betaFeedback

WorksPhysics and Astronomy

Composition dependent k⋅p\mathbf{k}\cdot\mathbf{p} band parameters for wurtzite (Al,Ga)N alloys from density functional theory

Abstract

UV emitters based on the semiconductor alloy aluminium gallium nitride, (Al,Ga)N, have attracted significant interest in recent years due to their potential for optoelectronic devices. To guide the design of such devices with improved efficiencies, theoretical frameworks based on so-called k.p methods have found widespread application in the literature. Given that k.p models are empirical in nature, parameters such as effective masses or crystal field splitting energies of (Al,Ga)N alloys have to be provided as input from first-principles calculations or experiment. Although these parameters are available for GaN and AlN, detailed information on their composition dependence is sparse. Here, we address this question and provide (Al,Ga)N band parameters for widely used k.p Hamiltonians. We start from density functional theory (DFT) to sample the electronic structure of (Al,Ga)N alloys over the full composition range. The k.p parameters are treated as free parameters to reproduce the DFT data. For GaN and AlN the parameters extracted here agree well with literature values. When turning to the composition dependence of the k.p parameters, our calculations show that most parameters deviate significantly from a linear interpolation of the GaN and AlN values, an approximation widely made in the literature. Moreover, to describe changes in the band parameters with Al content, composition dependent bowing parameters have to be considered for an accurate description of the DFT data. Finally, our analysis also provides initial insight into consequences of the nonlinear composition dependence of the k.p parameters for the electronic structure of (Al,Ga)N alloys. We find that in particular the band ordering is affected by the nonlinear evolution of the crystal field splitting energy with composition, an important aspect for the light polarization characteristics of high Al content (Al,Ga)N alloys.

Introduction

Optoelectronic devices operating in the ultraviolet (UV) spectral region are of interest for a variety of different applications, including biomedicine and disease diagnostics [1, 2, 3]. Furthermore, UV sterilization and disinfection [4, 5] received strong attention in fighting the coronavirus and its variants [6]. Current UV light emission relies to a large extent on bulky, inefficient and wavelength non-tunable gas lasers [7] or lamps utilizing toxic mercury [8]. In recent years, the semiconductor alloy aluminium gallium nitride, (Al,Ga)N, has been targeted for replacing these inefficient and environmental unfriendly devices as the direct (Al,Ga)N band gaps cover, in principle, the wavelength range from UV-A (315-400 nm) to UV-B (280-315) into the UV-C (200-280 nm) [9].

Although (Al,Ga)N-based UV light emitters present an attractive option for solid-state-based solutions, the efficiency of these devices is currently low, especially in the deep UV (UV-C) spectral range [10, 11]. Several aspects contribute to these low efficiencies, including low light extraction efficiencies (LEEs) and p-doping of (Al,Ga)N layers with high Al contents [9, 10]. In general, the low LEE is tightly linked to fundamental differences in the valence band structure of AlN and GaN systems [12]: The valence band edge (VBE) in AlN is of Γ7\Gamma_7 symmetry while in GaN it is of Γ9\Gamma_9 symmetry [13, 14]. Optical transitions involving the conduction band edge (CBE) and VBE result in transverse magnetic (TM) polarized light in AlN while in GaN the emitted light is of transverse electric (TE) polarization [15, 16]. Given that (Al,Ga)N based LEDs are mainly top or bottom emitting devices, TM polarized emission leads to a low LEE [17]. Moreover, in (Al,Ga)N QWs, which form the active region of an LED, several further factors impact the valence band ordering, including strain and carrier/quantum confinement, but also the magnitude and sign of the so-called crystal field splitting (CF) energy, Δcf\Delta_{\mathrm{cf}} [12]. For AlN, Δcf\Delta_{\mathrm{cf}} is negative (in the range −160 to −295 meV), while in GaN it is positive and significantly smaller in magnitude (< 40 meV) [18, 19, 20, 21, 22]. Understanding how Δcf\Delta_{\mathrm{cf}} changes with alloy content plays an important role in designing (Al,Ga)N-based light emitters, as it will affect the band ordering and thus polarization characteristics of the emitted light. However, as mentioned above, not only is the crystal field splitting energy important but also quantum confinement effects will impact the valence band ordering in (Al,Ga)N heterostructures. Therefore, the effective electron and hole masses of the alloy along, for instance, the growth direction of a quantum well are key.

Device-level modeling of, for example, (Al,Ga)N-based LEDs largely relies on self-consistent Schrödinger-Poisson solvers that are coupled with drift-diffusion simulations [23, 24]. For the wide-band gap system (Al, Ga)N, the Schrödinger equation is widely treated in the framework of 6+2 band k⋅pk \cdot p models [25, 26, 27]. Such an approach requires, e.g. effective masses as well as crystal field splitting energies as input. Parameters for binary AlN and GaN systems have been extracted from experiment but more often from first-principles calculations. However, information on the composition dependence of these parameters in an (Al,Ga)N alloy is sparse. Often a linear interpolation of the binary parameters with the Al content xx is employed in an Alx_xGa1−x_{1-x}N alloy [28, 29, 30, 31]. On the other hand, several studies have already indicated that, for example, the crystal field splitting energy exhibits deviations from such linear composition behavior [14, 32]. Building on this, it is unclear whether a linear interpolation of effective mass parameters is justified.

In this work, we provide information on the composition dependence of Alx_xGa1−x_{1-x}N band parameters. More specifically, we report hole and electron effective mass parameters, which are required for 6+2 band k⋅p\mathbf{k}\cdot\mathbf{p} Hamiltonians that underlie widely available device simulation packages. We determine Luttinger-like Ai(x)A_i(x) with i=1,…,6i=1,\ldots,6 parameters and effective electron mass me(x)m_e(x) as a function of the AlN content xx over the entire range of composition. In addition, we also study the composition dependence of the crystal field splitting energy Δcf(x)\Delta_{\mathrm{cf}}(x). To do so, we use density functional theory (DFT) to obtain effective electronic band structures of Alx_xGa1−x_{1-x}N alloys which are then fitted with our 6+2 band k⋅p\mathbf{k}\cdot\mathbf{p} Hamiltonian. Our procedure relies on a Sobol sequence that recently has been used to obtain k⋅p\mathbf{k}\cdot\mathbf{p} parameters for Hamiltonians with larger basis sets, e.g. a 16 band Hamiltonian, from DFT band structures [33]. In general, we follow the procedure established by Rinke et al. [22] to extract k⋅p\mathbf{k}\cdot\mathbf{p} parameters. Our determined k⋅p\mathbf{k}\cdot\mathbf{p} parameters for AlN and GaN agree well with previously published values [21, 22]. Moreover, our calculations further support that the crystal-field splitting energy deviates noticeably from a linear interpolation of the binary material values. In addition, we find that the effective mass parameters deviate significantly from linearly interpolated values and can be strongly composition dependent. Finally, when using the extracted composition dependent Ai(x)A_i(x) and Δcf(x)\Delta_{\mathrm{cf}}(x) parameters for band structure calculations, we observe noticeable deviations in ordering and orbital character of the valence bands when compared to band structures predicted from linearly interpolated Ai(x)A_i(x) and Δcf(x)\Delta_{\mathrm{cf}}(x) values. The difference in band ordering can be of strong importance when investigating light polarization characteristics of (Al,Ga)N-based devices. Thus, our extracted parameters present an ideal starting point for device-level simulations to further understand the impact of the material parameters for light-emitting devices.

The manuscript is organized as follows. We first describe the DFT approach underlying our calculations of (Al,Ga)N band structures in Sec. II A. The k⋅p\mathbf{k}\cdot\mathbf{p} model and the connected fitting procedure are discussed in Sec. II B. In Sec. III, we present the results of our study, namely the composition dependent me(x)m_e(x), Ai(x)A_i(x), and Δcf(x)\Delta_{\mathrm{cf}}(x) parameters for Alx_xGa1−x_{1-x}N alloys. Moreover, we examine the impact of the obtained parameters on electronic bulk band structures of (Al,Ga)N alloys.

Theory

In this section we outline the ingredients of the theoretical framework used to extract composition-dependent k⋅p\mathbf{k}\cdot\mathbf{p} parameters for (Al,Ga)N alloys. We start with details of the DFT simulations, which form the basis of our k⋅p\mathbf{k}\cdot\mathbf{p} fitting procedure. In Sec. II B, details of the k⋅p\mathbf{k}\cdot\mathbf{p} model are given before turning to the fitting approach.

Density functional theory calculations

The DFT calculations carried out in this work are based on the plane-wave-based Vienna ab initio Simulation Package (VASP) version 5.4.4 [34]. We employ the generalized gradient approximation (GGA), as implemented in the Perdew–Burke–Ernzerhof (PBE) exchange-correlation functional, using the projector augmented wave (PAW) scheme. The semi-core dd electrons of Ga are treated as valence electrons and the plane wave cut-off energy is 600 eV. The convergence criteria were 1×10−41\times10^{-4} eV for the energy minimisation and 1×10−31\times10^{-3} eV/Å for the forces in the ionic relaxations. DFT calculations have been performed for unstrained bulk systems, thus, we do not consider effects from an underlying substrate. Structural relaxation was achieved by minimizing the pressure on the supercell. As spin-orbit coupling (SOC) energies are small in nitrides when compared to other III–V materials (e.g. ΔsoGaN=17\Delta_{\mathrm{so}}^{\mathrm{GaN}}=17 meV and ΔsoGaAs=341\Delta_{\mathrm{so}}^{\mathrm{GaAs}}=341 meV) [20, 21], we neglect SOC effects as it significantly increases the computational load while introducing only small changes to the band structure. This is a widely used approach in DFT calculations for III-N materials [35, 36, 37, 38, 39].

The above outlined DFT framework has been chosen for several reasons for our k⋅p\mathbf{k}\cdot\mathbf{p} parameter extraction. Firstly, even though the PBE-GGA functional underestimates the band gap of GaN and AlN, previous literature studies show that effective electron masses obtained from this DFT setup agree well with data extracted from functionals that predict band gaps closer to experiment [40, 41]. Secondly, in comparison to state-of-the-art Heyd, Scuseria and Ernzerhof (HSE) hybrid functionals [42, 43], the PBE-GGA functional is computationally less demanding. Therefore, it allows for a numerically efficient way to screen larger numbers of alloy configurations (see below) and Al contents xx to determine k⋅p\mathbf{k}\cdot\mathbf{p} parameters for Alx_xGa1−x_{1-x}N alloys. Additionally, in case of HSE-DFT, the exact exchange mixing parameter, α\alpha, has to be determined and specified. Previous studies have shown that to obtain a good description of the bulk band gaps of GaN and AlN, the required α\alpha values for the two materials, GaN (α=0.294\alpha=0.294) and AlN (α=0.337\alpha= 0.337), are (very) different [44]. Thus, for an alloy, there is uncertainty on how to determine α\alpha. In contrast, in the PBE-GGA functional, there is no free parameter and therefore no further input is required for (Al,Ga)N alloys. Finally, and as we will show below, the valence band effective mass parameters extracted from our PBE-GGA calculations are in good agreement with literature values on GaN and AlN binary materials. Therefore, in terms of effective mass parameters, the PBE-GGA functional should provide a reliable framework over the full composition range in Alx_xGa1−x_{1-x}N alloys.

For the binary GaN and AlN systems a 4 atom primitive unit cell with a Γ\Gamma centered 6×6×46 \times6 \times4 kk-point mesh has been used. To treat Alx_xGa1−x_{1-x}N alloys we have employed a 72 atom supercell and considered Al contents of x=0.11,0.25,0.50,0.75x = 0.11, 0.25, 0.50, 0.75 and 0.890.89. In the supercell case, the Γ\Gamma centered kk-point mesh is 3×3×23 \times3 \times2. Starting from experimental findings [45, 46, 47], we treat (Al,Ga)N as a random alloy, thus Al and Ga atoms are randomly distributed in the supercell. To account for the impact of the alloy microstructure, for each alloy content xx, the electronic structure of 10 random alloy configurations has been investigated.

