WorksPhysics and Astronomy
Universal model of the growth of III-V tetrahedral semiconductors on polar surfaces: pyramids and glue atoms
Abstract
Ab initio extensive calculations were used to simulate the atomic stages of the vapor growth of III-V tetrahedral semiconductors (GaN, GaP, GaAs and GaSb) from their components. The structural and electronic properties of the stoichiometric polar surfaces of III-V tetrahedral semiconductors (GaN, GaP, GaAs and GaSb), i.e. metal- and nonmetal-terminated were determined. Adsorption of the components was simulated, shown to be atomic. At low coverages adatoms occupy H3 sites, considerably stabilizing the surfaces, at high coverage, the adatoms are located in the skewed on-top positions creating the first layer necessary for growth (with the exception of nitrogen admolecules). The second adlayer is adsorbed differently: on a Ga-terminated surface it does not create pyramid structures due to strong interaction within the first NM adlayer, thus preventing effective growth. On NM terminated Ga covered surface the second NM adlayer atoms create a stable pyramid structure which is an effective vehicle for surface diffusion. Growth of III-V semiconductors occurs via the attachment of the pyramids to steps by glue atoms, i.e. additionally attached NM adatoms. The NM glue atom connection process requires additional electrons, which leads to strong dependence of the process and the entire growth on the Fermi level at the NM terminated surface
Introduction
Despite the drastic differences within the family of III-V semiconducting compounds, the thermodynamic properties of a wide variety of these semiconductors exhibit some common features [1, 2, 3, 4]. These features are identified as the main factors leading to a specific growth mode that incorporates similarities in the atomic mechanisms of surface diffusion, sublimation by the creation of nonmetallic diatomic molecules and metallic atoms, and incorporation of Me-N pairs into the crystalline lattice [5, 6]. To illustrate these similarities, a series of gallium compounds, i.e. GaN, GaP, GaAs, and GaSb, will be discussed.
The common thermodynamic features of these systems are intrinsically related to the bonding of nonmetal atoms into diatomic or tetra-atomic molecules [2, 3, 4]. Metal vapors consist of atoms. The group III and V atoms differ primarily in the number of electrons in the outer shells: group III atoms have three electrons that occupy the two states, and one is located in state. Accordingly, orbitals are fully occupied, thus positive and negative overlap effects cancel. Therefore, the bonding is not strong enough for the metals to create molecules as it involves a single p-type electron only. The group III atoms do not create diatomic molecules in the vapor phase because bonding by two electrons does not provide sufficient energy gain. Consequently the saturated vapor pressures of group III metals above the condensed phases are low. Gallium compounds were selected because gallium is characterized by extremely low partial pressures in the vapor [7].
The different case are nonmetals as shown for the nitrides, where the nitrogen orbitals are not hybridized [5]. This hybridization is too costly because the difference between the energies of the nitrogen and states is extremely large. Therefore, nitrides are bonded differently: metal orbitals create bonding overlaps with orbitals of nitrogen, and separate metal orbitals create bonding overlaps with nitrogen orbitals. The latter contributes to isotropic cohesion, and the former determines lattice symmetry [6].
The group-V atoms have five electrons in their outer shells. Thus, their bonding to diatomic molecules involves five electrons. As a result, the molecular binding energies are high, and the saturated vapor pressures are correspondingly high. Among these compounds, nitrogen vapor consists exclusively of diatomic molecules, that is, . Phosphorus and arsenic form diatomic molecules: and , and can also form tetra-atomic molecules with a tetrahedral structure: and . In the case of antimony, the vapor predominantly consists of atoms [7].
The molecular binding energy is of primary importance both for understanding the bonding of nonmetal atoms and for verifying ab initio calculations. The dissociation energy of the nitrogen molecule () is extremely high, the highest of all diatomic molecules. The experimental value is [8, 9]. For phosphorus, the experimental value of the di-molecule () dissociation energy is [9]. For arsenic (), the diatomic molecular dissociation energy is [10]. Finally, antimony is atomic in the vapor; thus, the atomic state is used as the reference state in this study.
At a metal-terminated surface, the process leading to the adsorption of nonmetallic species is complex and includes the dissociation of the molecules. At present, this process is well understood in the case of the nitrides. The other systems have not been investigated. The investigations in this work include the adsorption of nonmetal molecules on metal-terminated surfaces. The adsorption of metal atoms on nonmetal-terminated surfaces will be considered separately. As shown already in the case of the nitrides, the results of the investigations of adsorption yield markedly different results at low and high coverages. For the AlN(0001) surface, adsorption at low coverage is dissociative, leading to the location of single N adatoms at H3 sites, i.e. triply coordinated [11, 12]. At higher coverage, adsorption is molecular. The adsorption energy differs markedly, in the low-coverage regime it is while in the second case it is only [11]. Similar differences are therefore expected for other surfaces and adsorbates, and therefore the investigations will consider separately low and high coverage.
The long-term objective of this research is to create ab initio based atomic-level model of the growth of III-V semiconductors that describes both growth and doping in these semiconductors. Among other factors, doping depends on the position of the Fermi level; therefore, it is necessary to include description of the electronic properties of these surfaces into dopant incorporation simulations. The present paper is devoted first to determining the crystallographic and electronic properties of polar surfaces. As shown, the two types of surfaces are drastically different. On the contrary, the surfaces of the same polarity show remarkable similarities within this set, which is the basis of the project aimed at developing a unified model of growth and doping applicable to all tetrahedral III-V semiconductors.
In this paper, we describe the entire sequence of problems involved in constructing an atomic-level quantum theory of the growth of III-V semiconductors on polar surfaces from the vapor phase. The considered examples include gallium compounds, i.e. GaN, GaP, GaAs and . Accordingly, we determine the structures of both polar, i.e. gallium (Ga) and nonmetal terminated (NM) stoichiometric surfaces. In the following, the adsorption of the first gallium and nonmetal layer on the respective NM- and Ga-terminated surfaces will be simulated at low and high coverages. During adsorption the NM di-molecules, i.e. , and , are used as the initial species. It is found that these molecules dissociate during adsorption on Ga-terminated surfaces. In the case of Sb the process is fully atomistic due to atomic nature of Sb vapor.
It is found that the high coverage of atomic nitrogen on Ga-terminated surface is unstable with respect to the formation of di-molecules and desorption. Therefore, nitrogen coverage of surface is not considered in the following. As growth requires the formation of double Ga-NM layer, the respective second supplementary layer is considered. In the course of these investigations it was discovered that the second layer of Ga on NM covered Ga-terminated surface creates structures that do not directly lead to the formation of tetrahedrally coordinated double III-V layer, therefore, growth on Ga-terminated surface is not investigated leaving this subject for future investigation.
It is found that the stable structure on NM-terminated double layer consists of pyramids, composed of Ga adatoms in the base, located on-top of NM topmost atoms. At the apex of each pyramid, a single NM atom is located attached by triple bonds to the Ga adatom base. It is shown below that the pyramids remain stable while moving by pivots around one base Ga adatom. In the course of this surface random wandering the pyramids approach the steps. The step structure naturally consists of Ga-NM layer with the Ga atoms in the first layer exposed in the front. Thus, direct attachment of the pyramid requires removal of Ga atoms in the base, either of the step or the pyramid. The process is energetically extremely costly, not efficient to sustain the growth.
A new scenario is therefore proposed in which the additional NM atoms, denoted as glue atoms are attached connecting the pyramids to the steps. The attachment of and by its dissociation into the pairs of glue atoms is simulated. Thus the process is presently directed to description of the vapor growth, further extensions to other processes are left for future investigations. It is shown that the glue-atom process depends on the mutual orientation of the step and the pyramid and also on the position of the Fermi level at the surface, as the process requires the acceptance of electrons from other quantum states. The overall conclusion is that the process is fundamental to the step dynamics of semiconductor growth in general. Further research is necessary to obtain a quantitative description of the growth and doping of semiconductors. In summary, the fundamental atomic model of the semiconductor growth is formulated.
The calculation procedure
Ab initio density functional theory (DFT) calculations were used in numerous simulations of semiconductor crystal surfaces representing the III-V family. The crystals considered were gallium-based compounds: GaN, GaP, GaAs, and GaSb. For these extensive calculations, we selected the Spanish Initiative for Electronic Simulations with Thousands of Atoms (SIESTA) ab initio shareware package [13]. The selected package, when properly parameterized, is sufficiently accurate to capture the main features of the processes within a quantum-mechanical formalism while being fast enough to perform the large number of extensive calculations required to create a complete picture of the growth. The main task is to solve the Kohn-Sham equations to determine the electronic properties of finite systems with periodic boundary conditions (PBC) imposed. The solutions are eigenvalues and the corresponding eigenfunctions, which are expressed as a series of specific finite-range numerical orbitals centered on the atoms present in the system [14, 15]. The orbitals are the products of angular and radial functions, the latter having finite cutoff radii. The angular dependence is expressed by standard spherical harmonics for , i.e., , , and orbitals. The and orbitals of gallium and nonmetals, i.e., nitrogen, phosphorus, arsenic, and antimony atoms, are represented by triple-zeta functions. In the case of gallium, the internal shell electrons are incorporated into the valence electrons, which are represented by single-zeta functions. The nonmetal atoms have electrons included in the atomic cores. The functional basis is greatly reduced by the application of Troullier-Martins pseudopotentials [16, 17].
Integration in reciprocal -space is approximated by summation over the Monkhorst-Pack grid, which was used in different density depending on the type of the simulated systems [18]. The method is an application of the 19{sup}th-century quadratures proposed by Johann Carl Friedrich Gauss [19]. The GGA-PBE (PBEJsJrLO) density functional was parameterized using the , , and values set by the jellium surface (Js), jellium response (Jr), and Lieb-Oxford bound (LO) criteria [20, 21].
The electrostatic-potential problem, tackled using the linear Poisson equation, is solved using the fast Fourier transform (FFT), which is superior to any other method. This imposes periodic boundary conditions that create coupling across the vacuum space owning to the surface-slab-dipole-induced potential difference [22]. Periodicity induces artificial interactions between slab copies that must be eliminated either by introducing a compensating dipole in the vacuum space [22, 23] or by adding a Laplace-equation solution to remove the field in the space between slab copies [24]. Both methods are applicable because they are essentially equivalent.
Under normal conditions, GaN has a hexagonal wurtzite structure with the following experimentally determined lattice parameters: Å, Å [25]. The calculated GaN lattice parameters were Å and Å in good agreement with the experimental values.

Figure 29.
The band correction scheme of Ferreira et al., known as GGA-1/2, was used to obtain improved band-gap energies, effective masses, and band structures [26, 27]. The electronic properties were obtained in the second stage using a modified Ferreira scheme, for which the positions of atoms and the periodic cell were first obtained using the PBEJsJrLO exchange-correlation functional. The band diagrams obtained using these two approximations for the four III-V compounds, GaN, GaP, GaAs, and GaSb, are presented in Figure 1 (PDF p. 7).

Figure 1 (PDF p. 7). Band diagrams and partial densities of states (PDOS) projected on metal (Ga) and nonmetal atomic states obtained directly using PBEJsJrLO exchange-correlation [13] (left column) and Ferreira correction [26, 27] (right) for the III-V compounds: (a) GaN, (b) GaP, (c) GaAs, (d) GaSb. In all electronic structure diagrams in this paper, the Fermi energy is set to zero.
The ab initio bandgap using PBEJsJrLO functional was while the corrected (GGA-1/2) value was . Thus, the Ferreira correction is essential to obtain good agreement with the low-temperature experimental data: [1, 28, 29]. These energy band data showing the bonding scheme are generally consistent with the recent results obtained for nitrides, indicating the existence of two subbands in the valence band (VB): an upper subband created by the overlap of hybridized states and nitrogen states, and a lower subband composed of and states [1, 6, 30, 31]. This division of the VB into upper and lower subbands will be used extensively below: the upper subband states and their counterparts in the case of adatoms create bonding/antibonding states that are occupied or empty, depending on the position of the Fermi level, and will therefore be considered explicitly. Thus, in the case of gallium these are , and also hybridized states. In the case of nitrogen, these will be nitrogen states. The lower GaN VB subband states and also the adatom and states will be tacitly assumed to be fully occupied, both in the crystal bulk and on the surface. Nevertheless, these states contribute to bonding both in the bulk and on the surface, stabilizing the system. Because these states are always occupied, their explicit consideration is not essential and therefore, they will not be shown in the following. The only exception will be antibonding which are additional states to be occupied and also higher in energy, so they will be considered exceptionally for this purpose.
Gallium phosphide has a cubic zinc-blende structure with the measured lattice parameter [1, 2]. The ab initio calculations gave which is in reasonable agreement with the experiment. Experimentally, GaP was determined to have an indirect bandgap [1]. The ab initio calculations using the PBEJsJrLO functional yielded an indirect gap , whereas the GGA-1/2 Ferreira scheme resulted in a direct gap . As in GaN, the VB in GaP is divided into two separate subbands: the upper composed of , and states and the lower due to and states. Accordingly, the latter are always occupied, so they will not be considered explicitly in the analysis of surface processes. The only exception will be made in the case of antibonding states that could potentially be occupied.
Gallium arsenide also has a cubic zinc-blende lattice with the parameter Å [1, 2]. The ab initio test calculations resulted in a lattice parameter: Å, i.e., in satisfactory agreement with experiment. The band gap of GaAs is direct eV at K [1, 2]. The ab initio calculations using the PBEJsJrLO functional resulted in a direct narrow gap eV. The Ferreira scheme increased the gap to eV. This result is in reasonable agreement with experimental data. In the case of GaAs, the pattern observed in the two preceding cases is again observed: the VB is divided into two separate subbands: the upper composed of , and states and the lower related to and states. By analogy with the preceding cases, the latter states will be routinely omitted.
Finally, gallium antimonide has, under standard conditions, a stable cubic zinc-blende structure with a lattice parameter: Å at K [1]. The ab initio calculations at 0 K gave similar results: Å. The band gap of GaSb is narrow and direct eV at K [1, 2]. The ab initio calculations based on the PBEJsJrLO functional yield a closed gap. The application of the GGA-1/2 correction (Ferreira scheme) opened the gap to yield a direct band gap eV. In summary, the Ferreira correction to the ab initio calculations provided satisfactory agreement of the structural and energetic parameters of all three zinc-blende III-V compounds suitable for the analysis of the most important properties of the polar surfaces. GaSb follows again the trends of separation into subbands: the upper due to overlap of , and states and the lower related to and states. The latter will not be considered in most of the investigations in the following. It is also important to consider the differences between the energies of the nonmetal and states. For GaN, the weighted energy difference was approximately 13 eV whereas for GaP, GaAs, and GaSb these differences were smaller (approximately 8 eV). Thus, the energy gain due to the increase in the proportion of states is much larger for gallium nitride than for the other compounds.
In these calculations, the self-consistent-field (SCF) loop was terminated when the difference for all elements of the density matrix between two consecutive iterations was smaller than . A real-space grid was used for the calculation of the multicenter overlap integrals which are controlled by the energy cutoff. The cutoff value is 410 Ry, which is equivalent to a grid spacing in real space of 0.08 Å.
For calculations of surface reactions with vapor species, the molecular energy is of primary importance, including both the experimental value and the theoretical estimate. Thus, the dissociation energy of the nitrogen diatomic molecule, obtained from ab initio calculations, is . This is in very good agreement with the experimental value [8, 9]. For the phosphorus diatomic molecule, the ab initio approach gave , which is again in good agreement with experiment [9]. Finally, the theoretical value for the arsenic diatomic molecule is , which is in agreement with the experimental data giving [10]. For antimony, the atomic state was used as the reference throughout these publications. For nitrogen, the tetratomic molecule has not been identified [32, 33, 34, 35]. For phosphorus and arsenic, the tetratomic molecules are stable. Their atomization energies are and , respectively [36, 37]. These tetratomic molecules will not be used as references throughout this work.
The Born-Oppenheimer approximation was enforced in the procedure for determining the energy barriers. The procedure uses the nudged elastic band (NEB) method to optimize the path whose endpoints are fixed in the optimization [38, 39, 40]. The minimum-energy pathway (MEP) is characterized by a saddle point that establishes the energy barrier. The NEB module was linked to SIESTA to determine the minimum barrier energy and path of the species. Atomic relaxation was terminated when the forces acting on the atoms were below . The opposite surfaces of all Ga-terminated slabs (i.e. NM-surfaces) have their bonds saturated by fractionally charged hydrogen () pseudoatoms. Similarly, the opposite surfaces of all NM-terminated slabs (i.e. Ga-surfaces) have their bonds saturated by fractionally charged hydrogen () pseudoatoms.
The results:
Stoichiometric surface structure
In this section, the structural and electronic properties of the metal- and nonmetal-terminated stoichiometric surfaces are investigated using ab initio calculations. These surfaces are considered on the assumption that the topmost atoms are not removed, i.e. no surface vacancies are created. Thus, the rich variety of the surface reconstruction will not be used in the design of the surface slabs [41, 42, 43, 44]. This assumption will be verified in the following sections where we will investigate the adsorption of the complementary species (i.e. Ga and NM) of the compound on metal- and nonmetal-terminated polar surfaces [45]. It will be shown that the adsorption stabilizes the polar surfaces and prevents their degradation be escape of the topmost surface atoms. Therefore, in the following, such covered surfaces are used in the atomic models of the vapor-growth-related processes.
Nonmetal (N, P, As, and Sb) side
The physical properties of nonmetal (group-V-atom)-terminated stoichiometric semiconductor surfaces, such as , , and are similar in terms of crystallographic structure and electronic properties. As shown in Figure 2, the results of ab initio slab simulations indicate the absence of any reconstruction on all surfaces.
Figure 2. Crystallographic structure of energetically stable nonmetal-terminated surfaces of III-V semiconductors obtained using slabs: (a) 8 double atomic layer (DALs) , (b) 9 DALs , (c) 9 DALs , (d) . In all calculations reported in this paper, the opposite surfaces of all slabs representing the nonmetal-terminated surfaces are saturated by fractionally charged hydrogen () pseudoatoms. Gallium, nitrogen, phosphorus, arsenic, antimony atoms, and hydrogen pseudoatoms are represented by blue, yellow, red, lavender orange and small sky-blue balls, respectively.
These results indicate that the energy reduction due to surface relaxation is limited, and it is insufficient to shift the atoms from the lattice registry. Nevertheless, the topmost nitrogen atoms are located below the plane of the gallium atoms. In the other cases, the nonmetal atoms were located above the Ga atoms.
Figure 2a.
Figure 85.
Consistent with the structural properties, the electronic properties of all four nonmetal-terminated surfaces are remarkably similar. The nonmetal broken-bond state is located slightly above the valence band maximum and is independent of the bandgap. The only difference is that, in the case of , this is the pure state in accordance with the lattice symmetry, in which N atoms are located in the plane of the Ga atoms, i.e., bonded through a hybridized state. For the other zinc-blende compounds, these are fully hybridized states. Naturally, these states are partially occupied, thereby pinning the Fermi level close to the valence band maximum (VBM). The result is consistent with the position of Fermi level (no band bending at the surface) in the bulk of p-type semiconductors.
Figure 86. (b) P —; GaP —. Color scale: , , , , . [eV]; wave vector ; [Å]; PDOS [arb. units]. Wave-vector labels: L, Γ, X, L. (c) As —; GaAs —. Color scale: , , , , . [eV]; wave vector ; [Å]; PDOS [arb. units]. Wave-vector labels: L, Γ, X, L. (d) Sb —; GaSb —. Color scale: , , , , . [eV]; wave vector ; [Å]; PDOS [arb. units]. Wave-vector labels: L, Γ, X, L.
Figure 3. Electronic properties of nonmetal-terminated surfaces represented by : (a) 8 double atomic layer (DALs) , (b) 9 DALs , (c) 9 DALs , (d) . for wurtzite and zinc blende crystalline structures respectively: band diagrams in momentum (left panel) and position space oriented along c-axis (middle panel). PDOS of nonmetal surface atoms are shown in the right panel.
The nonmetal-terminated surfaces do not exhibit separation into regions with and reconstructions. In the case of surface, the states have energies within valence band; therefore, they cannot be empty [31, 46, 47]. Thus, the total evacuation of states is not possible, and all states are fractionally occupied with all nitrogen surface atoms remain in hybridization. On the other NM surfaces, the effective, small separation of the energies of the valence band maximum states and the nonmetal broken-bond states of the GaP, GaAs, and GaSb surfaces leads to their partial occupation. Therefore, these states could not be fully occupied, as expected from the charge shift from to . Thus, they remain in hybridized state. In summary, these structural results are different from the nonstoichiometric reconstructions experimentally observed as these are reference surfaces suitable for growth simulation.
| Surface | Structure | Fermi level – VBM difference |
| - uniform | ~ 0.0 | |
| - uniform | 0.5 eV | |
| – uniform | 0.4 eV | |
| - uniform | 0.3 eV |
Table 1. Structural and electronic properties of nonmetal (NM)-terminated polar stoichiometric surfaces
As shown, these properties are relatively uniform across all considered compounds.
The opposite side of the slab, i.e., the metal-terminated surface, exhibits drastically different properties, both in terms of crystallographic structure and electronic properties, which are related to the charge-controlled and reconstruction[12, 48].
Metal (gallium) side
The structural and electronic properties of the gallium-terminated surfaces of all compounds, i.e., GaN, GaP, GaAs, and GaSb, are drastically different from those of the nonmetal side. In fact, this is related to metal states having a much higher energy. Therefore, the broken-bond states are located high in the band gap, close to the conduction band minimum (CBM). In addition, these states behave similarly for all members of the investigated group.
The atomistic structure and electronic properties of metal-terminated surfaces have been extensively investigated. The reported results changed considerably over time, the state primarily associated with the size of the simulated systems; hence, we only describe the final version [49, 50, 51, 52]. The surface was initially considered unreconstructed with Ga-BB states located close to CBM [49, 50, 51]. Subsequently, these results were considerably changed to identify - reconstructed pattern [48]. The structure of the slab representing the Ga-terminated surface is illustrated in Figure 4.
Figure 4. Metal-terminated GaN (0001) surface: represented by 8 Ga-N double atomic layers (8 DAL) slab, (a) side view, (b) top view. In the all simulation reported in this paper the opposite surfaces of all slabs representing the Ga-terminated surfaces are saturated by fractionally charged hydrogen () pseudoatoms. Gallium, nitrogen and pseudo-hydrogen atoms are represented by blue, yellow, and sky-blue balls. The topmost surface atoms in hybridization are represented by green balls.
The broken-bond states of the topmost Ga atoms (Ga-BB states) can be divided into two categories due to separation of topmost Ga surface atoms into and hybridization states. As shown in Figure 5, the latter have state empty due to the Peierls transition that opens the gap within the Ga broken-bond-state band that is located close to the conduction band minimum (CBM). This is because a fraction of the electrons is transferred to the bulk conduction band as the hybridized surface state is located partially above the Fermi level. Thus, the surface was identified as metallic because Ga-BB states exhibit considerable dispersion [49, 50, 51].

Figure 5. Electronic properties of the metal-terminated GaN (0001) surface, represented by 8 Ga-N DAL slab: band diagrams in momentum (left panel) and position space oriented along c-axis (middle panel). PDOS of gallium surface atoms are presented in the right panel.
According to recent investigations, the Ga-terminated GaN surface undergoes reconstruction [12, 46]. The periodicity of the reconstruction may be explained using the electron counting rule (ECR), originally proposed by Chadi [51]. According to the ECR argument, each Ga surface atom contributes electrons. These electrons may occupy a fraction of the Ga atom’s spin pair of hybridized states. The remaining Ga atoms have states empty. Hence, this rule yields , i.e. . The reconstruction pattern that includes the integer number of both hybridization states of the Ga atoms contains 16 atoms. The 10 atoms are in hybridization, the remaining 6 are in the standard hybridization [48]. This is in accordance with the simulated structure presented in Figure 4.
The electronic properties of the GaP(111) surface have not been determined by ab initio calculations. Such calculations were performed here using slab containing nine double Ga-P atomic layers, with a fractionally charged hydrogen pseudoatom termination at the opposite surface. As in the case of GaN the resulting GaP surface structure again adopted - reconstructions.
Figure 6. Metal-terminated GaP (111) surface: the atoms in 9 Ga-P DALs slab, (a) side view, (b) top view. Gallium, phosphorus and pseudo-hydrogen atoms are denoted by blue, red and sky-blue balls. The topmost surface atoms in hybridization are denoted by green balls.
The electronic states of (111) surface are shown in Figure 7. In contrast to the previous case, the energy difference between these two types of states is large, approximately 1 eV. Therefore, the states are clearly divided: states are located slightly below the midgap, whereas the states are degenerate with the conduction band (CB) and are therefore empty.

