Aluminium incorporation in polar, semi- and non-polar AlGaN layers: a comparative study of x-ray diffraction and optical properties

Abstract

Growth of Al_xGa_1−xN layers (0 ≤ x ≤ 1) simultaneously on polar (0001), semipolar ($10\bar{{\rm{1}}}$3) and ($11\bar{{\rm{2}}}2$), as well as nonpolar ($10\bar{{\rm{1}}}0$) and ($11\bar{{\rm{2}}}0$) AlN templates, which were grown on planar sapphire substrates, has been investigated by metal-organic vapour phase epitaxy. By taking into account anisotropic in-plane strain of semi- and non-polar layers, their aluminium incorporation has been determined by x-ray diffraction analysis. Optical emission energy of the layers was obtained from room-temperature photoluminescence spectra, and their effective bandgap energy was estimated from room-temperature pseudo-dielectric functions. Both x-ray diffraction and optical data consistently show that aluminium incorporation is comparable on the polar, semi- and non-polar planes.

Introduction

AlGaN-based light-emitting diodes (LEDs) operating in the deep ultraviolet (UV) spectral region (λ < 360 nm) have a wide range of potential applications such as water purification¹, disinfection of medical tools², as well as photo-therapy and medical diagnostics³. However, polar (0001) c-plane UV LEDs operating at below 240 nm still have very low external efficiency of less than 0.2%⁴. It is well-known that LEDs epitaxially grown along the [0001] c-direction have strong polarization fields across quantum-well (QW) structures⁵. These fields reduce the electron-hole wave-function overlap resulting in a reduction of the radiative recombination rate. Additionally, polar LEDs operating at λ < 250 nm, the light emission mode changes from transverse electric polarization (E ⊥ [0001]) to transverse magnetic polarization (E || [0001]) resulting in a reduced light extraction efficiency^6,7,8.

Growth on semi- and non-polar planes results in a reduction of built-in fields⁵, which should increase the radiative recombination efficiency of QW active regions. Additionally, the optical polarization of nonpolar ($10\bar{{\rm{1}}}0$) m-plane Al_xGa_1−xN QWs (0 ≤ x ≤ 1) grown on bulk AlN substrates has been found to be dominant along the c-direction over the entire range of composition⁹. However, despite these advantages, there are only a few reports about semi- and non-polar UV LEDs, e.g., a-plane AlN/non-UV-transparent-SiC LEDs operating at 210 nm¹⁰ and semipolar ($11\bar{{\rm{2}}}2$) AlGaN/UV-transparent-sapphire LEDs operating at 307 nm¹¹. One of the reasons is lack of high-quality high-UV-transparent AlGaN and AlN templates prepared on sapphire substrates. Recently, high-temperature thermal annealing has been used to improve the material and optical properties of semi- and non-polar AlN/sapphire templates^{12,13,14,15,16}.

In order to realize UV (visible) emitters, growth of AlGaN (InGaN) QWs with a desired Al (In) composition needs to be well-controlled. In contrast to widely compositional studies for up to twenty various InGaN surface orientations grown on GaN substrates using metal-organic vapour phase epitaxy (MOVPE)^{17,18,19,20,21,22} and ammonia molecular beam epitaxy (MBE)²³, very limited study has been performed for semi- and non-polar AlGaN grown on AlN substrates, e.g., m-plane²⁴, semipolar ($10\bar{{\rm{1}}}2$)²⁵ and ($20\bar{{\rm{2}}}1$)²⁶. This is mainly due to limit of available non-c-plane AlN substrates. Additionally, it should be noted that these substrates are very small and expensive, and UV-light transparency still remains a challenge^27,28. Semi- and non-polar AlGaN layers have also been heteroepitaxially grown on sapphire substrates, e.g., a-plane layers on ($10\bar{{\rm{1}}}2$) r-plane sapphire²⁹, m-plane³⁰ and ($11\bar{{\rm{2}}}2$) layers^31,32 on m-plane sapphire.

