculated within the present study. The ability of the halogens to act as surface stabilizing species was evaluated by investigating (i) the influence of halogen coverage on surface reconstruction, (ii) the process of adsorption/ abstraction of halogen to/from the surface, and (iii) the degree of hybridization of a surface radical site. Received: June 23, 2011 Revised: October 7, 2011 Published: October 10, 2011 22910

dx.doi.org/10.1021/jp205898w | J. Phys. Chem. C 2011, 115, 22910–22916

The Journal of Physical Chemistry C

Figure 2. Model showing the B-terminated c-BN(100)-(1  1) surface. (By interchanging the B and N atoms one obtains the N-terminated surface.) The coordinate system shows the (1  1) surface from above. In the text the surface atoms are referred to with a coordinate (x,y) according to this system.

Figure 3. Supercell for modeling the halogen-covered B-terminated c-BN(100)-(1  1) surface. (By interchanging the B and N atoms one obtains the halogen-covered N-terminated c-BN(100)-(1  1) surface.)

2. COMPUTATIONAL DETAILS The geometrical structures and total energies for the various systems were calculated using all-electron DFT, as implemented in the program package DMol3 (Materials Studio, v. 4.2) from Accelrys, Inc.13
17 The geometry was first optimized using the local (spin) density approximation (L(S)DA) with the PWC functional and, thereafter, further refined with the generalized gradient (spin density) approximation (GG(S)A) and the PW91 functional.18 All calculations were performed with the double numeric basis set with polarization functions (DNP). The surfaces were modeled as supercells (a = b = 10.22 Å, c = 30.85 Å, α = β = γ = 90°) under periodic boundary conditions (Figure 3).19 (For consistency, the energies of atomic (H, F, and Cl) and molecular (H2, F2, Cl2, HF, and HCl) species have been calculated by including them in a supercell of the same size as for the c-BN(100) surfaces.) The supercells consisted of 10 atomic layers, with 4  4 B (or N) atoms in each layer. To suppress the artificial charge transfer between the two polar ends of the slabs, the dangling bonds on the lower surfaces of the slabs were saturated with H atoms, and a large vacuum distance between the slabs was used (∼20 Å). The upper surface was covered with an F

ARTICLE

or Cl layer, containing 16 atoms. The positions of the B and N atoms within the bottom two layers (including the dangling bond-passivating H atoms) were fixed to simulate the structure of bulk c-BN. The rest of the atoms were allowed to fully relax using the Broyden
Fletcher
Goldfarb
Shanno (BFGS) algorithm.20 The Brillouin zone was sampled with a (2  2  1) Monkhorst-Pack grid.21 To assess the surface reactivity, the electrostatic potential22 (Epot) and Mulliken atomic charges23 (q) were used. Positive regions correspond to electron-deficient areas and are subject to nucleophilic attack; negative regions correspond to electron-rich areas and are subject to electrophilic attack. The potential was mapped onto a surface of the electron density (F) with an isovalue of 0.2 e/Å3. The deformation density (ΔF), i.e., the total density with the density of the isolated atoms subtracted, was also calculated. Regions of positive deformation density indicate the formation of bonds, whereas negative regions indicate electron loss. The convergence criteria within the calculations were 2  10
5 Ha for the maximum energy change per atom, 4  10
3 Ha/Å for the maximum force per atom, and 5  10
3 Å for the maximum displacement per atom. After optimization, the energies for the adsorption (1) and abstraction (2) processes were calculated according to eqs 3 and 4 BNð94% XÞ þ X f BNð100% XÞ ðX, Y ¼ H, F, or ClÞ

ð1Þ

BNð100% XÞ þ Y f BNð94% XÞ þ XY

ð2Þ

ΔEads ¼ EBNð100% XÞ
ðEBNð94% XÞ þ EX Þ

ð3Þ

ΔEabs ¼ ðEBNð94% XÞ þ EXY Þ
ðEBNð100% XÞ þ EY Þ

ð4Þ

where EBN(100% X), EBN(94% X), EX, EY, and EXY are the total energies for the 100% H-, F-, or Cl-covered surface, 94% (monoradical) H-, F-, or Cl-covered surface, an H, F, or Cl atom, and an H2, F2, Cl2, HF, or HCl molecule, respectively. The degree of hybridization of a surface radical site was measured by calculating (i) the surface N
B
N or B
N
B bond angle (ν) and (ii) the lowering of its position in the z direction (Δz) Δz ¼ zBNð94% XÞ
zBNð100% XÞ

ð5Þ

where zBN(94% X) and zBN(100% X) are the z coordinates for the radical site and the H-, F-, or Cl-covered radical site, respectively. A more elaborate description of the computational methods that has been used is presented and discussed in ref 6.

