InGaN/AlInN interface with enhanced holes to improve photoelectrochemical etching and GaN device release

We introduce a novel superlattice structure for releasing GaN-based devices with selective photo-electrochemical (PEC) etching by incorporating a lattice-matched AlInN barrier in an InGaN/GaN sacrificial stack. A dopant-free two-dimensional hole gas is formed at the InGaN/AlInN interface due to the band bending and strong polarization discontinuity, which is revealed in simulations. PEC etching using the four period InGaN/AlInN superlattice exhibits almost three times higher etch rate and smoother etched surfaces when compared to conventional InGaN/GaN release layers. A systematic investigation with different AlInN layer thicknesses shows that a thin AlInN layer is able to achieve smooth surface with uniform etch process during the PEC while thicker AlInN exhibits poorer surface morphology although the etch rate was faster. Furthermore, it is found that using HNO3 as the electrolyte improved the etched surface smoothness compared to KOH when followed by a post-release HCl treatment. This structure will enable the release of high quality GaN layers and the fabrication of novel optical devices.


Introduction
Over the past decades GaN-based alloys have been widely used in light-emitting diodes (LEDs) [1], lasers [2], and high power electronics [3] due to their unique optoelectronic properties such as wide range of bandgap energies, efficient light emission, high electron mobility, and as well as their * Authors to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. mechanical properties such as hardness, high thermal conductivity, and high mechanical strength. These materials are commonly grown on foreign substrates such as sapphire [4], SiC [5] and Si [6] due to their low cost. However, due to the lattice and thermal expansion mismatch between the nitrides and the substrates, such hetero-epitaxy leads to the formation of high-densities of dislocations and residual stresses in the epi-layers, which reduces the performance of devices [7,8]. To achieve higher performance, III-N devices can be grown on native GaN substrates but the cost of these substrates is a huge barrier for commercial applications. Therefore, the release of GaN-based epi-layers from free-standing GaN substrates or high-quality GaN/sapphire templates can be beneficial by reclaiming and reusing the original substrates to significantly reduce the production costs [9,10]. In addition, thin-film devices removed from bulk substrates exhibit superb optical performance due to their flexibility, thinness and transparency, which enables many new microcavity configurations [11]. It also facilitates the integration of thin-film devices with flexible or heat-conductive backplanes, as well as heterogeneous integration with other components, such as for multicolour emission on a single chip [12][13][14] using transfer printing. There are several methods for releasing III-N materials from their substrate, such as laser lift-off and chemical etching including electro-chemical (EC) etching and photoelectrochemical etching (PEC). Chemical etching, in particular, is a cost-effective and gentle approach that offers several benefits over laser lift-off as it avoids subjecting the GaN layers to thermal or mechanical stress which may degrade its quality. It also allows precise device registration and tethering during the releasing, which is important for the following integration process. But this usually cannot be easily achieved using laser lift-off technique unless multiple times of wafer bonding steps are used. While releasing thin-films of III-N heterostructures grown on Si can be performed easily using wet etching due to the high selectivity between nitrides and silicon [15], EC [16] or PEC etching [17,18] has been used for releasing GaN-based structures from substrates where selective chemical etching is difficult to achieve (e.g. sapphire, SiC, bulk III-N [19]. PEC etching relies on the generation of holes in the semiconductor material through photon absorption. The holes can diffuse to the surface where they act as the oxidizing agent, breaking the semiconductor bonds and creating volatile products that result in its etching. Therefore, the effectiveness of PEC etching depends heavily on the hole concentration and their ability to reach the surface with the electrolyte [16]. A lot of investigation has been done by using different sacrificial layers in PEC-etching. For instance, a thin (⩽50 nm) InGaN layer has been used as the release layer [20,21], to achieve uniform etching. However, such a thin layer usually does not allow the easy detachment of films as they can easily re-bond with the substrates by van-der-Waals forces. A thicker InGaN sacrificial layer is preferred but its growth is more difficult due to strain relaxation effects. The photo-generated holes cannot sufficiently diffuse and hence recombine before reaching the semiconductor-electrolyte interface [22]. As a solution, a superlattice structure of InGaN/GaN quantum wells (QWs) was implemented as a thick release structure [23,24]. Despite the increased confinement of holes, the material removal is rather non-uniform at the interface [25]. Therefore, a release layer for PEC-etching that can provide a smooth surface underneath the released devices is needed to both minimize scattering loss as well as to have a better adhesion onto the target substrates [26,27] for the transfer-print process is desired.
In this work, we introduce a band-engineering approach to enhance the hole carrier density inside the QWs. A latticematched AlInN layer is introduced as the lower barrier to the InGaN QW, which results in a high density of holes at the InGaN/AlInN interface due to band bending associated with the strong polarization discontinuity in the c-plane orientation. The structure is implemented in a relatively thick sacrificial superlattice structure for PEC etching with HNO 3 as the electrolyte. This leads to a root mean squared (RMS) roughness on the released surface of 2.5 nm, which is comparable to that of the as-grown epitaxial material (∼1.9 nm) and at the same time exhibits an increase in the etch rate by factor of three compared to an InGaN/GaN release structure. This study shows that thicker AlInN layers exhibit higher etch rate but much poorer etch homogeneity with the optimal thickness being 7.5 nm for having a smooth surface and a uniform etching. Different electrolytes were compared where HNO 3 resulted in smoother surfaces compared with potassium hydroxide (KOH).