Figure 7. Band diagrams in momentum (left panel) and position space (middle panel) of 9 -P DAL slab representing -terminated (111) surface. DOS and PDOS of gallium surface atoms are shown in the right panels.
These calculations confirmed the predictions based on the Chadi ECR hypothesis, which is also valid for the GaP surface [53]. A charge-based reconstruction was observed within the (111) GaP surface. Moreover, the large difference between the energies of these, differently hybridized states, is reflected in the reconstructed structure in which the difference in the positions of these two sets is large.
The subsequent, metal-terminated surface, investigated here is a gallium-terminated (111) surface, simulated using the slab plotted in Figure 8. As shown, a division into and hybridized gallium surface atoms is observed. Nevertheless, the number of atoms in the latter configuration is five, which differs from the previously observed 6 10 divisions.
Figure 8. Metal-terminated (111)GaAs surface represented by 9 Ga-As DALs slab. Gallium, arsenic and pseudo-hydrogen atoms are denoted by blue, pink and sky-blue balls. The topmost surface atoms in hybridization are denoted by green balls.
This behavior can be explained by analyzing of the quantum states calculated using slab presented in Fig. 9. The quantum states presented in Fig. 9 are again divided into two sets.

Figure 9.
Figure 9 (PDF p. 21). Band diagrams in momentum (left panel) and position space (middle panel) of (4 × 4) 9 Ga-As DAL slab representing (111) GaAs surface. DOS and PDOS of gallium surface atoms are presented in the right panels.
The states are degenerate with CB and are therefore empty. The states are located in the lower portion of the bandgap and are occupied. On the contrary, states have various energies which is related to the different relaxation of these atoms. In addition, the antibonding states pin the Fermi level, and accordingly they are partially occupied. This reduces the number of hybridized atoms to five. The number of atoms in these two sets is not compatible with the ECR because electrons are partially transferred to these antibonding states. In addition, the upward positions of these atoms are not as clearly divided as in the case of the GaP surface.
The final case considered is the Ga-terminated (111) surface, shown in Fig. 9. The division into and hybridized sets is again obeyed, in accordance with the ECR, in 6 ÷ 10 proportion as in GaN observed previously [48].
Figure 10. Metal-terminated (111) GaSb surface represented by (4 × 4) 9 Ga-Sb DALs slab. Gallium, antimony and pseudo-hydrogen atoms are denoted by blue, orange and sky-blue balls. The topmost surface atoms in hybridization are denoted by green balls.
This conclusion is supported by the slab band diagram shown in Figure 11. In fact, states are degenerate with the valence band (VB) and therefore they are fully occupied. The states are degenerate with the CB and are therefore empty. The bandgap is narrow; therefore, the surface states are located outside the gap, shifted into the bands.

Figure 11. Band diagrams in momentum (left panel) and position space (middle panel) of (4 × 4) 9 Ga-Sb DAL slab representing (111) GaSb surface. DOS and PDOS of gallium surface atoms are presented in the right panels.
The results of the simulations of Ga-terminated polar surfaces are summarized in Table 2 below. This summary shows that metal-terminated stoichiometric surfaces also exhibit unified behavior marked by charge-controlled reconstruction into (4 × 4) pattern. Thus, the hypothesis of charge-controlled reconstruction, first proposed in Ref. 48, was confirmed for a wider class of semiconductor surfaces [12].
| Surface | Structure ( ratio) | CBM - Fermi level difference |
| eV | ||
| 0.8 eV | ||
| eV | ||
| 0.4 eV |
Table 2. Structural and electronic properties of gallium (Ga)-terminated polar stoichiometric surfaces
The Table 2 below presents the summary of the structural and electronic properties of the simulated Ga terminated surfaces. The ( ratio) equal to is equivalent to charge controlled (4 × 4) reconstruction pattern.
Adsorption of nonmetal atoms (N, P, As, and Sb) on Ga-terminated surfaces
Adsorption of the III-V species being the components of the semiconductor compounds, relevant to growth, can be divided into the adsorption of metal and nonmetal atoms and/or molecules. On a metal-terminated surface, the attachment of the nonmetal (NM) species is the first stage of the vapor growth sequence followed by the attachment of metal atoms to create a complete double atomic layer. The first stage, considered here, is the attachment of NM species on Ga-terminated surfaces. Again, in all simulations the opposite surfaces (NM surfaces) of the Ga-terminated slabs have their bonds saturated by fractionally charged hydrogen () pseudoatoms.
Adsorption of nitrogen at Ga-terminated surface
Nitrogen adsorption on Ga-terminated GaN(0001) has already been intensively investigated. The nitrogen molecule has a very high binding energy, which was eV [54], in good agreement with the experimental value eV [8]. For nitrogen the tetratomic molecule has not been discovered [32, 33, 34, 35]. Despite the large dissociation-energy barrier of molecular nitrogen, the nitrogen di-molecule dissociates during attachment on a liquid-gallium surface [55, 56] and also on the Ga-terminated GaN(0001) surface [57] [58, 59]. The energetically stable location of a single N adatom at stoichiometric surface is in H3 site [51], as it is shown in Figure 12. In fact nitrogen adsorption at low coverage was investigated as the stability problem for surface in which the stable H3 site was identified [51] [59].
Figure 12. 8 Ga-N double atomic layers (DALs) slab used to simulate a nitrogen adatom attached to a clean (stoichiometric) (0001) surface representing low coverage: (a) and (b) single N adatom, corresponding to , (c) and (d) three N adatoms, corresponding to . Blue, yellow and small sky-blue balls represent gallium, nitrogen,
and hydrogen pseudoatoms, respectively. The small magenta and large green balls represent N adatoms and the topmost Ga atoms bonded to the N adatoms, respectively.
As shown, the nitrogen adatom, located at an H3 site, is bonded to the three topmost Ga atoms, i.e., it is located outside a standard site of the GaN crystal lattice. The band structure of the GaN slab and the quantum states of topmost Ga atoms and the N adatom are shown in Fig. 13.

Figure 13.
Figure 13 (PDF p. 26). Energies of the quantum states of the slab representing clean GaN(0001) surface with a single N adatom located at an H3 site: (a) single N adatom to (), (b) three N adatoms (). The panels represent, from left to right, the energies of the quantum states in the momentum and the position space, projected density of states (PDOS) of the topmost Ga atoms and the N adatom, and, in the rightmost panel, Crystal Orbital Hamilton Population (COHP) of the atoms distinguished in Figure 12 [60, 61]. Thus, the presented here COHP data denote the N adatom and the closest topmost Ga atoms. In all diagrams in this paper reporting ab initio calculations the positive COHP values correspond to the bonding overlap.
These data show that the states create bonding states with the nearest Ga atoms hybridized orbitals. Finally, nitrogen orbital has no overlap, i.e. it represents the nitrogen broken-bond state. This state is located slightly above the midgap. It is worth noting that the bonds between N adatom and the three nearest Ga atoms are severely strained as these atoms are not in the tetrahedral registry. Therefore, the overlap is reduced and the state energy increases. Nevertheless, these states are located approximately 0.5 eV above the valence band maximum (VBM). Thus, all nitrogen states are occupied. The Fermi level is not pinned, located between and hybridized atoms states.
The adsorption of a larger number of N adatoms could potentially lead to the Fermi level transition, but the results obtained do not prove that. In fact, the result is different, it is impossible to create high atomic N coverage. As shown in Figure 14, at high coverage, the N adatoms form molecules, weakly attached to the surface. Thus, the process does not provide a pathway for GaN growth at this surface using molecular nitrogen. At high coverage, the created nitrogen molecules are only weakly attached to the surface, in full analogy to nitrogen at surface [11]. As explained above, the energy gain related to electron transfer to the lowest energy state is large, thereby inducing decomposition on a bare surface or at low coverage on the surface. At higher coverage, the nitrogen atoms coalesce into the molecular form which is eventually detached from the surface. It has to be added that effective MOVPE/HVPE growth of GaN is possible due to presence on hydrogen stabilizing high nitrogen radicals/admolecules at high coverage creating the route to effective crystallization of GaN using ammonia source of active nitrogen [61, 62, 12]. These processes are outside of the scope of the present paper limited to growth of III-V compounds from their components, i.e. without hydrogen.
Figure 14. (4 × 4) Ga-N 8 DALs slab used for simulations of nitrogen attachment at (0001) surface: (a) side view, (b) top view. The symbols are identical to those used in Figure 12. In addition, the magenta balls represent nitrogen adsorbate (in molecular form).
In summary surface may be covered by nitrogen adatoms, up to critical coverage . Within this range, the surface has three types of broken bonds (unsaturated states), ordered according to their energy: the highest related to state, degenerate with the CB and empty, the second associated with the state, close to the CBM and occupied by electrons and the lowest associated with the state, close to the VBM and also occupied by electrons. These states and their electrons play an important role in stabilizing surface against evaporation, as it will be shown below.
An important factor is the difference in adsorption energy. Accordingly, the energy gain is sufficient to attach nitrogen at an H3 site at low, and insufficient to attach separate N adatoms at higher coverage. Nevertheless, the attachment of the N adatom at an H3 site leads to stabilization of the unreconstructed surface as the topmost Ga atoms are located in standard crystallographic positions. The only exception concerns the N adatoms located at H3 sites, outside the registry with the GaN lattice, which are a critical stabilizing factor, proving the earlier stability results [49] [59].
Adsorption of phosphorus at Ga-terminated surface
The adsorption of phosphorus at Ga-terminated surface may be potentially different because of the two factors: (i) GaP lattice symmetry is zinc blende, (ii) phosphorus forms molecules that are much more weakly bound than nitrogen. Their dissociation energies are and for P and P molecules, respectively [36]. Thus, the energies associated with a single bond are much lower for the tetra-molecule. Nevertheless, the two topmost atomic layers are the same, leading to an identical structure of zinc blende and wurtzite gallium-terminated surfaces as demonstrated above. Thus, the difference in lattice structure (i) is not essential. The second factor, i.e. weaker bonds (ii) is more important; therefore, the propensity toward the molecular state is considerably reduced and the atomic forms of the P adsorbates are more stable.
A single phosphorus atom is attached at a skewed H3 site at low coverage, as shown in Figure 15(a-Figure 15b). The phosphorus adatom creates bonds with the three nearest top-layer Ga atoms, maximizing the energy gain. Of the remaining top-layer Ga atoms, a fraction is located in the P plane, thus adopting hybridization while the others are located higher and preserve hybridization typical of the bulk.
Figure 15. (4 × 4) Ga-P 9 DALs slab (left panels - top view, right panels - side view) obtained in simulations of phosphorus attached at GaP (111) surface: (a-b) – single P adatom (), (c-d) sixteen P adatoms (). Blue, red and small sky-blue balls represent
gallium, phosphorus, and pseudo-hydrogen atoms, respectively. Magenta balls represent adsorbed phosphorus adatoms. The beige ball represents the adatom labeled 13 located in slightly skewed on-top position. The orange and brown balls represent adatoms labeled 2 and 8, located in heavily skewed on-top positions.
When the coverage increases to , the reconstruction of the surface is removed, thus the crystallographic surface is stabilized. It is also expected that a considerable increase of the phosphorus coverage leads to a change in adsorption sites, the phosphorus adatoms tending to move to the on-top positions. The crystallographic arguments indicate that the transition should occur for the critical coverage .
The structure of GaP(111) surface, fully covered with phosphorus atoms, is presented in Figure 15(c-Figure 15d) showing that the phosphorus atoms tend to occupy on-top positions with varying degrees of skew, creating irregular phosphorus-adatom chains. Thus, the phosphorus-adatom configuration is different from that on the previously described GaN surface: the phosphorus adatoms coalesce into chains of different lengths which are attached to the topmost Ga atoms. They remain in the connected disordered atomic states. The band structure of these slabs, shown in Figure 16, does not indicate any significant field within these slabs.

Figure 16. Energies of the quantum states of the (4 × 4) slab representing GaP(111) surface with: (a) single P adatom located at an H3 site, (b) covered by 16 P adatoms. The panels represent, from left to right, the energies of the quantum states in the momentum and the position space, PDOS of the topmost Ga atoms and P adatom, normalized PDOS of all 16 P adatoms, and COHP of the atoms distinguished in Figure 15, i.e. P adatom and the closest topmost Ga atoms or other P adatoms.
The data in Figure 16(a) indicate that the single phosphorus atom states form three bonding states with the nearest Ga atoms hybridized states, thereby saturating their broken bonds. The energies of these states are degenerate with the valence band; accordingly, these states are fully occupied. Because the Fermi level is pinned by the Ga broken-bond state located close to CBM, the phosphorus orbital broken-bond state is also fully occupied. It is also worth noting that the broken-bond state of phosphorus has an energy at the VBM. In fact, the energy difference between Ga-P saturated and broken P bonds is approximately 0.8 eV, i.e., it is not large, indicating that misalignment of the saturated bonds reduces overlap considerably. It is therefore expected that an increase in coverage would lead to a transformation to another structure and is associated with a Fermi-level pinning transition from the Ga broken-bond state (Ga-BB) to the valence band maximum (VBM). The coverage can be estimated using the extended electron counting rule (EECR) [12, 62, 63], which can be applied to the attachment of a single P atom at an H3 site ():
where and are the number of electrons on the quantum states of P adatom and empty site (), respectively. These values are and . and are the number of electrons in the adsorbed atom () and at an empty site (), respectively. These calculations show that the critical phosphorus coverage is . This is considerably less than the crystallographic limit .
In the slab, shown in Figure 15(b), sixteen phosphorus atoms are attached to the topmost Ga atoms, covering all surface sites. The electronic states of this slab are presented in Figure 16(b). The overlap of the phosphorus adatoms states create bonding and antibonding states, both located deep in the valence band, not shown on the graph. Naturally, they create bonding overlap with the topmost which is the main stabilizing effect of phosphorous coverage [6]. They are fully occupied, each accepting two electrons from the five electrons contributed by a single P adatom. Accordingly, the P3s states of all 16 P adatoms accept 32 electrons. Therefore, each P adatom contributes the remaining three electrons, i.e., these 16 P adatoms contribute 48 electrons in total. The Ga broken-bond (Ga-BB) state contributes electrons each [53]. Thus, the Ga-BB states contribute 12 electrons in total. Accordingly, the combined number of electrons that could occupy P3p states is: . These electrons must be redistributed among the 16 phosphorus adatoms P3p states.

Figure 16a.
Accordingly, the number of electrons for a single P adatom in states is three. Since the states are spin-degenerate, they can be neither fully occupied nor empty. The system may be divided into two groups or become disordered. As shown in Figure 15(b), the full P coverage adopts a disordered state resulting from the two interconnected factors: (i) the energies of different states are close, (ii) the interaction between the adatoms is sufficiently strong to shift their energies so that their order can be interchanged in the energy scale. This disorder-inducing factor determines the division into occupied and empty states.
All phosphorous adatoms create chains with the other P adatoms. These chains are of various lengths, including two, three, or four adatoms, reflecting strong interactions that modify their energies considerably. In Figure 15(c-d), the three distinguished P adatoms are labeled P2, P8, and P13. The atom labeled P2 is in a highly skewed position, the angle relative to the Ga-atom plane is 47.9°. The other two (P8 and P13) are less skewed, their angles are 60.5° and 78.3°, respectively. These angles determine the P-P bonding energy: P2 and P8 atoms create a bonding state with the energy located deep in the VB. The two other states are unsaturated with their energies located at the CBM. Thus, the energy gain from saturation of states is on the order of , i.e., it is considerable. These skewed positions are therefore controlling the energies and occupation of the states through redistribution of the electrons.
Another factor controlling the system energy is bonding between a P adatom and Ga topmost atoms. This is related to the bonding overlap states presented in the lower part of Figure 16(b). The highly skewed P4 atom has considerably reduced overlap with the Ga88 atom; thus, this state has an energy at the CBM, i.e., it is not occupied. The attachment of the phosphorus to the Ga topmost atoms is therefore only due to bonding. The bonds and create two states: one located in the VB and the other in the band gap, i.e., they are occupied and empty, respectively. Finally, the upward-positioned atoms P11, P13, and P15 due to their overlap with Ga97, Ga79 and Ga106 topmost surface atoms create states in the VB and therefore they are occupied.
In summary, the analysis reveals the origin of the states that are located in the VB, in the band gap and close to CBM as shown in the band diagram in Figure 16(b). These states may be affected by additional external factors, such as the deposition of a second layer of Ga adatoms, as it will be discussed below.
Adsorption of arsenic at Ga-terminated surface
The next member of the series of metal-terminated surfaces of III-V compounds is the GaAs(111) surface. The attachment of arsenic at Ga-terminated surface is similar to that of phosphorus due to the two factors: (i) the lattice symmetry is zinc blende, (ii) arsenic in the vapor phase forms molecules and analogous to phosphorus molecules in the vapor [37]. Therefore, the propensity toward the molecular state is reduced and the adatoms are more stable. Arsenic is attached in atomic form at a skewed H3 site even at low coverage, as shown in Figure 17a-Figure 17b). Nevertheless, there are considerable differences between these systems. The obtained configuration of a single As adatom on the GaAs(111) surface is similar to that of phosphorus. The As adatom is located at an H3 site, saturating the three broken Ga bonds.

(a)

(b)

(c)

(d)
Figure 17.
Figure 17 (PDF p. 35). (4 × 4) Ga-As 9 DALs slab (left panels - top view, right panels - side view) obtained in simulations of arsenic attached at GaAs (111) surface: (a-b) – single As adatom (),(c-f) – 15 As adatoms (), (c-d) 16 As adatoms (). Blue, lavender and small sky-blue balls represent gallium, arsenic and pseudo-hydrogen atoms, respectively. The adsorbed As adatoms are represented by red balls. In panel (b), As14, As11, and As12 are represented by green and yellow balls, respectively, in (c) As7 and As8 are represented by green and yellow balls, respectively,
The calculations were also used to determine the configuration of As adatoms at high coverage on the GaAs (111) surface. The results are shown in Figure 18. As with the previously observed adsorption of P at GaP(111) surface, at high coverage, arsenic remains in atomic form, attached by a single bond to the underlying topmost Ga atoms. In addition, the overlap of the lower energy states, i.e. and contributes to the attachment energy stabilizing the As adlayer at the surface. In contrast to P adatoms, the configuration of As adatoms varies at high coverage: for full coverage () the adatoms are arranged in highly ordered rows composed of the As-adatom pairs. At lower coverage (), the order is lost; the As adatoms are no longer arranged in rows, but some of them are attached in pairs forming bridges.

Figure 18. Energies of the quantum states of the slab representing clean GaAs(111) surface with: (a) single As adatom located at an H3 site (), (b) covered by 15 As adatoms (), (c) covered by 16 As adatoms (). The panel description is as in Figure 16.


Figure 87.
The band structures of these slabs are presented in Figure 18. As it is shown, the broken As bond state is close to the VBM. The difference between the Ga-As and As broken-bond energies is smaller than for Ga-P; it is approximately 0.5 eV. As before, it is expected that an increase in coverage would lead to a transformation to another structure and is associated with a Fermi-level pinning transition from the Ga broken-bond state (Ga-BB) to the valence band maximum (VBM). The coverage can be estimated using the extended electron counting rule (EECR) [62, 63, 12] by analogy with Ga-P. Again, these calculations show that the critical As coverage is .
The high-coverage 15-As-adatom () slab has configurational and electronic properties that are essentially similar to those of the slab representing fully covered GaP (111) surface (). The As adatoms are grouped into disordered As-As chains that are located on top of the topmost Ga atoms, with varying degrees of skew. Thus, these atoms have their states that may be saturated through the formation of bonds with Ga atoms or other As adatoms, or remain unsaturated, i.e., without overlap. Therefore, these states are divided into low-energy states degenerate with the valence band and the higher energy states with no overlap.
As shown in Figure 18(b), the most strongly skewed As adatom, e.g., the one labeled As14, has and states with no overlap, located at the CBM. The other state, which overlaps with Ga, is degenerate with the VB. The other, less skewed atoms, such as those labeled As11 and As12, are bound differently. They have no broken bonds, and all and states saturated by overlaps with Ga and As, or with As, respectively. These overlaps give different energies: the and are located at the VBM and deep in the VB, respectively.
Full coverage of the surface leads to an ordered arrangement of atoms. The As-As bonding states are located deep in the VB. The Ga-As bonding states are located at the VBM. In general, the structure order is fragile, removal of a single As adatom leads to complete disorder in the system as in the slab covered by 15 As adatoms, shown in Figure 17b.
Adsorption of antimony at Ga-terminated surface
Finally, the adsorption of antimony at surface is also simulated. The attachment of Sb at Ga-terminated surface differs from the preceding cases in one of the two aspects: (i) the lattice symmetry is zinc blende as in the previous two cases, (ii) antimony is atomic in the vapor. The latter accounts for the difference. Therefore, for GaSb the propensity toward a molecular state is absent and the adatoms are more stable. A single antimony atom is adsorbed in atomic form at a skewed on-top position as shown in Fig. 19.
Figure 19. Ga-Sb 9 DALs slab (left panels - top view, right panels - side view) used to simulate antimony adsorption at (111) surface: (a-b) – single Sb adatom (), (c-d) 16 Sb adatoms (). Blue, orange and small sky-blue balls represent gallium, antimony, and pseudo-hydrogen atoms, respectively. The green balls represent Sb adatoms. The magenta and red balls represent Sb2, Sb14, and Sb5, respectively.
The obtained configuration of the single Sb adatom on (111) surface is similar in some respects to the two preceding cases. The single Sb adatom is located in an H3 site, saturating three broken Ga bonds. At high coverage, antimony adatoms behave similarly to phosphorus, with the difference that at full coverage the Sb adatoms remain in a disordered arrangement, as shown in Fig. 19. Analogously to phosphorus, Sb is attached in atomic form to topmost Ga atoms. Sb adatoms are not ordered; nevertheless, they create Sb-Sb bonds.

Figure 19 (PDF p. 40).
The electronic properties of this system are presented in Figure 20. The band structure shows similarities to the previous cases. Despite the much smaller band gap, the broken Sb bond states are again close to the VBM. The difference between Ga-Sb and Sb bonds is similar to that for Ga-As and is approximately 0.5 eV. As before, it is expected that an increase in coverage would lead to a transformation to another structure and is related to a Fermi-level pinning transition from the Ga broken-bond state (Ga-BB) to the valence band maximum (VBM). The coverage can be estimated using the extended electron counting rule (EECR) [62, 63, 12] by analogy to Ga-P and GaAs. Again, these calculations show that the critical Sb coverage is .