3. RESULTS AND DISCUSSION 3.1. F-Covered B-Terminated c-BN(100). The calculations showed that a 100% surface coverage of on-top F on the initially clean B-terminated c-BN(100)-(1  1) surface will induce a (2  1) reconstruction of the surface (Figure 4). This restructuring of the surface plane will saturate the dangling bonds of the surface B atoms and, hence, yield a lower surface energy. The average B
B dimer bond length was 1.90 Å, while the B
B distance in the bulk was 2.56 Å. The B
B dimer rows were separated by troughs, approximately 3.21 Å apart. The surface B atoms were found to exhibit a tetrahedral coordination, i.e., a large degree of sp3 hybridization, with an average surface N
B
N bond angle and B
N bond length of 109° and 1.57 Å, respectively. These values are in very good agreement with the bulk values of 109.5° and 22911

dx.doi.org/10.1021/jp205898w |J. Phys. Chem. C 2011, 115, 22910–22916

The Journal of Physical Chemistry C

ARTICLE

Figure 4. (A) 100% F-covered B-terminated c-BN(100)-(2  1). (B) Electrostatic potential for the surface in (A). (C) Deformation density of a B
B dimer in a plane perpendicular to the surface in (A).

Figure 5. (A) 94% (monoradical) F-covered B-terminated c-BN(100)-(2  1). (B) Electrostatic potential for the surface in (A). (C) Deformation density of the B(2,1)-B(1,1) dimer in a plane perpendicular to the surface in (A).

1.57 Å, i.e., a 100% surface coverage of on-top F on the B-terminated c-BN(100)-(2  1) surface is able to uphold a bulk-like bond angle and bond length for the surface B atoms. However, the F atoms are not able to uphold the ideal bulk-like (1  1) structure. This is similar to the behavior of the 100% H-covered B-terminated c-BN(100)-(2  1) surface (νN
B
N = 110°, dB
N = 1.56 Å). The interplanar B
N spacing between the first (B) and second (N) layer was 0.87 Å, which is within 3% of the bulk value (ΔzB(bulk)‑N(bulk) = 0.90 Å). The B
F bond was tilted 17° from the surface normal, facing away from the other B
F bond of the same dimer, and it was coplanar with the B
B dimer bond. The B
F bond length was 1.35 Å (the experimentally determined B
F bond length in BF3 is 1.31 Å).24 The adsorption energy for an atomic F on a monoradical B surface site was calculated to be
471 kJ/mol. From these results it can be concluded that atomic F (ΔEads =
471 kJ/mol) chemisorbs significantly stronger than atomic H (ΔEads =
198 kJ/mol) to the B-terminated c-BN(100)-(2  1) surface. This difference in adsorption energy is most probably due to the more ionic character of the B
F bond (qB = +0.40 e, qF =
0.25 e) compared to that of the B
H bond (qB = +0.19 e, qH =
0.12 e); that is, there is a larger attractive electrostatic contribution to the B
F bond energy compared to that for the B
H bond energy. The more ionic character of the B
F bond is also in accordance with calculated electronegativity differences, which characterize the B
F bond (|ΔχB
F| = 1.94) as “mostly ionic”, whereas the B
H bond (|ΔχB
H| = 0.16) is “mostly covalent”.25 Because the surface-covering F atoms are nucleophilic (qF =
0.25 e), they should be highly susceptible toward attack by electrophilic radical species in the gas-phase. However, the process of F abstraction from the surface, with gaseous atomic F, was found to be strongly endothermic (ΔEabs = +252 kJ/mol). The abstraction process was, however, found to be exothermic if H radicals were used as abstracting species (ΔEabs =
130 kJ/mol). Similar results were obtained for the process of H abstraction from the 100% H-covered B-terminated c-BN(100)-(2  1)

