Fully Relaxed, Crack-Free AlGaN with upto 50% Al Composition Grown on Porous GaN Pseudo-Substrate

Abstract

Fully relaxed, crack free, smooth Al_xGa_1−xN layers with up to 50% Al composition were demonstrated on pseudo-substrates composed of dense arrays of 10 × 10 µm² compliant porous GaN-on-porous-GaN tiles. The AlGaN layers were grown in steps for a total of 1.3 µm. The growth conditions necessary to demonstrate high quality films at higher Al compositions also suppressed any sidewall growth.

1. Introduction

The wide bandgap tunability range of (Al,Ga)N from 3.4 eV for GaN to 6.1 eV for AlN allows for its use in deep ultraviolet (UV) light emitting diodes (LEDs) in fields such as medical disinfection, water purification, biochemical detection and UV curing [1,2,3,4,5,6]. In addition, high power UV laser diodes (LDs) are needed for solar-blind non-line-of-sight communications, UV spectroscopy, laser lithography, and optical data storage [7,8,9,10,11]. Recently, photonics integrated circuits have been of interest for developing self-powered UV photonics systems that can work sustainably and independently with a reduced size and power consumption for future on-chip optical communication [12,13,14,15,16,17,18]. For all these applications, high quality, large area AlGaN substrates are needed. Bulk AlN substrates with a low dislocation density are the ideal candidate [19,20,21,22,23,24]; however, they are currently very expensive and only available in small sizes [25,26]. When AlGaN is grown on foreign substrates such as silicon, sapphire, and silicon carbide, the lattice mismatch usually leads to films with a relatively high density of threading dislocations [27,28]. This is the main cause for non-radiative recombination, leading to the deterioration of the optical and electrical performance of AlGaN UV optoelectronic devices [1,4,27]. When grown hetero-epitaxially on GaN, the lower lattice constant of AlGaN leads to tensile strain and cracking with epilayers thicker than the critical thickness [29,30]. Using lateral overgrowth of buried cracks, relaxed Al_0.2Ga_0.8N layers on GaN templates were formed using plastic relaxation [31]. Lateral growth of AlGaN on porous GaN or AlN base layers has also been pursued in an effort to reduce the dislocation density [32,33]. Several research groups have also attempted to use a sacrificial layer for flip chip UV LEDs [34].

Recently, 10 × 10 µm² patterned compliant porous GaN pseudo-substrates (PSs) were formed [35]. Due to its high surface-to-volume ratio, porous GaN exhibits reduced mechanical stiffness with increased porosity [36], similar to porous silicon [37], facilitating the elastic relaxation of InGaN grown on porous GaN PSs [38,39]. On these patterned compliant pseudo-substrates, the growth of 1.3 µm thick crack free, elastically relaxed AlGaN with a low Al composition was also demonstrated [40]. In this work, we performed the deposition of fully relaxed AlGaN with an Al composition of up to 50% on similar GaN-on-porous-GaN PSs.

2. Materials and Methods

The layer structure used for porosification as the pseudo-substrate consisted of a 2.8 µm thick unintentionally doped (UID) GaN on c-plane sapphire followed by 800 nm Si doped GaN with a doping of 5 × 10¹⁸ cm⁻³ which was capped with 100 nm of UID GaN. The sample was patterned and dry etched with a 100 W BCl₃/Cl₂ etch, forming 10 × 10 µm² square tiles 2 µm apart and with an etch depth of 600 nm. The sidewalls of the tiles were aligned parallel to the orthogonal GaN [1$\overline{1}$00] and GaN [11$\overline{2}$0] planes, respectively. The GaN:Si layers were then porosified with a doping selective electrochemical (EC) etch in 0.3 M oxalic acid with metal contact to the sample with anode and Pt wire acting as the cathode [41]. The current flowing through the sample caused a redox reaction resulting in GaN:Si etching, which formed GaN tiles with embedded porous GaN as depicted in the schematic in Figure 1a and as reported previously [35].