Figure 20. Energies of the quantum states of the (4 × 4) slab representing clean GaSb(111) surface with: (a) single Sb adatom located at an H3 site (), (b) covered by 16 Sb adatoms (). The panel description is as in Figure 16.
At high coverage, the arrangement of Sb adatoms is disordered. Most Sb adatoms form bonds with topmost Ga atoms and with other Sb adatoms. In some cases, Sb adatoms form bonds with other Sb atoms in the crystal lattice. These atoms are typically located at skewed on-top positions. The bonds formed between Sb14 and Ga106 and between Sb15 and Ga125 atoms create bonding states with energies at the VBM, thereby pinning the Fermi energy. In addition, neighboring Sb atoms create bonding states with much lower energies, deep in the VB. For example, the Sb2 atom is located in the bridge position between an Sb223 atom in the GaSb lattice and a Ga31 top-layer atom. The Sb223 atom has a single bond with a Ga115 atom in addition to the bond with the Sb2 adatom. The remaining state is unsaturated, and its energy is located at the CBM, thus it is empty. This is rare; in most cases, Sb atoms have all their bonds saturated so that the number of broken bonds is very small. This is visible in the structure presented in Fig. 19(b), where all Sb atoms are locked in tight chains.
The results of the large-slab calculations related to Ga-terminated surfaces of III-V compounds are summarized in Table 3.
| Surface | Low coverage () | High coverage () | ||
| Structure | Fermi level | Structure | Fermi level | |
| 0.4 eV below CBM | None | |||
| 0.3 eV below CBM | – disorder | Midgap | ||
| CBM | - order/disorder | VBM | ||
| VBM | - disorder | VBM |
Table 3. Structural and electronic properties of gallium (Ga) terminated polar surfaces under nonmetal single layer coverage.
As summarized for all compounds the low and high NM coverage Ga terminated surfaces exhibit drastically different structural and electronic properties.
Adsorption energies and the thermodynamics of the adsorbates
The most important quantity relevant to the thermodynamic equilibrium between the vapor and the adsorbate is the adsorption-energy gain. As shown, the energy gain critically depends on the pinning of the Fermi level at the surface owing to the electron-transfer contribution [62, 63, 12]. The adsorption energy of species S is defined as:
where the DFT total energies are obtained for the separate slab and adsorbate, and the slab with the attached adsorbate (slab + adsorbate), respectively.
The selected species are , , and , respectively. Thus, the adsorption energy is negative when the energy of the system after attachment is increased. The adsorbed state is then is energetically stable with respect to the molecules/atoms in the vapor. The adsorption energies obtained for the considered processes is shown in Figure 21.
Figure 21. Adsorption energies of nonmetal species at Ga-terminated polar surfaces as a function of the surface coverage.
These data prove that adsorption of nitrogen at full coverage on GaN(0001) is not favored energetically, i.e., the energy change upon adsorption is positive at very high coverage. This results in the formation of molecules as shown in Figure 14. At very low coverage, the electron transfer contribution during adsorption of molecular nitrogen is large enough to obtain the energy gain. At the critical coverage the bond-formation-related energy gain approaches zero. At higher coverage, the adsorption energies show large energy loss, which is partially reduced due to the formation of molecules at higher coverage, as shown in Figure 14.
The N adatom in the on-top position is a high-energy, completely unstable position and is therefore prone to desorption upon the formation of the molecules. In all other cases, the adsorption energy is negative, i.e., the adsorbate is attached to the surface. Nitrogen adsorption is more favorable at low coverage, whereas for the other species adsorption at full coverage provides a larger energy gain.
At low coverage, i.e., for the attachment of a single atom at an H3 site, an increase in coverage leads to a reduction in the adsorption energy when the electron contribution is absent. At higher coverage, however, the energy gain is again increased due to the formation of adatom chains located at on-top positions as shown in Figure 15 and 17. Finally, Sb adsorption is entirely atomic. The electron-transfer energy gain is small because of the narrow gap; therefore, the energy is nearly constant at low coverage. Similarly, the bond strength is low, so the energy gain is nearly unchanged at high coverage. The overall energy gain is about 4 eV, which is related to an atomic vapor configuration and therefore lacks a molecular-energy contribution.
The equilibrium pressure of nonmetal atoms above the surface can be obtained by assuming equilibrium between the vapor and the adsorbate, i.e., with respect to the reaction
which requires equality of the chemical potentials of both species:
Assuming the ideal-gas equation of state and the standard dependence of the chemical potential of the adsorbate, we obtain the standard equation for phosphorus adsorption at the surface:
where the vapor part of the free energy of di-molecular phosphorus in the vapor phase is given by the free energy at normal pressure ( bar) , and the vibrational contribution to the free energy of the adsorbate . The adsorption energy is defined as follows (Eq. (5)). The temperature dependence of the free energy of the di-molecular phosphorus vapor, obtained from the thermodynamic properties of the elements [64], is plotted in Figure 22.
Figure 22. Free energy of di-molecular phosphorus at normal pressure, as a function of temperature: black line – thermodynamic data, red line – polynomial approximation (Eq. (6)).
The polynomial fit gives the following temperature dependence of the vibrational free energy (in eV):
where and the temperature is in kelvins. Similarly, the vibrational free energy was obtained for phosphorus adatoms, as plotted in Figure 23.
Figure 23. Vibrational free energy of single P adatom obtained as an average for a selected number of adatoms, as a function of temperature.
Note that the free energies of the P adatoms are much lower than those of the vapor molecules. This is observed both at low coverage for P adatoms located at H3 sites and at high coverage, where P adatoms are located at on-top positions. This result is also partially related to the contribution from surface reconstruction, which contributes to the energy change, increasing the free-energy offset at zero coverage. On the other hand, similar results were obtained at high coverage, where the atoms are relatively loosely attached to the surface Ga atoms. Nevertheless, P adatoms create tightly bound chains, which drastically increase the stiffness of the system to a level comparable to that of adatoms located at H3 sites.
Again, a polynomial fit was used to obtain the following temperature dependence of the vibrational free energy (in eV):
where and the temperature is in kelvins. These data can be used to obtain the minimum phosphorus pressure in the vapor, required to achieve the surface coverage threshold () i.e., the emergence of P adatoms located at on-top positions. This pressure is plotted in Figure 24.
Figure 24. Threshold pressure of molecular di-molecular phosphorus () for the vapor growth of GaP on the GaP(111) surface.
The temperature of the vapor growth of GaP depends on the method:
These pressures are extremely low; thus, for growth simulations the only valid assumption is a high-coverage configuration of the phosphorus atoms at on-top positions on the topmost Ga atoms. Similar assumptions can be made for the growth of GaAs and GaSb. For GaN, growth from the vapor in the N-rich regime is not possible.
Surface stability
An important question is related to the application of reconstructed surfaces to crystal-growth modeling. Detailed examination indicates that this is unnecessary in the case of Ga-terminated surfaces. Reconstructions may be divided into three categories that are induced by [41, 42, 43, 44]:
atom escape, i.e., the emergence of surface vacancies,
the attachment of adatoms
(iii) the motion of surface atoms leading to a reduction in surface symmetry.
The most critical process is the detachment of surface atoms from these surfaces. These investigations compare a clean surface with surfaces at low coverage in which the nonmetal atoms are located in the H3 positions. The ab initio calculations were used to determine the energy difference between a clean or nonmetal-covered surface and a slab with a single detached Ga atom:
These results are presented in Figure 25, where the negative values correspond to stability with respect to Ga detachment from the surface. As shown, these surfaces are highly stable against the formation of surface vacancies. This is partly related to the atomic nature of Ga vapor, so that the energy gain associated with molecule formation is absent.
Figure 25. Ga detachment energy from a Ga-terminated surface, calculated using Eq. (1), for the case of slab clean and also with four adatoms at H3 sites (). For example, GaN/Ga denotes detachment of a Ga atom from GaN surface.
As shown, the energy barrier for detachment of a single Ga atom from a clean surface is on the order of 3 – 4 eV. The lowest value occurs for GaAs, reaching eV. Thermal effects at typical process temperatures are much smaller (at least one order of magnitude), therefore, these surfaces are stable against Ga surface-vacancy formation at relevant temperatures. Adsorption of a nonmetal atom at an H3 site additionally stabilizes the surface by increasing the Ga desorption energy barrier by at least 1 eV. Nevertheless, this is of minor importance as the nonmetal atoms have stability similar to that of Ga surface atoms. In summary, the formation of Ga surface vacancies is unlikely. Thus, these investigations can be limited to the two different channels of surface reconstruction: (ii) and (iii).
Investigations of the GaN(0001) surface revealed a relaxation of the topmost Ga atoms leading to separation into and hybridized atoms, opening a Peierls gap and reducing surface symmetry [46]. The reconstruction is charge-controlled, leading to the emergence of symmetry. These findings were fully confirmed by the investigations above. In fact, Ga-terminated surfaces of all the considered semiconductors undergo such reconstruction, demonstrating the universality of the charge control argument. Thus, this type of reconstruction is fully accounted for by these ab initio calculations.
Nevertheless, the third channel (iii) only remains relevant for the vapor growth and therefore it is explicitly simulated in the following. The crystal-growth processes on the Ga-terminated surfaces of these compounds can be simulated without initially invoking surface reconstruction (i-ii), by starting with the structures considered above, i.e. without any surface vacancies.
Adsorption of gallium at nonmetal-terminated surfaces (B surfaces).
The stoichiometric nonmetal-terminated (NM-terminated) surfaces, investigated above have all essentially identical properties. They do not reconstruct, and their electronic properties are determined by the topmost nonmetal broken-bond states, which are partially occupied. Therefore, they pin the Fermi level on the surface at the corresponding energy. These states have their energies located at the VBM. Their dispersion is relatively small because of the small wavefunction overlaps between broken bond states of neighboring atoms. Technically, the simulation slabs of the opposite surfaces of all nonmetal-terminated slabs have their bonds saturated by fractionally charged hydrogen () pseudoatoms.
Adsorption of gallium on N-terminated surface
Adsorption of gallium on N-terminated surface has been investigated less intensively than the surface of the opposite polarity. This is because the growth of nitride layers. by MOVPE and MBE on this side has been relatively less successful. Thus, relatively few ab initio based studies were devoted to simulations of this system [31, 44, 45, 49, 50] [59]. The investigations showed that at low coverage the Ga adatoms are located at an H3 site, saturating N broken bond states [59]. At higher coverage, they are shifted to the on-top positions [59].
In parallel, aluminum adsorption on N-terminated surface was investigated [69]. These investigations also showed that a single Al atom is attached at an H3 site, creating bonding states with the three nearest-neighbor atoms at approximately $2.5$ eV above VBM. According to the analysis presented in Ref. 69, the Fermi level is pinned by N-broken bond (N-BB) states located above the VBM. At the critical coverage () the N-BB state is filled and the Fermi level is shifted to state near midgap. At this point, the Al adsorption energy is reduced from eV to eV$ [69].
Similar electronic properties related to Fermi level upward motion is observed for N-terminated surface under Ga coverage as shown in Ref 59 [59]. Analogous results were obtained in this study as presented in Fig. 26. Again, single gallium adatom is located at an H3 site, saturating three N-BB states.
Figure 26.
Figure 26 (PDF p. 51). (4 × 4) 8-double-atomic-layer (DALs) Ga-N slab with a single Ga adatom, used to simulate gallium adsorption on N-terminated GaN (000) surface at (a) low coverage (), (b) full () coverages, respectively. Top row - side view, bottom - top view. Blue, yellow and small sky-blue balls represent gallium, nitrogen, and hydrogen pseudoatoms, respectively. The magenta ball represents a single Ga adatom. The green balls represent the adatoms Ga4 and Ga9.
The electronic properties of the above system are shown in Figure 27. The topmost nitrogen bare atoms have their states located at the VBM. As shown above, these states are partially occupied, pinning the Fermi level. The Ga adatom and - states are not hybridized; they bond with the topmost states separately. These states are degenerate with states in the VB, located below the VBM, i.e., they are occupied. Similarly, the two states are occupied, located deep in the VB. It is worth noting that and states are hybridized, creating antibonding overlap with states. These states are located in the middle of the gap, and are unoccupied. Thus the total number of occupied states for the Ga adatom is 6.

Figure 27. Energies of the quantum states of the (4 × 4) slab representing stoichiometric N-terminated GaN(000) surface with: (a) single Ga adatom located at an H3 site (), (b) 16 attached Ga adatoms (). The panels represent, from the left: energy of the quantum states in the momentum and the position space (the surface is located on
The EECR may be applied to the analysis of coverage dependence of the Fermi level pinning in the formulation presented in a recent review on semiconductor surfaces [12]. In this case, a single unit (a Ga adatom) covers three sites; therefore, . The charge of the unit is calculated assuming that an N broken bond has the average charge thus [12]. A single Ga adatom contributes three electrons, i.e., . The number of states occupied after adsorption is six, including the two and four – and – states, i.e., . Because the Fermi level shifts upward during adsorption, all three nitrogen broken bonds are fully occupied, i.e., the number of electrons for uncovered unit is . The charge balance yields the critical fraction and the coverage [12] as:
Substitution of the above values gives the expected critical fraction so that the critical coverage is .
Alternatively, the transferred electrons can be counted directly. The adsorbed Ga atom contributes three electrons which combine with 15/4 electrons from three N-BBs, creating 3/4 electron surplus which is equal to the number of electrons required to fill the unsaturated N-BB state. Thus, the ratio of attached Ga adatoms to unsaturated N-BBs is 1:1. As each Ga covers 3 N sites, a single Ga adatom is allocated to four N sites, i.e., the critical coverage is .
the left-hand side), PDOS of the top atoms and adatom, and the rightmost panel: COHP of the adatom and the three nearest topmost atoms. The topmost atom located close to adatom is labeled N113, the topmost atom located away (without coverage) is labeled N81.
At high coverage, Ga adatoms are located in the on-top positions above the topmost N atoms, as reported in Ref 59 [59]. As shown in Fig 26, the Ga–Ga interaction plays a secondary role so that the ordering of the Ga atoms is nearly perfect.
The electronic properties of the slab representing surface at full coverage are presented in Figure 27. As shown, the gap is closed by the states arising from the overlap of states of neighboring adatoms: bonding and antibonding states in the lower and upper part, respectively. The bonding states are located deep in the valence band. The hybridized and states are located at the VBM due to the antibonding and bonding overlap, respectively. Of the total electrons ( from Ga and from N-BB state) four electrons are located in the above states. The remaining electron is located in bonding state. The upper part of the gap is occupied by states that are empty. The majority of states are empty; thus, the Ga–Ga interaction is weak and does not affect the distribution of Ga adatoms.
Therefore, the fully Ga covered surface behaves as a collection of individual atoms, and the interaction of the coverage with the additional N adatom is effectively limited to the three nearest-neighbor atoms only. These adatoms may create occupied states with additional N atoms located above the Ga adlayer.
Adsorption of gallium on P-terminated surface
A similar process was considered for the polar P-terminated surface. A single gallium atom was attached to the surface at the energetically most favorable H3 site, saturating three P-broken bond (P-BB) states. The slab representing surface with single adatom is presented in Figure 28.
Figure 28. 9-DAL Ga-P slab used for simulations of gallium adatoms attached at clean (stoichiometric) P-terminated surface at: (a) low coverage (), (b) full coverage (). Top row - side view, bottom row - top view. Blue, red and small sky-blue balls represent gallium, phosphorus, and hydrogen pseudoatoms, respectively. The magenta balls represent Ga adatoms, The green balls represent selected Ga adatoms: Ga8 (highly skewed position), Ga10 (no attachment to P), and Ga12 (on-top).
In general, Ga adsorption at surface of zinc-blende system is crystallographically identical to adsorption on the wurtzite (0001) surface. The electronic properties of the Ga-P slab representing the surface at low coverage are shown in Fig. 29.
The electronic properties of this slab are presented in Fig. 29, showing similarity to the results presented earlier in Figure 27(b). The phosphorus broken bond state (P-BB) energy is in the valence band, close to the VBM, fully occupied. The Fermi level is pinned at the VBM. The states related to the bonding overlap of and and states and are within the valence band, i.e., they are fully occupied. The states of antibonding overlap of states and are located near the middle of the gap. Thus, the analogy with the earlier analyzed surface is complete, i.e., the critical coverage is surface is complete, i.e., the critical coverage is . The Fermi level is pinned, located close to the VBM for both low and high coverages.

Figure 88 (PDF p. 55).
Fig. 29. Energies of the quantum states of the slab representing stoichiometric P-terminated surface with: (a) single Ga (Ga1) adatom located at an H3 site (), (b) surface with 16 Ga (Ga1) adatoms (). The panel description is as in Figure 27. The topmost P atom located close to Ga adatom is labeled P191, the topmost P atom located farther away (without overlap) is labeled P155. The topmost P atoms located close to Ga adatoms are labeled P215 and P260.
A slightly different atomic arrangement is obtained for the Ga atoms attached to surface. As shown in Figure 28, gallium adatoms interact strongly, creating a network of connected Ga adatoms. Part of the surface remains uncovered. This arrangement results from the interaction between Ga adatoms and the topmost P atoms, as well as among the Ga adatoms. The band diagrams illustrating the quantum states of Ga-P slab with 16 Ga adatoms are presented in Fig. 29. As shown, the gap is closed by states. Again, the number of electrons is electrons (3 from Ga and from P-BB state). As shown, the attraction is partially due to the binding overlap of states and states, accepting 2 electrons, located deep in the VB. The other two electrons are located on the hybridized state of and states having the antibonding and bonding overlap with the states, located at the VBM. The remaining electron occupies bonding states. This fractional occupation leads to disorder in the adsorbed Ga layer, as shown in Figure 28.
In summary, the adsorbed Ga layer has fractional occupation of the bonds and is therefore highly active and readily forms bonds with additional P adatoms contributing three electrons to the bonding. The Fermi level is located at the VBM for all coverage values.
Adsorption of gallium on As-terminated surface
Similar results were obtained for the case of Ga coverage of surface, represented by Ga-As slab shown in Figure 30. The surface is a structural analogue of the previously studied Ga-P slab.
Figure 30. (4 × 4) 9-DAL -As slab used for simulations of gallium adatoms attached at clean (stoichiometric) As-terminated surface at: low (), (b) full () coverage. Top row - side view, bottom row - top view. Blue, lavender and small sky-blue balls represent gallium, arsenic, and hydrogen pseudoatoms, respectively. The magenta balls represent adatoms, The green balls represent selected adatoms: Ga8 (highly skewed position), Ga10 (no attachment to As), and Ga12 (on-top).
As expected, the electronic properties of this slab, presented in Figure 31, are to some extent similar to the previous case, i.e., to Ga-P slab. Some differences are also observed; in fact, broken bond state is fully occupied. The remaining states are fully compatible with the previous analysis, i.e., the Fermi level is pinned up to $\theta_{\mathrm{As-cr}} = 1/4 coverage. For higher coverage, i.e., for , the Fermi level is pinned at state, which is located at the CBM, i.e., higher than in the case of surface.

Figure 31. Energies of the quantum states of the slab representing stoichiometric As-terminated surface with: (a) single Ga (Ga1) adatom located at an H3 site
The arrangement of Ga adatoms attached to surface is different. All Ga atoms are highly ordered into rows, positioned in skewed on-top positions. This is different from both cases discussed earlier.
(), (b) with 16 Ga (Ga1) adatoms (). The panel description is as in Figure 27. The topmost As atom located close to Ga adatom is labeled As191, the topmost As atom located away (without overlap) is labeled As155. The topmost As atoms located close to Ga adatoms are labeled As215 and As251.
This arrangement is related to the electronic properties of the system. As shown in Figure 31, the bonding is due to the overlap of and orbitals with orbital. The bonding state is located deep in the valence band. The second state, located below the VBM, is composed of and orbitals with orbital. The first contribution stems from antibonding, while the second is due to bonding overlap. The states are degenerate with the conduction band, so the band gap is partially empty. Some of the horizontal Ga states are at midgap and are therefore occupied; thus, the Fermi level is located at the CBM. The interaction between Ga adatoms and As topmost atoms is mixed: both bonding and antibonding states are occupied.
These data show that the system has an open gap and the Fermi level close to CBM. Thus, the system is different from the previous cases. Gallium states are located above midgap or within the conduction band (CB). Therefore Fermi level is located at the VBM for low coverage and moves rapidly upward to the CBM for high coverage.
Adsorption of gallium on Sb-terminated surface
Finally, the adsorption of a single Ga atom at stoichiometric surface was simulated. As shown in Figure 32, single Ga adatom is located at an H3 site, in complete analogy with the surfaces studied above. At high coverage, Ga adatoms are moved to skewed on-top positions, affected by strong interactions between Ga adatoms. The arrangement of Ga adatoms at surface is disordered, but in a manner different from the previous cases. Approximately half of them are located at H3 sites, including the atom labeled Ga6. The other gallium atoms form tightly bound chains, clustered in one region of the surface. Some of them are attached to topmost Sb atoms, but the others are not. This suggests a stronger interaction between Ga adatoms that dominates their arrangement.
Figure 32. 9-DAL-thick Ga-Sb slab used for simulations of gallium adatoms attached at clean (stoichiometric) Sb-terminated GaSb surface at: (a) low , (b) full coverage . Top row: side view, bottom row - top view. Blue, orange and small sky-blue balls represent gallium, antimony, and hydrogen pseudoatoms, respectively. The magenta balls represent Ga adatoms, The green balls represent selected Ga adatoms: Ga6 (H3 position), Ga10 (no attachment to Sb), and Ga12 (on-top).
The electronic properties of this slab system are presented in Figure 33. Despite the much smaller band gap, the basic features are the same as in two previous cases: for low coverage the Fermi level is pinned at the VBM, the states occupied after adsorption include two and four and states.