Band-diagram simulation
A dopant-free two-dimensional hole gas (2DHG) can be formed in III-nitride heterostructures due to spontaneous and piezoelectric polarization discontinuities at their interfaces. As a recent example, strained metal-polar GaN on AlN revealed that holes can be formed at the negative polarization difference interface without the need for acceptor doping [28]. In addition, high densities of 2DHG were observed at pseudomorphic (In)GaN/AlN heterostructures.
AlInN is one of the important nitride-based alloys, which can be lattice-matched to GaN at an In content of ∼18%. It exhibits a strong spontaneous polarization (P sp ) when grown on GaN, and is used to enhance the two-dimensional electron density for electronic devices [29]. Here, we investigate the interface of InGaN (nominal In content 17%) grown on nominally lattice-matched AlInN to enhance the hole concentration. Figure 1(a) illustrates the structures compared here: GaN (10 nm)/InGaN (2 nm)/(10 nm) GaN and GaN (10 nm)/AlInN (variable)/InGaN (2 nm)/(10 nm) GaN. AlInN thicknesses of 7.5, 15 and 30 nm were considered. The energy band diagrams for the structures above were simulated under an applied bias of 2 V using SiLENSe, a self-consistent one-dimensional Poisson solver [30], and are presented in figure 1(b) together with the calculated hole concentration for each structure.

Sample fabrication and PEC etching
Our simulations indicated that hole formation could be favored for a layer of InGaN grown on a layer of AlInN. We thus experimented with PEC etching on such structures. Four structures (labeled as A, B, C and D) were investigated in this work. The epi-layers consist of a 500 nm unintentionally doped GaN (u-GaN)/ 1.5 µm-thick n-doped GaN (n-GaN) buffer layer, four pairs of either InGaN/GaN or InGaN/AlInN/GaN stacks as the sacrificial layers, followed by 1 µm u-GaN layer to be separated. The sacrificial layer in sample A is a conventional InGaN (2 nm)/GaN (10 nm) configuration while samples B, C and D are /InGaN (2 nm) /AlInN/ GaN (60 nm) with AlInN thickness of (7.5, 15 and 30 nm) ±1 nm respectively. The nominal indium concentration in the InGaN layer was 17% which is designed to absorb the 405 nm light from the illumination source during the PEC etching, and 18% in AlInN layer to obtain lattice-match conditions with GaN. To prepare the samples for PEC etching, arrays of 50 µm × 50 µm and 20 µm × 100 µm mesas were formed using a 300 nm-thick SiN x as a hard mask, followed by inductively coupled plasma etching through the sacrificial layers to the top of the n-GaN current spreading layer. A mixture of In and Ga was used as the positive contact pad on the n-GaN while a Pt mesh in the electrolyte solution was used as the negative electrode. The PEC etching process uses an electrolyte solution, an applied bias and the illumination of sample with an external optical source [17]. Here, the samples were illuminated from the front side by a focused 405 nm LED with a power density of 105 mW cm −2 , and an applied bias of 2 V. The photon energy of the excitation source is higher than the bandgap of the sacrificial InGaN but less than the GaN in order to achieve the band-gap selective PEC etching. The electrolyte used was either KOH or HNO 3 solutions. The released films were picked up using polydimethylsiloxane (PDMS) stamps allowing characterization of the exposed N-polar GaN surface using atomic force microscopy (AFM) and scanning electron microscopy (SEM).