Even though semipolar ($\mathrm{10}\bar{{\rm{1}}}\bar{{\rm{3}}}$) AlN templates can be grown on m-plane sapphire¹⁶, crystal twinning has been observed for the layers that might lead to difficulties for AlGaN growth and composition determination. Recently, we have successfully produced untwinned semipolar ($10\bar{{\rm{1}}}3$) AlN templates on m-plane sapphire using directional sputtering³³. Therefore, to extend compositional study of AlGaN with different surface orientations, in this paper, we report on MOVPE-growth of Al_xGa_1−xN layers simultaneously on polar (0001), semipolar ($10\bar{{\rm{1}}}3$) and ($11\bar{{\rm{2}}}2$), as well as nonpolar ($10\bar{{\rm{1}}}0$) and ($11\bar{{\rm{2}}}0$) AlN/sapphire templates over the entire range of composition. Compositional study of these layers has been investigated by x-ray diffraction (XRD), room-temperature photoluminescence (RT-PL) and pseudo-dielectric functions (DFs) measurements.

Results

Determination of Al incorporation by XRD

To vary the aluminium mole fraction (x_AlN) of the grown AlGaN layers, different R_AlGaN = 2 · [TMAl]/(2 · [TMAl] + [TMGa]) gas phase ratios (0 ≤ R_AlGaN ≤ 1) were employed while keeping NH₃ flow rate constantly. Figure 1(a) shows a symmetric ω-2θ XRD scan of an AlGaN layer grown on a ($10\bar{{\rm{1}}}3$) AlN/sapphire template with R_AlGaN = 0.2. Besides the ($\mathrm{30}\bar{{\rm{3}}}0$) diffraction peak of sapphire at 34.1°, there are only the ($10\bar{{\rm{1}}}3$) AlGaN and AlN peaks, indicating that this layer is indeed single phase.

Figure 1

(a) Symmetric ω-2θ XRD scan of an AlGaN layer grown on a ($10\bar{{\rm{1}}}3$) AlN/m-plane sapphire template with R_AlGaN = 0.2. The inset shows azimuthal scans of the {$\mathrm{20}\bar{{\rm{2}}}4$} sapphire and {0002} Al_xGa_1−xN diffraction peaks performed in skew symmetry. (b) 2θ scans of the ($10\bar{{\rm{1}}}3$) Al_xGa_1−xN diffraction peaks of the layers grown on ($10\bar{{\rm{1}}}3$) AlN templates with different R_AlGaN.

To investigate crystal twinning and the epitaxial in-plane relationship of the grown ($10\bar{{\rm{1}}}3$) Al_xGa_1−xN layers and m-plane sapphire, XRD off-axis ϕ-scans were measured. The skew-symmetric {$20\bar{{\rm{2}}}4$} sapphire diffraction peak of m-plane sapphire substrate was measured with a tilt angle of: 32.4°, which indicates [0001]_sapphire. To indicate [0001]_AlGaN/AlN of the ($10\bar{{\rm{1}}}3$) layers, the skew-symmetric {0002} AlGaN peak was measured with a tilt angle of: 31.6°. The inset of Fig. 1(a) shows ϕ-scans of the {$20\bar{{\rm{2}}}4$} sapphire and {0002} AlGaN diffraction peaks of a ($10\bar{{\rm{1}}}3$) layer. Only one peak of {0002} AlGaN is found, which tilts exactly 90° with respect to the {$20\bar{{\rm{2}}}4$} sapphire peak, indicating that this layer is untwinned. The relationship is found to be [$\mathrm{30}\bar{{\rm{3}}}\bar{{\rm{2}}}$]_AlGaN/AlN || [$11\bar{{\rm{2}}}0$]_sapphire and [$11\bar{{\rm{2}}}0$]_AlGaN/AlN || [0001]_sapphire.

Figure 1(b) shows 2θ scans of the ($10\bar{{\rm{1}}}3$) AlGaN layers grown with different R_AlGaN. Various diffraction peak positions indicate different x_AlN of these layers. A similar result has been found for the other layers with different surface orientations.