surface: abstraction with an F radical (ΔEabs =
403 kJ/mol) was found to be significantly more exothermic than abstraction with an H radical (ΔEabs =
258 kJ/mol). The removal of the F(2,1) atom from the 100% F-covered B-terminated c-BN(100)-(2  1) surface resulted in a pronounced, but very local, collapse of the B(2,1) radical site (ΔzB(2,1) =
0.35 Å) to a more trigonal planar coordination (νN
B(2,1)‑N = 116°; Figure 5). The collapse was accompanied by a buckling of the B(2,1)
B(1,1) dimer. The buckled dimer bond length and buckling height in the z direction were 1.71 and 0.32 Å, respectively. The change in bonding coordination for the radical site indicates a change in hybridization, from sp3 to sp2, for the B(2,1) atom; this is most probably due to (i) an inductive electron withdrawal from the less electronegative B(2,1) atom (χB = 2.04, qB(2,1) = +0.31 e) toward the more electronegative second layer N atoms (χN = 3.04, qN(second layer) =
0.71 e) and (ii) a hyperconjugation interaction between the unhybridized 2p orbital on B(2,1) and adjacent σ bonds (as explained in ref 6). This behavior of the 94% (monoradical) F-covered B-terminated c-BN(100)-(2  1) surface is very similar to the behavior of the 94% (monoradical) H-covered B-terminated c-BN(100)-(2  1) surface (ΔzB(2,1) =
0.26 Å, νN
B(2,1)‑N = 115°). 3.2. F-Covered N-Terminated c-BN(100). The calculations showed that a 100% surface coverage of on-top F on the initially clean N-terminated c-BN(100)-(1  1) surface will have the capacity to maintain the (1  1) structure (Figure 6). The surface N atoms were found to exhibit a trigonal planar coordination, i.e., a large degree of sp2 hybridization, with an average surface B
N
B bond angle and N
B bond length of 117° and 1.50 Å, respectively. These values are substantially different from the bulk values of 109.5° and 1.57 Å; that is, a 100% surface coverage of on-top F on the N-terminated c-BN(100)-(1  1) surface is not able to uphold a bulk-like bond angle and bond length for the surface N atoms. However, the F atoms are able to uphold the ideal bulk-like (1  1) structure. This is similar to the behavior of the 100% H-covered N-terminated c-BN(100)-(1  1) surface (νB
N
B = 116°, dB
N = 1.51 Å). The average surface and bulk 22912

dx.doi.org/10.1021/jp205898w |J. Phys. Chem. C 2011, 115, 22910–22916

The Journal of Physical Chemistry C

ARTICLE

Figure 6. (A) 100% F-covered N-terminated c-BN(100)-(1  1). (B) Electrostatic potential for the surface in (A). (C) Deformation density of the N atoms in a plane perpendicular to the surface in (A).

Figure 7. (A) 94% (monoradical) F-covered N-terminated c-BN(100)-(1  1). (B) Electrostatic potential for the surface in (A). (C) Deformation density of the N(2,1)-N(1,1) dimer in a plane perpendicular to the surface in (A).

N
N distances were 2.55 and 2.56 Å, respectively. The interplanar N
B spacing between the first (N) and second (B) layer was 0.79 Å, which is within 12% of the bulk value (ΔzB(bulk)‑N(bulk) = 0.90 Å). The N
F bond was parallel to the surface normal and had a length of 1.38 Å (the experimentally determined N
F bond length in NF2 is 1.35 Å). The adsorption energy for an atomic F on a monoradical N surface site was calculated to be
283 kJ/mol (Figure 7). From these results it can be concluded that atomic F chemisorbs strongly to both the B-terminated c-BN(100)-(2  1) surface (ΔEads =
471 kJ/mol) and the N-terminated c-BN(100)-(1  1) surface (ΔEads =
283 kJ/mol). However, the F species binds significantly stronger to the B-terminated surface compared to that for the N-terminated one. The lower bond strength of the N
F bond, compared to that for the B
F bond, is most probably due to repulsion between nonbonding electrons on the N and F atoms. In addition, calculated Mulliken charges predict partial negative charges on both the F (qF =
0.17 e) and N atoms (qN =
0.20 e) for the N
F bond, whereas there will be a partial negative charge on F (qF =
0.25 e) and a partial positive charge on B (qB = +0.40 e) for the B
F bond. This results in a repulsive (destabilizing) electrostatic contribution to the N
F bond energy, whereas the B
F bond will be stabilized by an attractive electrostatic contribution. It can also be concluded that atomic H (ΔEads =
431 kJ/mol) chemisorbs significantly stronger than atomic F (ΔEads =
283 kJ/mol) to the N-terminated c-BN(100)-(1  1) surface. This is reasonable since (i) H has no nonbonding electrons and (ii) the N
H bond is electrostatically stabilized (qN =
0.48 e, qH = +0.12 e). The preferential adsorption of H on the N-terminated surface, and of F on the B-terminated surface, is also in agreement with experimental studies by Zhang et al.12 Even though the F atoms on the 100% F-covered N-terminated c-BN(100)-(1  1) surface are nucleophilic (qF =
0.17 e), the process of F abstraction from the surface, with gaseous atomic F, was found to be endothermic (ΔEabs = +63 kJ/mol). However,