The discussed DFT framework can now provide electronic band structures for extracting k⋅pk \cdot p parameters. Details of the fitting procedure are given in the following section, where we first introduce the k⋅pk \cdot p model employed.

kp model and band structure fitting approach

Following the study by Rinke et al. [22], our k⋅pk \cdot p Hamiltonian is expanded in the basis states ∣S↑⟩|S \uparrow\rangle, ∣X↑⟩|X \uparrow\rangle, ∣Y↑⟩|Y \uparrow\rangle, ∣Z↑⟩|Z \uparrow\rangle, ∣S↓⟩|S \downarrow\rangle, ∣X↓⟩|X \downarrow\rangle, ∣Y↓⟩|Y \downarrow\rangle, ∣Z↓⟩|Z \downarrow\rangle. The explicit form of the Hamiltonian is given in Appendix A. Although in general the chosen basis states lead to a so-called 8-band model, thus accounting for the coupling between ss-like conduction and px,y,zp_{x,y,z}-like valence bands, device simulations of (Al,Ga)N heterostructures often treat valence and conduction bands separately [48, 49, 50]. This approximation is, in general, justified by the large band gap of the alloy (>3.5> 3.5 eV). Decoupling conduction and valence bands helps reducing the computational burden when connecting self-consistent Schrödinger-Poisson solvers with drift-diffusion models for device simulations. Bearing this in mind, we consider in the following a 2+62+6 band k⋅pk \cdot p model. However, it should be noted that the DFT band structures used for k⋅pk \cdot p parameter extraction account intrinsically for conduction and valence band coupling effects. Therefore, even though the fitting is carried out with a 2+62+6 band k⋅pk \cdot p model, coupling effects are implicitly built into the parameters obtained.

When it comes to fitting k⋅pk \cdot p to DFT band structures, we treat the effective electron mass, mem_{\mathrm{e}}, valence band effective mass parameters, AiA_i with i=1,…,6i = 1, \ldots, 6, and crystal field splitting energy, Δcf\Delta_{\mathrm{cf}}, as free parameters. Here, it should be noted that in an (Al,Ga)N alloy, disorder will also lead to band mixing effects [51, 52]. On the one hand, it is important to capture these effects in the fitting procedure, as this can affect the polarization of the emitted light from (Al,Ga)N systems, and potentially the LEE of a device [52]. On the other hand, no extra terms can be added to the k⋅pk \cdot p Hamiltonian discussed above and in Appendix A. This restriction is introduced to ensure that our model parameters are directly applicable to widely available (commercial) software packages, which, in general, only allow defining input parameters for the Hamiltonian but not a “new” Hamiltonian. To account for this constraint, but also to capture alloy disorder-induced band mixing effects and lifting of degeneracies, we proceed as follows. In general, SOC introduces band mixing effects but also a lifting of degeneracies in bulk k⋅pk \cdot p band structures. Therefore, we use this parameter to mimic the effect of alloy disorder on the band structure of (Al,Ga)N alloys. For example, at k=0k = 0 the energies of the three highest lying valence bands states in a semiconductor with a wurtzite crystal structure can be written as follows [53]:

EVB1=Ev+Δcf+Δsp3,(1)E_{\mathrm{VB1}} = E_{\mathrm{v}} + \Delta_{\mathrm{cf}} + \frac{\Delta_{\mathrm{sp}}}{3}, \tag*{(1)}
EVB2=Ev+3Δcf−Δsp6+(3Δcf−Δsp6)2+2(Δsp3)2,(2)E_{\mathrm{VB2}} = E_{\mathrm{v}} + \frac{3\Delta_{\mathrm{cf}} - \Delta_{\mathrm{sp}}}{6} + \sqrt{\left(\frac{3\Delta_{\mathrm{cf}}-\Delta_{\mathrm{sp}}}{6}\right)^2 + 2\left(\frac{\Delta_{\mathrm{sp}}}{3}\right)^2}, \tag*{(2)}
EVB3=Ev+3Δcf−Δsp6−(3Δcf−Δsp6)2+2(Δsp3)2.(3)E_{\mathrm{VB3}} = E_{\mathrm{v}} + \frac{3\Delta_{\mathrm{cf}} - \Delta_{\mathrm{sp}}}{6} - \sqrt{\left(\frac{3\Delta_{\mathrm{cf}}-\Delta_{\mathrm{sp}}}{6}\right)^2 + 2\left(\frac{\Delta_{\mathrm{sp}}}{3}\right)^2}. \tag*{(3)}

Compared to Ref. [53], we assumed that Δsp=Δ2=Δ3\Delta_{\mathrm{sp}} = \Delta_2 = \Delta_3, which describes the SOC energy. Moreover, our Δcf\Delta_{\mathrm{cf}} corresponds to Δ1\Delta_1 in Ref. [53] and is the crystal field splitting energy. In the absence of SOC and alloy disorder, two of the three valence bands (neglecting spin) are degenerate EVB1=EVB2≠EVB3E_{\mathrm{VB1}} = E_{\mathrm{VB2}} \ne E_{\mathrm{VB3}}. In this case, the band-ordering depends on the sign of Δcf\Delta_{\mathrm{cf}}. For our purpose, Δsp\Delta_{\mathrm{sp}} and Δcf\Delta_{\mathrm{cf}} are introduced as free parameters to account for alloy disorder induced valence band splitting effects in (Al,Ga)N alloys, especially at the Γ\Gamma-point. Thus, in conjunction with DFT data on the energetic separation of the highest lying three valence bands at the Γ\Gamma-point, Eqs. (1)–(3) are used to obtain an initial guess for Δsp\Delta_{\mathrm{sp}} and Δcf\Delta_{\mathrm{cf}} when targeting data away from the Γ\Gamma-point.

In terms of the numerical routine used to fit k⋅pk \cdot p to DFT band structures and thus to extract the k⋅pk \cdot p parameters outlined above, we follow Marquardt et al. [33]. Here, we use a Monte Carlo fitting scheme [54] based on a Sobol sequence method [55, 56]. Sobol sequences have been shown to yield superior results as compared to several other low-discrepancy point fitting techniques and as a natural substitute to usual random sequences; further details on the employed approach can be found in Refs. [33, 57]. The k⋅pk \cdot p Hamiltonian and the Monte Carlo fitting scheme are implemented in the flexible, generalized multi-band k⋅pk \cdot p formalism within the plane-wave framework of the SPHInX library [57, 58, 59].

When it comes to selecting DFT data for k⋅pk \cdot p fitting, we follow largely the work by Rinke et al. [22]. Thus, for our k⋅pk \cdot p parameter extraction we chose the directions Σ\Sigma, Λ\Lambda, TT, and Δ\Delta in the first Brillouin zone. More specifically, we select a) Γ→K20\Gamma\to\frac{K}{20} (Σ\Sigma path), b) Γ→L20\Gamma\to\frac{L}{20} (Λ\Lambda path), c) Γ→K20\Gamma\to\frac{K}{20} (TT path) and d) Γ→A10\Gamma\to\frac{A}{10} (Δ\Delta path). Given the underlying assumptions of the k⋅pk \cdot p method used here, i.e. perturbation theoretical approach at the Γ\Gamma-point, our fitting procedure focuses on the region close to the Γ\Gamma-point. For the k⋅pk \cdot p parameter extraction, equal path lengths have been covered. Furthermore, a high density of DFT kk-points has been chosen along the different directions for accurate parameter extraction (≥30\ge30 points). To determine the k⋅pk \cdot p parameters for AlN and GaN, the parameter sets of Vurgaftman and Meyer [21] have been used as an initial guess. To reduce the likelihood that the fitting procedure converges to a local minimum, the k⋅pk \cdot p parameters are allowed to be varied by plus minus the magnitude of the respective initial guess value. We have tested the impact of both changing initial guess values as well as the range over which the parameters are allowed to be varied on the resulting final parameter set. In general, both factors (initial guess and parameter range) had a negligible impact on the final k⋅pk \cdot p parameter set. For (Al,Ga)N alloys we generally employ the same procedure, except that the initial guess for our alloy k⋅pk \cdot p parameters is obtained as a linear interpolation of our GaN and AlN values.

Results

In this section we present (i) results of the above outlined parameter extraction procedure and (ii) consequences of the obtained composition dependent k⋅pk \cdot p parameters for the valence band structure of (Al,Ga)N alloys. In Sec. III A we summarise our k⋅pk \cdot p parameters for GaN and AlN systems and compare the obtained data with available literature values. The k⋅pk \cdot p parameters extracted for (Al,Ga)N alloys are discussed in Sec. III B. Finally, in Sec. III C we compare electronic band structures of (Al,Ga)N alloys calculated using our composition dependent k⋅pk \cdot p parameters with band structures obtained when employing a linear interpolation of the values for GaN and AlN found in Sec. III A.

GaN and AlN parameters

Figure 1 displays the DFT band structure together with the fitted k⋅pk \cdot p band structure for two of the four considered directions in the first Brillouin zone for (a) GaN and (b) AlN. The very good agreement between k⋅pk \cdot p and the DFT band structure highlights that both GaN and AlN are accurately described by our k⋅pk \cdot p model in

the chosen kk range. The extracted valence band k⋅pk \cdot p parameters AiA_i, electron effective mass values parallel, me∥m_e^{\parallel}, and perpendicular, me⊥m_e^{\perp}, to the cc-axis, as well as Table I. Band parameters for AlN, GaN and (Al,Ga)N alloys when accounting for bowing parameters bb, α\alpha and β\beta. Definitions of these parameters are given in the main text. Effective electron mass values parallel and perpendicular to the cc-axis have been denoted by superscripts ∥\parallel and ⊥\perp, respectively. Literature values are from Rinke et al. [22] and Vurgaftman and Meyer [21] are given for comparison. Percentage differences between values obtained in the present study and literature data are given in brackets.

Band ParametersMaterials
GaNAlNbbα\alphaβ\beta
PresentRef. [22]Ref. [21]PresentRef. [22]Ref. [21]
A1A_1-7.438-5.947 (20.0%)-7.21 (3.1%)-4.101-3.991 (2.7%)-3.86 (5.9%)--5.483.46
A2A_2-0.485-0.528 (8.9%)-0.44 (9.3%)-0.211-0.311 (47.4%)-0.25 (18.5%)--1.076.35
A3A_36.9655.414 (22.3%)6.68 (4.1%)3.8223.671 (4.0%)3.58 (6.3%)-5.465.00
A4A_4-2.963-2.512 (15.2%)-3.46 (16.8%)-1.487-1.147 (22.9%)-1.32 (11.2%)--0.29-0.89
A5A_5-2.998-2.510 (16.3%)-3.40 (13.4%)-1.632-1.329 (18.6%)-1.47 (9.9%)--2.400.33
A6A_6-4.352-3.202 (26.4%)-4.90 (12.6%)-2.11-1.952 (7.5%)-1.64 (22.3%)--3.840.88
Δcf\Delta_{\mathrm{cf}} (eV)0.0400.034 (15.0%)0.010 (75.0%)-0.177-0.295 (66.7%)-0.169 (4.52%)--0.19 (eV)-0.22
Δsp\Delta_{\mathrm{sp}} (eV)--------0.07 (eV)-0.91
me∥m_e^{\parallel} (m0m_0)0.1660.209 (25.9%)0.20 (20.0%)0.300.329 (9.7%)0.30 (0.0%)-0.060--
me⊥m_e^{\perp} (m0m_0)0.1500.186 (24.0%)0.20 (33.3%)0.2870.322 (12.2%)0.32 (11.5%)-0.052--

Table 1.

the crystal field splitting energy Δcf\Delta_{\mathrm{cf}} are summarized in Table I. To compare our data with existing literature data, Figure 1 also displays k⋅p\mathbf{k}\cdot\mathbf{p} band structures when using the AiA_i and mem_e parameter sets from Vurgaftman and Meyer [21] as well as Rinke et al. [22]. To focus on the band dispersion, the band structures are always plotted with respect to CBE and VBE. Moreover, the crystal field splitting energies from our calculations have been used for all parameter sets. Our DFT calculations give ΔcfAlN=−177\Delta_{\mathrm{cf}}^{\mathrm{AlN}}=-177 meV for AlN and ΔcfGaN=40\Delta_{\mathrm{cf}}^{\mathrm{GaN}}=40 meV for GaN. The obtained AlN value is in good agreement with the value reported by Vurgaftman and Meyer [21] (ΔcfAlN,VM=−169\Delta_{\mathrm{cf}}^{\mathrm{AlN,VM}}=-169 meV) but magnitude-wise it is noticeably smaller when compared to the data reported by Rinke et al. [22] (ΔcfAlN,Rinke=−295\Delta_{\mathrm{cf}}^{\mathrm{AlN,Rinke}}=-295 meV) [22]. However, for GaN, our data on the crystal splitting energy agree well with the value reported by Rinke et al. ($\Delta_{\mathrm{cf}}^{\mathrm{GaN,Rinke}}=34$ meV) [22]. Vurgaftman and Meyer [21] suggest a smaller value (ΔcfGaN,VM=10\Delta_{\mathrm{cf}}^{\mathrm{GaN,VM}}=10 meV). In general, one should note that the work in Ref. [21] is a summary and average of parameters reported in the literature.