Semi- and non-polar AlGaN layers hetero-epitaxially grown on sapphire substrates generally have triclinic and orthorhombic distortions of their wurtzite unit cells, respectively. Anisotropy in the lattice and thermal expansion mismatches along two in-plane directions results in anisotropic in-plane strain causing these distortions. This makes lattice parameter measurements, and thus x_AlN determination, difficult. By taking into account these distortions, XRD methods have been developed to determine x_AlN of nonpolar³⁴ and semipolar AlGaN layers³⁵.

For the differently oriented AlGaN layers studied here, their a and c lattice constants have been calculated by measuring different symmetric, skew-symmetric and asymmetric 2θ diffraction peaks, as shown in Table 1. An example of lattice measurements for m-plane AlGaN can be seen in ref. 30. Figure 2(a) shows the measured lattice constants of the layers as a function of R_AlGaN. The lattice constants of all the layers show a linear behaviour with R_AlGaN. Additionally, all layers in this study have an expected ratio of a to c lattice constant with a corresponding composition, indicating that they are fully relaxed.

Table 1 2θ XRD peaks used to measure lattice constants of the AlGaN layers with different surface orientations.

Figure 2

(a) Measured lattice constants of the differently oriented Al_xGa_1−xN layers grown with different R_AlGaN. (b) Calculated x_AlN from XRD data of the differently oriented Al_xGa_1−xN layers plotted as a function of R_AlGaN. Error bars in (a) and (b) are standard errors estimated from the calculations.

Based on these measured lattice constants, x_AlN of all the layers with different surface orientations has been estimated, as shown in Fig. 2(b). At each growth condition (R_AlGaN), x_AlN values of these layers are slightly different. For example, maximum differences (Δx) of 0.02/0.08/0.03 are estimated for the layers grown with R_AlGaN = 0.1/0.4/0.8, respectively. Given these scattered data points, x_AlN values of all the layers can be considered to be comparable. In contrast to a linear behaviour of R_AlGaN-x_AlN observed for c- and m-plane layers grown at 1050°C reported in ref. 30, for the samples studied here grown at 1150°C, a non-linear behaviour has been observed. This is attributed to TMAl:NH₃ pre-reactions and gallium desorption^36,37.

Optical properties

Optical bandgap energy

For wurtzite nitrides, the valence band maximum is split both by spin-orbit interaction and non-cubic crystal field, resulting in three valence-band states (i.e., ${\Gamma }_{7-}^{v}$, ${\Gamma }_{7+}^{v}$ and ${\Gamma }_{9}^{v}$) at the Brillouin zone centre³⁸. For m-plane AlN³⁹ and ($11\bar{{\rm{2}}}2$) AlGaN⁴⁰, the absorption origin for E || [0001] indicates transitions from ${\Gamma }_{7+}^{v}$, while the one for E ⊥ [0001] indicates transitions mainly from ${\Gamma }_{7-}^{v}$ and/or ${\Gamma }_{9}^{v}$. Figure 3(a) exemplifies real parts (<ε₁>) of the DFs measured on the differently oriented AlGaN layers grown with R_AlGaN = 0.2 (x_AlN ≈ 0.1). The <ε₁> parts of m-plane and ($10\bar{{\rm{1}}}3$) layers were measured along [$11\bar{{\rm{2}}}0$]_AlGaN, while they were measured along [$1\bar{{\rm{1}}}00$]_AlGaN for the a-plane and ($11\bar{{\rm{2}}}2$) layers. Compared to the c-plane, a-plane and ($11\bar{{\rm{2}}}2$) layers, the interference fringes of the <ε₁> parts of the m-plane and ($10\bar{{\rm{1}}}3$) layers have weaker amplitudes because of their rougher interfaces³⁰. From these <ε₁> parts, the fundamental bandgap energy (${E}_{{\rm{g}}}^{{\rm{AlGaN}}}$) of the grown layers is approximately estimated from a sharp excitonic E₀ peak^39,40,41.