Figure 8. Deformation density for H2 (bottom), F2 (middle), and HF (top).

the abstraction process for the N-terminated surface was found to be less endothermic than the abstraction process for the Bterminated surface (ΔEabs = +252 kJ/mol). This is in agreement with the lower strength of the N
F bond (ΔEads =
283 kJ/mol) compared to that of the B
F bond (ΔEads =
471 kJ/mol). As for the B-terminated surface, the abstraction process was found to be significantly more favorable if H radicals were used as abstracting species (ΔEabs =
318 kJ/mol). Similar results were obtained for the process of H abstraction from the 100% H-covered N-terminated c-BN(100)-(1  1) surface: abstraction with an F radical (ΔEabs =
170 kJ/mol) was found to be significantly more exothermic than abstraction with an H radical (ΔEabs =
25 kJ/mol). This is most probably due to the significantly more ionic character of the H
F bond (|ΔχH
F| = 1.78 “mostly ionic”) compared to those for the H
H and F
F bonds (|ΔχF
F| = |ΔχH
H| = 0 “nonpolar covalent”; Figure 8). The unequal sharing of electrons in the H
F bond makes the F end of the bond partially negative (qF =
0.35 e) and the H end partially positive (qH = +0.35 e), and the attraction 22913

dx.doi.org/10.1021/jp205898w |J. Phys. Chem. C 2011, 115, 22910–22916

The Journal of Physical Chemistry C

ARTICLE

Figure 9. (A) 100% Cl-covered B-terminated c-BN(100)-(2  1). (B) Electrostatic potential for the surface in (A). (C) Deformation density of a B
B dimer in a plane perpendicular to the surface in (A).

Figure 10. (A) 94% (monoradical) Cl-covered B-terminated c-BN(100)-(2  1). (B) Electrostatic potential for the surface in (A). (C) Deformation density of the B(2,1)-Cl(3,1)-B(3,1) bridge in a plane perpendicular to the surface in (A).

between these partial charges increases the energy required to break the bond. In addition, the F
F bond is weakened by strong repulsion between nonbonding electrons on the F atoms (ΔEH
F =
601 kJ/mol, ΔEH
H =
456 kJ/mol, ΔEF
F =
219 kJ/mol; the experimentally determined bond energies for H
F, H
H, and F
F, in HF, H2, and F2, are
570,
436, and
159 kJ/mol, respectively). The exchange of the H
H and F
F bonds with the H
F bond will therefore provide a large energy gain for abstraction of F or H. These observations indicate that only F as surface stabilizing species is unfavorable for growth of c-BN(100). A better choice is to use only H. However, the best alternative is to use a mixture of H and F. Such a mixture of gases would make it possible to grow c-BN exclusively through surface reactions, i.e., there is no need for surface damaging ion bombardment, which will allow the production of electronic grade, single-crystal c-BN wafers. However, it has been found experimentally that energetic particle bombardment is necessary for F abstraction from the B-terminated surface.12 The removal of the F(2,1) atom from the 100% F-covered N-terminated c-BN(100)-(1  1) surface resulted in a mediumsized collapse of the N(2,1) radical site (ΔzN(2,1) =
0.16 Å; Figure 7). However, the relative collapse is very small since the other surface atoms relaxed downward with approximately 0.11 Å. The relaxation of the radical site was accompanied by dimerization of N(2,1) and N(1,1), with a bond length of 1.45 Å. The dimerization induced a dramatic change in bonding coordination for N(2,1) and N(1,1), going from a trigonal planar coordination for the 100% F-covered surface (νB
N(2,1)‑B = 117°, νB
N(1,1)‑B = 117°) to a more tetrahedral coordination for the 94% (monoradical) F-covered surface (νB
N(2,1)‑B = 109°, νB
N(1,1)‑B = 107°). There are also indications of interactions between N(2,1) and the neighboring adsorbate F(3,1), which results in a significant change in F-to-surface normal angle for F(3,1), going from 0° for the 100% F-covered surface to 30° for the 94% (monoradical) F-covered surface. This behavior of the 94%