Image not extracted

Figure 1. Band structure of wurtzite (a) GaN and (b) AlN in the vicinity of the Γ\Gamma point. Our PBE DFT data are presented by filled black circles and the fitted k⋅pk \cdot p band structures by the solid black line. Band structures using the k⋅pk \cdot p parameter sets from Rinke et al. [22] and Vurgaftman and Meyer (VM) [21] are given by dashed green lines and red open diamonds, respectively. To focus on differences in the dispersion, the band structures are plotted with respect to the conduction and valence band edge, respectively. Moreover, the crystal field splitting energy of our PBE-DFT calculations are used for all k⋅pk \cdot p calculations.

Looking at the energy dispersions of AlN, Figure 1 (b) we find that over the k\mathbf{k}-point range considered, all three parameter sets give band structures in very good agreement with each other. The different k⋅p\mathbf{k}\cdot\mathbf{p} parameter sets are summarized and compared in Table I. For GaN, Figure 1 (a), we observe some deviations between three the k⋅p\mathbf{k}\cdot\mathbf{p} parameter sets, however, very close to the Γ\Gamma-point the band structures do not differ significantly. In general, numbers may also vary slightly depending on the lengths of the path chosen for the fitting procedure. Given the general good agreement between our data and the literature values for GaN and AlN, in the next section we extract k⋅p\mathbf{k}\cdot\mathbf{p} parameters for (Al,Ga)N alloys.

(Al,Ga)N alloy parameter

In this section we focus on the composition dependence of the k⋅p\mathbf{k}\cdot\mathbf{p} parameters of Alx_xGa1−x_{1-x}N alloys. We start by investigating the effective electron mass, both me∥(x)m_e^{\parallel}(x) and me⊥(x)m_e^{\perp}(x), of (Al,Ga)N alloys. Subsequently, we study the valence band effective mass parameters Ai(x)A_i(x). As discussed in Sec. II A, we use 10 different random alloy configurations per Al content xx in the considered AlxGa1−xN\mathrm{Al}_{x}\mathrm{Ga}_{1-x}\mathrm{N} alloys and fit each individual band structure with the k⋅p\mathbf{k}\cdot\mathbf{p} model. The final parameter set is then obtained as an average over the 10 different configurations. Figure 2 displays an example of our k⋅p\mathbf{k}\cdot\mathbf{p} fit to an effective Al0.75Ga0.25N\mathrm{Al}_{0.75}\mathrm{Ga}_{0.25}\mathrm{N} DFT band structure (arbitrarily selected alloy configuration). As Figure 2 shows, the k⋅p\mathbf{k}\cdot\mathbf{p} band structure is in excellent agreement with the DFT data over the k\mathbf{k}-point range considered. This highlights that (i) the k⋅p\mathbf{k}\cdot\mathbf{p} model is suitable for describing the alloy band structure and (ii) that the SOC-like term in our k⋅p\mathbf{k}\cdot\mathbf{p} Hamiltonian captures the alloy-induced lifting of degeneracies in the band structure well. We note that alloy disorder breaks the translational symmetry of the crystal, which results in the situation that Bloch’s theorem is in principle no longer applicable. Thus, defining a band structure is not strictly valid. However, k⋅p\mathbf{k}\cdot\mathbf{p} descriptions of alloys assume that perturbations of disorder are weak and that an effective band structure can still be assumed. To bridge the gap between disorder and effective band structures, we have constructed supercells that are large enough to

Image not extracted

Figure 2. (a) Conduction and (b) valence band structure of an Al0.75_{0.75}Ga0.25_{0.25}N random alloy in the vicinity of the Γ\Gamma point. The black dotted lines denote our PBE-DFT results while the open green diamonds show the fitted k⋅p\mathbf{k}\cdot\mathbf{p} band structures.

capture disorder effects on an atomistic level but small enough so that electrons with low k\mathbf{k}-vectors move in the periodic potential due to the periodic boundary conditions of the plane-wave DFT approach used. In doing so, effects of alloy disorder are included in a limit that is consistent with the underlying assumptions of a k⋅p\mathbf{k}\cdot\mathbf{p} model.

Effective electron masses

To gain insight into the composition dependence of the effective electron mass in (Al,Ga)N alloys, we follow here previous literature, and distinguish between effective electron masses perpendicular, me⊥(x)m_e^{\perp}(x), and parallel, me∥(x)m_e^{\parallel}(x), to the cc-axis. The effective electron masses are extracted according to the above outlined fitting procedure considering the in-plane directions Σ\Sigma, Λ\Lambda, TT, and the out-of-plane direction Δ\Delta in the first Brillouin zone. The obtained masses are displayed in Figure 3 as a function of the Al content xx in AlxGa1−xN\mathrm{Al}_{x}\mathrm{Ga}_{1-x}\mathrm{N} alloys. In Figure 3 (a) the mass me⊥m_e^{\perp} perpendicular to the cc-axis is shown while (b) displays the mass me∥m_e^{\parallel} parallel to the cc-axis. To evaluate the impact of alloy content xx on the effective electron mass, Figure 3 also shows the effective mass values for the two directions obtained by a linear interpolation of the effective mass values of GaN and AlN via:

Image not extracted

Figure 3. Conduction band effective masses of (Al,Ga)N alloys for (a) perpendicular, me⊥m_e^{\perp}, and (b) parallel, me∥m_e^{\parallel}, to the wurtzite cc-axis.

me,AlGaNα=xme,AlNα+(1−x)me,GaNα,(4)m_{e,\mathrm{AlGaN}}^{\alpha}=x m_{e,\mathrm{AlN}}^{\alpha}+(1-x)m_{e,\mathrm{GaN}}^{\alpha}, \tag*{(4)}

where α={⊥,∥}\alpha=\{\perp,\parallel\} denotes the directions. The corresponding data are given in Figure 3 by the blue line (open blue diamonds). We note that different approaches have been used in the literature to estimate the effective electron mass of an alloy [60, 61, 62]. However, to compare our results with recent literature data by Balestra et al. [40], we use Eq. (4) as their DFT data indicate a Vegard law-type behavior for the composition dependence of the effective electron masses in (Al,Ga)N. From Figure 3 we infer that both for me⊥m_e^{\perp} and me∥m_e^{\parallel} our data deviate from a linear interpolation of the effective electron masses via Eq. (4). It should be noted that our extracted electron masses do not differ significantly when extracted from the 10 different microscopic configurations at each Al content xx (the largest deviation is 1.65% from the average value). The differences between our results and Ref. [40] may be due to several factors. Although in general the approach used in Ref. [40] is similar to ours, which means that it also utilizes PBE-DFT, the region over which the band structure is fitted to extract the effective electron masses may be different. Furthermore, the supercell employed in Ref. [40] is smaller (32-atoms) when compared to ours (72 atoms). Further studies would be required elucidate on the differences between our work and Ref. [40].

To provide an improved description of the composition dependence of the effective electron masses but also to extrapolate to composition values not explicitly studied here, we adapt the widely used expression for the band gap evolution [63, 64] of semiconductor alloys to effective electron masses and introduce a bowing parameter:

me,AlGaNα=xme,AlNα+(1−x)me,GaNα−b(1−x)x.(5)m_{e,\mathrm{AlGaN}}^\alpha= x m_{e,\mathrm{AlN}}^\alpha+ (1-x)m_{e,\mathrm{GaN}}^\alpha- b(1-x)x . \tag*{(5)}

Figure 3 reveals that me⊥(x)m_e^\perp(x) and me∥(x)m_e^\parallel(x) are indeed well described by Eq. (5), thus when introducing the bowing parameter bb. The effective electron mass bowing parameters obtained are summarized in Table I.

Valence band kp parameters

Having discussed the composition dependence of the electron effective mass in (Al,Ga)N alloys, we turn now and focus on the composition dependence of the valence band effective mass parameters AiA_i with i=1…6i=1\ldots6, crystal field splitting energy, Δcf\Delta_{\mathrm{cf}}, and the alloy disorder parameter, Δsp\Delta_{\mathrm{sp}}. We start the analysis with the parameters Ai(x)A_i(x), which are shown in Figure 4 as a function of the Al content xx in AlxGa1−xN\mathrm{Al}_x\mathrm{Ga}_{1-x}\mathrm{N} alloys. The black open circles give the average values of Ai(x)A_i(x), including error bars, from the fitting to PBE-DFT band structures. Again, to visualize the impact of alloy disorder on Ai(x)A_i(x), Figure 4 also shows Ai(x)A_i(x) parameters obtained (blue open diamonds) when employing the often used linear interpolation (Vegard’s law) of the GaN and AlN parameters with composition xx [28, 29, 30, 31]. Except for A4(x)A_4(x), the data obtained from a linear interpolation noticeably deviates from the PBE-DFT values.

Figure 4. Luttinger-like parameter Ai(x)A_i(x) with i=1,…,6i = 1,\ldots,6 as a function of the alloy content xx of AlxGa1−xN\mathrm{Al}_x\mathrm{Ga}_{1-x}\mathrm{N} alloys. The black circles represent the average value of AiA_i parameters over 10 random alloy configurations per xx; the blue diamonds denote the values obtained from a linear interpolation of the AlN and GaN AiA_i values. The solid green lines are obtained using Eq. (7); see main text for further details.

Moreover, when trying to describe the composition dependence of the Ai(x)A_i(x) parameters similar to the effective electron mass values via a composition independent bowing parameter:

AiAlGaN=xAiAlN+(1−x)AiGaN−biA(1−x)x(6)A_i^{\mathrm{AlGaN}} = xA_i^{\mathrm{AlN}} + (1-x)A_i^{\mathrm{GaN}} - b_i^{A}(1-x)x \tag*{(6)}

the PBE-DFT is poorly approximated. An example for such a poor fitting to the PBE-DFT data via Eq. (6) is given in Figure 5 for A1(x)A_1(x) as an example. A similar behavior is observed for the other Ai(x)A_i(x) parameters (not shown here). In general, this finding suggests that the bowing parameter itself is composition dependent. A similar behavior has been observed for the band gap bowing in (In,Ga)N [65, 66] but in particular in (Al,In)N alloys [67, 68, 69]. In case of the band gap bowing of (Al,In)N, Sakalauskas et al. [67] introduced an empirical formulae to account for the composition dependence of the bowing parameter. We have adopted this approach here for the valence effective mass parameters Ai(x)A_i(x):

Image not extracted

Figure 5. Luttinger-like parameter A1(x)A_1(x) as a function of the Al content xx of AlxGa1−xN\mathrm{Al}_x\mathrm{Ga}_{1-x}\mathrm{N} alloys. The black circles represent the average value of A1A_1 over 10 random alloy configurations per xx while the blue diamonds denote the values obtained from a linear interpolation of the AlN and GaN A1A_1 values. The solid orange line is obtained by fitting the data with a composition independent bowing parameter, Eq. (6). See main text for further details.