Figure 33. Energies of the quantum states of the slab representing stoichiometric Sb-terminated surface with: (a) single Ga (Ga1) adatom located at an H3 site
(), surface with 16 Ga (Ga1) adatoms (). The panel description is as in Figure 27. The topmost Sb atom located close to Ga adatom is labeled Sb191, the topmost Sb atom located away (without overlap) is labeled Sb155. The topmost Sb atoms located close to Ga adatoms are labeled Sb206 and Sb223.
Accordingly, for low coverage the Fermi level is located at the VBM. The state, state has antibonding overlap with the closest topmost Sb atoms so it is located at the CBM and therefore not occupied. In summary, Ga adatoms on nonmetal-terminated surfaces exhibit essentially identical coordination at H3 sites and electronic properties; consequently, the Fermi level is pinned at the VBM at low coverage.
At higher coverage, the Fermi level is shifted to state located much higher. The difference in the arrangement of Ga adatoms is reflected in the electronic properties of the system shown in Figure 33. The Ga adatoms at H3 sites have six occupied states. The electrons originate from the Ga adatom (3) and from three Sb-BB states (15/4), thereby creating a surplus of (3/4) electrons. Thus, the Fermi level is located at the CBM. The states are in the conduction band and are therefore empty.
The surplus electrons are transferred from Ga adatoms located at H3 sites to the other adatoms arranged in chains. The atoms in the chains are bonded to topmost Sb atoms via bonding states located deep in the VB and hybridized states creating combined bonding and antibonding overlaps at the VBM. This small surplus, together with the electrons from Ga adatoms at H3 sites, occupies all bonding states so that the Fermi level is shifted to the CBM. Thus, these Ga adatoms are less likely to bond with additional Sb adatoms.
The results of the above investigations are summarized in Table 4. They prove that the structural and electronic properties of NM-terminated surfaces at low and high Ga coverage are drastically different. At low Ga coverage, the structural and electronic properties of all surfaces are identical: Ga adatoms are located at an H3 sites with the Fermi level located at the VBM. At high Ga coverage, the system shows some variation: the adatoms are located on top, but the adsorbate may be clustered. The stronger clustering is observed in case of GaSb which leads to island creation with the remaining Ga adatoms scattered, located at an H3 sites. The electronic properties are more uniform: the gap is predominantly closed, with the Fermi level high: in the case of GaN it is at midgap, for all other materials, it is at the CBM.
| Surface | Low coverage () | High coverage () | ||
| Structure | Fermi level | Structure | Fermi level | |
| VBM | – order | Midgap | ||
| VBM | –disorder/cluster | VBM | ||
| VBM | - order | CBM | ||
| VBM | & mixed | CBM |
Table 4. Structural and electronic properties of nonmetal (NM) terminated polar surfaces under single-layer Ga coverage
The structural and electronic properties of Ga-covered NM terminated surfaces have identical structural properties while the electronic properties are different as shown in Table 4. This is especially visible for high coverage, important for the growth of these semiconductors. Thus Fermi level position is potentially important for the growth at NM terminated surfaces.
Adsorption energies and surface stability
The adsorption energy is calculated using the standard method, i.e., as the total energy difference between the slab with the adsorbate S attached and the sum of the total energies of the slab and the adsorbate S separated:
For an incremental increase in coverage of the slab, the data correspond to the number of attached adatoms divided by the number of sites (16). The first case is a single Ga atom, i.e., . The adsorption energy of gallium atoms as a function of Ga coverage on nonmetal-terminated surface is presented in Figure 34.
Figure 34. Adsorption energy of Ga at nonmetal-terminated surfaces of III-V semiconductors. The plotted coverage corresponds to the state of the surface after adsorption, i.e., it is the number of Ga adatoms divided by 16 (number of sites) after the process.
As shown in Figure 34, the dependence is different for various surfaces. For , this follows strictly the predictions based on EECR analysis. For low coverage () Ga adatoms are located at H3 sites, as shown in Fig. 26. The electrons are transferred to the states of Ga adatoms, located deep in the VB, providing a considerable energy gain so that the adsorption energy is high, close to 6.4 eV. At higher coverage, this contribution vanishes, reducing the energy gain by about 3 eV. At higher coverage, the Ga adatoms, located in the on-top positions, interact weakly as shown in Fig. 26, where the Ga adatoms occupy nearly perfect positions. This provides a small increase in the adsorption energy gain of about 1 eV.
Adsorption of gallium at surface at low coverage is similar. Gallium is located at H3 sites, saturating the phosphorus broken bonds located in the VB. The Fermi level is pinned by broken bond states located at the VBM. Thus state, in the midgap, is empty. The energy gain is small because the energy difference is small. Therefore, the difference between low coverage () and higher coverage is smaller by approximately 1 eV. Ga adsorption at high coverage () places Ga adatoms in the skewed on-top positions. As shown in Figure 28, the Ga–Ga states are not occupied because they are in the gap and the Fermi level is shifted to the VBM. Therefore, the bonding and antibonding overlaps are created between and states which cancels the energy gain. As a result, the energy gain is reduced for high coverage ().
The adsorption of gallium adatoms on As-terminated GaAs and Sb-terminated GaSb surfaces is similar to each other but differ from the two previous cases. At low Ga coverage, the Fermi level is pinned at the broken-bond states of the surface atoms, located close to the VBM. In fact, the bonding, due to overlap with the states is weak, so that the electron transfer energy gain is small. Therefore, the effect is negligible. At higher coverage, the dependence is different: for GaAs the Fermi level shifts upward, which is related to occupation of antibonding states as shown in Figure 30. The adsorption energy is reduced for higher coverages. For the GaSb surface, the Fermi energy is again shifted to the CBM. Since the occupied states have bonding overlap, the adsorption energy increases slightly for high coverage.
In summary, adsorption of gallium at low coverage provides a considerable energy gain, of the order of 4 eV. Therefore, the Ga adatoms are very stable at H3 sites and their removal is practically impossible, even at high temperatures. The thermal contributions to the free energy are on the order of 1 eV for GaP, GaAs, and GaSb. For GaN, exceptionally, it is larger, but not higher than 2 eV, while the Ga adsorption energy at surface is higher than 6 eV. Therefore, in all cases, the thermal effects cannot change the stability of such a Ga adlayer. Thus, it may be assumed that under typical growth conditions, these adatoms are always present, provided that they are present in the vapor. The only way to cause their absence is to cleave the crystals under ultrahigh vacuum so that these adatoms are absent from the outset [39, 40, 41, 42].
To verify this assumption, the equilibrium pressure of gallium above the surface can be obtained by assuming equilibrium between the vapor and the adsorbate, i.e., with respect to evaporation
which entails equality of the chemical potentials of gallium in the vapor and on the surface:
Assuming the ideal-gas equation of state and the standard dependence of the chemical potential of the adsorbate, we obtain the standard equation for adsorption of the atomic species on the surface:
where the vapor part of the free energy of gallium in the vapor is given by the free energy at normal pressure () , and the vibrational contribution to the free energy of the adsorbate . The adsorption energy is obtained from ab initio calculations via Eq. (13). The free energy of the gallium vapor is obtained from the thermodynamic properties of the elements [64, 70, 71, 72, 73] and is plotted in Figure 35.
Figure 35. Threshold pressure of gallium for the vapor growth on nonmetal-terminated surfaces.
whereas the vibrational free energy of the Ga adsorbate was obtained by scaling the free energy of P adsorbate using the bond strength and mass
and is used to obtain the threshold gallium-pressure dependence presented in Figure 35.
In general, these pressures are very low, indicating that during growth the surface is covered by Ga adatoms located in on-top positions. Specifically, the MBE growth temperatures are:
These pressures are extremely low, below technical limits; thus, for simulations of growth the only valid assumption is that at high Ga coverage, the adatoms are located in the on-top positions.
Finally, the important problem of applying ideally terminated semiconductor surfaces to simulations of vapor crystal growth must be analyzed. It is well recognized that semiconductor surfaces undergo reconstruction to decrease surface energy associated with broken bonds [41]. The reconstructions were widely investigated, providing a large body of data summarized in the review paper by Srivastava [42]. These results are numerous, they include most discussed and famous (7 × 7) reconstruction of Si(111) surface [78]. The results more relevant to the present study are those related GaP, GaAs and GaSb.
In the case of gallium phosphide, a rich variety of reconstructions is observed. The P-terminated GaP(111)B surface shows two different reconstruction units [79]. Recently, an extensive investigation of the surface reconstructions of the GaP(111)B surface established a phase diagram comprising , phases as a function of flow and the temperature [80]. The opposite Ga-terminated GaP(111)A surface has not been extensively investigated.
In the case of GaAs, the surface chemical state is controlled by Ga or As evaporation, which leads to surface reconstruction associated with Ga vacancies or adsorbed As trimers at GaAs A and B surfaces, respectively [81]. Depending on the As pressure, a number of reconstructions are possible on the GaAsB surface, such as , and symmetries [81, 82, 83].
Finally, the gallium antimonide metal-terminated surfaces exhibit the most complex reconstruction patterns because of the relatively narrow band gap and weak bonding, which entail relatively high mobility of surface atoms. Owing to the large atomic radius of Sb, the creation of Sb dimers, trimers, and tetramers is favored. Accordingly, this mechanism is responsible for the Ga-terminated surface or , and reconstructions [84]. Similarly, the Sb-terminated surface undergoes reconstructions [84].
In summary, these results could imply that the reconstructions should be used as a starting point for the simulations of adsorption and vapor growth of III-V semiconductors which could create formidable obstacles. In such a case, efficient ab initio simulations of the growth would be extremely difficult. However, detailed examination indicates that the use of reconstructed structures is unnecessary because these configurations are not realized in thermodynamic conditions of vapor crystal growth.
Reconstructions may be divided into three categories according to their origin [41, 42, 43, 44]:
atom escape, i.e., the emergence of surface vacancies,
the attachment of adatoms
the motion of surface atoms leading to a reduction of surface symmetry.
In case (i), a critical factor is the energy barrier that must be overcome for an atom to escape. It is expected that the presence of adatoms may affect this barrier. Specifically, the presence of adatoms at H3 sites may stabilize the surface by saturating the bonds and filling the empty states on the surface. For a nonmetal-terminated surface, these states are close to the VBM, thus the stabilizing effect may be considerable. This was investigated by calculating, ab initio, the difference between the energies of the nonmetal-terminated slabs optimized in Paper I and the slab with a missing atom plus the separated molecule (atom):
where is the total energy of the slab, the total energy of the slab with the single atom missing, and or is the total DFT energy of the separate atom or the molecule, respectively. Naturally, for N, P, and As, half the total energy of the molecule was subtracted (Eq. 16b), whereas for Sb, the energy of a single atom was used (Eq. 16a).
These calculations generally show that the detachment of a single nonmetal atom is an energetically costly process as shown in Table 5. Nevertheless, this is an atomic process. To assess surface stability, the considerable molecular dissociation energy must be subtracted:
| Energy (eV) | Clean surface | Covered by Ga at H3 sites | ||
| () | ||||
| Surface | Atomic | Molecular | Atomic | Molecular |
| GaN | -3.33 | 1.60 | -5.39 | -0.43 |
| GaP | -1.78 | 0.80 | -4.81 | -2.23 |
| GaAs | -1.56 | 2.04 | -3.37 | -1.30 |
| GaSb | -1.56 | - | -3.98 |
Table 5. Energy gain during detachment of group-V species into atomic (Eq. 16a) and molecular (Eq. 16b) state in the vapor from the clean and Ga-covered surfaces.
For antimony, the atomic state is the most stable.
Positive values indicate a lower energy for the separated system, i.e., for a system that is unstable against separation, i.e., the creation of surface vacancies. Because of the large molecular energies , and . The energy cost is positive for molecules. These values show that the clean surfaces are unstable against molecular escape. Paradoxically, the most stable surface is the Sb-terminated GaSb surface because the energy gain from molecule formation is absent. Coverage of the surface by a Ga adatom at an H3 site has a substantial stabilizing effect. The molecular detachment is unfavorable for all surfaces, i.e., these surfaces are stable.
Thus, adsorption of Ga adatoms, located at H3 sites, completely changes the outcome. As shown in Figure 36, these adatoms are very stable, and remain fixed at H3 sites. On the other hand, their triple bonding and associated electron transfer increase their detachment energy considerably. Therefore, these structures are stabilized with a low probability of spontaneous creation of surface vacancies at the initial stage of growth. From the perspective of crystal-growth simulations, the above results exclude the application of reconstructed surface structures generated by atom loss, i.e., by category (i). The only possible exception is for GaN. The other three cases are fully tractable in simulations starting from unreconstructed surface structures.
Figure 36. Nonmetal-atom detachment energy , from nonmetal-terminated surface, calculated using Eq. 16a. The calculations use slabs, both clean and covered by 4 Ga adatoms at H3 sites ().
The other two mechanisms of surface reconstruction are fully tractable because symmetry changes can be simulated explicitly. This is the approach that will be used in the following. The proposed approach is devoted exclusively to developing the vapor growth model. It is therefore assumed that the surfaces are covered by both species which, as will be shown, stabilizes the surface against evaporation. Therefore, the thermodynamic properties of the system leading to the loss (evaporation) of matter will not be considered because these are typical of a high-vacuum environment [41]. This assumption remains in accordance with the earlier surface stability investigations showing the minimum-energy configuration at an H3 site for single adatom [11, 12, 49, 51, 52, 62, 63, 69] [59]. The above results supplemented these data with information on the stability of topmost layer atoms against evaporation.
Adsorption of gallium on top of an adlayer of nonmetal atoms on a Ga-terminated surfaces (A surfaces).
In the simulations reported in this Section, we assume that the first adsorbate layer is composed of nonmetal adatoms located in strongly skewed on-top sites due to strong interactions with other nonmetal adatoms. This excludes GaN because, at higher coverage, nitrogen preserves a diatomic molecular structure that is not attached to a Ga-terminated surface. Therefore, the three cases considered are GaP, GaAs and GaSb. It has to be repeated again that effective MOVPE/HVPE growth of GaN is possible due to presence on hydrogen stabilizing high nitrogen radicals/admolecules at high coverage creating the route to effective crystallization of GaN using ammonia source of active nitrogen [61, 62, 12]. These processes are outside of the scope of the present paper limited to growth of III-V compounds from their components, i.e. without hydrogen. Similarly, the growth processes including presence of hydrogen used for transport of phosphorous or arsenic are not considered in this paper.
In calculations of the attachment of the second layer, an important issue is the structure of the first layer. As shown in the case of phosphorus, the P vapor pressure corresponding to the transition to the on-top regime is very low. Accordingly in the MBE growth of GaP the phosphorous pressure is below [65, 66]. Higher pressures are associated with MOVPE and HVPE growth temperatures [67, 68]. These processes involve hydrogen in the system, which is not considered here. For other systems, i.e. GaAs and GaSb these pressures are even lower, due to lower growth temperatures [85, 86, 87, 88, 89]. Thus, it may be assumed that growth occurs at high coverage of nonmetal adatoms, located in the skewed on-top positions. This assumption will be used in the study below, which focuses on the attachment of the second Ga layer. Since the present series is limited to vapor-phase growth, these cases of double liquid Ga layers will not be discussed.
The most technically demanding task is the construction of atomically precise structures, necessary for electronic and optoelectronic devices. At present, this task is accomplished using vapor-phase techniques. In the case of gallium phosphide, the subject with special emphasis on nonlinear optoelectronic applications, was reviewed recently by Wang et al. [90]. The growth techniques applied include Molecular Beam Epitaxy (MBE) [65, 66, 80, 91]. Metal-Organic Vapor Phase Epitaxy (MOVPE) [92, 93], and Hydride Vapor Phase Epitaxy (HVPE) [94]. Similarly, the growth of a large number of GaAs optoelectronic devices is based on vapor growth techniques, both MBE and MOVPE [85, 86, 87]. GaSb structures for optoelectronics are also grown by MBE [88, 89]. Therefore, the adsorption of the second Ga layer on top of NM adlayers attached to Ga-terminated surfaces of GaP, GaAs and GaSb is discussed below.
Adsorption of gallium above a phosphorus adlayer attached to a Ga-terminated GaP(111) surface
Attachment of a single Ga atom at full P coverage () on the Ga-terminated surface was simulated using slab. The slab was nine Ga-P double atomic layers thick with the opposite surface terminated by fractionally charged hydrogen () pseudoatoms. The ab initio obtained structure is presented in Fig. 37.
These diagrams show that the Ga adatom does not attract the P nonmetal adatoms strongly enough to form a pyramid. In fact, the P adatoms appear to be more strongly bound to other P adatoms. Thus, the Ga adatom is attached to a pair of P adatoms, labeled P15 and P8. The distances between the Ga adatom and these two atoms are: Å and Å. These two P adatoms are bonded in a P-P chain. The third-nearest P9 atom, located in the second chain, is located at a distance Å.

(a)