The epitaxial layers in this study were grown by metal organic vapor phase epitaxy (MOVPE) in a close coupled showerhead reactor using the precursors trimethylgallium (TMGa), trimethylaluminum (TMAl), and ammonia (NH₃) on the c-plane porous GaN pseudo-substrates at a pressure of 100 Torr. In our previous work, a total of 1.3 µm thick elastically relaxed AlGaN with x_Al = 0.18 was demonstrated using a growth temperature of 1080 °C and a mixture of hydrogen and nitrogen as the carrier gas [40]. To obtain good quality AlGaN films with a higher Al composition, the deposition temperature was increased to 1155 °C [42,43], and the NH₃ flow was decreased to 45 mmol/min from 178 mmol/min [40]. The molar flow of TMGa varied at 8.8–14.6 µmol/min and TMAl was in the range of 0.98–2.4 µmol/min, respectively. However, growth under these conditions on the GaN-on-porous-GaN PSs resulted in the formation of circular islands on the sample surface (Figure 2a). While the higher growth temperature of 1155 °C was necessary to achieve high quality AlGaN at the targeted composition, it was not suitable for deposition on the GaN-on-porous-GaN PSs. In order to achieve smooth layers, a two-step process was implemented, where the initial 120–130 nm of AlGaN was grown at 1080 °C in a nitrogen ambient atmosphere, after which the temperature was increased to 1155 °C for the rest of the run and the subsequent AlGaN growth was performed in a hydrogen ambient atmosphere.

The 1.3 µm thick AlGaN layers were grown in steps, and before every regrowth the sample was solvent cleaned with acetone and isopropanol which was followed by a buffered HF dip and rinsing with deionized water to remove contaminants and the oxide layer introduced during characterization. The surface morphology was characterized using an optical microscope and atomic force microscope (AFM). High resolution x-ray diffraction (XRD) was used for rocking curve measurements to evaluate the film quality. Reciprocal space maps (RSMs) were taken using XRD around the GaN ($\overline{1}\overline{1}$24) reflection. Composition and relaxation were determined from these RSMs using X-Pert Epitaxy software [44,45].

3. Experiment and Results

In the first experiment, the growth was initiated with 10 nm of GaN, which acted as a transition layer before introducing any strain in the structure, followed by 30 nm thick AlGaN layers where the composition was increased in steps with target al. composition (x_Al) = 0.05, 0.09, 0.13 and 0.17 at 1080 °C. Afterwards, the temperature was increased to 1155 °C and 30 nm each of AlGaN with target x_Al = 0.21 and 0.25 was grown followed by 330 nm of AlGaN with a target al. mole fraction of 0.30. The target x_Al was obtained from fully strained AlGaN on GaN calibration samples. The sample was characterized and a further 400 nm of AlGaN was regrown with adjusted TMAl and TMGa flows, so as to decrease x_Al and match the lattice constant to the existing epilayer. An additional regrowth was then performed for another 400 nm of AlGaN to bring the total thickness of the AlGaN layer to 1.3 µm. The schematic of the full epitaxial layer structure grown on the pseudo-substrate is shown in Figure 3a.

Evaluation of the initial ~500 nm thick AlGaN structure revealed smooth, crack free layers (Figure 3b) with x_Al = 0.35 and a relaxation of ~85%. The increase in measured Al composition over the targeted composition was expected due to the composition pulling effect [40,46,47,48,49,50]. The surface morphology remained smooth after the first and second regrowth (Figure 3c,d respectively). After the first regrowth (which brought the total thickness to ~900 nm), the AlGaN peak corresponded to x_Al = 0.32 with ~96% relaxation. The final structure with ~1.3 µm thick AlGaN was fully relaxed and showed x_Al = 0.32 (Figure 3e).