AiAlGaN=xAiAlN+(1−x)AiGaN−αi1+βix2(1−x)x=xAiAlN+(1−x)AiGaN−b~iA(x)(1−x)x.(7)\begin{aligned} A_i^{\mathrm{AlGaN}} &= xA_i^{\mathrm{AlN}} + (1-x)A_i^{\mathrm{GaN}} - \frac{\alpha_i}{1+\beta_i x^2}(1-x)x \\ &= xA_i^{\mathrm{AlN}} + (1-x)A_i^{\mathrm{GaN}} - \widetilde{b}_i^{A}(x)(1-x)x . \tag*{(7)} \end{aligned}

Here, αi\alpha_i and βi\beta_i are free parameters which are used to fit the averaged PBE-DFT data for each Ai(x)A_i(x). Figure 4 shows indeed that Eq. (7) allows to capture the composition dependence of the AiA_i parameters over the full composition range. Moreover, Eq. (7) enables us also to predict AiA_i parameters at compositions not explicitly targeted in the PBE-DFT calculations. The obtained αi\alpha_i and βi\beta_i values for the different AiA_i parameters are summarized in Table I.

Having presented the results for the effective mass parameters Ai(x)A_i(x), we turn in a second step to the composition dependence of the crystal field splitting energy, Δcf\Delta_{\mathrm{cf}}, and the alloy disorder parameter, Δsp\Delta_{\mathrm{sp}}. Figure 6 (a) shows Δcf(x)\Delta_{\mathrm{cf}}(x) as a function of the Al content xx in AlxGa1−xN\mathrm{Al}_x\mathrm{Ga}_{1-x}\mathrm{N} while Figure 6 (b) depicts the data for Δsp(x)\Delta_{\mathrm{sp}}(x). Looking at Figure 6 (a), one observes that the Δcf(x)\Delta_{\mathrm{cf}}(x) values extracted from PBE-DFT (black open circles) noticeably deviate from a linear interpolation (blue open diamonds) of the GaN and AlN crystal field splitting values. The evolution of Δcf\Delta_{\mathrm{cf}} with xx is again well described by a composition-dependent bowing parameter of the form bcf(x)=αcf/(1+βcfx2)b_{\mathrm{cf}}(x)=\alpha_{\mathrm{cf}}/(1+\beta_{\mathrm{cf}}x^2), as used for the AiA_i parameters in Eq. (7). The extracted values αcf\alpha_{\mathrm{cf}} and βcf\beta_{\mathrm{cf}} are given in Table I. As discussed above, the sign of Δcf\Delta_{\mathrm{cf}} plays an important role for the polarization characteristics of light emitted from (Al,Ga)N alloys. Assuming that carriers only occupy the highest valence band and in the absence of band mixing effects, a positive (negative) crystal field splitting energy Δcf(x)\Delta_{\mathrm{cf}}(x) suggests TE (TM) polarized light emission. Given the non-linear behavior of Δcf(x)\Delta_{\mathrm{cf}}(x) with composition xx, Figure 6 (a) indicates that the onset of a polarization switching from TE to TM occurs at higher Al contents xx when compared to a linear interpolation of Δcf(x)\Delta_{\mathrm{cf}}(x) with composition xx. The dashed lines show the critical Al content xcrx_{\mathrm{cr}} at which the sign of Δcf\Delta_{\mathrm{cf}} changes. In case of a linear interpolation of Δcf\Delta_{\mathrm{cf}}, xcrlin≈0.18x_{\mathrm{cr}}^{\mathrm{lin}}\approx0.18. When accounting for the composition-dependent bowing parameter b~cf(x)\widetilde{b}_{\mathrm{cf}}(x) for predicting Δcf(x)\Delta_{\mathrm{cf}}(x), xcrbow≈0.41x_{\mathrm{cr}}^{\mathrm{bow}}\approx0.41.

Image not extracted

Figure 6. Composition dependence of (a) crystal field splitting energy Δcf(x)\Delta_{\mathrm{cf}}(x) and (b) the alloy disorder parameter Δsp(x)\Delta_{\mathrm{sp}}(x) in (Al,Ga)N alloys.

In general, previous studies also come to the conclusion that Δcf(x)\Delta_{\mathrm{cf}}(x) deviates from the linear interpolation of the binary values of GaN and AlN [14, 32]. However, a direct comparison with existing literature data is difficult. Firstly, experimental studies providing information on the polarization switching are usually carried out on samples that are strained to a substrate. One can in principle account for strain effects when estimating the crossover for the unstrained case, although, it requires information on (i) the valence band deformation potentials and their composition dependence and (ii) crystal field splitting energies of the binary materials. As already discussed in § 2 A, in the literature there is a large spread in AlN and GaN values for Δcf\Delta_{\mathrm{cf}} and the valence and conduction band deformation potentials [70, 71]. The closest to our study here are the empirical tight-binding model (ETBM) calculations by Coughlan et al. [14], which predict that Δcf\Delta_{\mathrm{cf}} changes sign at xcrETBM=0.25x_{\mathrm{cr}}^{\mathrm{ETBM}}=0.25. However, the calculations in Ref. [14] use different binary Δcf\Delta_{\mathrm{cf}} values for GaN, ΔcfGaN,ETBM=23\Delta_{\mathrm{cf}}^{\mathrm{GaN,ETBM}}=23 meV, and AlN, ΔcfAlN,ETBM=−240\Delta_{\mathrm{cf}}^{\mathrm{AlN,ETBM}}=-240 meV, when compared to the values obtained here from PBE-DFT calculations (ΔcfGaN=40\Delta_{\mathrm{cf}}^{\mathrm{GaN}}=40 meV and ΔcfAlN=−177\Delta_{\mathrm{cf}}^{\mathrm{AlN}}=-177 meV). Using our compositiondependent bowing parameter formulae with ΔcfGaN=23\Delta_{\mathrm{cf}}^{\mathrm{GaN}} = 23 meV and ΔcfAlN=−240\Delta_{\mathrm{cf}}^{\mathrm{AlN}} = -240 meV from Ref. [14], we obtain that Δcf\Delta_{\mathrm{cf}} changes sign at xcr=0.20x_{\mathrm{cr}} = 0.20, which is close to the value obtained by Coughlan et al. [14]. Thus, the here obtained composition-dependent bowing parameter reflects well the ETBM results in Ref. [14] when accounting for differences in the crystal field splitting energies.

As already mentioned above, the sign of the crystal field splitting energy gives a first indication for changes in the valence band structure. However, alloy disorder also leads to band mixing effects and a lifting of degeneracies. We mimicked this effect through an SOC-like interaction term, introducing a single free parameter, Δsp\Delta_{\mathrm{sp}}, in the Hamiltonian. In doing so, the developed method is compatible with k⋅pk \cdot p Hamiltonians underlying widely available (commercial) software packages. Figure 6 (b) displays Δsp\Delta_{\mathrm{sp}} as a function of the Al content xx. The data is again fitted with a bowing parameter of the form b~cf(x)=αcf/(1+βcfx)\tilde{b}_{\mathrm{cf}}(x) = \alpha_{\mathrm{cf}}/(1+\beta_{\mathrm{cf}}x); αcf\alpha_{\mathrm{cf}} and βcf\beta_{\mathrm{cf}} values for Δsp\Delta_{\mathrm{sp}} are summarized in Table I. As there is no alloy disorder in GaN and AlN, for the binary systems Δsp=0\Delta_{\mathrm{sp}} = 0 meV. Overall, we observe that Δsp\Delta_{\mathrm{sp}} can be of similar magnitude to Δcf\Delta_{\mathrm{cf}} in the Al content composition range up to 50-60%. For higher Al contents Δsp<∣Δcf∣\Delta_{\mathrm{sp}} < |\Delta_{\mathrm{cf}}|. We also observe that Δsp\Delta_{\mathrm{sp}} is largest between x=0.75x = 0.75 and x=0.89x = 0.89, thus indicating that alloy disorder impacts the electronic structure in the high AlN content regime more strongly. A similar observation has been made by Finn et al. when studying the electronic structure of (Al,Ga)N QWs by means of an atomistic ETBM [51].

Impact of composition dependent k⋅pk \cdot p parameter on (Al,Ga)N band structures

As discussed above, Δcf\Delta_{\mathrm{cf}} and Δsp\Delta_{\mathrm{sp}} will impact both valence band ordering and mixing, which will affect the light polarization characteristics of (Al,Ga)N alloys. In the following, we compare k⋅pk \cdot p band structures of (Al,Ga)N alloys where the underlying parameters are determined by using (i) the composition-dependent bowing parameters, b~\tilde{b}, obtained in the present work and (ii) when linearly interpolating the AlN and GaN parameters and thus also neglecting alloy disorder-induced band mixing effects. Our aim is a general study of changes in the band structure rather than providing detailed insight into light polarization characteristics as this requires accounting for carrier densities and temperature effects, e.g., via a Fermi-Dirac distribution of carriers in the different valence bands. However, we can provide initial information on how the k⋅pk \cdot p parameters can affect the light polarization characteristics of (Al,Ga)N alloys by studying not only the energy dispersion of the different valence bands but also by resolving the orbital character of the different bands as a function of kk. For this analysis, we have selected two alloy compositions, namely Al0.25_{0.25}Ga0.75_{0.75}N and Al0.50_{0.50}Ga0.50_{0.50}N.

In a first step, we study the k⋅pk \cdot p band structure of an Al0.25_{0.25}Ga0.75_{0.75}N alloy. For this Al content, Δcf<0\Delta_{\mathrm{cf}} < 0 when using a linear interpolation of the GaN and AlN values, as shown in Figure 6 (a). However, when accounting for the composition-dependent bowing parameter, b~(x)\tilde{b}(x), Δcf>0\Delta_{\mathrm{cf}} > 0. As mentioned above, the sign of Δcf\Delta_{\mathrm{cf}} gives a first insight into the valence band ordering but does not provide information about band mixing effects or energetic separation of the different valence bands due to disorder. Figure 7 (a) shows the band structure of an Al0.25_{0.25}Ga0.75_{0.75}N alloy in the vicinity of the Γ\Gamma-point for a linear interpolation of the k⋅pk \cdot p parameters with composition, while Figure 7 (b) depicts the band structure when accounting for composition-dependent bowing parameters, b~(x)\tilde{b}(x). The color coding of the bands gives the orbital character. We have grouped pxp_x and pyp_y-like orbital contributions together as they both facilitate TE polarized light emission in contrast to pzp_z-like states which contribute to, in general, TM polarization. Looking at the band structure obtained from the linearly interpolated k⋅pk \cdot p parameters, Figure 7 (a), one finds that the topmost valence band, denoted as VB1_1 in the following, is entirely pzp_z-like in character. The lower lying valence states, denoted as VB2_2 and VB3_3, are pxp_x and pyp_y-like in character. The valence bands VB2_2 and VB3_3 are degenerate here since neither alloy disorder-related nor small SOC effects are considered in the linearly interpolated k⋅pk \cdot p parameter set. Overall, the band ordering seen in Figure 7 (a) is consistent with a negative Δcf\Delta_{\mathrm{cf}} value (cf. Figure 6 (a)), which means that the topmost valence band is expected to be pzp_z-like in character. However, even though VB1_1 is pzp_z-like in character and thus suggesting TM polarized light emission, VB2_2 and VB3_3 are only approximately 20 meV below VB1_1. Given that the thermal energy at room temperature (T=300T = 300 K) is around 24 meV, one can expect some occupation of VB2_2 and VB3_3 with holes at room temperature. This in turn will impact the degree of optical polarization, ρ\rho, which can be defined as the ratio of intensities of TE to TM polarized light emission, ρ=(ITE−ITM)/(ITE+ITM)\rho= (I_{\mathrm{TE}} - I_{\mathrm{TM}})/(I_{\mathrm{TE}} + I_{\mathrm{TM}}) [14, 72, 73] and potentially the LEE in a device. However, in general, the band structure depicted in Figure 7 (a) is indicative of predominantly TM polarized light emission in an Al0.25_{0.25}Ga0.75_{0.75}N alloy when using a linear interpolation of the k⋅pk \cdot p parameters with Al content.