(b)
Figure 37.
Figure 37 (PDF p. 73). (4 × 4) 9 DALs Ga-P slab obtained from simulations of a single gallium adatom above a layer of 16 phosphorus adatoms () attached to the GaP (111) surface: (a) top view; (b) side view. Blue, red and small sky-blue balls represent gallium, phosphorus and hydrogen pseudoatoms, respectively. The magenta balls represent the phosphorus adatoms, while the orange and yellow balls represent selected P adatoms labeled P8, P15 and P9, respectively. The green ball represents the adsorbed gallium adatom.
The electronic properties of this system, presented in Figure 38, are related to the properties investigated previously. Accordingly, the electronic properties of GaP(111) surface, fully covered by P adatoms, are dominated by disorder arising from the incompatibility of Ga-P and P-P interactions. The Ga-P overlap may be reduced by displacement from an on-top site due to P-P interaction. Therefore, some Ga-P states are empty, whereas others are occupied. P-P overlap may reduce the energies of these states by about 3 eV, and these states control the position of the Fermi level.
Figure 38. Energies of the quantum states of the (4 × 4) slab representing GaP(111) surface with 16 P adatoms attached and an additional single Ga adatom above them shown in Fig. 37. The panels description is as in Figure 27.
The adsorption energy of gallium in the second layer is , i.e. the process provides a considerable energy gain. As shown in Figure 38, the Ga adatom forms bonds by overlap with P15 and P8 adatoms. Bonding with the more distant P9 atom is much smaller and is overcompensated by antibonding overlap. Thus, effective bonding occurs between Ga and a pair of P adatoms leading to the aforementioned energy gain. In addition, electron transfer from states to phosphorus adatom states contributes to the energy gain which is considerable. This configuration cannot be treated as a segment of crystalline GaP lattice, therefore, growth on this orientation is difficult if possible at all. Since this is the most favorable atomic configuration owing to the abundance of P adatoms, this conclusion may be extended to other configurations, leading to the overall conclusion that vapor growth simulations on this surface should be abandoned.
Adsorption of gallium above a arsenic adlayer attached to a Ga-terminated surface
Similar simulations were conducted for the composite coverage of surface that was represented by slab of nine Ga-As double atomic layers. The surface was fully covered by As adatoms with the opposite surface terminated by fractionally charged hydrogen pseudatoms. The obtained configuration of the system is shown in Fig. 39.
(a)
(b)
Figure 39.
Figure 39 (PDF p. 75). (4 × 4) 9 DALs Ga-As slab representing (111) surface with 16 attached As adatoms and a single Ga atom above them: (a) top view; (b) side view. Blue, lavender and small sky-blue balls represent gallium, arsenic and hydrogen pseudoatoms, respectively. The As adatoms are represented by red balls while selected As9, As14 and As7 adatoms are represented by orange and yellow balls, respectively. The Ga adatom is represented by a green ball.
Again, the results show that the Ga adatom interacts differently with the nonmetal As adatoms. In fact, the Ga adatom seems to be more strongly attached to a pair of As adatoms (As9 and As14). The distances between the Ga adatom and these two As adatoms are: Å and Å. These two As adatoms are bonded within a single As-As chain and are thus closely positioned. The third-nearest As7 atom, located in the second chain, is located at a larger distance Å.
The electronic properties presented in Figure 40 differ from those of the previously considered GaP surface. In fact, at high As coverage, the Fermi level is strongly pinned at the VBM. The details differ for high coverage, but the Ga-As states have a strong overlap so that these states are located at the VBM, i.e. partially occupied. The in-plane states create overlaps with the states of other As atoms, located deep in the VB; thus, they dominate the energy landscape.
Figure 40. Energies of the quantum states of the (4 × 4) slab representing (111) surface with 16 As adatoms attached and an additional single Ga adatom above them. The panels description is as in Figure 27.
Attachment of a Ga atom does not affect the As adlayer considerably, which is related to strong interactions between As adatoms. Thus, the Ga adatom is attached to a pair of As adatoms, without forming the growth pyramid. This is related to the compensation of bonding and antibonding with the nearest As7 atom. The adsorption energy of gallium is . This represents a considerable energy gain, which is related to the formation of a pair of bonds and also to the electron transfer from states to states. Despite the much smaller band gap of GaAs, the electron transfer contribution is considerable. As before, this configuration will not be considered because it is not likely to be suitable for standard vapor crystal growth simulations.
Adsorption of gallium above antimony adlayer attached to a Ga-terminated surface
Similar simulations were conducted for the composite coverage of surface that was represented by slab of nine Ga-Sb double atomic layers. The surface was fully covered by Sb adatoms with the opposite surface terminated by fractionally charged hydrogen pseudoatoms. The obtained configuration of the system is shown in Fig. 41.
Figure 41.
Figure 41 (PDF p. 77). 9 DAL Ga-Sb slab representing (111) surface with 16 attached adatoms and a single Ga adatom: (a) top view; (b) side view. Blue, orange and sky-blue balls represent gallium, antimony and hydrogen pseudoatoms, respectively. The magenta balls represent the adatoms, while the red and yellow balls represent selected , and adatoms. The green ball represents Ga adatom.
As expected, the Ga adatom does not attract the nonmetal Sb adatoms strongly. In fact, they appear to be more strongly bound to other Sb adatoms. Therefore, the properties presented in Figure 42 are related to the full Sb adlayer. These properties follow the trends observed above for other surfaces. The Sb-Sb interaction is stronger than the Ga-Sb interaction. The Ga-Sb state is located at the VBM while Sb-Sb interaction-related states are in the VB. This is related to the overall reduction of the bonding energy observed in GaSb system in accordance with the general trend. Therefore, the Ga adatom is attached to a pair of Sb adatoms (Sb6 and Sb15). The distances between the Ga adatom and these two Sb adatoms are: Å and Å. These two Sb adatoms are bonded in the same chain, and are therefore close together. The third-nearest Sb atom, located in the second chain, is located at a much larger distance Å.
Figure 42. Energies of the quantum states of the slab representing (111) surface with 16 Sb adatoms attached and an additional single Ga adatom above them. The panels description is as in Figure 27.
Bonding to the two nearest Sb adatoms is related to overlap. In addition, an electron is transferred from Ga4p to Sb5p states, contributing to the adsorption energy. As a result, the adsorption energy of gallium is . This is a considerable energy gain stabilizing the Ga adatom. Nevertheless, this configuration is not likely to be suitable for building the crystalline lattice of GaSb therefore it will not be considered below in vapor growth simulations.
Nonmetal-terminated surfaces (B surfaces): adsorption of nonmetal atoms above a Ga adlayer.
In the simulations of the second layer on top of Ga first adlayer, the limited number of configurations will be considered. This is partially related to the fact that we exclude growth from the liquid in our investigations. Thus attachment of the second layer of gallium atoms is excluded. A natural criterion is the role of coverage in the electronic and configurational properties of the surfaces. For a stoichiometric, nonmetal-terminated surface, the Fermi level is pinned at a nonmetal broken-bond state which is located in the vicinity of the valence-band maximum (VBM) [47]. Adsorption of a large number of gallium adatoms closes the gap at the surface almost completely. In addition, the Fermi level is shifted upward so that the fraction of the states in the gap is occupied. Thus, a natural choice is to simulate the systems at high Ga coverage with a small coverage of nonmetal atoms. This is consistent with the estimate of the Ga pressure in equilibrium with the adsorbate. The temperatures corresponding to the growth of GaP are relatively low, below 1200 K [65, 66, 67]. The corresponding Ga pressures are far below the practical requirements for growth, in the range . Therefore, a surface with at least a relatively high fractional Ga coverage is attained, and accordingly this configuration corresponds to the metal-dominated system, i.e. with a small number of nonmetal adatoms.
The second possible state is a nonmetal-dominated surface. This could occur in either of two limits: at very low Ga coverage where the system is nonmetal-dominated with only a small contribution from Ga adatoms, i.e. dominated essentially by the topmost nonmetal atoms or at limited Ga coverage balanced by additional nonmetal atoms on top. This is possibly way to obtain high antisite defect density, which is not typical growth. The second case is investigated below.
Adsorption of nitrogen above gallium adlayer attached to a N-terminated surface (B surface)
The first process considered was the adsorption of nitrogen on top of a Ga adlayer attached to N-terminated surface (or, equivalently, surface). As shown in Fig. 26, the full Ga adlayer is a highly ordered structure with the gap closed. As shown in Figure 27, the gap is closed by and states located in the lower part of the gap and by in the upper part. The former are occupied, whereas the latter are empty.
The attachment of a single N adatom forms bonds to the three nearest Ga adatoms forming a partial tetrahedral structure as presented in Fig. 43. The simulated systems include two configurations: (i) fractional coverage by Ga adatoms () and coverage of N adatoms (), bond-compatible (), (ii) full Ga coverage () and small N coverage (). In the () slab, these coverages were translated into: (i) 9 Ga and 3 N adatoms, (ii) 16 Ga and 1 N adatom.
(a)
(b)
Figure 43.
Figure 43 (PDF p. 80). (4 × 4) 8 double atomic layers (DALs) Ga-N slab obtained from simulations of the attachment of gallium and nitrogen adatoms at N-terminated GaN surface with: (a) 9 Ga and 3 N adatoms, (b) 16 Ga and 1 N adatom: upper row, side view; lower row, top view. Blue, yellow and small sky-blue balls represent gallium, nitrogen and hydrogen pseudoatoms, respectively. The magenta ball represents N adatom, and the green balls represent selected Ga adatoms (a) Ga4, Ga8 and Ga11, (b) Ga5, Ga6 and Ga11.
In both cases, the tetrahedral N-Ga structure is located in the layer of Ga adatoms without substantially disturbing the layer, thus confirming the earlier assumption that the Ga adatoms behave as a collection of independent species.
The electronic properties of this system are presented in Figure 44. As shown, the N adatom is attached to three Ga adatoms through bonding states and also through − and − bonding and antibonding states. These states are located in the valence band (VB) and are therefore all occupied. Thus, this bonding is extremely strong; during diffusion jumps, these bonds are not likely to be broken. In addition, the broken-bond state is located in the VB close to the valence-band maximum (VBM), and is also fully occupied. The Fermi level is not pinned as the gap is opened by the N coverage.
Figure 44. Energies of the quantum states of the slab representing N-terminated surface with: (a) 9 Ga and 3 N adatoms; (b) 16 Ga and 1 N adatoms. The panels description is as in Figure 27. The attached N atom is labeled N1, the closest Ga adatoms are Ga5, Ga6 and Ga11.
An increase in Ga coverage and a reduction in the number of N adatoms shifts the system toward a metallic state. The gap is closed by Ga states. The Fermi level is not shifted, and only some of these Ga states in the bandgap are occupied. The Ga-N states are located deep in the VB.
These properties of the system indicate that the arrangement in the form of a pyramid is extremely stable and is unlikely to be destroyed by the growth temperature, even at the high temperatures typical for high-nitrogen-pressure solution growth (HNPSG), which is similar in some respects to the processes simulated here [95, 96, 97]. The attachment of additional N adatoms above the adlayer of Ga atoms leads to the formation of tetrahedral complexes, attached to the topmost N atoms. As shown, N adatoms are attached to the three Ga adatoms through and bonding and antibonding states with energies located at the VBM.
The energy gain due to adsorption of a single N adatom is . Using the extremely high dissociation energy of nitrogen molecules equal to [8, 9], this yields a negative value for the molecule . Therefore, adsorption of molecular nitrogen at a Ga adlayer of GaN is energetically unfavorable. This agrees with the results obtained for adsorption on liquid Ga [55, 56]. The adsorption of molecular nitrogen on the Ga(l) surface must overcome a high energy barrier, associated with the dissociation of an molecule [55]. This barrier is high, but considerably smaller than the dissociation energy of molecular nitrogen . In addition, the energy of separated atoms at the surface is about 1.5 eV higher than that of an molecule in the vapor. The desorption of nitrogen atoms in diatomic molecular form is still difficult as it has to combine atoms into pairs and overcome a high energy barrier. In addition, other paths are not attainable because the formation of a tetratomic nitrogen molecule is excluded [32, 33, 34, 35]. Still, growth of GaN on this surface using molecular nitrogen is technically impossible The effective growth requires use of an activated nitrogen source, such as plasma-activated molecular beam epitaxy (PA-MBE) [74, 75].
Adsorption of phosphorous above gallium adlayer attached to a P-terminated surface (B surface)
The next system analyzed involved the adsorption of a Ga adlayer and an additional P layer at P-terminated or, equivalently, surface. Two systems were simulated: (i) fractional coverage by Ga adatoms () and coverage of P adatoms (), bonding compatible (), (ii) full Ga coverage () and small P coverage (). In the slab, these coverages were translated into: (i) 9 Ga and 3 P adatoms, (ii) 16 Ga and 1 P adatom.
Figure 89 (PDF p. 83).
Figure 45. (4 × 4) 9 DALs Ga-P slab obtained from simulations of composite adlayers of gallium and phosphorus adatoms at P-terminated surface with (a) 9 Ga and 3 P adatoms, (b) 16 Ga and 1 P adatom; upper row, side view; lower row, top view. Blue, red and small sky-blue balls represent gallium, phosphorus and hydrogen pseudoatoms, respectively. The magenta ball represents the P adatom, and the green balls represent selected Ga adatoms; (a) Ga7, Ga8, and Ga9, (b) Ga10, Ga11, and Ga12.
The electronic properties of these slabs are shown in Figure 46. These results can be compared with the results for the adsorbed Ga layer at full coverage on surface. The Ga adlayer is disordered with Ga adatoms located in randomly skewed on-top positions. As shown previously, the Ga adatoms are bound mainly by and bonding and antibonding states, with their energies degenerate with the valence bands. The Ga atoms create groups that are more tightly bound by bonding states located in the gap close to the VBM. These states are fractionally occupied; thus, the Ga-Ga clustering is weak. The attachment of additional P adatoms above the adlayer of Ga atoms leads to the formation of tetrahedral complexes, i.e. pyramids attached to the topmost P atoms. Phosphorous adatoms are attached to the three Ga adatoms through and bonding and antibonding states with energies located in the VB, thus with relatively strong bonds. The energy gain due to adsorption of a single P adatom is . Using the energy of dissociation of phosphorus di-molecules equal to [9, 36], this yields an adsorption-energy gain for a P molecule of . Therefore, the pyramid structure at the surface is stable.
Figure 46. Energies of the quantum states of the slab representing P-terminated surface with (a) 9 Ga and 3 P adatoms and (b) 16 Ga and 1 P adatom. The panel descriptions are as in Figure 27. The Ga adatoms located close to the P adatom are labeled Ga10, Ga11 and Ga12.
It is also found that the attachment of an additional P adatom on top of Ga adlayer has a relatively strong influence on the electronic properties of the system. In both cases, the gap is open, with the Fermi level located at the VBM. A second gap occurs near midgap within the surface band of unoccupied horizontal P states. The strong influence is related to the fact that part of the surface horizontal P states are in the VB; therefore, the Fermi energy is always at the VBM. The P-Ga states are shifted into the VB, which opens the gap at the Fermi energy. This arrangement of the states and the Fermi energy is to some degree similar to surface at low coverage. The only difference is that the gap is empty as the gallium atoms are not present. Additional P atoms have a weaker influence also due to the fact that the states within the gap are created due to horizontal overlap of the states whereas bonding to the P adatom is along the vertical direction.
Adsorption of arsenic above gallium adlayer attached to a As-terminated surface (B surface)
The next case considered in this work was double-layer adsorption of Ga (first layer) and As (second layer) at As-terminated or, equivalently, surface. Two systems were simulated: (i) fractional coverage by Ga adatoms () and coverage of As adatoms (), bonding compatible () and (ii) full Ga coverage () and small As coverage (). For the slab, these coverages were translated into: (i) 9 Ga and 3 As adatoms, (ii) 16 Ga and 1 As adatom.
These systems show a strong tendency to create As-Ga pyramids for both full and fractional Ga coverage. This is consistent with the data shown in Figure 30, showing that the adsorbed Ga layer at full coverage is disordered with Ga adatoms located in skewed on-top positions. As shown in Figure 30, these Ga adatoms are bound to the underlying topmost As atoms by bonding and antibonding states. Their energies are within the valence band. These four atoms form typical tetrahedral configurations similar to those in the previous cases, i.e. in the form of pyramids.
Figure 47. 9 DALs Ga-As slab used for simulations of an arsenic adatom attached to a Ga adlayer at As-terminated GaAs surface with (a) 9 Ga and 3 As adatoms, (b) 16 Ga and 1 As adatom; upper row, side view, lower row, top view. Blue, lavender and small sky-blue balls represent gallium, arsenic and hydrogen pseudoatoms, respectively. The magenta ball represents the As adatom, and the green balls represent selected Ga adatoms; (a) Ga7, Ga8 and Ga9, (b) Ga6, Ga11, and Ga15.
The electronic properties of this system are presented in Figure 48. As shown, the three As adatoms are attached to the nine Ga adatoms by bonding and antibonding states located in the VB and and are therefore occupied. Nevertheless, Ga-Ga bonding is much stronger for the GaAs surface than for the GaP surface considered previously; thus, the Fermi level is not pinned; it lies in the band gap above the VBM.
Figure 48.
Fig. 48. Energies of the quantum states of the slab representing As-terminated surface with (a) 9 Ga and 3 As, (b) 16 Ga and 1 As adatom. The panel descriptions are as in Figure 2. The Ga adatoms located close to the As adatom are labeled Ga6, Ga11 and Ga15.
As shown before, the increase of Ga coverage and a reduction in the number of As adatoms shifts the system toward a metallic state. The Fermi energy is pinned by an antibonding state located in the upper part of the gap. The bonding state is located deep in the VB. Therefore, the surface is able to attach other As adatoms. Nevertheless the gap is partly closed, whereas the Fermi level is shifted toward the CBM, and is pinned by horizontally - bonded states in the gap. In summary, the system at higher secondary As coverage to some extent recovers properties of the surface under low coverage.
The energy gain due to adsorption of a single As adatom is . Using the dissociation energy of an As molecule, [37] this yields the molecular value: . Therefore, the pyramid structure at the surface serves as a vehicle for surface diffusion as will be discussed below.
Adsorption of antimony above gallium adlayer attached to a Sb-terminated surface (B surface)
Finally, the adsorption of Sb atoms on top of a layer of Ga adatoms covering surface was analyzed. Two systems were simulated: (i) fractional coverage by Ga adatoms () and coverage of Sb adatoms (), bonding compatible (), (ii) full Ga coverage () and small Sb coverage (). In the slab, these coverages were translated into: (i) 9 Ga and 3 Sb adatoms, (ii) 16 Ga and 1 Sb adatoms.
The obtained configuration, shown in Figure 49, proves that Sb atoms are attached to underlying Ga atoms in different configurations. In the present case, the Sb adatom is attached to four underlying Ga adatoms, and therefore does not form a standard tetrahedral pyramid. As shown in Figure 49, the Ga adatoms at surface are attached differently: half of them are bound in Ga-Ga chains located at skewed on-top sites above the topmost Sb atoms, whereas the other half are located at H3 sites, typical for low coverage. The Sb adatom is attached to four Ga atoms located at on-top sites. The other atoms are locked in chains and the remaining atoms occupy H3 sites.
Figure 49. (4 × 4) 9 DALs Ga-Sb slab obtained from simulations of composite Ga-Sb layers at Sb-terminated GaSb () surface with: (a) 9 Ga and 3 Sb, (b) 16 Ga and 1 Sb shown in the top row as a side view and in the bottom row as a top view. Blue, orange and sky-blue balls
represent gallium, antimony and hydrogen pseudoatoms, respectively. The magenta ball represents the Sb adatom, and the green balls represent selected Ga adatoms: (a) Ga10, Ga11 and Ga12; (b) Ga10, Ga13, Ga14 and Ga15.
The electronic properties of the system are shown in Fig. 50. Sb adatoms are bound differently to Ga14 and Ga15 adatoms, which are positioned at shorter distances than Ga10 and Ga13. This is visible in the states: in the first case, the attachment is due to bonding states and and also to antibonding . The attachment of a small secondary coverage of Sb shifts the Fermi energy down to VBM for both small and large numbers of Sb adatoms.
In the second case, the bonding state is missing. At higher secondary Sb coverage the system has the Fermi level shifted downward toward the vicinity of the VBM and pinned by the state. The second Ga-rich configuration has slightly different properties. The Fermi level is pinned by an antibonding state located at the VBM. Thus, the Fermi energy is not shifted. The bonding between Ga adatoms opens the gap so that the system is semiconducting despite the small band gap of GaSb. Thus, in all cases, a gap remains open, although the Fermi level is pinned at lower Sb coverage and unpinned at higher coverage.
Figure 50.
Figure 50 (PDF p. 92). Energies of the quantum states of the (4 × 4) slab representing Sb-terminated surface with: (a) 9 Ga and 3 Sb, (b) 16 Ga and 1 Sb adatoms. The panel descriptions are as in Figure 27. The Ga adatoms located close to the Sb adatom are labeled Ga6, Ga11 and Ga15.
The energy gain due to adsorption of a single Sb adatom is . This value is relatively high because the vapor state is atomic. Therefore, the created structure may be modified during surface processes; nevertheless, Sb desorption is unlikely. These processes will be investigated below.
The structural and electronic properties of NM-terminated surfaces under combined high Ga coverage, balanced and metal-dominated, summarized in Table 6, are essentially identical: these surfaces are covered by skewed pyramids. The remaining Ga adatoms are in on-top positions. In contrast, the electronic properties are complex: the gap is closed in the metal-dominated case while at balanced coverage it has open windows within part of the gap. The Fermi level depends mainly on the compound, but not on the coverage type. For GaN and GaAs, it is close to midgap. In case of GaP and GaSb is close to VBM. Where possible, the Fermi energy tends to lie within a gap window.
Table 6. Structural and electronic properties of nonmetal (NM) terminated polar surfaces under a double layer: high coverage of the first Ga adlayer and limited coverage of second NM adlayer.
| Surface | Balanced coverage | Metal dominated coverage | ||
| & | & | |||
| First-second adlayer | Structure | Electronic | Structure | Electronic |
| − − | pyramids skewed | Half open gap FE - Midgap | pyramid | Closed gap FE - Midgap |
| − − | pyramids skewed | Some window in gap FE - VBM | pyramid skewed | Closed gap FE - VBM |
| − − | pyramids skewed | Window in gap FE – 0.4 eV above VBM | pyramid skewed | Closed gap FE – above midgap |
| − − | pyramids skewed | Closed gap FE - VBM | pyramid irregular | Window in gap FE – VBM |
Table 8.
Surface diffusion on a nonmetal-terminated surfaces.
Growth of semiconductor structures for electronic and optoelectronic applications requires atomic-precision control of the deposited layers. This is indispensable for the active layers of optoelectronic devices, such as laser diodes (LDs) and light-emitting diodes (LEDs), such as multiquantum wells (MQWs) [98, 99]. Essentially, the same requirements apply to electronic devices (field-effect transistors (FETs) and high-electron-mobility transistors (HEMTs)) based on a two-dimensional electron gas (2DEG) [100, 101]. In the future, the creation of nanometer-scale devices for artificial intelligence (AI) applications will be based on the growth of designed structures, as these structures will be devoid of structural and chemical defects that may degrade their operating parameters [102]. Thus, the atomic-scale control of the growth of semiconductor structures is of great importance.
The creation of atomically flat structures during epitaxy is possible in the step-flow growth mode [103, 104]. The model assumes that growth units are attached on the surface terraces, diffuse toward the steps, and are finally incorporated in the crystal. The essential quantity is the fraction of these units that is incorporated, which depends on the energy barrier for surface diffusion jumps and the energy barrier for desorption. An important parameter is the surface diffusion length , i.e. the average distance of the attached growth unit before detachment, given as [103, 104]
where is the energy barrier for desorption of species S from the surface. For the present investigation, it was assumed that the desorption energy barrier is equal to the adsorption energy calculated using Eq. (2). The jump parameters a and are the jump length and the jump barrier, respectively. For the estimate presented in this work, it was assumed that the diffusion length is the lattice parameter a for both zinc blende and wurtzite lattices. The energy barriers are determined by assuming the basic pyramid configuration of the two-layer adsorbed species.
Energy barrier for N - Ga pyramid jumps at highly Ga-covered nonmetal-terminated surface
The results were obtained by simulating the motion of pyramids composed of Ga adatoms at the base and the nitrogen adatom at the top. We used slabs covered by 15 Ga adatoms in the first adsorbed layer and a single nonmetal adatom in the second adsorbed layer. Gallium-adatom jumps at highly Ga-covered nonmetal-terminated surface occur through the transition of a Ga adatom between two sites located at skewed on-top locations above the topmost N surface atoms. In these simulations, the 8 double atomic Ga-N layer slab with 15 Ga adatoms attached was used. Thus, the jump is a transition to an unoccupied site. Thus, the simulated motion is essentially vacancy-mediated surface diffusion.