Band-diagram and carrier concentration
The net polarization field is the sum of P sp and the lattice-mismatch induced piezoelectric polarization (P pz ), i.e. P = P sp + P pz . At the InGaN/GaN interface, the polarization discontinuity is P = P InGaN sp + P InGaN pz − P GaN sp , while for the InGaN/AlInN interface, assuming that AlInN is lattice matched with GaN, P = P InGaN sp + P InGaN pz − P AlInN sp . The sign of the net polarization of InGaN is opposite to AlInN resulting in a net increase in charge at the interface. The calculated net polarization field in the InGaN layer is 1.7 MV cm −1 for the InGaN/GaN interface and increases to 2.5 MV cm −1 for InGaN/AlInN interface. As depicted in figures 1(d)-(f), the presence of AlInN together with the polarization induced at the interfaces results in an electric field across the AlInN layer lifting the valence band maximum of the InGaN QW to the Fermi level resulting in calculated 2DHG densities of 1.3 × 10 13 cm −2 at the InGaN/AlInN (7.5 nm) interface, while the hole density is only 37 cm −2 for the conventional InGaN/GaN structure. A low hole density for a thin InGaN QW agrees with previous results reported [28]. It seen that, with increasing AlInN thickness, the corresponding hole density increased to 1.9 × 10 13 cm −2 (15 nm) and 2.3 × 10 13 cm −2 (30 nm). The value for hole densities have been calculated by integrating the peak area of hole concentration in figure 1. Thus, inserting an AlInN lower barrier is beneficial to enhancing the hole density for a fixed thickness of the InGaN layer. Moreover, the induced triangular shape of the energy bands in the InGaN layer leads to a reduced overlap of the electron and hole wave functions. Hence, electron and holes are separated to opposite sides of the QW reducing the recombination probability. Therefore, due to band bending, large number of holes are available to reach the InGaN electrolyte interface and finally contribute to the etching process.

PEC etch comparison without and with AlInN.
To investigate how the different sacrificial structures affect the releasing process, PEC etching with sample A (InGaN/GaN) and B with AlInN (7.5 nm)/InGaN (2 nm) was carried out under identical conditions, i.e. using 0.01 M HNO 3 with a bias of 2 V. As depicted in figure 2(a), the photocurrent initially decreases exponentially and then declines slowly towards a steady value due to the complete undercut of films in the focused region. Since a large area of the sample was in the solution, the non-zero current here is attributed to the photogenerated carriers by the scattered light outside of the focused region. It has been found that the etch rate for sample B was about three times higher than that in sample A, which was also reflected by the photocurrent level during etching as well as the total etching time. Note that photocurrent is proportional to the reaction rate at semiconductor /electrolyte interface, according to Faraday's law of electrolysis.
The fully undercut devices were picked up by a PDMS stamp and the backside morphology was examined by AFM, as shown in figure 2(b). The surface of sample A after PEC etching exhibits a high density of islands, with the RMS roughness of 5.3 nm, whereas sample B exhibits terrace-like morphology, with RMS roughness of 2.5 nm. Taking into account that the as-grown surface exhibits a roughness of 1.9 nm, the etched surface from sample B is quite smooth. This is attributed to the high density of polarization charges in combination with the photo-induced holes improving the uniformity of the etching, leading to a smoother released surface. a higher etch rate than that without AlInN (sample A), which is consistent with the previous observations when HNO 3 was used as the electrolyte. With increasing AlInN layer thickness the etch rate increased (1.8 → 2.3 µm min −1 ) and then saturated (2.3 → 2.5 µm min −1 ). This result agrees with simulations exhibiting increasing hole densities versus the AlInNthickness. Based on the optical and SEM images shown in figure 3, the best surface morphology after releasing was observed in sample B with a 7.5 nm thick AlInN layer. The etching homogeneity became worse with increasing AlInN thickness, where a rougher surface was observed. A possible reason might relate to the challenges in growing thick AlInN layers. Due to the different optimal growth conditions for AlN and InN, the AlInN layer can suffer from compositional inhomogeneity and poor surface morphology [31]. Such issues become more severe with increasing layer thickness, which eventually has implication on the homogeneity of PEC etching. One observation is that 'whiskers' are formed on the etched surface and their density becomes higher for thicker AlInN samples. These dislocation-related whiskers indicate the degradation of the crystalline quality in the samples with thicker AlInN layers. Therefore, it can be concluded that although thicker AlInN is beneficial for enhancing the PEC etch rate, it results in rougher interfaces and more inhomogeneous etching.