Image not extracted

Figure 7. Band structure of an Al0.25_{0.25}Ga0.75_{0.75}N alloy obtained from (a) linearly interpolated k⋅pk \cdot p parameters and (b) when accounting for alloy disorder effects, Δsp\Delta_{\mathrm{sp}}, and composition dependent bowing parameters, b~(x)\tilde{b}(x).

When accounting for the extracted composition-dependent bowing parameters, Figure 7 (b) reveals that the resulting k⋅pk \cdot p valence band structure noticeably deviates from the band structure obtained within the linear interpolation scheme, cf. Figure 7 (a). Firstly, the topmost valence band, VB1_1, is mainly px+pyp_x+p_y-like in character at the Γ\Gamma point when accounting for composition dependent bowing parameters. Also VB2_2 is predominantly of px+pyp_x+p_y-like character near Γ\Gamma. The degeneracy of VB1_1 and VB2_2 is lifted due to alloy disorder, here introduced by the Δsp\Delta_{\mathrm{sp}} term in the k⋅pk \cdot p Hamiltonian. The energetically lowest valence band, VB3_3, is of pzp_z-like character. It is to note that VB3_3 lies more than 30 meV below the valence band edge. Therefore, thermal occupation of this band at room temperature is unlikely, at least at low carrier densities. Even though there is some band mixing between VB2_2 and VB3_3 along the Γ→K\Gamma\to K direction away from Γ\Gamma, predominantly TE polarized light emission is expected when accounting for the composition-dependent bowing parameters, even at elevated temperatures.

We note again that our discussion is based on unstrained bulk systems. Strain or quantum confinements in, e.g., QWs can lead to further changes in the band structure and band-ordering as discussed in the literature [52, 74, 75]. The effective masses of the different bands play an important role in understanding the impact of quantum confinement on, for instance, the valence band ordering in a QW system and thus the optical properties. In general, UV light emitters that utilize (Al,Ga)N alloys are grown along the wurtzite cc-axis [76, 77, 78]. Therefore, the effective masses associated with the different valence bands toward the AA point (along the wurtzite cc-axis) are important and we can now gain information on how these masses depend on the interpolation of the k⋅pk \cdot p parameters. For the calculated band structure, we employed both the analytic expressions provided in Appendix B, and numerical differentiation of the band dispersion in the vicinity of the Γ\Gamma point to extract the effective masses. Both approaches yield consistent results. Figure 7 (a) reveals a low effective hole mass (mVB1=0.15 m0m_{\mathrm{VB}_1} = 0.15\,m_0) in this direction for the highest valence band, VB1_1, when using linearly interpolated parameters for the band structure calculation. The energetically lower-lying bands, VB2_2 and VB3_3, exhibit much higher effective masses (mVB2=2.35 m0m_{\mathrm{VB}_2} = 2.35\,m_0; mVB3=2.35 m0m_{\mathrm{VB}_3} = 2.35\,m_0). This suggests that quantum confinement effects will significantly impact the valence band ordering, as the topmost valence (VB1_1) will shift to lower energies more strongly when compared to VB2_2 and VB3_3. Given the differences in the orbital character of the different valence bands, strong band mixing effects may be expected in a cc-plane Al0.25_{0.25}Ga0.75_{0.75}N QW depending on e.g. the well width. This will also affect the degree of optical polarization and potentially the LEE. However, when accounting for nonlinearities in the k⋅p\mathbf{k}\cdot\mathbf{p} parameter interpolation the situation is different, as Figure 7 (b) shows. In this case, VB1_1 and VB2_2 exhibit heavier effective masses (mVB1=2.78 m0m_{\mathrm{VB1}} = 2.78\,m_0, mVB2=1.04 m0m_{\mathrm{VB2}} = 1.04\,m_0) toward the AA point, while VB3_3 has a low effective mass (mVB3=0.19 m0m_{\mathrm{VB3}} = 0.19\,m_0) in this direction. This means that quantum confinement along the wurtzite cc-axis will shift VB3_3 away from the valence band edge formed by VB1_1 and VB2_2. Given that VB1_1 and VB2_2 are predominately pxp_x and pyp_y-like in orbital character, the energetically highest valence states in an Al0.25_{0.25}Ga0.75_{0.75}N QW will be formed by such orbitals, indicating mainly TE polarized light emission. This indicates a very different result compared to the conclusions drawn from the linearly interpolated k⋅p\mathbf{k}\cdot\mathbf{p} parameter set presented in Figure 7 (a).

In a second step, we turn our attention to the band structure of Al0.50_{0.50}Ga0.50_{0.50}N alloys, which often form the active region or the cladding layer of UV light emitters in the 250-300 nm wavelength range [79, 80, 81, 82]. Figure 8 (a) shows the band structure resulting from a linear interpolation of the k⋅p\mathbf{k}\cdot\mathbf{p} parameters with Al content x=0.5x = 0.5. The band structure is largely identical to the band structure displayed in Figure 7 (a) for the Al0.25_{0.25}Ga0.75_{0.75}N case: The energetically highest valence band, VB1_1, is of pzp_z-like character, while the energetically lower-lying valence bands, VB2_2 and VB3_3, are px+pyp_x + p_y-like in character and degenerate at the Γ\Gamma-point. The main difference to the Al0.25_{0.25}Ga0.75_{0.75}N case is that the energetic separation between VB1_1 and VB2,3_{2,3} is much greater at the Γ\Gamma point. In the Al0.50_{0.50}Ga0.50_{0.50}N alloy, the energetic separation is >60> 60 meV, compared to only 20 meV for Al0.25_{0.25}Ga0.75_{0.75}N. This behavior is expected when looking at the linear interpolation of the crystal field splitting energy with composition (cf. Figure 6 (a)). As a consequence and in contrast to Al0.25_{0.25}Ga0.75_{0.75}N, thermal occupation of VB2_2 and VB3_3 is unlikely in the bulk case. Thus, the light emitted from a bulk Al0.50_{0.50}Ga0.50_{0.50}N alloy is expected to be TM polarized, at least when using linearly interpolated k⋅p\mathbf{k}\cdot\mathbf{p} parameters.

Image not extracted

Figure 8. Band structure of an Al0.50_{0.50}Ga0.50_{0.50}N alloy obtained from (a) linearly interpolated k⋅p\mathbf{k}\cdot\mathbf{p} parameters and (b) when accounting for alloy disorder effects, Δsp\Delta_{\mathrm{sp}}, and composition dependent bowing parameters, b~(x)\widetilde{b}(x).

Turning to the effective masses of the different valence bands toward the AA-point, we observe a situation similar to Al0.25_{0.25}Ga0.75_{0.75}N, Figure 7 (a), in the sense that VB1_1 (mVB1=0.17 m0m_{\mathrm{VB1}} = 0.17\,m_0) exhibits a low effective mass while VB2_2 and VB3_3 (mVB2=2.66 m0m_{\mathrm{VB2}} = 2.66\,m_0, mVB3=2.66 m0m_{\mathrm{VB3}} = 2.66\,m_0) have a high effective mass along this direction. Thus, quantum confinement in cc-plane Al0.50_{0.50}Ga0.50_{0.50}N QWs can reduce the energetic separation between pzp_z and px+pyp_x+p_y-like valence bands. However, given the large crystal field splitting energy, even in a strong carrier confinement regime, e.g. narrow wells, and when focusing on energetic shifts in the band structure resulting from the different effective masses of the valence bands only (no strain contributions), TM polarized emission may still be expected in the case of the linearly interpolated k⋅p\mathbf{k}\cdot\mathbf{p} parameters. Strain effects may change this behavior and would have to be further analyzed.

The situation is again different when accounting for composition-dependent bowing parameters in determining the k⋅p\mathbf{k}\cdot\mathbf{p} parameters of the Al0.50_{0.50}Ga0.50_{0.50}N alloy, as the band structure in Figure 8 (b) clearly shows. Although the energetically highest valence band, VB1_1, also has a larger pzp_z orbital character (87%) near the Γ\Gamma point, the energy separation to VB2_2 is much smaller (≈20\approx20 meV) when compared to the band structure obtained from a linearly interpolated set of k⋅p\mathbf{k}\cdot\mathbf{p} parameters; see Figure 8 (a). Thus, at room temperature, occupation of VB2_2 with holes can be expected. Moreover, the orbital character of VB2_2 is mainly px+pyp_x+p_y-like. Therefore, in the case of a non-linear interpolation of the k⋅p\mathbf{k}\cdot\mathbf{p} parameters, the degree of optical polarization will be determined by the orbital character in the valence bands VB1_1 and VB2_2. Looking at the effective masses of the different valence bands near the Γ\Gamma point and along the direction of the AA-point, we find that VB1_1 (mVB1=0.23 m0m_{\mathrm{VB1}} = 0.23\,m_0) has a low effective mass, while VB2 and VB3 (mVB2=4.03 m0m_{\mathrm{VB2}} = 4.03\,m_0, mVB3=1.14 m0m_{\mathrm{VB3}} = 1.14\,m_0) exhibit high effective masses. However, given the smaller crystal field splitting energy in the case of a nonlinear interpolation of the k⋅p\mathbf{k}\cdot\mathbf{p} parameters, quantum confinement in a cc-plane Al0.50_{0.50}Ga0.50_{0.50}N QW may now give rise to strong band mixing effects as VB1, VB2 and VB3 may be energetically very close and QW states are then formed from all three bands as linear combinations. As such, the degree of optical polarization may be strongly impacted by this effect and may differ from the results predicted when linearly interpolated parameters are used to calculate the QW electronic structure.

Conclusion

In summary, we have investigated the composition dependence of k⋅p\mathbf{k}\cdot\mathbf{p} parameters for wurtzite (Al,Ga)N alloys. These parameters are key ingredients for widely available theoretical and simulation models to analyze electronic, optical and carrier transport properties of (Al,Ga)N-based heterostructures and devices.

To gain insight into the composition dependence of the k⋅p\mathbf{k}\cdot\mathbf{p} band parameters, we use GGA-DFT calculations to sample the electronic structure of (Al,Ga)N alloys over the full composition range. The DFT data were fitted with a k⋅p\mathbf{k}\cdot\mathbf{p} model widely used in the literature. Thus, the obtained k⋅p\mathbf{k}\cdot\mathbf{p} parameter set is compatible with, for instance, commercially available software packages, which, in general, do not allow to define ”arbitrary” k⋅p\mathbf{k}\cdot\mathbf{p} Hamiltonians.

Our investigations show that the extracted band parameters for GaN and AlN agree well with literature data. However, our study also reveals that the widely used approximation of a linear interpolation of the GaN and AlN parameter sets with composition breaks down in (Al,Ga)N alloys. We find that to describe the DFT data accurately, composition-dependent bowing parameters are required for most of the (Al,Ga)N k⋅p\mathbf{k}\cdot\mathbf{p} parameters. We provide here an empirical expression to predict (Al,Ga)N k⋅p\mathbf{k}\cdot\mathbf{p} band parameters at desired alloy compositions.