Figure 90 (PDF p. 94).
At the initial position (the T4 site of the N adatom) the distances between the N adatom and the three Ga adatoms within the pyramid are: , and . At the bridge position, these distances change to: , and . At the final position (the H3 site) these distances are: , and . Thus, it may be concluded that, during the jump,, the pyramid remains geometrically intact with relatively little deformation of its shape. It should be noted that the plot presents half of the path between the identical sites, whereas the complementary part of the T3-H4 path has an identical energy profile; therefore, it was not simulated.
The diffusion rates depend on the magnitude of the migration barrier. The latter can be obtained from the energy profile calculated for the selected positions along the migration path presented in Figure 52. From these calculations, the energy barrier for the jump was estimated to be . Therefore, the migration barrier is not high, and at typical temperatures the complexes are highly mobile and readily arrive at the steps.
Figure 52. Energy of the system during motion of the pyramid of three Ga adatoms and a single N adatom along the jump path on GaN(000) surface under high Ga coverage represented by 15 attached Ga adatoms. The coordinates , and correspond to Figure 1 (PDF p. 7)a, b and c, respectively.
The energies of the quantum states of the Ga adatoms in the pyramid, in both the initial and activated-complex (bridge) states, depend on the bonding between these Ga adatoms and the topmost N atom of the GaN lattice and the other Ga adatoms. These bonds, their energies, densities of states, and overlaps are shown in Fig. 53.
Figure 53.
Figure 53 (PDF p. 97). Energies of the quantum states of the (4 × 4) slab representing N-terminated surface with 15 Ga adatoms attached. The panel descriptions are as in Figure 27. The upper and lower rows represent the system properties at the NEB path (Figure 52) pyramid coordinate equal to (Figure 1 (PDF p. 7)a) and (Figure 51b), respectively.
Figure 51. 8 double atomic layer (DALs) Ga-N slab used for simulations of diffusion of gallium adatom attached at highly Ga-covered () N-terminated GaN surface in top and side views, shown in the upper and lower rows, respectively: (a) initial position (), (b) bridge position () (c) final position (). Blue, yellow and small sky-blue balls represent gallium, nitrogen and hydrogen pseudoatoms, respectively. The magenta balls represents Ga adatoms. The green balls represent selected Ga adatoms, red small ball represents single N adatom in the second layer.
At this coverage, the bandgap of the system is closed at the surface by Ga adatom quantum states, these of () overlap, whereas states are. The Fermi level is located in the middle of the bulk bandgap. Thus, some of these Ga-adatom states are occupied, whereas the remaining states are empty. As shown, the N1 adatom is bonded to Ga5 adatom within the pyramid by the overlap between and also by states. These states are degenerate with the valence bands (VB) and therefore they are occupied. It has to be noted that the first overlap has composite bonding and antibonding sign while the latter is bonding only. In addition, the N adatom states. These states are degenerate with the valence bands (VB) and therefore they are occupied. It should be noted that the first overlap has composite bonding and antibonding sign while the latter is bonding only. In addition, the N adatom state energy is located below the Fermi level. Therefore, this state is also fully occupied.
The gallium adatoms in the pyramid are bonded through the combined bonding-antibonding overlap with the underlying topmost N atoms, creating states that are degenerate with VB. Thus, they are fully occupied. In addition, Ga adatoms in the pyramid have overlap with the neighboring Ga adatoms. The energies of these states lie within the band gap, and they are partially occupied.
The jump does not change the overlap between the states of N adatom and Ga adatoms in the pyramid, and also their energies. This demonstrates that the pyramid remains intact during the jump, indicating that surface diffusion proceeds through the motion of the entire N-Ga pyramids. The overlap between the shifted Ga adatom and the underlying topmost N atom is fractionally reduced. This results in a relatively low energy barrier for the pyramid motion equal to eV.
It should be added that the pyramid motion is relatively easy, thus allowing this structure to arrive at a step on surface. The attachment of the pyramid structure to the step is not straightforward, therefore it will be investigated in the next Section in mor detail.
Energy barrier for P - Ga pyramid jumps at highly Ga-covered nonmetal-terminated surface
The migration of the pyramid (three Ga adatoms and one P adatom) at highly Ga-covered nonmetal-terminated surface occurs through the transition of a Ga adatom between two sites located at skewed on-top sites above the topmost P surface atoms. In the simulations the 9 double atomic Ga-P layer slab with 15 Ga and a single P adatom was used.
Figure 91.
At the initial position, the distances between the P adatom and three Ga adatoms in the pyramid are: Å, Å and Å. In the intermediate bridge state these distances are: Å, Å and Å. At the final position, these distances are: Å, Å and Å. From Figure 54 and the above data it may be concluded that, during the jump,, the pyramid remains geometrically intact with relatively little deformation of its shape.
Figure 54. 9 DALs Ga-P slab used for simulations of diffusion of the pyramid (three gallium adatoms and a single P adatom) at highly Ga-covered () P-terminated surface in top view: (a) initial position (), (b) bridge position (), (c) final position (). Blue, red and small sky-blue balls represent gallium, phosphorus and hydrogen pseudoatoms, respectively. The magenta ball represents Ga adatom, the green balls represent selected Ga adatoms, the blue ball represents P adatom in the second layer.
The diffusion rates critically depend on the heights of the migration barriers. The latter can be obtained from the energy profile corresponding to several locations along the migration path, presented in Figure 55. The energy barrier for the jump was assessed to be eV. Therefore, the energy barrier for surface diffusion over surface is considerably higher, thereby somewhat limiting the Ga-P pyramid mobility over the surface.
Figure 55. Energy of the system obtained for the movement of Ga-P pyramid along the jump path on GaP surface with 15 Ga adatoms attached. The coordinates , and correspond to Figure 4a, and , respectively.
The energies of the Ga adatoms in both the initial and activated-complex (bridge) states depend on the bonding between the Ga atom and the topmost atom of GaP lattice and the other Ga adatoms. Naturally, they are affected by the deformation presented above. These bonds, their energies, densities of states, and overlaps are shown in Fig. 56.
Figure 56.
Figure 56 (PDF p. 101). Energies of the quantum states of the (4 × 4) slab representing P-terminated GaP surface with 15 Ga adatoms attached. The panel descriptions are as in Figure 27. The upper and lower rows represent the system properties at the NEB path (Figure 54) pyramid coordinate equal to (Figure 54a) and (Figure 54b), respectively.
At this coverage, the gap is partially closed by bonding states located in the gap close to the VBM, thus they are occupied. The higher are empty. The obtained data confirm previous results demonstrating the existence of the bonding between P adatom and Ga adatoms in the pyramid by overlap. In addition, the Ga atoms in the pyramid are bonded to other Ga adatoms due to overlap. A third type of bonding occurs between Ga adatoms in the pyramid and the topmost P atom via overlap.
During the motion, the structure remains intact as confirmed by the overlap within the pyramid. The connection between the pivoting Ga adatom and the underlying P topmost atoms remains unchanged whereas the overlap between the displaced Ga adatom and the atoms underneath is drastically reduced. Thus, as in the previous case, the Ga-P pyramid remains intact throughout the simulated trajectory. The entire pyramidal structure arrives at the step and is finally attached and incorporated into the crystal body. The atomic stages of attachment of such structures at the step will be simulated in the next Section.
Energy barrier for As - Ga pyramid jumps at highly Ga-covered nonmetal-terminated surface
The motion of the pyramidal structure, composed of three Ga adatoms and one As adatom, over highly Ga-covered nonmetal-terminated surface occurs through the transition of a Ga adatom between two skewed on-top sites above the topmost As surface atoms. In these simulations, the 9 double atomic Ga-As layer slab with 15 Ga adatoms was used. As shown in Figure 57, the structure moves by rotation with a single Ga adatom attached at the site.
Figure 57. 9 DALs Ga-As slab used for simulations of diffusion of gallium adatom attached on a highly covered () As-terminated in top view: (a) initial position (), (b) bridge position (), (c) final position (). Blue, lavender and small sky-blue balls represent gallium, arsenic and hydrogen pseudoatoms, respectively. The magenta ball represents As adatoms, the green balls represent selected Ga adatoms. The red ball represents Ga adatom in the second layer.
At the initial position along the path, the distances between the As adatom and three Ga adatoms within the pyramid were: Å, Å and Å. At the bridge position, these distances change to: Å, Å and Å. At the final position, these distances are: Å, Å and Å. Thus, it may be concluded that, during the jump,, the pyramid remains intact with relatively little deformation of its shape.
The diffusion rates critically depend on the height of the migration barrier. This can be obtained from the energy profile along the path presented in Figure 58. The energy barrier for the jump was assessed to be eV, i.e. comparable to the barrier at GaP surface. Therefore, the energy barrier for surface diffusion on the GaAs surface is only slightly lower, similarly limiting the pyramid mobility over the surface.
Figure 58. Energy of the system obtained for the movement of Ga adatom along the jump path on GaAs surface with 15 Ga adatoms attached. The coordinates , and correspond to Figure 57 (a), (b) and (c), respectively.
The energies of the Ga adatoms in both the initial and activated-complex (bridge) states depend on the bonding of the atoms to the topmost As atoms and also to other Ga adatoms. These bonds, their energies, densities of states, and overlaps are shown in Figure 59. The data shown in Figure 59 indicate that the band gap is only partially closed at the surface. The presented data show that the atoms in the pyramid are tightly bound by a covalent interaction between As and Ga adatoms antibonding state, shown by the overlap of these quantum states. The Fermi level is pinned by antibonding states of the pyramid located in the gap of the bulk system. In addition, the interaction between Ga adatoms is strong as demonstrated by the overlap between these atoms' states; both and . These states are also degenerate with the valence band and are occupied.
Figure 59.
Fig. 59. Energies of the quantum states of the slab representing As-terminated surface with 15 Ga adatoms attached and 1 As adatom. The three Ga and single As adatom create the pyramid. The upper and lower rows present the system for the pyramid located in the initial and the maximal energy locations, i.e. in Figure 57 a and b, respectively The panel descriptions are as in Figure 27.
During this shift by rotation (pivot) around the Ga adatom in the pyramid, a single Ga adatom does not move whereas the other two Ga adatoms are displaced. Therefore, the overlap interaction of the first Ga adatom and the underlying As topmost surface atoms remains unchanged. The two shifted atoms have their overlap reduced as shown in Figure 58, which leads to an increase in the system energy, contributing to the emergence of the jump barrier.
Arrival at a step is not equivalent to incorporation at the step as the pyramid configuration is not compatible with the typical step configuration obtained by attachment of the pyramids. Therefore, the process is complicated, requiring detailed simulation of the atomic motion. These processes will be discussed in the following.
Energy barrier for Sb - Ga pyramid jumps at highly Ga-covered nonmetal-terminated surface
The complex motion of the Ga-Sb pyramid at highly Ga-covered nonmetal-terminated surface occurs through the transition of a Ga adatom between two sites located at skewed on-top sites above the topmost Sb surface atoms. In these simulations, the 9 double atomic Ga-Sb layer slab with 15 Ga adatoms was used.
Figure 92.
Figure 60. 9-DAL-thick Ga-Sb slab used for simulations of diffusion of gallium adatom attached on a highly covered () Sb-terminated GaSb surface in top and side views, shown in the upper and lower rows, respectively: (a) initial position (), (b) bridge position (), (c) final position (). Blue, orange and small sky-blue balls represent gallium, antimony and hydrogen pseudoatoms, respectively. The magenta balls represents Ga adatoms, the green balls represent selected Ga adatoms. The red ball represents Sb adatom in the second layer.
At the initial position, the single Sb adatom is located above a pyramid of four Ga adatoms, i.e. differently from the previous cases. Thus, the distances between the Sb adatom and the four Ga adatoms in the pyramid are: Å, Å, Å and Å. At the bridge position, these distances are: Å, Å, Å and Å. At the final position, the four-atom-base pyramid is recovered with these distances: Å, Å, Å and Å. Thus, it may be concluded that, during the jump, the four-atom-base pyramid changes to a three-atom base but recovers its shape in the final position.
The diffusion rates critically depend on the magnitude of the migration barrier. This parameter can be obtained from the energy profile based on the energies obtained for the points distributed along the path. These energies and the approximated profile are presented in Figure 61. Thus, the energy barrier for the depicted jump was assessed to be eV. Therefore, the energy barrier for surface diffusion on the GaSb surface is of the same order as those for GaP and GaAs, limiting the pyramid mobility over the surface only partially. The calculated energy barriers for the jump of the pyramid are listed in Table 7.
Figure 61. Energy of the system obtained for the movement of Ga adatom along the jump path on surface with 15 Ga adatoms attached. The coordinates , and correspond to Fig. 10 a, b and c, respectively.
| Ga-NM | T | (Å) | () | () | |||
| GaN | 0.198 | 5.922 | 0.995 | 923<br>1073 | 4.784 | <br> | 0.477<br>0.237 |
| GaP | 0.642 | 5.569 | 2.994 | 773<br>823 | 8.235 | <br> | <br> |
| GaAs | 0.581 | 4.485 | 2.789 | 823<br>923 | 8.565 | <br> | <br> |
| GaSb | 0.628 | 5.552 | - | 723<br>823 | 9.240 | <br> |
Table 7.
The energies of the Ga adatoms in both the initial and activated-complex (bridge) states depend on the bonding of the atoms to the topmost Sb atom and also to other Ga adatoms. These bonds, their energies, densities of states, and overlaps are shown in Fig. 62.
Figure 62.
Figure 62 (PDF p. 108). Energies of the quantum states of the slab representing Sb-terminated surface with 15 Ga adatoms attached. The panel descriptions are as in Figure 27. The upper and lower rows represent the properties of the system at the NEB path (Figure 61) pyramid coordinate equal to and , respectively.
As shown, the gap is closed due to the small band gap of GaSb and the presence of Ga states. The bonding between the Sb adatom and Ga adatoms is related to the covalent overlap of and states. The first overlap has both bonding and antibonding character while the second is purely bonding. These states are degenerate with the valence band states and are therefore fully occupied. The bonding states between Ga adatoms are located within the bandgap. The energies of these states lie above the Fermi level; therefore, they are empty. Thus, it may be concluded that the Ga-Ga bonding is inactive. The bonding states between Ga adatoms and Sb topmost atoms are degenerate with the valence band states and therefore they are occupied.
During the jump, the structure consisting of Sb and Ga adatoms remains intact, as demonstrated by the overlap integrals of and states. In addition, the overlap between the pivoting Ga adatom and the underlying Sb topmost atom remains unaffected. The overlap between neighboring Ga adatoms is preserved, thus confirming the overall picture of the entire pyramid as the basic mode of surface diffusion.
Energy barrier for desorption
As indicated above, an important parameter is the energy barrier for desorption of the species attached at the surface. From the above investigations it follows that the configuration of particular interest to growth of III-V semiconductors is the NM terminates surface with the double-layer pyramid with a base composed of Ga adatoms and the nonmetal adatom at the apex. The desorption energy barrier was estimated using Eq. (2):
where S denotes a nonmetal adatom and slab denotes a crystal-surface slab with the attached Ga adatoms. Effectively, the configuration denotes the pyramid. The desorption energies obtained using Eq. (19) are listed in Table 7. The final state for nitrogen, phosphorus and arsenic is molecular. In the case of nitrogen the only stable molecular form in the vapor is diatomic [32, 33, 34, 35]. In the case of phosphorus and arsenic, both diatomic and tetratomic forms are present in the vapor [36, 37]. For antimony, the only form is atomic.
From the previously obtained results and also for the date in this paper it follows that the nonmetal atoms in H3 sites on a gallium-terminated surface are extremely stable. This is partially related to the large distance between these adatoms. Similarly, the closest distance between two adatoms in neighboring pyramids is considerable. According to a crystallographic estimate it is equal to: . The obtained surface-diffusion-length estimates are listed in Table 7. This distance is considerable; therefore, the dominant desorption mode of the nonmetal atoms from the pyramids is atomic. The energy difference in Eq. (19) was calculated assuming that the separated form (S) is a single atom. The atomic length is denoted by . In second scenario, the diffusion length was calculated assuming that the atoms create a molecule. The energy barrier is lower by half of the energy of dissociation of the molecule, i.e.
The molecular length is denoted by . The second process data are also listed in Table 7.
Table 7. Surface diffusion parameters: jump barrier (eV), desorption barrier to the atomic state (eV), desorption barrier to molecular state (eV), temperature T (K) minimal and maximal, closest distance between nonmetal adatoms in two pyramids (Å), diffusion length – atomic (), diffusion length - molecular (). The selected temperature was based on the minimum and maximum MBE growth: GaN [74, 75], GaP[65, 66], GaAs [76] and GaSb [77].
The obtained values of the jump and the desorption barriers were used to calculate the surface diffusion length using Eq. 16. Naturally, the length depends on the temperature; therefore, the typical temperatures are listed in Table 7. The obtained surface diffusion lengths are listed in Table 7. Both molecular and atomic diffusion lengths are macroscopically large, indicating that the entire flux of nonmetal atoms arrives at the steps at the surface.
Step kinetics
Vapor growth requires the attachment of the species at the surface supplying material for step motion [103]. In case of III-V semiconductors the step structure is the termination of a double-atom terrace. The terrace consists of two atomic layers: a first (lower) layer of Ga atoms located in the on-top positions in the crystalline lattice, and the second (top) layer of the nonmetal adatoms, forming bonds between these Ga adatoms [104]. The topmost layer of NM atoms are triply bonded to the lower layer Ga atoms in the terrace. The step is the termination of the layer of the topmost NM atoms, that are connected by three bonds to the Ga atoms below. Thus a row of Ga adatoms protrudes from the step above the lower terrace. When a pyramid composed of the triple Ga adatoms in the base and the nonmetal adatom at the top arrives at the step, the structure cannot be attached directly. The Ga adatoms from the pyramid and from the step face each other thus the pyramid cannot be connected to the step. The direct removal of a Ga adatom either from the step or from the pyramid base is difficult, i.e. energetically very costly. Thus this is not viable scenario of the effective step growth kinetics.
A new scheme for step kinetics is therefore proposed which is denoted as glue-atom kinetics. When the Ga atoms, from both the step and the pyramid, are located at neighboring lattice sites the incorporation into the terrace proceeds through the attachment of an additional nonmetal atom (glue atom) forming a connection. Therefore, such step kinetics is considered here the preferential route for the growth of semiconductors at polar surfaces.
It has to be noted that, this attachment requires eight electrons to occupy all quantum states of a single glue atom: six for NM-Ga bonds and two for NM broken bond which is usually located low, in the vicinity of valence band maximum (VBM), i.e. occupied. NM adatom brings five electrons. The other three have to be brought from the surface states, i.e. from the surface states located close to the Fermi level. Therefore it is expected that the process energy characteristics depend on the position of the Fermi level at the surface [12].
Naturally, the number of atomic configurations of the step and the pyramids is large and would require considerable effort to identify and categorize these processes. This field extends far beyond the scope of the research planned in this paper. Here we would like to establish the main features of these atomic processes. Further investigations providing atomic details of the incorporation of both native and foreign defects are a matter for future research. The results could be very valuable because they would provide a unified atomic-level picture of the growth of these semiconductors and also the incorporation of all point defects, both native and foreign. The advantage is that the entire research program described above need not be repeated. Consideration of step kinetics is sufficient. This is the main benefit from the research reported in this paper. Therefore we present only a limited set of simulations of the attachment of the glue atoms from diatomic molecules in the vapor: at and at surfaces.
Ga-N pyramid at the step on nonmetal-terminated surface
The glue atom scheme at a wurtzite surface is presented using a GaN slab with steps and a nearby pyramid, suitably located near the step with an arriving nonmetal atom playing the role of atomic glue, thus, mass incorporation at the step is referred to as glue atom step kinetics. The mechanism of these processes is presented in Figure 63 using surfaces as an example.
Figure 63. Ga-N slab used to illustrate of the glue N atoms attachment at surface: (a) glue atoms in their initial configurations, (b) glue atoms after full attachment (left panel side view, right panel – top view). The two different orientations of the pyramid with respect to the step are possible due to wurtzite lattice symmetry: (i) case 2 – the two N-Ga pyramid bonds are directed towards the step (upper terrace), (ii) the single N-Ga pyramid bond is directed towards the step. Blue and yellow balls represent gallium and nitrogen atoms, respectively. The red ball represents N adatom in the pyramid apex (second layer), and the green balls represent Ga adatoms in the base of the pyramid (first layer). The magenta balls represent N atoms in the molecule that will become glue N atoms, the brown balls represent Ga atoms in the step connected to N glue atoms.
As it is shown, the case 1 and 2 differ by orientation of the pyramid and the step. In case 1, the single NM-Ga bond within the pyramid is pointing towards the step. In the case 2 the pair of the NM-Ga bonds within the pyramid are pointing towards the step. Note that in the wurtzite lattice, the two consecutive pyramid-step structures have different (1 and 2) arrangement, indicating possible difference in step kinetics.
The above assumptions do not specify the origin of the nonmetal atom playing the role of atomic glue. These atoms may originate from the vapor phase (Gilmer mechanism [105, 106]), or through jumps from other surface sites. In particular, direct arrival may occur from the vapor phase thus involving a diatomic molecular process.
Therefore, the first case analyzed is the dissociative adsorption of molecular nitrogen leading to the formation of a pair of glue N adatoms. Diatomic nitrogen molecules are the sole molecular species of nitrogen vapor [32, 33, 34, 35]. Thus the obtained energy profiles for the dissociative pathways of nitrogen diatomic molecules leading to adsorption at the surface are simulated and the results presented in Figure 64. As shown, the obtained energy profiles have some common features and also show substantial differences. The common feature is the presence of a high energy barrier for the decomposition of nitrogen () diatomic molecule, the phenomenon already observed for nitrogen adsorption on group III metal surfaces [55, 56]. The dissociation energy barriers depend on the local environment of the adsorption site and the surface coverage. In the case 2 the barriers for the full glue N atom attachment are and , for low and high coverages, respectively. In the case 1 these barriers are: and for low and high Ga coverage, respectively. Thus these barrier heights indicate that they depend not only on the bonding to the underlying Ga adatoms but also on the position of the Fermi level. Nevertheless, these barriers are drastically lower than the molecular dissociation energy equal to [54].
Figure 64. Energy profiles for the paths of the pair of nitrogen glue atoms that are finally attached at surface, plotted for different Ga coverages of the surface: 12 and 7 Ga adatoms. The case 1 and case 2 correspond to the pyramid orientation of the single and the pair of N-Ga bonds directed towards the step, respectively.
The modification of the diatomic nitrogen molecule is confirmed by the plot of the N-N distance as a function of the path length shown in Figure 65. As shown, for case 2 the distance is increased from Å, typical for nitrogen diatomic molecule in vacuum to the distance Å and Å for the 16 site slab with Ga coverages of 7 and 12 adatoms, respectively. The interatomic N-N distances attain the values close to the lattice parameter of GaN Å [25]. This confirms the dissociation of the diatomic molecules and attachment of the pair of N atoms as glue atoms connecting the pyramid to the GaN layer. As shown, the increase of the N-N interatomic distance occurs at the NEB path point at which the energy is increased to the barrier value. Subsequently, the energy is decreased to attain the energy gain for the pair of N glue atoms equal to eV and eV for Ga coverage of 7 and 12 adatoms. This corresponds to the energy gain of a single N glue atom energy gain of eV and eV. Such a considerable difference in these energies will be analyzed below through the electronic contribution to the energy gain, using the methods presented in Ref. 12 [12].
Figure 65. N-N interatomic distance change along the paths nitrogen atoms, that are finally attached at GaN(000) surface, plotted for different Ga coverages of the surface: 12 and 7 Ga adatoms The case 1 and case 2 correspond to the pyramid orientation of the single and the pair of Ga-N bonds towards the step, respectively.