Figure 3

(a) Real part (<ε₁>) of the pseudo-dielectric functions of semi- and non-polar AlGaN layers (E ⊥ [0001]), and of a c-plane co-loaded layer (E || [0001]) grown with R_AlGaN = 0.2. Effective bandgap (E₀) of the band structure is indicated by an arrow. (b) Bandgap of these layers plotted as a function of x_AlN. The dashed line is a bandgap-bowing fitting of the experimental data with a bowing parameter of 0.9 eV.

${E}_{{\rm{g}}}^{{\rm{AlGaN}}}$ of all the AlGaN layers is plotted as a function of x_AlN in Fig. 3(b). Their ${E}_{{\rm{g}}}^{{\rm{AlGaN}}}$ values are comparable over the entire range of composition. This indicates comparable x_AlN values, consistent with the values estimated by XRD (Fig. 2(b)). The dependence of ${E}_{{\rm{g}}}^{{\rm{AlGaN}}}$ on x_AlN can be described as:

$${E}_{g}^{{{\rm{Al}}}_{x}{{\rm{Ga}}}_{{\rm{1}}-x}{\rm{N}}}=x\cdot {E}_{g}^{{\rm{AIN}}}+(1-x)\cdot {E}_{g}^{{\rm{GaN}}}-b\cdot x\cdot (1-x),$$

where b denotes the bandgap bowing parameter. To fit the experimental data, a measured ${E}_{{\rm{g}}}^{{\rm{AlN}}}$ of 6.11 eV and a measured ${E}_{{\rm{g}}}^{{\rm{GaN}}}$ of 3.42 eV on the ($11\bar{{\rm{2}}}2$) AlN and GaN samples along [$1\bar{{\rm{1}}}\mathrm{00}$]_AlGaN were used, respectively. The shift of ${E}_{{\rm{g}}}^{{\rm{AlGaN}}}$ with x_AlN is well reproduced with a bowing parameter of about 0.9 eV. This value is in good agreement with values reported for a-plane²⁹, m-plane³⁰, ($11\bar{{\rm{2}}}2$)⁴⁰, and c-plane AlGaN layers^30,41.

Photoluminescence

A correlation between the bandgap energy with optical emission properties has also been investigated. Due to the excitation energy of laser (E_ex = 5 eV), only samples grown with R_AlGaN < 0.8 (i.e., x_AlN < 0.7) can be measured. Figure 4(a) exemplifies RT-PL spectra measured on the differently oriented AlGaN layers grown with R_AlGaN = 0.4. The near band-edge (NBE) emission energy of these samples, which was estimated from a Gaussian fit of the corresponding band, is following ($11\bar{{\rm{2}}}0$)_3.92 eV < ($11\bar{{\rm{2}}}2$)_3.98 eV = ($10\bar{{\rm{1}}}3$) < (0001) = ($10\bar{{\rm{1}}}0$)_4.04 eV. This order is slightly different from the order shown in XRD data (Fig. 2(b)). However, the maximum NBE difference is of about 120 meV, which is equal to about a difference of 0.06 in x_AlN. This composition difference is comparable with the maximum Δx of 0.08 estimated from the XRD data.

Figure 4

(a) RT-PL spectra of the differently oriented Al_xGa_1−xN layers grown with R_AlGaN = 0.4. (b) PL peak energy of the layers plotted as a function of x_AlN. A near band-edge energy of AlN at 6.035 eV (⊠) is taken from ref. 42. The dashed line is a bowing fitting of the experimental data with a bowing parameter of 0.9 eV.

The PL emission energy vs x_AlN is also well reproduced with a bowing parameter of about 0.9 eV, as shown in Fig. 4(b). For the bowing fitting, an NBE of 3.42 eV obtained from the grown GaN layers and an NBE of 6.035 eV of c-, a- and m-plane AlN homo-epilayers taken from ref. 42 were used. The PL data correlates very well with the optical bandgap data indicating a random alloy and almost negligible Ga clustering. This is different from the case of InGaN QWs, where a strong In clustering has often been reported, which results in a large Stokes shift between NBE and effective bandgap^43,44,45,46.