Moreover, we furnish initial insight into consequences of the nonlinear behavior of the k⋅p\mathbf{k}\cdot\mathbf{p} parameters with Al content for the bulk band structure of (Al,Ga)N alloys. We observed, for example, that band-ordering and energetic separation of the different valence bands are strongly impacted by the nonlinear evolution of the k⋅p\mathbf{k}\cdot\mathbf{p} parameter set with composition. The obtained results indicate that optical properties in (Al,Ga)N bulk alloys but also QW systems can differ significantly depending on whether a linear or nonlinear interpolation scheme to determine the k⋅p\mathbf{k}\cdot\mathbf{p} parameters is used.

Appendix A

The k⋅p\mathbf{k}\cdot\mathbf{p} Hamiltonian used in the present work is based on the work of Ghersoni et al. [83], and is also described in Refs. [53, 84]. For a wurtzite structure the 8×88\times8 Hamiltonian in the basis (∣S↑⟩|S\uparrow\rangle, ∣X↑⟩|X\uparrow\rangle, ∣Y↑⟩|Y\uparrow\rangle, ∣Z↑⟩|Z\uparrow\rangle, ∣S↓⟩|S\downarrow\rangle, ∣X↓⟩|X\downarrow\rangle, ∣Y↓⟩|Y\downarrow\rangle, ∣Z↓⟩|Z\downarrow\rangle) is described in matrix form as:

H^=[G(k)Γ−Γ∗G∗(k)](8)\hat{H} = \begin{bmatrix} G(\mathbf{k}) & \Gamma\\ -\Gamma^* & G^*(\mathbf{k}) \end{bmatrix} \tag*{(8)}

Here, G(k)G(\mathbf{k}) and Γ\Gamma are 4×44\times4 matrices and G(k)G(\mathbf{k}) can be further decomposed into 4×44\times4 matrices:

G(k)=G1(k)+G2(k)+Gsp+Gcf.(9)G(\mathbf{k}) = G_1(\mathbf{k}) + G_2(\mathbf{k}) + G_{\mathrm{sp}} + G_{\mathrm{cf}} . \tag*{(9)}

The matrix G1(k)G_1(\mathbf{k}) is given by

G1(k)=[EciP2kxiP2kyiP1kz−iP2kxEv00−iP2ky0Ev0−iP1kz00Ev](10)G_1(\mathbf{k}) = \begin{bmatrix} E_c & iP_2k_x & iP_2k_y & iP_1k_z \\ -iP_2k_x & E_v & 0 & 0 \\ -iP_2k_y & 0 & E_v & 0 \\ -iP_1k_z & 0 & 0 & E_v \end{bmatrix} \tag*{(10)}

where EcE_c and EvE_v are the conduction and valence band edge, respectively. G2(k)G_2(\mathbf{k}) has the following form:

G2(k)=[A2′(kx2+ky2)+A1′kz2B2kykzB2kxkzB1kxkyB2kykzL1′kx2+M1ky2+M2kz2N1′kxkyN2′kxkz−N3′kxB2kxkzN1′kxkyM1kx2+L1′ky2+M2kz2N2′kykz−N3′kyB1kxkyN2′kxkz+N3′kxN2′kykz+N3′kyM3kx2+M3ky2+L2′kz2](11)G_2(\mathbf{k}) = \begin{bmatrix} A'_2(k_x^2+k_y^2)+A'_1k_z^2 & B_2k_yk_z & B_2k_xk_z & B_1k_xk_y \\ B_2k_yk_z & L'_1k_x^2+M_1k_y^2+M_2k_z^2 & N'_1k_xk_y & N'_2k_xk_z-N'_3k_x \\ B_2k_xk_z & N'_1k_xk_y & M_1k_x^2+L'_1k_y^2+M_2k_z^2 & N'_2k_yk_z-N'_3k_y \\ B_1k_xk_y & N'_2k_xk_z+N'_3k_x & N'_2k_yk_z+N'_3k_y & M_3k_x^2+M_3k_y^2+L'_2k_z^2 \end{bmatrix} \tag*{(11)}

where terms Li′L'_i, Ni′N'_i and MiM_i are associated with the Luttinger-like parameters AiA_i with i=1...6i=1...6, while the terms A1′A'_1 and A2′A'_2 are determined by conduction band effective masses me∥m_e^{\parallel} and me⊥m_e^{\perp}, respectively:

A1′=ℏ22me∥(12)A'_1 = \frac{\hbar^2}{2m_e^{\parallel}} \tag*{(12)}
A2′=ℏ22me⊥(13)A'_2 = \frac{\hbar^2}{2m_e^{\perp}} \tag*{(13)}
L1′=ℏ22m0(A2+A4+A5),(14)L'_1 = \frac{\hbar^2}{2m_0}(A_2 + A_4 + A_5), \tag*{(14)}
L2′=ℏ22m0(A1),L'_2 = \frac{\hbar^2}{2m_0}(A_1),
M1=ℏ22m0(A2+A4−A5),(15)M_1 = \frac{\hbar^2}{2m_0}(A_2 + A_4 - A_5), \tag*{(15)}
M2=ℏ22m0(A1+A3),M_2 = \frac{\hbar^2}{2m_0}(A_1 + A_3),
N1′=ℏ22m0(2A5),N'_1 = \frac{\hbar^2}{2m_0}(2A_5),
N2′=ℏ22m0(2A6),(16)N'_2 = \frac{\hbar^2}{2m_0}(\sqrt{2}A_6), \tag*{(16)}
N3′=i(2A7).N'_3 = i\left(\sqrt{2}A_7\right).

Following Ghersoni et al. [83], we set the BB parameter to zero. This is consistent with the work of Rinke et al. [22] where it is discussed that including this term did not improve the fitting of the band structures. Moreover, the A7A_7 parameter is set to zero in our calculations, which is a widely used approximation and makes the model compatible with, e.g., k⋅p\mathbf{k} \cdot\mathbf{p} Hamiltonians underlying commercially available software packages such as nextnano [85] and Crosslight APSYS [86]. Furthermore, the parameters P1/2P_{1/2} that describe conduction band valence band coupling are set to zero, which is also a widely used approximation for wide-band gap materials such as (Al,Ga)N alloys [25, 26, 87]. The reduced form of the Hamiltonian is often referred to as a 6 + 2 band model.

In the standard definition of the Hamiltonian, Eq. (8), the matrices GspG_{\mathrm{sp}} and Γ\Gamma describe the spin-orbit coupling. As discussed in the main text, we use these terms here to account for alloy disorder-induced band mixing and lifting of degeneracies in the band structure of (Al,Ga)N alloys. In general, we use the standard definitions of GspG_{\mathrm{sp}} and Γ\Gamma in our fitting procedure:

Gsp=Δsp3(000000−i00i000000),(17)G_{\mathrm{sp}} = \frac{\Delta_{\mathrm{sp}}}{3} \begin{pmatrix} 0 & 0 & 0 & 0 \\ 0 & 0 & -i & 0 \\ 0 & i & 0 & 0 \\ 0 & 0 & 0 & 0 \end{pmatrix}, \tag*{(17)}

and

Γ=Δsp3(00000001000−i0−1i0).(18)\Gamma= \frac{\Delta_{\mathrm{sp}}}{3} \begin{pmatrix} 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & -i \\ 0 & -1 & i & 0 \end{pmatrix}. \tag*{(18)}

Finally, the crystal field splitting contribution, present in the wurtzite crystal structure, is described by:

Gcf=Δcf(0000010000100000).(19)G_{\mathrm{cf}} = \Delta_{\mathrm{cf}} \begin{pmatrix} 0 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 0 \end{pmatrix}. \tag*{(19)}

Appendix B

Expressions for the effective hole effective masses parallel and perpendicular to the wurtzite cc-axis have been given in Refs. [22, 53] in terms of the Luttinger-like parameters AiA_i with i=1...6i = 1...6. Following Refs. [22, 53] the masses can be calculated close to or far from the Γ\Gamma-point, and the corresponding equations are summarized below.

a. Hole masses close to Γ\Gamma-point

m0mVB1∥=−(A1+A3)\frac{m_0}{m_{\mathrm{VB1}}^{\parallel}} = -(A_1 + A_3)
m0mVB1⊥=−(A2+A4)(20)\frac{m_0}{m_{\mathrm{VB1}}^{\perp}} = -(A_2 + A_4) \tag*{(20)}
m0mVB2∥=−[A1+(EVB2EVB2−EVB3)A3]\frac{m_0}{m_{\mathrm{VB2}}^{\parallel}} = -\left[A_1 + \left(\frac{E_{\mathrm{VB2}}}{E_{\mathrm{VB2}} - E_{\mathrm{VB3}}}\right) A_3\right]
m0mVB2⊥=−[A2+(EVB2EVB2−EVB3)A4](21)\frac{m_0}{m_{\mathrm{VB2}}^{\perp}} = -\left[A_2 + \left(\frac{E_{\mathrm{VB2}}}{E_{\mathrm{VB2}} - E_{\mathrm{VB3}}}\right) A_4\right] \tag*{(21)}
m0mVB3∥=−[A1+(EVB3EVB3−EVB2)A3]\frac{m_0}{m_{\mathrm{VB3}}^{\parallel}} = -\left[A_1 + \left(\frac{E_{\mathrm{VB3}}}{E_{\mathrm{VB3}} - E_{\mathrm{VB2}}}\right) A_3\right]
m0mVB3⊥=−[A2+(EVB3EVB3−EVB2)A4](22)\frac{m_0}{m_{\mathrm{VB3}}^{\perp}} = -\left[A_2 + \left(\frac{E_{\mathrm{VB3}}}{E_{\mathrm{VB3}} - E_{\mathrm{VB2}}}\right) A_4\right] \tag*{(22)}

Here, EVB2E_{\mathrm{VB2}} and EVB3E_{\mathrm{VB3}} are the energies of the three valence bands at the Γ\Gamma-point described by Eqs. (2) and (3) in § 2 B.

b. Hole masses far from Γ\Gamma-point

m0mVB1∥=−(A1+A3)\frac{m_0}{m_{\mathrm{VB1}}^{\parallel}} = -(A_1 + A_3)
m0mVB1⊥=−(A2+A4−A5)(23)\frac{m_0}{m_{\mathrm{VB1}}^{\perp}} = -(A_2 + A_4 - A_5) \tag*{(23)}
m0mVB2∥=−(A1+A3)\frac{m_0}{m_{\mathrm{VB2}}^{\parallel}} = -(A_1 + A_3)
m0mVB2⊥=−(A2+A4+A5)(24)\frac{m_0}{m_{\mathrm{VB2}}^{\perp}} = -(A_2 + A_4 + A_5) \tag*{(24)}
m0mVB3∥=−A1\frac{m_0}{m_{\mathrm{VB3}}^{\parallel}} = -A_1
m0mVB3⊥=−A2(25)\frac{m_0}{m_{\mathrm{VB3}}^{\perp}} = -A_2 \tag*{(25)}

Acknowledgments

This work was financially supported by the Marie Sklodowska-Curie Actions programme (EU) and Taighde Éireann – Research Ireland, formerly Science Foundation Ireland (SFI), co-fund programme SPARKLE (Grant No. 847652), by Research Ireland (Grant No. 21/FFP-A/9014), and by the Research Ireland-funded Irish Photonic Integration Centre (IPIC; Grant No. 12/RC/2276_P2). ams-OSRAM International GmbH gratefully acknowledges support by the Federal Ministry for Economic Affairs and Climate Action on the basis of a decision by the German Bundestag and by the Bavarian Ministry of State for Economic Affairs, Regional Development and Energy as well as by the European Union – NextGenerationEU for the Important Project of Common European Interest on microelectronics and communication technologies (IPCEI ME/CT) - OptoSuRe (16IPCEI221).