(monoradical) F-covered N-terminated c-BN(100)-(1  1) surface is very similar to the behavior of the 94% (monoradical) H-covered N-terminated c-BN(100)-(1  1) surface and is most probably due to (i) inductive electron withdrawal from the second layer B atoms (χB = 2.04, qB(second layer) = +0.69 e) toward the surface radical N(2,1) atom (χN = 3.04, qN(2,1) =
0.43 e) and (ii) the electron-rich dangling bond on N(2,1) (as explained in ref 6). 3.3. Cl-Covered B-Terminated c-BN(100). The calculations showed that a 100% surface coverage of on-top Cl on the initially clean B-terminated c-BN(100)-(1  1) surface will induce a (2  1) reconstruction of the surface (Figure 9). The average B
B dimer bond length was 1.71 Å, and the B
B distance in the bulk was 2.56 Å. The B
B dimer rows were separated by troughs, approximately 3.40 Å apart. The surface B atoms were found to exhibit a tetrahedral coordination, i.e., a large degree of sp3 hybridization, with an average surface N
B
N bond angle and B
N bond length of 109° and 1.57 Å, respectively. These values are in very good agreement with the bulk values of 109.5° and 1.57 Å, i.e., a 100% surface coverage of on-top Cl on the B-terminated c-BN(100)-(2  1) surface is able to uphold a bulk-like bond angle and bond length for the surface B atoms. However, the Cl atoms are not able to uphold the ideal bulk-like (1  1) structure. This is similar to the behavior of the 100% H(νN
B
N = 110°, dB
N = 1.56 Å) and F-covered (νN
B
N = 109°, dB
N = 1.57 Å) B-terminated c-BN(100)-(2  1) surfaces. The interplanar B
N spacing between the first (B) and second (N) layer was 0.82 Å, which is within 9% of the bulk value (ΔzB(bulk)‑N(bulk) = 0.90 Å). The B
Cl bond was tilted 13° from the surface normal, facing away from the other B
Cl bond of the same dimer, and it was coplanar with the B
B dimer bond. The B
Cl bond length was 1.87 Å (the experimentally determined B
Cl bond length in BCl3 is 1.74 Å). The adsorption energy for an atomic Cl on a monoradical B surface site was found to be endothermic (ΔEads = +262 kJ/mol). 22914