Discussion

Of the twenty different semi- and non-polar InGaN QWs MOVPE-grown on bulk GaN substrates investigated so far, compositional study shows different results^{17,18,19,20,21,22}. For example, it has been reported that In incorporation in (0001) < ($11\bar{{\rm{2}}}2$)¹⁹ or (0001) ≈ ($11\bar{{\rm{2}}}2$)^{18,19,20,21,22}, ($10\bar{{\rm{1}}}0$) < ($11\bar{{\rm{2}}}2$)^{17,18,19,20,21} or ($10\bar{{\rm{1}}}0$) ≈ ($11\bar{{\rm{2}}}2$)¹⁸, ($10\bar{{\rm{1}}}0$) < (0001)^20,21 or ($10\bar{{\rm{1}}}0$) ≈ (0001)^18,19, and ($11\bar{{\rm{2}}}0$) < (0001)^21,22 or ($11\bar{{\rm{2}}}0$) ≈ (0001)¹⁸. This is possible that these discrepancies are due to divergence in strain relaxation in the investigated samples and/or indium clustering. For InGaN layers grown by MBE²³, which has completely different growth kinetics from MOVPE, it has been reported that In incorporation in ($11\bar{{\rm{2}}}2$) < (0001) < ($10\bar{{\rm{1}}}0$). So far, there is only one compositional study for ($10\bar{{\rm{1}}}3$) InGaN QWs by MOVPE²¹, indicating that this plane has the lowest In-content among all the aforementioned planes.

To study In incorporation in different InGaN surface orientations, a few theoretical calculations have also been performed by taking into account surface kinetics^47,48 or strain energy dependent surface orientations^49,50. However, they also show contrary results, e.g., In incorporation in m-plane InGaN was found to be smaller⁴⁷ or higher than c-plane InGaN^49,50. Additionally, it has been theoretically^49,50 and experimentally^19,23 reported that different growth conditions (e.g., pressure, temperature, and V/III) might result in different In incorporations on different surface orientations.

Of the relaxed AlGaN layers with five different surface orientations studied here, their x_AlN is comparable over the entire range of composition, as consistently confirmed by XRD and optical data. The comparable x_AlN of the m-plane and c-plane layers is in good agreement with a previous report³⁰, even though the growth temperature used here is 100°C higher. Given the slightly scattered data points of the a-plane and c-plane layers, their comparable x_AlN also can be considered as a consistent result with a previous report²⁹, where only a slightly higher x_AlN of c-plane layers was found (Δx_AlN ≤ 0.05).

For the c-plane and ($11\bar{{\rm{2}}}2$) AlGaN layers studied here, their comparable x_AlN is contrary to previous results reported for ($11\bar{{\rm{2}}}2$) vs c-plane layers, where x_AlN of ($11\bar{{\rm{2}}}2$) layers was found to be lower (Δx ≤ 0.1)³² or higher (Δx ≤ 0.2)³¹ than that of c-plane layers. This might be due to different growth conditions and/or calculation methods used. So far no theoretical study about composition dependent surface orientations has been done for AlGaN. In case of InGaN, most experimental data seems to indicate a higher In incorporation for orientations with almost upright metal dangling bonds. This can indicate that the bonding and incorporation of In versus In desorption are the most important step. Since the AlGaN layers studied here have a similar Al incorporation for all orientations, one may argue that the strong polarity of Al(Ga)N together with the lower total strain facilitates Ga incorporation and makes Ga desorption the less likely process. Further investigations and calculations need to be performed to clarify this.