A drastically different result was obtained for case 1, i.e. the arrangement in which the pyramid is oriented with a single Ga-N bond pointing towards the step. In this case the N-N interatomic distance increases from its value at a large distance Å to Å and Å at Ga coverages of 7 and 12 adatoms, respectively. This increase of merely Å is the result of the repulsive interaction with the nearest atom while keeping the molecule intact. This is supported by an energy loss in the process and at Ga coverages of 7 and 12 adatoms, respectively. Thus these data show that the attachment of an N glue atoms in this configuration is not possible and that, ultimately, the molecule is detached from the surface. Therefore the growth of GaN at this step is unlikely. The analysis of the quantum states presented below confirms this result. Therefore the results are illustrated using the plots of the band profiles, and quantum states for the special points of the dissociation paths.
The first example analyzed in detail is the Case 2 with 7 Ga adatoms attached, i.e. the pyramid orientation of the pair of Ga-N bonds pointing towards the step. The analysis will use the earlier obtained results for attachment of nitrogen at low coverage on GaN(0001) surface, where an N adatom is located at an H3 site. Actually, the arrangement is analogous to the structure of the pyramid. In this case the state is located high in the band gap and the Fermi level is located 0.4 eV below the conduction band minimum (CBM). Nevertheless, as shown above, the increase in Ga coverage from 7 to 12 can shift the Fermi level from the VBM to midgap. As the attachment of an N glue atom brings 5 electrons creating 8 occupied states, the missing electrons are shifted from Ga states close to the Fermi level. Naturally, the energy gain is more substantial for higher Fermi energy, i.e. for 12 Ga adatoms. Since the transfer involves 3 electrons for a single glue atom, the energy gain difference of approximately 0.5 eV is related to a relatively small Fermi-level shift of order of 0.2 eV. The configuration of the slab with 7 Ga adatoms attached is presented in Fig. 66.
Figure 66.
Figure 66 (PDF p. 117). (8 × 4) 8 double atomic layer (DALs) Ga-N slab with a single step used to simulate the attachment of the Ga-N pyramid to the step pointing the pair of Ga-N pyramid bonds towards the step. The N-terminated GaN (000) surface with 7 Ga adatoms attached is shown in side and top views. The N glue atoms are shown at: (a) maximum-energy (barrier) position, (b) final position. Blue, yellow and sky-blue balls represent gallium, nitrogen, and hydrogen pseudoatoms, respectively. The N adatom in the pyramid (N10) and the N glue atoms (N1 and N2) are represented by red and magenta balls, respectively. The Ga adatoms in the pyramid: Ga 35 (single), Ga23 and Ga38 (connected to glue atom) are represented by green balls. The step Ga atoms connected to glue atoms, Ga27, Ga26 and Ga25 are represented by orange balls.
The energies of quantum states associated with N glue atoms are presented in Figure 67 along with the band structure in momentum and real space. As shown, the N-N bonding states present at the barrier position (a) are not present at the final position (b). Thus, this confirms the distance data in Figure 65 indicating the attachment of the pair of single glue N atoms. The bonding between the glue atoms and Ga atoms, in both the step and the pyramid is absent at the barrier position. The bonding between N10 and Ga23 within the pyramid is present in both barrier and final positions. On the contrary, the bonding between N1 glue atom and Ga23 (pyramid) and Ga26 and 27 (step) are located in the valence band, i.e. they are occupied. In addition, broken bond state of the glue atom is located in VB, i.e. it is occupied. In summary, all eight quantum states associated with both N glue atoms are occupied. This is in agreement with the data presented in Figure 64 in which the final state corresponds to large energy gain.
Figure 67. Energies of the quantum states of the slab representing N-terminated surface with case 2 position of the pyramid with 7 Ga adatoms and the pair of glue
atoms in: (a) maximum-energy (barrier) position, (b) final position The panel description is as in Figure 27.
The second example considered is the Case 2 with 12 Ga adatoms attached to the slab. The two configuration of the slab attached are presented in Fig. 68 for the two positions of pair of adatoms: maximal energy and final.
Figure 68.
Figure 68 (PDF p. 120). (8 × 4) 8 double atomic layer (DALs) Ga-N slab with a single step used to simulate the attachment of the Ga-N pyramid to the step pointing the pair of Ga-N pyramid bonds towards the step (case 2). The N-terminated surface with 12 Ga adatoms attached is shown in the side and top views. The N glue atoms are shown at: (a) maximum-energy (barrier) position, (b) final position. Blue, yellow and sky-blue balls represent gallium, nitrogen, and hydrogen pseudoatoms, respectively. The N adatom in the pyramid (N15) and the N glue atoms (N1 and N2) are represented by red and magenta balls, respectively. The Ga adatoms in the pyramid: Ga 40 (single), Ga31 and Ga43 (connected to glue atom) are represented by green balls. The step Ga atoms connected to glue atoms, Ga32, Ga31 and Ga30 are represented by orange balls.
The energies of quantum states associated with N glue atoms are shown in Fig. 69 along with the band structure in momentum and real space. As shown, the increase of the number of attached Ga adatoms leads to a higher density of the states in the bandgap, especially in the lower part. At the barrier position, the nitrogen pins the Fermi level, i.e., it is partially occupied. Again, the bonding between N glue adatom and the Ga adatoms both in the pyramid and in the step is not present. The N-N bonding typical for molecule is present.
Figure 69.
Figure 69 (PDF p. 121). Energies of the quantum states of the slab representing N-terminated surface with case 2 position of the pyramid with 12 Ga adatoms and the pair of glue N atoms in: (a) maximum-energy (barrier) position, (b) final position. The panel description is as in Figure 27.
The attachment of a glue atom in the final position leads to the disappearance of the N-N bonding state. On the contrary, the bonding between glue N atoms and the Ga atoms in the step and in the pyramid is present in VB, i.e. these states are occupied. The nitrogen glue atom state is located above the valence band maximum (VBM) but it is still occupied. In summary, the electron transfer to glue atom state is the same as for 7 Ga adatoms. Nevertheless, a higher Ga coverage leads to a higher Fermi level which leads to a larger energy gain ( and ), respectively. The difference in Fermi level position is not large, nevertheless this leads to an energy-gain difference for the pair of N glue adatoms, i.e. for 6 electrons.
The third example considered is the Case 1 with 7 Ga adatoms attached to the slab. The case 1 corresponds to the pyramid orientation with a single Ga-N bond towards the step. The arrangement of this slab with 7 Ga adatoms attached is presented in Figure 70.
Figure 70. (8 × 4) 8 double atomic layer (DALs) Ga-N slab with a single step used to simulate the attachment of the pyramid oriented so that single Ga-N bond points towards the step (case 1). The N-terminated surface with 7 Ga adatoms attached is shown in side and top views. The N glue atoms are shown at: (a) maximum-energy (barrier) position, (b) final position. Blue, yellow and small sky-blue balls represent gallium, nitrogen, and hydrogen pseudoatoms, respectively. The N adatom in the pyramid (N10) and the N glue atoms (N1 and N2) are represented by red and magenta balls, respectively. The Ga adatoms in the pyramid: Ga 23, Ga34 (single) and Ga31 (connected to glue atom) are represented by brown balls. The step Ga atoms connected to glue atoms, Ga24 and Ga25 are represented by orange balls.
The data presented in Figure 65 indicate an essentially unchanged N-N interatomic distance and those in Figure 64 indicate no energy gain. These findings are confirmed by the plots in Figure 70. These findings are examined through an analysis of quantum states shown in Fig. 71. As shown, the N-N overlap is large for nitrogen atoms, dominating other bonds. These bonding states are not occupied at the barrier position. In the final position the N-N bonds are partially occupied, thus bringing some energy gain. The bonding between the pair of N adatoms and the Ga adatoms in both the step and the pyramid is absent. Thus the pair of N atoms does not become glue atoms, the preserve intact molecule. This is in agreement with the data in Figure 64 where no energy gain is observed in this process.
Figure 71.
Figure 71 (PDF p. 124). Energies of the quantum states of the slab representing -terminated surface with case 1 position of the pyramid with 7 Ga adatoms and the pair of glue atoms in: (a) maximum-energy (barrier) position, (b) final position. The panels description is as in Figure 27.
As shown in Figure 65, the increase in Ga coverage, and accordingly the Fermi level at the surface has a negligible influence on the N-N interatomic distance at the surface. Similarly, as shown in Figure 64, the electron contribution has limited impact on the energy of the system, the energy loss under attachment is and for low and high coverages. Thus, the dependence is considerable. Nevertheless, these data indicate that the high Ga coverage is not sufficient for the decomposition of diatomic molecule as shown in Fig. 72 below.
Figure 72.
Figure 72 (PDF p. 125). 8 double atomic layer (DALs) Ga-N slab with a single step used to simulate the attachment of the pyramid oriented so that single Ga-N bond points towards the step (case 1). The N-terminated surface with 12 Ga adatoms attached is shown in side and top views. The N glue atoms are shown at: (a) maximum-energy (barrier) position, (b) final position. Blue, yellow and sky-blue balls represent gallium, nitrogen, and hydrogen pseudoatoms, respectively. The N adatom in the pyramid (N15) and the N glue atoms (N1 and N2) are represented by green and red balls respectively. The Ga adatoms in the pyramid: Ga
38 and Ga28 (single) and Ga36 (connected to glue atom) are represented by brown balls. The step Ga atoms connected to glue atoms, Ga29 and Ga30 are represented by orange balls.
The properties of this system and the role of the electronic contribution are analyzed using the data presented in Fig. 73. As shown, the increase in the number of Ga adatoms does not change the results: molecule remains intact. The bonding between these N atoms and Ga is absent.
Figure 73.
Figure 73 (PDF p. 127). Energies of the quantum states of the slab representing -terminated surface with case 1 position of the pyramid with 12 Ga adatoms and the pair of glue atoms in: (a) maximum-energy (barrier) position, (b) final position. The panels description is as in Figure 27.
The general summary of this research indicates that the attachment of glue atoms depends on the arrangement of the pyramids and the steps: in case 2 it is possible with a high energy gain, while in case 1 the molecule remains intact. The energy gain depends strongly on the electronic structure, being larger for higher Ga coverage, i.e. for higher Fermi level. In the second case the growth is highly inefficient, essentially stopped. As the steps are intertwined, the growth of GaN layers using a molecular nitrogen source without activation is not possible.
Ga-P pyramid at the step on nonmetal-terminated surface
Similarly, the GaP surface slab is presented to show the adatom pyramid configurations on a zinc-blende nonmetal-terminated surface to be attached to the step by glue P adatoms. This example is shown in Figure 74. Attachment of nonmetal glue atoms of the Ga-P pyramid to the step on a nonmetal-terminated surface is presented for the two orientations of the bonds within the pyramid: (i) case 1 – a single Ga-P bond directed towards the step, (ii) case 2 – a pair of Ga–P bonds directed towards the step.
Figure 74. Ga-P slab used for simulations of the glue P atoms attachment to surface: (a) initial configurations of the glue atoms, (b) fully attached glue atoms (left panel: side view; right panel: top view). The three consecutive steps shown correspond to zinc blende ABC. This orientation corresponds to case 2 – the two P-Ga pyramid bonds are directed to the step. This is possible in the zinc-blende lattice. The alternative orientation, denoted as case 1, i.e. the single P-Ga pyramid bond is directed towards the step is not shown. Blue and red balls represent gallium and phosphorus atoms, respectively. The magenta ball represents a P adatom in the apex (second layer), and the green balls represent Ga adatoms in the base of the pyramid (first layer). The yellow balls represent P atoms in the molecule that are to become glue P atoms. The brown balls represent Ga atoms in the step connected to P glue atoms.
Thus, the first case analyzed is the dissociative adsorption of molecular phosphorus leading to the formation of a pair of glue P adatoms. Phosphorus in the vapor remains in the form of diatomic molecules and tetraatomic molecules [36]. The energy profiles, obtained for the dissociation pathways of phosphorus diatomic molecules leading to adsorption at the surface are presented in Figure 75. As shown, the obtained energy profiles are different from the previously discussed case of nitrogen adsorption. There are small dissociation barriers, but they are at lower energy than the value at large separation and therefore do not effectively block adsorption. In summary, it is expected that the barrier is overcome due to the kinetic energy gain so that, effectively, no barrier limits attachment of phosphorus glue atoms to connect the pyramid. In fact, the energy of dissociation of phosphorus diatomic molecules equal to , therefore the bonding effect of the glue atoms is very strong.
Figure 75. Energy profiles for the paths of a pair of phosphorus glue atoms that are finally attached to GaP() surface, plotted for different Ga coverages of the surface: 12 and 7 Ga
In case 2, the energy gain is: and for Ga coverage of 7 and 12 adatoms. This corresponds to an energy gain per P glue atom of and . Thus, the energy gain shows some additional dependence on Ga coverage which will be discussed below.
In case 1, these energy gains are much lower and for Ga coverage of 7 and 12 adatoms. This corresponds to an energy gain per P glue atom of and . These results indicate that the bonding is different in these two cases as discussed below using the methods presented in Ref. 12 [12].
adatoms. Cases 1 and 2 correspond to the pyramid orientations with a single bond and a pair of Ga-P bonds towards the step, respectively.
Despite the considerable difference between these two cases, the outcome of these processes is identical: diatomic molecules dissociate and are attached in atomic form as P glue atoms. This is supported by the plot of the P-P interatomic distance showing its increase from Å value at large separation to about Å, i.e. coinciding with the GaP lattice parameter. It should be added that the interatomic distance is increased rapidly at short distances from the surface.
Figure 76. P-P interatomic distance variation along the paths of a pair of phosphorus glue atoms, that are finally attached to GaP() surface, plotted for different Ga coverages of the surface: 12 and 7 Ga adatoms Cases 1 and 2 correspond to the pyramid orientations with a single bond and a pair of Ga-P bonds towards the step, respectively.
The analysis begins with the first example, i.e. Case 2 with 7 Ga adatoms attached, i.e. the pyramid orientations with a single bond and a pair of Ga-P bonds towards the step. As shown before an increase in Ga coverage from 7 to 12 adatoms shifts the Fermi level only slightly. Although the attachment of a P glue atom introduces five electrons creating eight occupied states, and that the missing electrons are shifted from Ga states close to the Fermi level, i.e. the energy difference is small. Accordingly, the P glue adatom energy gain varies by only 0.15 eV. The configuration of the slab with 7 Ga adatoms attached is presented in Figure 77.
Figure 77. 9 double atomic layer (DALs) Ga-P slab with a single step used for simulations of the attachment of the Ga-P pyramid to the step with the pair of Ga-P pyramid bonds towards the step (case 2). The P-terminated surface with 7 Ga adatoms attached is shown in the side and top views. The P glue atoms are shown at: (a) maximum-energy (barrier) position, (b) final position. Blue, red and sky-blue balls represent gallium, phosphorus and hydrogen pseudoatoms, respectively. The Ga adatoms in the pyramid: Ga231 – single and Ga138 and Ga162 – connected to glue atoms are represented by green balls. The P adatom in the pyramid (P137) and the P glue atoms (P1 and P2) are represented by magenta and yellow balls, respectively. The step adatoms connected to glue P atoms, i.e. Ga114, Ga139 and Ga163 are denoted by orange balls.
The energies of quantum states of the Ga and P atoms are presented for two configurations: (i) maximal energy at and (ii) final at . As shown, the gap is closed by the states of Ga adatoms both for maximal energy and final positions. The Ga 231 state in the pyramid is located in VB, i.e. it is occupied for both cases. The states are antibonding for the final state and bonding in the maximal energy state confirming the dissociation of molecule during adsorption. The bonding states between the states of the P glue atoms and the Ga atoms in the pyramid and in the step are not present in the maximal energy state. Differently, the states are present in the maximal energy position, located in the VB and therefore occupied, thus confirming the formation of bonds between the pyramid and the step via glue atoms.
Figure 93 (PDF p. 132).
Figure 78. Energies of the quantum states of the slab representing P-terminated surface with case 2 position of the pyramid with 7 Ga adatoms and the pair of glue P atoms in: (a) maximum-energy (barrier) position, (b) final position. The panels description is as in Figure 27.
An increase in surface coverage to 12 Ga adatoms is presented in Fig. 79 showing the slab with molecule in the maximum-energy and final minimum-energy configurations. As shown, the results are similar, during adsorption the molecule dissociates and the two P glue atoms attach, linking the pyramid and the step.
Figure 79.
Figure 79 (PDF p. 134). 9 double atomic layer (DALs) Ga-P slab with a single step used for simulations of the attachment of the Ga-P pyramid to the step with the pair of Ga-P pyramid bonds towards the step (case 2). The P-terminated surface with 12 Ga adatoms attached is shown in the side and top views. The P glue atoms are shown in: (a) maximal energy (barrier) position, (b) final position. Blue, red and small sky-blue balls represent gallium, phosphorus and hydrogen pseudoatoms, respectively. The Ga adatoms in the pyramid: Ga236 – single and Ga143 and Ga167 – connected to glue atoms are represented by green balls. The P adatom in the pyramid (P142) and the P glue atoms (P1 and P2) are represented by magenta and yellow balls, respectively. The step adatoms connected to glue P atoms, i.e. Ga119, Ga144 and Ga168 are denoted by orange balls.
An increase of Ga coverage to 12 adatoms in the slab results in a higher density of the Ga states in the bandgap. This leads to a higher Fermi level at the surface. The general process of attachment of a P glue atoms are unaffected. The molecule decomposes into a pair of P glue atoms which are bonded to the surrounding Ga atoms both in the step and in the pyramid. This bonding is absent in the maximal energy position in agreement with the previous data. Thus, the impact of larger Ga coverage is limited as indicated by the energy plot in Figure 80.
Figure 80. Energies of the quantum states of the slab representing P-terminated GaP() surface with case 2 position of the pyramid with 12 Ga adatoms and the pair of glue
P atoms in: (a) maximum-energy (barrier) position, (b) final position. The panels description is as in Figure 27.
The subsequent case investigated is the one denoted as case 1, in which the pyramid is oriented so that a single P-Ga bond in the pyramid is directed towards the step. This is analyzed at low and high Ga coverage represented by 7 and 12 Ga adatoms, respectively. The first is the low-coverage case shown in Figure 81. As shown, the pyramid shows considerable misalignment, stretching Ga-P bonds of glue atoms. Nevertheless, the dissociation of the molecule occurs and the glue atoms link the pyramid and the step albeit much more weakly as shown in Fig. 82.
Figure 81. Fig. 81. (8 4) 9 double atomic layer (DALs) Ga-P slab with a single step used for simulations of the attachment of the Ga-P pyramid to the step with a single Ga-P pyramid bond towards the step (case 1). The P-terminated GaN surface with 7 Ga adatoms attached is
shown in the side and top views. The P glue atoms are shown at: (a) maximum-energy (barrier) position, (b) final position. Blue, red and sky-blue balls represent gallium, phosphorus and hydrogen pseudoatoms, respectively. The Ga adatoms in the pyramid: Ga148 – single and Ga125 and Ga219 – connected to glue atoms are represented by green balls. The P adatom in the pyramid (P124) and the P glue atoms (P1 and P2) are represented by magenta and yellow balls, respectively. The step adatoms connected to glue P atoms, i.e. Ga219, Ga218 and Ga243 are denoted by orange balls.
The energies of the quantum states in the maximum-energy and final minimum-energy configurations are plotted in Fig. 82. As shown, the maximum-energy configuration is characterized by the overlap between P atoms in molecule. This bonding is absent for the final position demonstrating that, unlike nitrogen,, in this case decomposition of the molecule takes place and the glue atoms are attached. The P-Ga bonding states are located at the VBM, thus they are occupied, but are higher in energy; therefore, the energy gain is limited as shown in Figure 75.
Figure 82. shown in the side and top views. The P glue atoms are shown at: (a) maximum-energy (barrier) position, (b) final position. Blue, red and sky-blue balls represent gallium, phosphorus and hydrogen pseudoatoms, respectively. The Ga adatoms in the pyramid: Ga148 – single and Ga125 and Ga219 – connected to glue atoms are represented by green balls. The P adatom in the pyramid (P124) and the P glue atoms (P1 and P2) are represented by magenta and yellow balls, respectively. The step adatoms connected to glue P atoms, i.e. Ga219, Ga218 and Ga243 are denoted by orange balls.
Figure 82 (PDF p. 138). Energies of the quantum states of the slab representing P-terminated surface with case 1 position of the pyramid with 7 Ga adatoms and the pair of glue P atoms in: (a) maximum-energy (barrier) position, (b) final position. The panels are described as in Figure 5.
The final slab considered is the case 1 with high Ga coverage represented by 12 Ga adatoms shown in Figure 83. The maximum-energy and final configurations show better alignment than in the previous case; thus, the energy gain shown in Figure 84 is much larger.
Figure 83. (8 × 4) 9 double atomic layer (DALs) Ga-P slab with a single step used for simulations of the attachment of the Ga-P pyramid to the step with a single Ga-P pyramid bond towards the step (case 1). The P-terminated surface with 12 Ga adatoms attached is shown in the side and top views. The P glue atoms are shown at: (a) maximum-energy (barrier) position, (b) final position. Blue, red and sky-blue balls represent gallium, phosphorus and hydrogen pseudoatoms, respectively. The Ga adatoms in the pyramid: – single and and – connected to glue atoms, are represented by green balls. The P adatom in the pyramid () and the P glue atoms ( and ) are represented by magenta and yellow balls, respectively. The step adatoms connected to glue P atoms, i.e. , and are denoted by orange balls.
Figure 84. Energies of the quantum states of the slab representing P-terminated surface with case 1 position of the pyramid with 12 Ga adatoms and the pair of glue P atoms in: (a) maximum-energy (barrier) position, (b) final position. The panels description is as in Figure 27.
The energies of the system in the maximal energy position and in the final lowest energy state are shown in Figure 84. As it is shown, in the maximal energy position overlap exists only between P atoms in molecule. The P atoms do not have overlap with Ga atoms in the pyramid and at the step. The Fermi level is pinned by the states from molecule.
Figure 94 (PDF p. 140).
The final configuration is characterized by the presence of bonding states between P glue atoms and Ga adatoms both in the pyramid and in the step. Thus the pyramid is attached to the step by glue atoms. These states are located at lower energies than in the low-Ga-coverage slab, so that the energy gain is larger. This is supported by the results presented in Figure 75, showing much lower energy in this case.
In summary, the case 1 process leads to the attachment of a P glue atoms, but their positions are energetically less favorable so they could lead to difference in the vapor growth and contribute to the possible multiplication of native defects.
Summary and conclusions
The ab initio based investigations reported in this paper were directed to the formulation of an atomic model of the vapor growth of semiconductors, belonging to the III-V family, i.e. Ga-NM on polar surfaces, both Ga- and nonmetal NM-terminated. The model was constructed assuming the presence of the components in the gas phase only, i.e. Ga and NM vapor. These investigations included wurtzite GaN and zinc blende GaP, GaAs, and GaSb polar surfaces.
These results are summarized by presenting:
a) the state of the art before the present investigations
b) the results reported in this paper,
c) the state of the art after completion of this investigation
a) the state of the art before the research presented in this work was made, can be summarized as follows :
(i) The properties of polar metal- and nonmetal-terminated clean surfaces had been investigated only in the case of nitrides.
(ii) The Ga-terminated GaN(0001) surface undergoes ECR charge-controlled reconstruction with the Fermi level located between these states. (iii) Ga broken bond state of surface is located close to the CBM,
(iv) N-terminated stoichiometric surface is not reconstructed,
(v) The energy of the N broken-bond state of GaN surface is located close to the VBM, pinning the Fermi level,
(vi) Ga and N adatoms at low coverage are adsorbed at H3 sites on and surfaces, respectively,
(vii) Polar surfaces of GaP, GaAs and GaSb are reconstructed, partially through evaporation of atoms from these surfaces, creating sophisticated reconstruction patterns,
(viii) Both Ga- and N-terminated surfaces are covered by a disordered high-Ga-coverage adlayer
b) The main results reported in this paper are:
i) The properties of both polar stoichiometric surfaces are drastically different,
ii) The properties of the same-polarity surfaces of chemically different compounds are virtually almost identical,
iii) Stoichiometric Ga-terminated polar surfaces, i.e. , , and all undergo ECR charge-controlled reconstruction,
iv) The Fermi level at stoichiometric Ga-terminated polar surfaces is located in the bandgap created between empty and occupied hybridized states (Peierls effect), close to the CBM,
v) Stoichiometric N-terminated surface atoms in the topmost layer remain in hybridization, without any reconstruction,
vi) Nonmetal-terminated stoichiometric surfaces , and are not reconstructed, with topmost NM atoms in hybridization,
vii) The Fermi level at NM-terminated stoichiometric polar surfaces is pinned by a broken-bond state, located in the vicinity of VBM,
viii) At very low coverage NM atoms adsorbed at Ga-terminated surfaces are located in H3 sites with a considerable energy gain, despite the fact that the reference states are energetically strongly bound diatomic molecules (, , ). The adsorption energy for Sb is very high owing to the atomic reference state.
ix) The stable state for the Ga-terminated surface is the surface with fractional NM coverage () located in H3 sites,
x) The presence of NM adatoms in H3 sites stabilizes Ga terminated surfaces, preventing escape of the topmost Ga surface atoms and creation of Ga surface vacancies,
xi) H3 site NM adatoms in low coverage of Ga-terminated surfaces do not create suitable structure for the growth,
xii) Higher N coverage of Ga-terminated surface is not stable; the N adatoms create molecules that detach from the surface, thus excluding the growth of GaN from molecular vapor on this surface,
xiii) At higher coverage the P, As and Sb adatoms are located in the highly skewed on-top positions due to strong attractive interactions within the adlayer
xiv) The high-coverage adsorption energy gain is small but positive for P and As which is related to high bonding energy of diatomic molecules in the vapor, in the case of atomic Sb vapor the energy gain is large, above 4 eV,
xv) At low coverage () Ga atoms on NM-terminated surfaces are attached in H3 sites (III), with relatively high energy gain, of the order of 3 eV,
xvi) Fermi level at low Ga coverage NM-terminated surface is located at the VBM,
xvii) NM-terminated surface of GaN, GaP and GaAs are energetically unstable with respect to detachment of NM topmost atoms to create diatomic molecules and surface vacancies owing to the high bonding energies of diatomic molecules,
xviii) The Sb surface is stable with respect to creation of surface vacancies owing to the atomic nature of the vapor,
xix) The low coverage adsorption of Ga adatoms in H3 positions stabilizes NM-terminated surfaces with respect to escape of NM topmost atoms, thus, the creation of surface vacancies is not energetically favorable, and Ga-vacancy reconstructions of these surfaces need not be considered in crystal growth models,
xx) At high coverage Ga adatoms are located in the on-top positions of NM-terminated surface, thus creating first layer necessary for the growth,
xxi) The adsorption energy gain of Ga adatoms at high coverage is above 4 eV, thus creating stable structures, the first step in growth,
xxii) At high Ga coverage of the NM-terminated surface the bandgap is closed by Ga states, with the Fermi level located differently (GaN – midgap, GaP – VBM, GaAs and GaSb – VBM),
xxiii) Adsorption of the second layer on NM adatoms over the first layer of high Ga covered (on-top) NM-terminated surface leads to the creation of a stable pyramidal structure with NM atoms located at the apex,
xxiv) The limited presence of a second NM adlayer does not change the position of Fermi level at the surface,
xxv) The pyramid structure is fundamental, stable structure necessary to the growth of semiconductors at NM – terminated surfaces,
xxvi) Adsorption of a Ga atom on top of a nonmetal-adatom layer on a Ga-terminated surface does not create a pyramidal structure, necessary for efficient growth, thus the growth on this face was not simulated,
xxvii) A Ga adatom is attached to a pair of nonmetal adatoms on a Ga-terminated surface, which is related to the strong bonding within the nonmetal chains,
xxviii) The second layer Ga adsorption energy is high, on the order of 3.5 eV which is partially related to electron transfer from Ga to nonmetal-atom states,
xxix) Surface diffusion across NM-terminated surface proceeds via pivotal motion of the stable pyramids,
xxx) During the motion of the pyramid by rotation (pivot), all Ga adatoms remains stably attached to the NM atom in the apex,
xxxi) The energy barrier for the motion is relatively small and different for GaN and the other pyramids: it is about 0.2 eV for GaN and above 0.5 eV for the others,
xxxii) The detachment energy of NM adatoms is high; accordingly, the surface diffusion length is macroscopically large and all attached NM adatoms arrive at the steps,
xxxiii) Additional atoms (Glue atoms) attachment was proved to be effective way of connecting the pyramids to the steps at NM-terminated surface,
xxxiv) Attachment of glue atoms critically depends on the orientation of the pyramid and the step,
xxxv) In the wurtzite lattice, the orientation of the pyramid with respect to two consecutive steps is different, thus the two consecutive steps have different kinetic properties (VIII),
xxxvi) In the zinc-blende lattice, the pyramids are identically oriented with respect to three consecutive steps, thus all steps have identical kinetic properties, although there are two different families of these steps,
xxxvii) Glue nonmetal (NM) atom attachment requires donation of electrons from other surface, thus the energy gain in the process depends on position of Fermi level at the surface,
xxxviii) The growth of GaP from the vapor is possible, based on Ga and vapor mixture, the growth of GaN from the Ga and vapor mixture is not.
c) the state of the art after completion of this investigation:
i) The structural and electronic properties of both polar stoichiometric surfaces of GaN, GaP, GaAs and GaSb were determined,
ii) Low coverage adatoms on polar surfaces are located in H3 sites, drastically stabilizing their structures,
iii) High coverage of polar surfaces is atomic (with the exception of ) located in the skewed on-top positions, partially due to interaction with the adlayer,
iv) A second Ga adlayer on a Ga-terminated surface covered by first NM adatoms does not create pyramid structure due to strong attractive interaction of the first NM adlayer,
v) Second layer NM adatoms create stable pyramid structure on NM-terminated surfaces which is capable of rotational (pivoting) motion, i.e. surface diffusion,
vi) Growth of III-V semiconductors occurs via NM glue atom attachment of the pyramids to the steps,
vii) The glue atom attachment process depends critically on the Fermi level at the surface as it requires the acceptance of several electrons from the other quantum states,
viii) The model does not include H in the system, therefore these conclusion do not apply to MOVPE/HVPE growth of III-V compounds using , , and other similar NM compounds.
In conclusion, the quantum theory ab initio atomic level model of the vapor growth of III-V semiconductors on NM polar surfaces via step kinetics based on glue atom connection of the pyramids to the steps was developed. This is a considerable step forward which is essentially complete; nevertheless, it still requires verification by independent research. The open part is that of the further investigation of the pyramid - glue atom step kinetics which requires considerable effort to determine complex dynamics of the layered growth and the incorporation of point defects which remain topics for future investigations.
Acknowledgements
The calculations reported in this paper were performed using the computing facilities of the Interdisciplinary Centre for Mathematical and Computational Modelling of Warsaw University (ICM UW) under grant GB84-23.
References
- [1]I. Vugraftman, J. R. Meyer, L. R. Ram-Mohan, Band parameters for III-V compound semiconductors and their alloys. J. Appl. Phys. 89, 5815 (2001). https://doi.org/10.1063/1.1368156
- [2]CRC Handbook of Chemistry and Physics, W. M. Haynes (ed). CRC Press (97th ed.) 2016
- [3]P. Atkins, J. de Paula, J. Keeler, Atkins’ Physical Chemistry, Oxford Univ. Press, Oxford 2022.DOI
- [4]N. N. Greenwood, A. Earnshaw, Chemistry of the Elements, Elsevier 1997.
- [5]H. Ibach, H. Luth, Solid-State Physics. An Introduction to Principles of Materials Science, Springer-Verlag, Berlin Heidelberg 2009. p. 21DOI
- [6]P. Strak, K. Sakowski, P. Kempisty, I. Grzegory, A. Kaminska, S. Krukowski, Ab initio elucidation of the nature of bonding of tetrahedral nitrides (BN, AlN, GaN, InN), hexagonal BN and graphene, Materials 18, 2876 (2025). https://doi.org/10.3390/ma18122875
- [7]I. R. Grant, III-V Compound Semiconductors: Growth, in Encyclopedia of Materials: Science and Technology (Second Edition) 2001. https://doi.org/10.1016/B0-08-043152-6/00264-3
- [8]A. Stirling, I. Papai, J. Mink, D. R. Salahub, Density Functional Theory Study of Nitrogen Oxides. J. Chem. Phys. 100, 2910 (1994). https://doi.org/10.1063/1.466433
- [9]B. Ruscic, Active Thermochemical Tables (ATcT), Argonne National Laboratory, https://atct.anl.gov/
- [10]K. A. Gingerich, D. L. Cocke, J. Kordis, Gaseous Phosphorus Compounds, X. Mass Spectrometric Determination of the Dissociation Energies of Arsenic and Bismuth Monophosphides, J. Phys. Chem. 78, 603 (1974). https://doi.org/10.1021/j100599a009
- [11]P. Strak, K. Sakowski, P. Kempisty, S. Krukowski, Structural and electronic properties of AlN(0001) surface under partial N coverage as determined by ab initio approach, J. Appl. Phys. 118, 095705 (2015). http://dx.doi.org/10.1063/1.4929823
- [12]S. Krukowski, P. Kempisty, P. Strak, Charge control of semiconductor surfaces as elucidated by ab initio calculations – a review. Prog. Surf. Sci. 101, 100809 (2026). https://doi.org/10.1016/j.progsurf.2025.100809
- [13]A. Garcia et al. SIESTA: Recent developments and applications. J. Chem. Phys. 152, 204108 (2020). https://doi.org/10.1063/5.0005077
- [14]J. Junquera, O. Paz, D. Sanchez-Portal, and E. Artacho, Numerical atomic orbitals for linear-scaling calculations. Phys. Rev. B 64, 235111 (2001). https://doi.org/10.1103/PhysRevB.64.235111
- [15]E. Anglada, J. M. Soler, J. Junquera, and E. Artacho, Systematic generation of finite-range atomic basis sets for linear-scaling calculations. Phys. Rev. B 66, 205101 (2002). https://doi.org/10.1103/PhysRevB.66.205101
- [16]N. Troullier, and J. L. Martins, Efficient pseudopotentials for plane-wave calculations. Phys. Rev. B, 43, 1993 (1991). https://doi.org/10.1103/PhysRevB.43.1993
- [17]N. Troullier, and J. L. Martins, Efficient pseudopotentials for plane-wave calculations. II. Operators for fast iterative diagonalization. Phys. Rev. B, 43, 8861 (1991). https://doi.org/10.1103/PhysRevB.43.8861
- [18]H.J Monkhorst, and J. D. Pack, Special points for Brillouin-zone integrations. Phys. Rev. B, 13, 5188 (1976). https://doi.org/10.1103/PhysRevB.13.5188
- [19]M. Abramowitz, I. A. Stegun, Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, Dover 1983 N. York 1983, p 883.
- [20]L. S. Pedroza, A. J. R. da Silva, and K. Capelle, Gradient-dependent density functionals of the Perdew-Burke-Ernzerhof type for atoms, molecules, and solids, Phys. Rev. B 79, 201106(R) (2009). https://doi.org/10.1103/PhysRevB.79.201106
- [21]M. M. Odashima, K. Capelle, and S. B. Trickey, Tightened Lieb-Oxford Bound for Systems of Fixed Particle Number, J. Chem. Theory Comput. 5, 798 (2009) . https://doi.org/10.1021/ct8005634
- [22]J. Neugebauer, M. Scheffler, Adsorbate-substrate and adsorbate-adsorbate interactions of Na and K adlayers on Al(111). Phys. Rev. B 46, 16067 (1992). https://doi.org/10.1103/PhysRevB.46.16067
- [23]L. Bengtsson, Dipole correction for surface supercell calculations. Phys. Rev. B 59 (1999) 12301 – 12304. https://doi.org/10.1103/PhysRevB.59.12301
- [24]S. Krukowski, P. Kempisty, P. Strak, Electrostatic condition for the termination of the opposite face of the slab in DFT simulations of semiconductor surfaces. J. Appl. Phys. 105, 113701 (2009). http://dx.doi.org/10.1063/1.3130156
- [25]M. Leszczynski, H. Teisseyre, T. Suski, I. Grzegory, M. Bockowski, J. Jun, S. Porowski, K. Pakula, J.M. Baranowski, C.T. Foxon, T.S. Cheng, Lattice parameters of gallium nitride Appl. Phys. Lett. 69, 73 (1996). https://doi.org/10.1063/1.118123
- [26]L.G. Ferreira, M. Marques, and L.K. Teles, Approximation to density functional theory for the calculation of band gaps of semiconductors. Phys. Rev. B 78, 125116, (2008). https://doi.org/10.1103/PhysRevB.78.125116
- [27]M. Ribeiro, L.R.C. Fonseca, and L.G. Ferreira, Accurate prediction of the Si/SiO2 interface band offset using the self-consistent ab initio DFT/LDA-1/2 method. Phys. Rev. B, 79, 241312 (2009). https://doi.org/10.1103/PhysRevB.79.241312
- [28]B. Monemar, J.P. Bergman, I. A. Buyanova, H. Amano, I. Akasaki, T. Detchprohm, K. Hiramatsu, and N. Sawaki, The excitonic bandgap of GaN: Dependence on substrate. Solid State Electron. 41, 239 (1997). https://doi.org/10.1016/S0038-1101(96)00208-0
- [29]Y.C. Yeo, T.C. Chong, M.F. Li, Electronic band structures and effective-mass parameters of wurtzite GaN and InN. J. Appl. Phys. 83, 1429 (1997). https://doi.org/10.1063/1.366847
- [30]M. Magnuson, F. Mattesini, C. Hoglund, J. Birch, L. Hultman, Electronic structure of GaN and Ga investigated by soft x-ray spectroscopy and first-principles methods. Phys. Rev. B 82, 085125 (2010). https://doi.org/10.1103/PhysRevB.81.085125
- [31]M. Ptasinska, J. Soltys, J. Piechota, S. Krukowski, Electronic properties on GaN(0001) surface. Vacuum 99, 166 (2014) https://doi.org/10.1016/j.vacuum.2013.05.006
- [32]A. van der Avoird, P. E. S. Wormer, A. P. J. Jansen, An improved intermolecular potential for nitrogen, J. Chem. Phys. 84, 1629 (1986). https://doi.org/10.1063/1.450457
- [33]P. Strak, S. Krukowski, Molecular nitrogen – N₂ properties: The intermolecular potential and equation of state. J. Chem. Phys. 126, 194501 (2007) http://dx.doi.org/10.1063/1.2733651
- [34]S. Krukowski, P. Strak, Equation of state of nitrogen (N₂) at high pressures and high temperatures, J. Chem. Phys. 124, 16103 (2006). http://dx.doi.org/10.1063/1.2185096
- [35]P. Strak, S. Krukowski, Determination of Shear Viscosity of Molecular Nitrogen: Molecular Dynamics Hard Rotor Methodology and the Results, J Phys. Chem. B 115, 4359 (2011). http://dx.doi.org/10.1021/jp111071v
- [36]A. Karton, J. M. L. Martin, “W4 thermochemistry of P₂ and P₄. Is the CODATA heat of formation of the phosphorus atom correct?”, Mol. Phys. 105, 2499 (2007). https://doi.org/10.1080/00268970701543572
- [37]R. K. Yoo, B. Ruscic, J. Berkowitz, Photoionization of As₂ and As₄: Implication for group V clusters. J. Chem. Phys., 96, 6696 (1992). https://doi.org/10.1063/1.462579
- [38]G. Henkelman, B.P. Uberuaga and H. Jonsson, A climbing image nudged elastic band method for finding saddle points and minimum energy paths. J. Chem. Phys. 113, 9901 (2000). https://doi.org/10.1063/1.1329672
- [39]G. Henkelman, H. Jonsson, Improved tangent estimate in the nudged elastic band method for finding minimum energy paths and saddle points. J. Chem. Phys. 111, 9978 (2000). https://doi.org/10.1063/1.1323224
- [40]D. Sheppard, R. Terrell, G. Henkelman, Optimization method for finding minimum energy path. J. Chem. Phys. 128, 134106 (2008). https://doi.org/10.1063/1.2841941
- [41]W. Mönch, Semiconductor Surfaces and Interfaces. Springer Berlin 1993
- [42]G. P. Srivastava, Theory of semiconductor surface reconstruction. Rep. Prog. Phys. 60, 561 (1997). https://doi.org/10.1088/0034-4885/60/5/002
- [43]C. B. Duke, Principles of Semiconductor Surface Reconstruction. Scanning Microscopy 8, 753 (1994). https://digitalcommons.usu.edu/microscopy/vol8/iss4/1
- [44]M. C. Desjonqueres, D. Spanjaard. Concept in Surface Physics Springer Berlin, 1998.
- [45]V. Bermudez, The fundamental surface science of gallium nitride. Surf. Sci. Rep. 72, 147 (2017). https://doi.org/10.1016/j.surfrep.2017.05.001
- [46]M. Ptasinska, J. Piechota, S. Krukowski, Adsorption of hydrogen at the GaN(000ī) surface – ab initio study, J. Phys. Chem. C 119, 11563 (2015). https://doi.org/10.1021/acs.jpcc.5b00836
- [47]P. Strak, P. Kempisty, K. Sakowski, S. Krukowski, Ab initio determination of electron affinity of polar nitride surfaces, clean and under Cs coverage, J. Vac. Sci. Technol. A 35, 021406 (2017). http://dx.doi.org/10.1116/1.4975332
- [48]P. Strak, W. Miller, S. Krukowski, Charge-Controlled Energy Optimization of the Reconstruction of Semiconductor Surfaces: sp³-sp² Transformation of Stoichiometric GaN(0001) Surface to (4 × 4) Pattern, Materials 17, 2614 (2024). https://doi.org/10.3390/ma17112614
- [49]F-H Wang, P. Kruger, J. Pollmann, Electronic structure of 1 x 1 GaN(0001) and GaN(000ī) surfaces, Phys. Rev. B 64, 035305 (2001) . https://doi.org/10.1102/PhysRevB.64.035305
- [50]A. L. Rosa, J. Neugebauer, First-principles calculations of the structural and electronic properties of clean GaN(0001) surfaces, Phys. Rev. B 73, 203346 (2006). https://doi.org/10.1103/PhysRevB.73.203346
- [51]C. G. Van de Walle, D. Segev, Microscopic origin of surface states on nitride surfaces, J. Appl. Phys. 101, 081704 (2007). https://doi.org/10.1063/1.2722731
- [52]N. Takeuchi, A. Selloni, T. H. Myers, A. Dolittle, Adsorption and diffusion of Ga and N adatoms on GaN surfaces: Comparing the effects of Ga coverage and electronic excitation, Phys. Rev. B 72, 115307 (2005). https://doi.org/10.1103/PhysRevB.72.115307
- [53]D. J. Chadi, Atomic structure of GaAs(100)-(2 × 1) and (2 × 4) reconstructed surfaces. J. Vac. Sci. Technol. A 5, 834 (1987). https://doi.org/10.1116/1.574366
- [54]J. Piechota, S. Krukowski, P. Sadovyi, B. Sadovyi, S. Porowski, I. Grzegory, Nitrogen dissolution in Liquid Ga and Fe: comprehensive ab initio analysis. Relevance for crystallization of GaN. Materials 14, 1306 (2021). https://dx.doi.org/10.3390/ma14051306
- [55]S. Krukowski, Z. Romanowski, I. Grzegory, S. Porowski, Interaction of N₂ molecule with liquid Ga surface – quantum mechanical calculations (DFT). J. Cryst. Growth 189-190, 159 (1998). http://dx.doi.org/10.1016/S0022-0248(98)00195-X
- [56]Z. Romanowski, S. Krukowski, I. Grzegory, S. Porowski, Surface reaction of nitrogen with liquid group III metals. J. Chem. Phys. 114, 6353 (2001). http://dx.doi.org/10.1063/1.1355984
- [57]R. M. Feenstra, Y. Dong, C. D. Lee, J. E. Northrup, Recent developments in surface studies of GaN and AlN. J. Vac. Sci. Technol. B23, 1174 (2005). https://doi.org/10.1116/1.1881612
- [60]R. Dronskowski, P. E. Bloechl, Crystal orbital Hamilton populations (COHP): energy-resolved visualization of chemical bonding in solids based on density-functional calculations. J. Phys. Chem. 97, 8617 (1993). https://doi.org/10.1021/j100135a014
- [61]V. L. Deringer, A. L. Tchougreeff, R. Dronskowski, Crystal Orbital Hamilton Population (COHP) Analysis as Projected from Plane-Wave Basis Sets. J. Phys. Chem. 115, 5461 (2011). https://doi.org/10.1021/jp202489s
- [62]P. Kempisty, S. Krukowski, Adsorption of ammonia at GaN(0001) surface in the mixed ammonia/hydrogen ambient - a summary of ab initio data, AIP. Adv. 4, 117109 (2014). http://dx.doi.org/10.1063/1.4901922
- [63]P. Kempisty, P. Strąk, K. Sakowski, S. Krukowski, Thermodynamics of GaN(s)-NH₃(v)+N₂(v)+H₂(v) system - electronic aspects of the processes at GaN(0001) Surface. Surf. Sci. 662, 12 (2017). https://doi.org/10.1016/j.susc.2017.03.007
- [64]CCDCBD, NIST Database 101, https://cccbdb.nist.gov Release 22 (May 2022).
- [65]N. Kafi, S. Kang, C. Golz, A. Rodrigues-Weisensee, L. Perischetti, D. Ryzhak, G Cepellini, D. Sprinto, M. Schmidbauer, A. Kwasniewski, C. Netzel, O. Skibitzki, F. Hatani, Selective Growth of GaP Crystals on CMOS-Compatible Si Nanotip Wafers by Gas Source Molecular Beam Epitaxy, Cryst. Growth Des. 24, 2724 (2024). https://doi.org/10.1021/acs.cgd.3c01337
- [66]M. Sadeghi, S. Wang, Growth of GaP on Si substrates by Solid source molecular beam epitaxy, J. Cryst Growth 227-229, 279 (2001). https://doi.org/10.1016/S0022-0248(01)00705-9
- [67]T. Suzuki, T. Saga, T. Jimbo, M. Umeno, Growth mechanism of GaP on Si substrates by MOVPE, J. Cryst. Growth 115, 158(1991). https://doi.org/10.1016/0022-0248(91)90731-J
- [68]V. Tassev, D. Bliss, C. Lynch, C. Yapp, W. Goodhue, K. Termkoa, Low pressure- low temperature growth of GaP for nonlinear optical frequency conversion devices. J. Cryst. Growth 312, 1146 (2010). https://doi.org/10.1016/j.jcrysgro.2009.12.053
- [69]P. Kempisty, P. Strak, K. Sakowski, Y. Kangawa, S. Krukowski, Ab initio and thermodynamic picture of Al adsorption of AlN(0001) surface – role of bond creation and electron transition contributions, Appl. Surf. Sci. 532, 147719 (2020). https://doi.org/10.1016/j.apsusc.2020.147419
- [70]W. P. Glushko (ed.) Termodinamiczeskije swoistva individualnych wescestw. Nauka, Moscow, 1978
- [71]I. Barin, Thermochemical Data of Pure Substances, VCH, Weinheim 1994.DOI
- [72]NIST-JANAF Thermochemical Tables. NIST Standard Reference Database 13, https://doi.org/10.18434/T42S31
- [73]S. Krukowski, P. Kempisty and A. Jalbout, Thermodynamic and kinetic approach in DFT studies of microscopic structure of GaN(0001) surface in ammonia-rich conditions, J. Chem. Phys. 129, 234705 (2008). http://dx.doi.org/10.1063/1.3037218
- [74]T. D. Moustakas, B. Paiella, Optoelectronic device physics and technology of nitride semiconductors from the UV to terahertz. Rep. Prog. Phys. 80, 106501 (2017). https://doi.org/10.1088/1361-6633/aa7bb2
- [75]C. Skierbiszewski, Z. R. Wasilewski, I. Gregory, S. Porowski, Nitride-based laser diodes by plasma-assisted MBE- from violet to green emission. J. Cryst. Growth 311, 1632- 1639 (2008). https://doi.org/10.1016/j.jcrysgro.2008.12.040
- [76]A. Y. Cho, J. R. Artur, Molecular beam epitaxy, Prog. Solid State Chem. 10, 157 (1975). https://doi.org/10.1016/0079-6786(75)90005-9
- [77]H. Kroemer, The family (InAs, GaSb, AlSb) and its heterostructures: a selective review, Physica E 20, 196 (2004) https://doi.org/10.1016/j.physe.2003.08.003
- [78]K. Takayanagi, Y. Tanishiri, S. Takahashi, M. Takahashi, Structure analysis of Si(111)-7 × 7 reconstructed surface by transmission electron diffraction. Surf. Sci. 164, 367 – 392 (1985). https://doi.org/10.1016/0039-6028(85)90753-8
- [79]K. Hattori, K. Ishikara, Y. Miyatake, F. Matsui, S. Takeda, H. Daimon, F. Komori, GaP(111) reconstructed surface studied with STM and LEED, Surf. Sci. 525, 57 (2003). https://doi.org/10.1016/S0039-6028(02)02564-5
- [80]K. Hestofer, K. Rivoire, J. Vučković, F. Hatami, High-Quality GaP(111) Grown by Gas Source MBE for Photonic Crystals and Advanced Nonlinear Optical Applications. Nanomaterials 15, 619 (2025). https://doi.org/10.3390/nano15080619
- [81]T. Komura, T. Hanada, T. Yao, H. Kuramochi, M. Ozeki, Surface structures of GaAs{111}A,B-(2x2). Phys. Rev. B 64, 045318 (2001). https://doi.org/10.1103/PhysRevB.64.045318
- [82]J. M. C. Thornton, D. A. Wolf, P. Weightman, Reconstructions of the GaAs(111)B Surface. Appl. Surf. Sci. 123/124, 115 (1998). https://doi.org/10.1016/S0169-4332(97)00433-9
- [83]H.H. Farell, J. Lu, B. D. Schultz, A. D. Denison, C. J. Palmerston, GaAs(111)B(√19 × √19) surface reconstruction. J. Vac. Sci. Technol. B 19, 1597 (2001). https://doi.org/10.1116/1.1387460
- [84]A. Proessdorf, F. Grosse, W. Braun, F. Katmis, H. Riechert, O. Romanyuk, Analysis of GaSb and AlSb reconstructions on GaSb(111) A- and B-oriented surfaces by azimuthal-scan reflection high-energy electron diffraction. Phys. Rev. B 83, 155317 (2011). https://doi.org/10.1103/PhysRevB.83.155317
- [85]D. C. Look, Molecular-beam epitaxial GaAs grown at low-temperatures, Thin Sol. Films 231, 61 (1993) https://doi.org/10.1016/0040-6090(93)90703-R
- [86]M. R. Melloch, J. M. Wodall, E. S. Harmon, N. Otsuka, F. H. Pollak, D. D. Nolte, R. M. Feenstra, M. A. Lutz, Low-Temperature-Growth III-V Materials, Ann. Rev. Mater. Res. 25, 547 (1995). https://doi.org/10.1146/annurev.ms.25.080195.002555
- [87]K. M. Yu, M. Kaminska, Z. Lilliental-Weber, Characterization of GaAs layers grown by low-temperature molecular beam epitaxy using ion-beam techniques, J. Appl. Phys. 72, 2850 (1992). https://doi.org/10.1063/1.351538
- [88]S. J. Eglash, H. K. Choi, G. W. Turner, MBE growth of GaInSAsSb/AlGaAsSb double heterostructures for infrared diode-lasers, J. Crystal Growth 11, 669 (1991) https://doi.org/10.1016/0022-0248(91)91061-E
- [89]Y. B. Li, Y. Zhang, Y.W. Zhang, B.Q. Wang, Z. Z. Zhu, Y. P. Zeng, Molecular beam epitaxial growth and characterization of GaSb layers on GaAs(001) substrates, Appl. Surf. Sci. 258, 6571 (2012). https://doi.org/10.1016/j.apsusc.2012.03.081
- [90]Y. Wang, Z. Pan, Y. Yan, Y. Yang, W.Zhao, N. Ding, X. Tang, P. Wu, Q. Zhao, Y. Li. A review of gallium phosphide nanophotonics towards omnipotent nonlinear devices, Nanophotonics, 13, 3207 (2024). https://doi.org/10.1515/nanoph-2024-0172
- [91]L. Töben, T. Hannapel, K. Moller, H.-J Crawack, C. Pettenkofer, F. Willig, RDS, LEED and STM of the P-rich and Ga-rich surfaces of GaP(100). Surf. Sci. 494, L755 (2001). https://doi.org/10.1016/S0039-6028(01)01492-3
- [92]P. Kleinschmidt, P. Mutombo, T. Berthold, A. Paszuk, M. Steidl, G. Ecke, A. Nägelein, C. Koppka, O. Supplie, S. Krischok, O. Romanyuk, M. Himmeliich, T. Hannapel, Atomic surface structure of MOVPE prepared GaP(111)B. Appl. Surf. Sci. 534, 147346 (2020). https://doi.org/10.1016/j.apsusc.2020.147346
- [93]T. Suzuki, T. Saga, T. Jimbo, M. Umeno, Growth mechanism of GaP on Si substrates by MOVPE, J. Cryst. Growth 115, 158 (1991). https://doi.org/10.1016/0022-0248(91)90731-J
- [94]V. Tassev, D. Bliss, C. Lynch, C. Yapp, W. Goodhue, K. Termkoa, Low-pressure, low-temperature growth of GaP for nonlinear optical frequency conversion devices. J. Cryst. Growth 312, 1146 (2010). https://doi.org/10.1016/j.jcrysgro.2009.12.053
- [95]S. Krukowski, I. Grzegory, M. Bockowski, B. Lucznik, T. Suski, G. Nowak, J. Borysiuk, M. Wroblewski, M. Leszczynski,, P. Perlin, S. Porowski, J. L. Weyher, Growth of AlN, GaN and InN from the solution, Int. J. Mater. Product Technol. 22, 226 (2005) http://dx.doi.org/10.1504/IJMPT.2005.005766
- [96]I. Grzegory, S. Krukowski, M. Leszczynski, P. Perlin, T. Suski, S. Porowski, The application of high pressure in physics and technology of II-V nitrides. Acta Phys. Pol. A 100, 57 (2001). http://dx.doi.org/10.12693/APhysPolA.100.57
- [97]J. Piechota, S. Krukowski, B. Sadovyi, P. Sadovyi, S. Porowski, I. Grzegory, Melting vs Decomposition of GaN: ab initio Molecular Dynamics Study and comparison to the experimental data. ACS. Chem. Mater. 35, 7694 (2023). https://doi.org/10.1021/acs.chemmater.3c01477
- [98]T. Taki, M. Strassburg, Visible LEDs: More than Efficient Light. ECS J. Solid State Sci. Technol. 9, 015017 (2020). https://www.doi.org/10.1149/2.0402001JSS
- [99]M. Kneissl, T. Y. Seong, J. Han, H. Amano, The emergence and prospects of deep-ultraviolet light-emitting diode technologies. Nature Photonics 13, 233-244 (2019). https://www.doi.org/10.1038/s41566-019-0359-9DOI
- [100]N. Islam, M. F. P. Mohamed, M. F. A. J. Khan, S. Falina, H. Kawarada, M. Syamsul, Reliability, Applications and Challenges of GaN HEMT Technology for Modern Power Devices: A Review. Crystals, 12, 1581 (2022). https://doi.org/10.3390/cryst12111581
- [101]S. Mohanty, K. Khan, E. Ahmadi, N-polar GaN: Epitaxy, properties and device applications, Prog. Quant. Electron. 87, 100450 (2023). https://doi.org/10.1016/j.pquantelec.2022.100450
- [102]K. Karimi, A. Fardoost, M. Javanmard, Comprehensive Review of FinFET Technology: History, Structure, Challenges, Innovations, and Emerging Sensing Applications, Micromachines, 15, 1187 (2024). https://doi.org/10.3390/mi15101187
- [103]W.K. Burton, N. Cabrera, F. C. Frank, The growth of crystals and the equilibrium structure of their surfaces. Phil. Trans. Roy. Soc. (London) A243, 299 (1951).DOI
- [104]S. Krukowski, K. Sakowski, P. Strak, P. Kempisty, J. Piechota, I. Grzegory, Macrostęp dynamics and the growth of crystals and epitaxial layers. Prog. Cryst. Growth Charact. Mater. 68, 100581 (2022). https://doi.org/10.1016/j.pcrysgrow.2022.100581
- [105]G. H. Gilmer, R. Ghez, N. Cabrera, An analysis of combined surface and volume diffusion in crystal growth, J. Cryst. Growth 8, 79 (1971). https://doi.org/10.1016/0022-0248(71)90027-3
- [106]J. P. van der Eerden, The advance velocity of steps under influence of volume and surface diffusion, by direct and indirect incorporation of growth units, J. Cryst. Growth 56, 174 (1982). https://doi.org/10.1016/0022-0248(82)90027-6


