dx.doi.org/10.1021/jp205898w |J. Phys. Chem. C 2011, 115, 22910–22916

The Journal of Physical Chemistry C This is most probably due to a high stabilization of the monoradical surface site (Figure 9) and to repulsive sterical interactions due to the overlap of electron density on adjacent bulky Cl atoms (the shortest distance between the Cl atoms on the surface is 2.54 Å, which is significantly smaller than the van der Waals distance of 3.50 Å). That there is sterical hindrance on the surface is also confirmed by the fact that the adsorption energy for a Cl atom on the clean surface is strongly exothermic (ΔEads =
471 kJ/mol), whereas the adsorption energy for a Cl atom on the unrelaxed monoradical surface is only weakly exothermic (ΔEads =
17 kJ/mol). It can therefore be concluded that the energy of adsorption becomes less exothermic as the coverage increases; indicating repulsive interactions between the Cl atoms. However, even though there are repulsive interactions between the adsorbed Cl atoms on the 100% Cl-covered surface, the B
Cl bond is strong enough to hinder atomic desorption of Cl (the average B
Cl bond strength was
140 kJ/mol). The shortest distance between the Cl atoms on the surface (dCl
Cl = 2.54 Å) is also too large for recombinative molecular desorption of Cl2 to occur (the calculated and experimentally determined Cl
Cl bond lengths, in Cl2, is 2.02 and 1.99 Å, respectively). The process of Cl abstraction from the surface, with gaseous atomic Cl, was found to be strongly exothermic (ΔEabs =
503 kJ/mol). The abstraction process was found to be even more exothermic if H radicals were used as abstracting species (ΔEabs =
696 kJ/mol); most probably due to the higher bond strength of the more polar H
Cl bond (ΔEH
Cl =
434 kJ/mol, |ΔχH
Cl| = 0.96 “polar covalent”, qH = +0.14 e, qCl =
0.14 e) compared to the Cl
Cl bond (ΔECl
Cl =
241 kJ/ mol, |ΔχCl
Cl| = 0 “nonpolar covalent”; the experimentally determined bond energies for H
Cl and Cl
Cl, in HCl and Cl2, are
431 and
242 kJ/mol, respectively). The removal of the Cl(2,1) atom from the 100% Cl-covered B-terminated c-BN(100)-(2  1) surface resulted in an initial, pronounced, but very local, collapse of the B(2,1) radical site (ΔzB(2,1) =
0.32 Å) to a more trigonal planar coordination (νN
B(2,1)‑N = 121°); that is, the B(2,1) radical initially adopts an sp2 hybridization. This initial behavior of the 94% (monoradical) Cl-covered B-terminated c-BN(100)-(2  1) surface is similar to the behavior of the 94% (monoradical) H- (ΔzB(2,1) =
0.26 Å, νN
B(2,1)‑N = 115°) and F-covered (ΔzB(2,1) =
0.35 Å, νN
B(2,1)‑N = 116°) B-terminated c-BN(100)-(2  1) surfaces. However, the electron-deficiency of the surface radical is thereafter removed by donation of lone pair electrons on Cl(3,1) into the unhybridized 2p orbital on B(2,1), giving rise to a chloride bridge bond between B(2,1) and B(3,1) (Figure 10). Such higher order coordination sites are common for halogen atoms. The B
Cl bridge bonds were tilted 51° from the surface normal and had a bond length of 1.96 Å. As the Cl(3,1) atom approached the B(2,1) radical site, the hybridization of the B(2,1) atom changed from sp2 to sp3, which resulted in a tetrahedral coordination and an elevation of the B(2,1) atom in the z direction with +0.21 Å. However, compared to the 100% Cl-covered surface, both B(2,1) and B(3,1) exhibited a small collapse (ΔzB(2,1) =
0.11 Å, ΔzB(3,1) =
0.09 Å), and formed a somewhat distorted tetrahedral coordination (νN
B(2,1)‑N = 114°, νN
B(3,1)‑N = 114°). The collapse of the B(2,1) and B(3,1) surface atoms was accompanied by an +0.10 Å elevation of B(1,1) and B(4,1) in the z direction, resulting in a buckling of the B(1,1)-B(2,1) and B(3,1)-B(4,1) dimer bonds. The average buckled dimer bond length and buckling height in the z direction were 1.76 and 0.20 Å, respectively.

ARTICLE

Figure 11. Clean N-terminated c-BN(100)-(1  1) and desorbed Cl atoms.

It can from these calculations also be seen that the radical site can be stabilized by an adjacent Cl atom, whereas adjacent H and F atoms are not able to interact with the radical site. This is most probably due to the longer B
Cl bond length (dB
Cl = 1.87 Å) compared to the B
H (dB
H = 1.20 Å) and B
F (dB
F = 1.35 Å) bond lengths. 3.4. Cl-Covered N-Terminated c-BN(100). A 100% surface coverage of on-top Cl on the initially clean N-terminated c-BN(100)-(1  1) surface was found to be unstable, and the Cl atoms atomically desorbed from the surface (Figure 11). The desorption is most probably due to repulsive sterical interactions between the Cl atoms (on-top Cl atoms would be separated by a distance of 2.56 Å, which is significantly smaller than the van der Waals distance of 3.50 Å). Such interactions were also found to be present on the 100% Cl-covered B-terminated surface. However, in contrast to the polar covalent B
Cl bond (|ΔχB
Cl| = 1.12), the N
Cl bond (|ΔχN
Cl| = 0.12) is almost nonpolar and therefore less electrostatically stabilized. The low stability of the N
Cl bond is also in accordance with the limited stability of NCl3. In the early stages of desorption, there were interactions between adjacent Cl atoms, indicating that recombinative molecular desorption of Cl2 also might be possible. That desorption occurs by an atomic mechanism, rather than by a molecular one, is probably due to the fact that the distance between the Cl atoms on the 100% Cl-covered N-terminated c-BN(100)-(1  1) surface is 2.56 Å, which is significantly larger than the calculated bond length in the Cl2 molecule (dCl
Cl = 2.02 Å). Initially, the surface adopted an amorphous structure. However, as the desorption process proceeded, the bulk-like (1  1) structure was re-established (the average surface and bulk N
N distances were 2.58 and 2.56 Å, respectively). The average surface B
N
B bond angle and N
B bond length were 120° and 1.47 Å, respectively, i.e., the clean surface exhibits a higher degree of sp2 hybridization than both the H- (νB
N
B = 116°, dB
N = 1.51 Å) and F-covered (νB
N
B = 117°, dB
N = 1.50 Å) surfaces. From these results for Cl adsorption on the B- and N-terminated c-BN(100)-(1  1) surface, it is reasonable to assume that adsorption of Br and I on the B- and N-terminated c-BN(100)-(1  1) surfaces will be highly unfavorable, due to the fact that: (i) Br and I have larger van der Waals radii (rBr(vdW) = 1.85 Å, rI(vdW) = 1.98 Å) than Cl (rCl(vdW) = 1.75 Å,), i.e., Br- and I-covered surfaces will be more sterically hindered than Cl-covered surfaces. (ii) the B
Cl (N
Cl), B
Br (N
Br), and B
I (N
I) bonds are formed by the overlap of an hybrid orbital of B (N) with a 3p, 4p, or 5p orbital of Cl, Br, and I, 22915