Conclusions

Compositional study of relaxed co-loaded AlGaN layers with polar (0001), semipolar ($10\bar{{\rm{1}}}3$) and ($11\bar{{\rm{2}}}2$), as well as nonpolar ($10\bar{{\rm{1}}}0$) and ($11\bar{{\rm{2}}}0$) surface orientations has been investigated. By taking into account the compositional effects of anisotropic in-plane strain, aluminium incorporation in semi- and non-polar layers was determined by x-ray diffraction analysis. The AlN mole fraction of all the co-loaded layers estimated by x-ray diffraction is comparable. This is consistent with their comparable optical bandgap energy and near band-edge emission energy, which were determined from room-temperature pseudo-dielectric functions and photoluminescence measurements, respectively. The dependence of the bandgap and emission energy on composition indicates a bowing parameter of 0.9 eV.

Experimental Methods

Growth was performed in an EpiQuest 3 × 2-inch close-coupled showerhead MOVPE reactor. Ammonia (NH₃), trimethylgallium (TMGa) and trimethylaluminium (TMAl) were used as precursors. Differently oriented AlN templates grown on sapphire substrates were used to grow AlGaN layers, including (0001) AlN (d ≈ 800 nm) on c-plane sapphire, ($11\bar{{\rm{2}}}2$) AlN (d ≈ 1000 nm) on m-plane sapphire, ($10\bar{{\rm{1}}}0$) AlN (d ≈ 350 nm) on m-plane sapphire, and ($11\bar{{\rm{2}}}0$) AlN (d ≈ 350 nm) on r-plane sapphire. The ($11\bar{{\rm{2}}}0$) AlN template was grown simultaneously with the ($10\bar{{\rm{1}}}0$) AlN template. Growth parameters of these templates are reported elsewhere³⁰. To produce an Al-polar ($10\bar{{\rm{1}}}3$) AlN template, about 10-nm-thick ($10\bar{{\rm{1}}}3$) AlN layer was initially sputtered onto a 2-inch m-plane sapphire wafer using directional sputtering³³. Afterwards, this wafer was loaded into the reactor chamber to grow a 300-nm-thick AlN layer at a surface temperature of 1290°C.

All the 2-inch AlN/sapphire wafers were diced into 1 × 1 cm² pieces. These pieces were then co-loaded into the reactor chamber for AlGaN epitaxy. Initially, about 100-nm-thick AlN layer was grown on these templates at 1290°C at a reactor pressure of 27 hPa. Afterwards, AlGaN layers with a nominal thickness of 1.5 μm were grown on these templates at 1150°C at a reactor pressure of 100 hPa. To vary x_AlN, different R_AlGaN ratios were employed (0 ≤ R_AlGaN ≤ 1), while keeping NH₃ flow rate constantly. Growth parameters of these layers are reported in ref. 30; however, the AlGaN growth temperature at 1050°C was used in that study.

The crystal orientation of the AlGaN/AlN samples was characterized using a PANalytical X’pert triple-axis high-resolution X-ray diffraction (HR-XRD) system equipped with an asymmetric four-crystal monochromator (4 × Ge220) for CuK_α1 source. On-axis ω-2θ scans have been measured using an open detector without any receiving slit to distinguish between all possible orientations of the epilayers. For lattice calculations, different 2θ diffraction peaks of the samples were measured using an HR analyzer detector, as shown in Table 1.

For room-temperature photoluminescence (RT-PL) measurements, the samples were excited by a Krypton Fluoride (KrF) excimer laser (ExciStar XS-200) with excitation wavelength of 248 nm (E_ex = 5 eV) and a spot size of 50 × 500 μm². During PL measurements, a pulse energy of 7 mJ and a repetition rate of 200 Hz were employed, giving a power density of 5.6 kW/cm². PL signals were recorded by a high-sensitivity Ocean Optics spectrometer (QE65 Pro).

The fundamental bandgap energy of the layers was estimated from real and imaginary parts of the pseudo-dielectric functions (DFs). DFs were recorded at RT using a Horiba UVISEL 2 spectroscopic ellipsometer at an incident angle of 70° and a spot size of 705 × 2030 μm². The photon energy was varied from 1.45 to 6.45 eV with the spectral resolution of 0.02 eV.