With encouraging results having been obtained so far, a second porous PS sample was used to grow AlGaN with a higher x_Al, following the structure depicted in the schematic in Figure 4a. A 10 nm GaN transition layer was again grown followed by a step-graded AlGaN layer composed of 25 nm AlGaN layers with target x_Al = 0.05. 0.11, 0.17, 0.23, and 0.29 at 1080 °C under nitrogen. As before, the temperature was then increased to 1155 °C and 25 nm AlGaN each with target x_Al = 0.35, and 0.41 were grown followed by 325 nm of AlGaN with a target x_Al of 0.47 in hydrogen. AFM images demonstrated a smooth surface morphology, and RSM showed an AlGaN peak at x_Al = 0.53 with 89% relaxation (Figure 4b). After the solvent was clean and buffered in an HF dip, an additional 400 nm of AlGaN was grown with a slightly reduced TMAl flow to lattice match with the already grown AlGaN layer. The aluminum composition was 50% with 90% relaxation (Figure 4c). A further 400 nm regrowth increased the relaxation to 97% while maintaining the same aluminum content (Figure 4d,e).

4. Discussion

Both the 1.3 µm thick AlGaN samples with x_Al = 0.32 and 0.50 showed no signs of any cracking, and were fully relaxed. For the sample with x_Al = 0.32, AFM images showed a slight increase in roughness, presented as the root mean square (RMS) value in Figure 5 after the individual regrowth steps. The RMS roughness values increased slightly from 1.1 nm for the first ~500 nm stack to 1.4 nm after the first regrowth and decreased to 1.2 nm after the second regrowth, which is likely due to the AlGaN layer being already fully relaxed. The sample with x_Al = 0.50 showed an appreciable increase in roughness with the RMS value rising from 1.2 nm for the initial AlGaN stack to 2.7 nm after the first regrowth (with a total of ~900 nm AlGaN). The second regrowth stage showed no further increase in roughness. While the increase in surface roughness after the first regrowth step was unclear, the circumstance that the roughness did not further increase during the second regrowth suggests that the relaxed underlying AlGaN layer (after the first regrowth) prevented any stress-driven increase in surface roughness in this study, which is often observed when growing strained AlGaN layers.

Further XRD characterization to determine the ω-FWHM of the (002) and ($\overline{2}$01) reflexes for the sample with x_Al = 0.32 showed 418 and 544 arc seconds, respectively, on par with good quality hetero-epitaxial growth [51,52]. Nevertheless, these values were higher than the underlying GaN layer for which ω-FWHM values of ~280 and 386 arc seconds for the (002) and ($\overline{2}$01) reflexes, respectively, had been recorded. While the somewhat higher FWHM values could be the result of partial plastic relaxation, they could also be caused by local variations in relaxation and in the Al composition, between the edge and center region of an individual tile, or tile to tile variations in the array. In addition, the elastic relaxation process could have led to tile tilt causing an increase in the ω-FWHM values.

For the sample with x_Al = 0.50, the (002) ω-FWHM amounted to 385 arc seconds, which was narrower than the value measured for the sample with x_Al = 0.32. In contrast, the ($\overline{2}$01) ω-FWHM increased to 677 arc seconds. This is a much more significant increase compared to the value of 386 arc seconds measured for the GaN base layer and the 544 arc seconds measured for the sample with x_Al = 0.32, which may be due to partial plastic relaxation under the formation of edge dislocations. Further investigations are currently underway to clarify the origin of the increase in FWHM, as well as to develop strategies to suppress it.

There was little evidence of growth on the tile sidewalls (Figure 6) which was observed previously when the AlGaN layers were grown with a higher NH₃ flow of 178 mmol/min of NH₃ [38]. Lower NH₃ flow rates are known to suppress lateral growth [53] and could have been the cause for the lack of sidewall growth in this work. Further studies are planned to investigate this aspect, as sidewall growth may allow for tile coalescence to create large-area planar relaxed AlGaN templates.

5. Conclusions

Fully relaxed, crack-free 1.3 µm thick AlGaN layers with x_Al = 0.32 and 0.50 were formed on pseudo-substrates composed of 10 × 10 µm² compliant GaN-on-porous GaN tile arrays. The epilayers were smooth for x_Al = 0.32, while an increase in surface roughness and potential onset of partial plastic relaxation were observed for the sample with x_Al = 0.50. Further optimization of the step-graded layer and lateral growth can open up a pathway to obtain high quality, large area AlGaN substrates as well as obtain fully relaxed AlGaN with even higher values of x_Al.