dx.doi.org/10.1021/jp205898w |J. Phys. Chem. C 2011, 115, 22910–22916

The Journal of Physical Chemistry C respectively. Since the average electron density is less in a 4p and 5p orbital than in a 3p orbital, the B
Br (N
Br) and B
I (N
I) bonds are expected to be weaker than the B
Cl (N
Cl) bond.26 (iii) the B
Br (χBr = 2.96, |ΔχB
Br| = 0.92 “polar covalent”) and B
I (χI = 2.66, |ΔχB
I| = 0.62 “polar covalent”) bonds are less electrostatically stabilized than the B
Cl bond (χCl = 3.16, |ΔχB
Cl| = 1.12 “polar covalent”). (iv) there is no significant electrostatic contribution to the N
Br (χBr = 2.96, |ΔχN
Br| = 0.08 “mostly covalent”) and N
I (χI = 2.66, |ΔχN
I| = 0.38 “mostly covalent”) bond energies. (v) NBr3 and NI3 are less stable than NCl3.

4. CONCLUSIONS In the present study, the surface stabilizing effect of F and Cl on the B- and N-terminated c-BN(100)-(1  1) surfaces has been investigated with DFT calculations. Both 100% and 94% (monoradical) halogen-covered surfaces were considered. The following conclusions were reached: (1) A 100% surface coverage of on-top F on the initially clean B-terminated c-BN(100)-(1  1) surface will induce a (2  1) reconstruction of the surface. However, a 100% surface coverage of on-top F on the initially clean N-terminated c-BN(100)-(1  1) surface will maintain the (1  1) structure. (2) The adsorbed F atoms on the 100% F-covered B-terminated c-BN(100)-(2  1) surface are able to uphold a bulk-like bond angle (νN
B
N = 109°) and bond length (dB
N = 1.57 Å) for the surface B atoms. However, opposite observations were made for the 100% F-covered N-terminated c-BN(100)-(1  1) surface (νN
B
N = 117°, dB
N = 1.50 Å). (3) F chemisorbs strongly to both the B-terminated c-BN(100)-(2  1) surface (ΔEads =
471 kJ/mol) and the N-terminated c-BN(100)-(1  1) surface (ΔEads =
283 kJ/mol) . (4) F abstraction from the 100% F-covered B-terminated c-BN(100)-(2  1) surface and the 100% F-covered N-terminated c-BN(100)-(1  1) surface is significantly more favorable if H radicals (ΔEabs =
130 kJ/mol (B-terminated), ΔEabs =
318 kJ/mol (N-terminated)) are used as abstracting species instead of F radicals (ΔEabs = +252 kJ/mol (B-terminated), ΔEabs = +63 kJ/mol (N-terminated)). (5) H abstraction from the 100% H-covered B-terminated c-BN(100)-(2  1) surface and the 100% H-covered N-terminated c-BN(100)-(1  1) surface is significantly more favorable if F radicals (ΔEabs =
403 kJ/mol (B-terminated), ΔEabs =
170 kJ/mol (N-terminated)) are used as abstracting species instead of H radicals (ΔEabs =
258 kJ/mol (B-terminated), ΔEabs =
25 kJ/mol (N-terminated)). (6) A 100% surface coverage of on-top Cl on the initially clean B-terminated c-BN(100)-(1  1) surface will induce a (2  1) reconstruction of the surface. However, a 100% surface coverage of on-top Cl on the initially clean N-terminated c-BN(100)-(1  1) surface is unstable, and the Cl atoms will atomically desorb from the surface. (7) The adsorbed Cl atoms on the 100% Cl-covered B-terminated c-BN(100)-(2  1) surface are able to uphold a bulk-like bond angle (νN
B
N = 109°) and bond length (dB
N = 1.57 Å) for the surface B atoms.