Effect of electrolyte on PEC etching.
To compare the effect of the different electrolytes on the etch rate and etched surface morphology, KOH and HNO 3 were employed using samples with structure A while keeping the other conditions identical. It has been reported that a lower concentration of KOH results in a smoother surface [32], therefore, a concentration of 0.01 M for both KOH and HNO 3 was used. As seen in figure 4(a), using KOH resulted in the formation of pyramid-like surface which is related to the exposure and etching of N-polar GaN. This phenomenon was also observed in PEC etching with structure B as shown in SEM cross-sectional images in figure 3. However, it was not observed when etching with HNO 3 as shown in figure 4(b). Note that KOH can chemically etch N-polar GaN in addition to the photo-assisted etching. The acidic electrolyte does not chemically etch the material leading to different morphologies in an acid and a base solution. Figures 4(c) and (d), show that the RMS roughness of the etched surface was reduced from 7 nm with KOH to 5.3 nm with HNO 3 , revealing that using an acidic electrolyte can result in smoother surface. SEM images in figure 4(d), shows small particles on the etched surface after etching by HNO 3 . These particles were also observed in PEC etching with other acid solutions such as H 2 SO 4 and HCl (not shown here). Based on the literature and our energy dispersive xray analysis, suggests that these particles might be Ga oxides which could be formed during PEC etching process. To further reduce RMS roughness, a post treatment can be beneficial to remove Ga oxides. The solubility of those oxides varies between acidic or basic solutions depending on whether they dissolves to form a soluble complex [33]. Although KOH or tetramethylammonium hydroxide (TMAH) is reported as a post treatment to remove the residuals [34], these solutions could result in a rough surface due to the chemical etch of Npolar GaN. Here, HCl (60 • C) for 15 min was used to dissolve these residuals, followed by dipping into acetone for 20 min to remove the created GaCl 3 particles. As seen in figure 4(e) the particles were sufficiently removed with the measured roughness reduced to 4.5 nm after post treatment.

Conclusion
In summary, a sacrificial stack with InGaN/AlInN/GaN configuration is introduced for enhanced hole carrier confinement at the InGaN/AlInN interface to enhance the PEC etching process. Simulations indicate a high-density of holes at the interface of InGaN/AlInN due to the strong polarization difference. It should be noted that reducing the In content in the AlInN alloy results in an increased spontaneous polarization charge together with tensile strain and could be used to balance the compressive strain of the InGaN QW. Correspondingly an increase in the In content of the AlInN layer reduces the spontaneous polarization charge and adds additional compressive strain.
We found samples with sacrificial layers containing AlInN exhibit higher etch rates and much smoother RMS roughness on the etched surface when compared to the conventional InGaN/GaN release structure. PEC etching with different AlInN thicknesses shows that the thin layer (i.e. 7.5 nm AlInN) is better for obtaining smooth etched surface and uniform etch profile while thicker AlInN exhibits poorer surface morphology although etch rate was even faster. Moreover, an HNO 3 electrolyte results in smoother surface by preventing the formation of pyramid-like features on the N-polar GaN layer. A post treatment in HCl can be used to remove the residual oxides after PEC etching, to further reduce the roughness.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).