Data Availabilty

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

References

  1. [1]Wiete Westerhof and Ludmila Nieuweboer-Krobotova. Treatment of vitiligo with UV-B radiation vs topical psoralen plus UV-A. Archives of Dermatology, 133(12):1525–1528, 1997.
  2. [2]Vladimir Panov and Tsvetelina Borisova-Papancheva. Application of ultraviolet light (UV) in dental medicine. Journal of Medical and Dental Practice, 2(2):194–200, 2015.DOI
  3. [3]Dinesh Dhankhar, Runze Li, Anushka Nagpal, Jie Chen, Arjun Krishnamoorthi, and Peter M Rentzepis. A novel approach for remote detection of bacteria using simple charge-coupled device cameras and telescope. Review of Scientific Instruments, 91(7), 2020.DOI
  4. [4]Hideki Hirayama, Norihiko Kamata, and Kenji Tsubaki. AlGaN-based deep-ultraviolet light-emitting diodes. In III-Nitride Based Light Emitting Diodes and Applications, pages 267–299. Springer, 2017.
  5. [5]Steven Clarke and William Bettin. Ultraviolet light disinfection in the use of individual water purification devices. US Army Public Health Command: Aberdeen Proving Ground, MD, USA, pages 1–12, 2006.DOI
  6. [6]Kevin W Houser. Ten facts about UV radiation and COVID-19. Leukos, 16(3):177–178, 2020.DOI
  7. [7]JM Green. A new generation of ultra-violet/visible gas lasers. Optics & Laser Technology, 10(6):289–300, 1978.DOI
  8. [8]Yoshihiko Muramoto, Masahiro Kimura, and Suguru Nouda. Development and future of ultraviolet light-emitting diodes: UV-LED will replace the UV lamp. Semiconductor Science and Technology, 29(8):084004, 2014.
  9. [9]Michael Kneissl, Tae-Yeon Seong, Jung Han, and Hiroshi Amano. The emergence and prospects of deep-ultraviolet light-emitting diode technologies. Nature Photonics, 13(4):233–244, 2019.DOI
  10. [10]Hiroshi Amano, Ramón Collazo, Carlo De Santi, Sven Einfeldt, et al. The 2020 UV emitter roadmap. Journal of Physics D: Applied Physics, 53(50):503001, 2020.
  11. [11]Jing Lang, Fujun Xu, Jiaming Wang, Lisheng Zhang, Xuzhou Fang, Ziyao Zhang, Xueqi Guo, Chen Ji, Chengzhi Ji, Fuyun Tan, et al. Progress in performance of AlGaN-based ultraviolet light emitting diodes. Advanced Electronic Materials, 11(1):2300840, 2025.
  12. [12]Cheng Liu and Jing Zhang. Influence of quantum well design on light polarization switching in AlGaN ultraviolet emitters. AIP Advances, 8(8):085125, 2018.DOI
  13. [13]Masakatsu Suzuki, Takeshi Uenoyama, and Akira Yanase. First-principles calculations of effective-mass parameters of AlN and GaN. Physical Review B, 52(11):8132, 1995.
  14. [14]Conor Coughlan, Stefan Schulz, Miguel A Caro, and Eoin P O’Reilly. Band gap bowing and optical polarization switching in Al₁₋ₓGaₓN alloys. Physica Status Solidi (B), 252(5):879–884, 2015.
  15. [15]KB Nam, J Li, ML Nakarmi, JY Lin, and HX Jiang. Unique optical properties of AlGaN alloys and related ultraviolet emitters. Applied Physics Letters, 84(25):5264–5266, 2004.DOI
  16. [16]RG Banal, M Funato, and Y Kawakami. Optical anisotropy in [0001]-oriented AlₓGa₁₋ₓN/AlN quantum wells (x>0.69). Physical Review B, 79(12):121308, 2009.DOI
  17. [17]Martin Guttmann, Frank Mehnke, Bettina Belde, Fynn Wolf, Christoph Reich, Luca Sulmoni, Tim Wernicke, and Michael Kneissl. Optical light polarization and light extraction efficiency of AlGaN-based LEDs emitting between 264 and 220 nm. Japanese Journal of Applied Physics, 58(SC):SCCB20, 2019.DOI
  18. [18]Samuel Poncé, Debdeep Jena, and Feliciano Giustino. Hole mobility of strained GaN from first principles. Physical Review B, 100(8):085204, 2019.DOI
  19. [19]Su-Huai Wei and Alex Zunger. Valence band splittings and band offsets of AlN, GaN, and InN. Applied Physics Letters, 69(18):2719–2721, 1996.
  20. [20]Igor Vurgaftman, JR Meyer, and L Ramdas Ram-Mohan. Band parameters for III–V compound semiconductors and their alloys. Journal of Applied Physics, 89(11):5815–5875, 2001.DOI
  21. [21]I Vurgaftman and JR Meyer. Band parameters for nitrogen-containing semiconductors. Journal of Applied Physics, 94(6):3675–3696, 2003.DOI
  22. [22]Patrick Rinke, Momme Winkelkemper, Abdallah Qteish, Dieter Bimberg, Jörg Neugebauer, and Matthias Scheffler. Consistent set of band parameters for the group-III nitrides AlN, GaN, and InN. Physical Review B, 77(7):075202, 2008.DOI
  23. [23]Mikołaj Chlipala and Henryk Turski. Harnessing III-Nitride built-in field in multi-quantum well LEDs. ACS Applied Materials & Interfaces, 16(18):24021–24028, 2024.DOI
  24. [24]Zhongqiu Xing, Yao Wang, Fang Wang, Juin J Liou, and Yuhuai Liu. Improvement of the optoelectronic characteristics in deep-ultraviolet laser diodes with tapered p-cladding layer and triangular electron blocking layer. Applied Physics B, 128(11):197, 2022.DOI
  25. [25]D Fu, R Zhang, B Liu, ZL Xie, XQ Xiu, H Lu, YD Zheng, and G Edwards. Exploring optimal UV emission windows for AlGaN and AlInN alloys grown on different templates. Physica Status Solidi (B), 248(12):2816–2820, 2011.DOI
  26. [26]Marta Gladysiewicz, Detlef Hommel, and Robert Kudrawiec. Material gain engineering in staggered polar AlGaN/AlN quantum wells dedicated for deep UV lasers. IEEE Journal of Selected Topics in Quantum Electronics, 25(6):1–8, 2019.
  27. [27]Zhongqui Xing, Yongjie Zhou, Fang Wang, and Yuhai Liu. Improving hole injection with the polarization effect for 279 nm AlGaN-based deep ultraviolet light-emitting diodes. Applied Physics B, 130(7):122, 2024.DOI
  28. [28]A Atsushi Yamaguchi. Theoretical investigation of optical polarization properties in Al-rich AlGaN quantum wells with various substrate orientations. Applied Physics Letters, 96(15), 2010.DOI
  29. [29]Weiying Wang, Huimin Lu, Lei Fu, Chenguang He, Mingxing Wang, Ning Tang, Fujun Xu, Tongjun Yu, Weikun Ge, and Bo Shen. Enhancement of optical polarization degree of AlGaN quantum wells by using staggered structure. Optics express, 24(16):18176–18183, 2016.DOI
  30. [30]Marta Gladysiewicz, Czeslaw Skierbiszewski, and Robert Kudrawiec. Material gain in polar GaInN and AlGaN quantum wells: How to overcome the ’dead’ width for light emitters in these QW systems? IEEE Journal of Selected Topics in Quantum Electronics, 28(1: Semiconductor Lasers):1–9, 2021.DOI
  31. [31]Huan-Ting Shen, Claude Weisbuch, James S Speck, and Yuh-Renn Wu. Three-dimensional modeling of minority-carrier lateral diffusion length including random alloy fluctuations in (In,Ga)N and (Al,Ga)N single quantum wells. Physical Review Applied, 16(2):024054, 2021.DOI
  32. [32]B Neuschl, J Helbing, M Knab, H Lauer, M Madel, K Thonke, T Meisch, K Forghani, F Scholz, and M Feneberg. Composition dependent valence band order in c-oriented wurtzite AlGaN layers. Journal of Applied Physics, 116(11), 2014.DOI
  33. [33]Oliver Marquardt, Miguel A Caro, Thomas Koprucki, Peter Mathe, and Morten Willatzen. Multiband k·p model and fitting scheme for ab initio based electronic structure parameters for wurtzite GaAs. Physical Review B, 101(23):235147, 2020.
  34. [34]Jürgen Hafter. Ab-initio simulations of materials using VASP: Density-functional theory and beyond. Journal of Computational Chemistry, 29(13):2044–2078, 2008.
  35. [35]Qimin Yan, Patrick Rinke, Anderson Janotti, Matthias Scheffler, and Chris G Van de Walle. Effects of strain on the band structure of group-III nitrides. Physical Review B, 90(12):125118, 2014.
  36. [36]Pawel Strak, Pawel Kempisty, Konrad Sakowski, Agata Kaminska, Dawid Jankowski, Krzysztof P Korona, Kamil Sobczak, Jolanta Borysiuk, Mark Beeler, Ewa Grzanka, et al. Ab initio and experimental studies of polarization and polarization related fields in nitrides and nitride structures. AIP Advances, 7(1), 2017.DOI
  37. [37]Justin C Goodrich, Chee-Keong Tan, Damir Borovac, and Nelson Tansu. Prospects for hole doping in dilute-anion III-nitrides. Applied Physics Letters, 118(7), 2021.DOI
  38. [38]Ashfaq Ahmad, Pawel Strak, Pawel Kempisty, Konrad Sakowski, Jacek Piechota, Yoshihiro Kangwa, Izabella Grzegory, Michal Leszczynski, Zbigniew R Zytkiewicz, Grzegorz Muziol, et al. Polarization doping—ab initio verification of the concept: Charge conservation and non-locality. Journal of Applied Physics, 132(6), 2022.DOI
  39. [39]Zheng Liu, Si-Yuan Xu, Su-Huai Wei, and Xie Zhang. Chemical trend of radiative recombination in III-nitrides. Physical Review Materials, 8(4):044602, 2024.DOI
  40. [40]Luigi Balestra, Elena Gnani, and Susanna Reggiani. Electron effective masses of ScₓAl₁₋ₓN and AlₓGa₁₋ₓN from first-principles calculations of unfolded band struc-DOI
  41. [41]Qimin Yan, Emmanouel Kioupakis, Debdeep Jena, and Chris G Van de Walle. First-principles study of high-field-related electronic behavior of group-III nitrides. Physical Review B, 90(12):121201, 2014.DOI
  42. [42]Jochen Heyd, Gustavo E Scuseria, and Matthias Ernzerhof. Hybrid functionals based on a screened coulomb potential. The Journal of Chemical Physics, 118(18):8207–8215, 2003.
  43. [43]Aliaksandr V Krukau, Oleg A Vydrov, Artur F Izmaylov, and Gustavo E Scuseria. Influence of the exchange screening parameter on the performance of screened hybrid functionals. The Journal of Chemical Physics, 125(22), 2006.
  44. [44]Poul Georg Moses, Maosheng Miao, Qimin Yan, and Chris G Van de Walle. Hybrid functional investigations of band gaps and band alignments for AlN, GaN, InN, and InGaN. The Journal of Chemical Physics, 134(8), 2011.DOI
  45. [45]Giuliano Coli, KK Bajaj, J Li, JY Lin, and HX Jiang. Excitonic luminescence linewidths in AlGaN alloys with high aluminum concentrations. Applied Physics Letters, 80(16):2907–2909, 2002.DOI
  46. [46]AN Westmeyer, S Mahajan, KK Bajaj, JY Lin, HX Jiang, DD Koleske, and RT Senger. Determination of energy-band offsets between GaN and AlN using excitonic luminescence transition in AlGaN alloys. Journal of Applied Physics, 99(1), 2006.DOI