ARTICLE

(8)

The adsorption energy for an Cl atom on the B-terminated c-BN(100)-(2  1) surface is strongly dependent upon the surface coverage of Cl atoms (ΔEads =
471 kJ/mol (0% Cl-covered), ΔEads =
17 kJ/mol (94% Cl-covered, unrelaxed), ΔEads = +262 kJ/mol (94% Cl-covered)). (9) Cl abstraction from the 100% Cl-covered B-terminated c-BN(100)-(2  1) surface is more favorable if H radicals (ΔEabs =
696 kJ/mol) are used as abstracting species instead of Cl radicals (ΔEabs =
503 kJ/mol). (10) Br and I, will, most probably, induce large sterical hindrance on both the B- and N-terminated c-BN(100) surface. These observations indicate that the optimal gas-phase composition for growth of c-BN consists of a mixture of H and F.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT This work was supported by the Swedish Foundation for Strategic Research (SSF); “Materials Science for Nanoscale Surface Engineering”. The computational results were obtained using DMol3 from Accelrys, Inc. 3D molecular models were made with ArgusLab 4.0.1 from Planaria Software LLC. ’ REFERENCES (1) Wentorf, R. H. J. Chem. Phys. 1957, 26, 956. (2) Pease, R. S. Acta Crystallogr. 1952, 5, 356. (3) Yamamoto, H.; Matsumoto, S.; Okada, K.; Yu, J.; Hirakuri, K. Diamond Relat. Mater. 2006, 15, 1357. (4) Larsson, K. Thin Solid Films 2006, 515, 401. (5) Olander, J.; Larsson, K. Diamond Relat. Mater. 2002, 11, 1286. (6) Karlsson, J.; Larsson, K. J. Phys. Chem. C 2010, 114, 3516. (7) Arvidsson, I.; Larsson, K. Diamond Relat. Mater. 2007, 16, 131. (8) Ruuska, H.; Larsson, K. Diamond Relat. Mater. 2007, 16, 118. (9) Olander, J.; Larsson, K. Phys. Rev. B 2003, 68, 075411–1.  (10) Marlid, B.; Larsson, K.; Carlsson, J.-O. Phys. Rev. B 1999, 60, 16065. (11) Larsson, K.; Carlsson, J.-O. J. Phys. Chem. B 1999, 103, 6533. (12) Zhang, W. J.; Chan, C. Y.; Meng, X. M.; Fung, M. K.; Bello, I.; Lifshitz, Y.; Lee, S. T.; Jiang, X. Angew. Chem., Int. Ed. 2005, 44, 4749. (13) Hohenberg, P.; Kohn, W. Phys. Rev. 1964, 136, B864. (14) Kohn, W.; Sham, L. J. Phys. Rev. 1965, 140, A1133. (15) Delley, B. J. Chem. Phys. 1990, 92, 508. (16) Delley, B. J. Chem. Phys. 2000, 113, 7756. (17) Accelrys, Materials Studio Release Notes, release 4.1; Accelrys Software, Inc.: San Diego, CA, 2006. (18) Perdew, J. P.; Wang, Y. Phys. Rev. B 1992, 45, 13244. (19) Payne, M. C.; Teter, M. P.; Allan, D. C.; Arias, T. A.; Joannopoulos, J. D. Rev. Mod. Phys. 1992, 64, 1045. (20) Pfrommer, B. G.; C^ote, M.; Louie, S. G.; Cohen, M. L. J. Comput. Phys. 1997, 131, 233. (21) Monkhorst, H. J.; Pack, J. D. Phys. Rev. B 1976, 13, 5188. (22) Delley, B. J. Phys. Chem. 1996, 100, 6107. (23) Mulliken, R. S. J. Chem. Phys. 1955, 23, 1833. (24) Lide, D. R. CRC Handbook of Chemistry and Physics, 90th ed.; Internet Version 2010; CRC Press/Taylor and Francis: Boca Raton, FL, 2009. (25) Silberberg, M. S. Chemistry
The Molecular Nature of Matter and Change, 3rd ed.; McGraw-Hill: New York, 2003. (26) Bruice, P. Y. Organic Chemistry, 3rd ed.; Prentice-Hall: New York, 2001.

22916

dx.doi.org/10.1021/jp205898w |J. Phys. Chem. C 2011, 115, 22910–22916