  47. [47]Lorenzo Rigutti, B Bonef, J Speck, F Tang, and RA Oliver. Atom probe tomography of nitride semiconductors. Scripta Materialia, 148:75–81, 2018.
  48. [48]Bernd Witzigmann, Friedhard Römer, Martin Martens, Christian Kuhn, Tim Wernicke, and Michael Kneissl. Calculation of optical gain in AlGaN quantum wells for ultraviolet emission. AIP Advances, 10(9), 2020.DOI
  49. [49]Gregor Hofmann, Anton Muhin, Norman Susilo, Friedhard Römer, Tim Wernicke, Michael Kneissl, and Bernd Witzigmann. Simulation of carrier injection efficiency in AlGaN-based UV-light-emitting diodes. IEEE Photonics Journal, 2022.
  50. [50]Martin Guttmann, Jakob Höpfner, Christoph Reich, Luca Sulmoni, Christian Kuhn, Pascal Röder, Tim Wernicke, and Michael Kneissl. Effect of quantum barrier composition on electro-optical properties of AlGaN-based UVC light emitting diodes. Semiconductor Science and Technology, 34(8):085007, 2019.DOI
  51. [51]Robert Finn and Stefan Schulz. Impact of random alloy fluctuations on the electronic and optical properties of (Al,Ga)N quantum wells: Insights from tight-binding calculations. The Journal of Chemical Physics, 157(24), 2022.DOI
  52. [52]Robert Finn, Michael O’Donovan, Thomas Koprucki, and Stefan Schulz. Theoretical study of the impact of carrier density screening on urbach tail energies and optical polarization in (Al,Ga)N quantum well systems. ArXiv Preprint ArXiv:2501.16808, 2025.arxiv.org/abs/2501.16808
  53. [53]SL Chuang and CS Chang. k·p method for strained wurtzite semiconductors. Physical Review B, 54(4):2491, 1996.DOI
  54. [54]Harald Niederreiter. Random number generation and quasi-Monte Carlo methods. Society for Industrial and Applied Mathematics, 1992.
  55. [55]IM Sobol. The distribution of points in a cube and the accurate evaluation of integrals. USSR Computa- tional Mathematics and Mathematical Physics, 7(4):86– 112, 1967.DOI
  56. [56]Harald Niederreiter. Low-discrepancy and low-dispersion sequences. Journal of Number Theory, 30(1):51–70, 1988.DOI
  57. [57]Oliver Marquardt. Simulating the electronic properties of semiconductor nanostructures using multiband k·p models. Computational Materials Science, 194:110318, 2021.DOI
  58. [58]Oliver Marquardt, Sixten Boeck, Christoph Freysoldt, Tilmann Hickel, and Jörg Neugebauer. Plane-wave implementation of the real-space k·p formalism and continuum elasticity theory. Computer Physics Communications, 181(4):765–771, 2010.
  59. [59]Oliver Marquardt, Sixten Boeck, Christoph Freysoldt, Tilmann Hickel, Stefan Schulz, Jörg Neugebauer, and Eoin P O’Reilly. A generalized plane-wave formulation of k·p formalism and continuum-elasticity approach to elastic and electronic properties of semiconductor nanostructures. Computational Materials Science, 95:280–287, 2014.
  60. [60]MPCM Krijn. Heterojunction band offsets and effective masses in III-V quaternary alloys. Semiconductor Science and Technology, 6(1):27, 1991.DOI
  61. [61]S Schöche, P Kühne, T Hofmann, M Schubert, Daniel Nilsson, A Kakanakova-Georgieva, Erik Janzén, and Vanya Darakchieva. Electron effective mass in Al₀.₇₂Ga₀.₂₈N alloys determined by mid-infrared optical hall effect. Applied Physics Letters, 103(21), 2013.DOI
  62. [62]Chi-Kang Li, Marco Piccardo, Li-Shuo Lu, Svitlana Mayboroda, Lucio Martinelli, Jacques Peretti, James S Speck, Claude Weisbuch, Marcel Filoche, and Yuh-Renn Wu. Localization landscape theory of disorder in semiconductors. III. application to carrier transport and recombination in light emitting diodes. Physical Review B, 95(14):144206, 2017.DOI
  63. [63]Y Koide, H Itoh, MRH Khan, K Hiramatu, N Sawaki, and I Akasaki. Energy band-gap bowing parameter in an AlₓGa₁₋ₓN alloy. Journal of Applied Physics, 61(9):4540–4543, 1987.DOI
  64. [64]Yen-Kuang Kuo and Wen-Wei Lin. Band-gap bowing parameter of the AlₓGa₁₋ₓN derived from theoretical simulation. Japanese Journal of Applied Physics, 41(1R):73, 2002.
  65. [65]Chris G Van de Walle, MD McCluskey, CP Master, LT Romano, and NM Johnson. Large and composition-dependent band gap bowing in InₓGa₁₋ₓN alloys. Materials Science and Engineering: B, 59(1-3):274–278, 1999.DOI
  66. [66]Miguel A Caro, Stefan Schulz, and Eoin P O’Reilly. Theory of local electric polarization and its relation to internal strain: Impact on polarization potential and electronic properties of group-III nitrides. Physical Review B, 88(21):214103, 2013.
  67. [67]E Sakalauskas, H Behemburg, C Hums, P Schley, G Rossbach, C Giesen, M Heuken, H Kalisch, RH Jansen, J Bläsing, et al. Dielectric function and optical properties of Al-rich AlInN alloys pseudomorphically grown on GaN. Journal of Applied Physics, 43(36):365102, 2010.DOI
  68. [68]Stefan Schulz, Miguel A Caro, Lay-Theng Tan, Peter J Parbrook, Robert W Martin, and Eoin P O’Reilly. Composition-dependent band gap and band-edge bowing in AlInN: A combined theoretical and experimental study. Applied Physics Express, 6(12):121001, 2013.DOI
  69. [69]Shahab N Alam, Vitaly Z Zubialevich, Bijan Ghafary, and Peter J Parbrook. Bandgap and refractive index estimates in AlN and related nitrides across their full composition ranges. Scientific Reports, 10(1):16205, 2020.
  70. [70]Qimin Yan, Patrick Rinke, Matthias Scheffler, and Chris G Van de Walle. Strain effects in group-III nitrides: Deformation potentials for AlN, GaN, and InN. Applied Physics Letters, 95(12), 2009.DOI
  71. [71]Thomas P Sheerin and Stefan Schulz. Strain effects in wurtzite boron nitride: Elastic constants, internal strain, and deformation potentials from hybrid functional density functional theory. Physica Status Solidi (RRL)–Rapid Research Letters, 16(6):2200021, 2022.
  72. [72]Yi Li, Mei Ge, Meiyu Wang, Youhua Zhu, and Xinglong Guo. Effect of surface plasmon coupling with radiating dipole on the polarization characteristics of AlGaN-based light-emitting diodes. Chinese Physics B, 31(7):077801, 2022.
  73. [73]Shiqiang Lu, Xinjun Jiang, Yaozeng Wang, Kai Huang, Na Gao, Duanjun Cai, Yinghui Zhou, CC Yang, Junyong Kang, and Rong Zhang. Enhancing deep-UV emission at 234 nm by introducing a truncated pyramid AlN/GaN nanostructure with fine-tuned multiple facets. Nanoscale, 14(3):653–662, 2022.DOI
  74. [74]Masakatsu Suzuki and Takeshi Uenoyama. Strain effect on electronic and optical properties of GaN/AlGaN quantum-well lasers. Journal of Applied Physics, 80(12):6868–6874, 1996.
  75. [75]Huan-Ting Shen, Yu-Chieh Chang, and Yuh-Renn Wu. Analysis of light-emission polarization ratio in deep-ultraviolet light-emitting diodes by considering random alloy fluctuations with the 3d k·p method. Physica Status Solidi (RRL)–Rapid Research Letters, 16(2):2100498, 2022.DOI
  76. [76]P Sohi, D Martin, and N Grandjean. Critical thickness of GaN on AlN: impact of growth temperature and dislocation density. Semiconductor Science and Technology, 32(7):075010, 2017.DOI
  77. [77]Prashant Tyagia, Ch Ramesha, SS Kushvahaa, and M Senthil Kumara. AlGaN nanowall network structure grown on sapphire (0001) substrate by laser molecular beam epitaxy. Materials Science in Semiconductor Processing, 89:143–148, 2019.
  78. [78]Chung-Chi Chen, Ting-Chun Huang, Yu-Wei Lin, Yu-Ren Lin, Ping-Hsiu Wu, Ping-Wei Liou, Hao-Yu Hsieh, Yang-Yi Huang, Shaobo Yang, Yuh-Renn Wu, et al. Hole mobility behavior in Al-gradient polarization-induced p-type AlGaN grown on GaN template. Applied Physics Letters, 120(2), 2022.
  79. [79]Zi-Hui Zhang, Sung-Wen Huang Chen, Yonghui Zhang, Luping Li, Sheng-Wen Wang, Kangkai Tian, Chun-shuang Chu, Mengqian Fang, Hao-Chuk Kuo, and Wengang Bi. Hole transport manipulation to improve the hole injection for deep ultraviolet light-emitting diodes. ACS Photonics, 4(7):1846–1850, 2017.
  80. [80]Fatima Asif, Hung-Chi Chen, Antwon Coleman, Mohamed Lachab, Iftikhar Ahmad, Bin Zhang, Khalid Fareed, Vinod Adivarahan, and Asif Khan. Substrate lifted-off AlGaN/AlGaN lateral conduction thin-film light-emitting diodes operating at 285 nm. Japanese Journal of Applied Physics, 52(8S):08JG14, 2013.
  81. [81]Chunshuang Chu, Kangkai Tian, Jiamang Che, Hua Shao, Jianquan Kou, Yonghui Zhang, Zi-Hui Zhang, and Hao-Chung Kuo. On the impact of electron leakage on the efficiency droop for AlGaN based deep ultraviolet light emitting diodes. IEEE Photonics Journal, 12(3): 1–7, 2020.
  82. [82]Muhammad Ajmal Khan, Noritoshi Maeda, Yuri Itokazu, Masafumi Jo, Kazuki Iimura, and Hideki Hirayama. Milliwatt-power AlGaN deep-UV light-emitting diodes at 254 nm emission as a clean alternative to mercury deep-UV lamps. Physica Status Solidi (A), 220(1):2200621, 2023.
  83. [83]D Gershoni, CH Henry, and GA Baraff. Calculating the optical properties of multidimensional heterostructures: Application to the modeling of quaternary quantum well lasers. IEEE Journal of Quantum Electronics, 29(9):2433–2450, 1993.
  84. [84]Momme Winkelnkemper, Andrei Schliwa, and Dieter Bimberg. Interrelation of structural and electronic properties in InₓGa₁₋ₓN/GaN quantum dots using an eight-band k·p model. Physical Review B, 74(15):155322, 2006.
  85. [85]Stefan Birner, Tobias Zibold, Till Andlauer, Tillmann Kubis, Matthias Sabathil, Alex Trelakis, and Peter Vogl. Nextnano: general purpose 3-D simulations. IEEE Transactions on Electron Devices, 54(9):2137–2142, 2007.
  86. [86]Crosslight Software Inc. Crosslight Device Simulation Software - General Manual. 2024.
  87. [87]Vladimir A Fonoberov and Alexander A Balandin. Excitonic properties of strained wurtzite and zinc-blende GaN/AlₓGa₁₋ₓN quantum dots. Journal of Applied Physics, 94(11):7178–7186, 2003.

Paper details

vlnc:W_4unc9a7x
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 6

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents