Strongly reduced V pit density on InGaNOS substrate by using InGaN/GaN superlattice

Highlights

• InGaN/GaN superlattice was used as a buffer layer on InGaN pseudo-substrate.

• The V pit density was reduced by one order of magnitude.

• (0 0 2) XRD ω scan linewidth reduces from 3000 arcsec to 780 arcsec.

• InGaN based quantum wells have an internal quantum efficiency of 6.5% at 624 nm.

Abstract

The InGaN pseudo-substrate, namely InGaNOS (InGaN On Sapphire), is used to enhance the In incorporation rate in In_yGa_1-yN/In_xGa_1-xN multiple quantum wells (MQWs) to get red emission for micro-display applications. However, the starting material for the InGaNOS fabrication is a non-optimized In_0.08Ga_0.92N layer grown on GaN on sapphire substrate which exhibits V shaped defects (V pits). Such V pits remain afterwards in the final InGaNOS substrate. We demonstrate here that In_xGa_1-xN/GaN superlattice has the potential to cover or fill the native V pits while maintaining a pseudomorphic growth. A combination of a thin GaN interlayer and an InGaN layer in a slight tensile strain state for each pair of the superlattice is necessary to achieve this goal. In addition, it is shown that the presence of GaN interlayers improves the material quality and the surface roughness. (0 0 2) X-ray diffraction rocking curve linewidth reduces to 780 arcsec compared to 3000 arcsec for the substrate. Finally, In_yGa_1-yN/In_xGa_1-xN multiple quantum wells grown on In_xGa_1-xN/GaN superlattice buffer layer on InGaNOS 3.205Ȧ substrate shows a central emission wavelength, measured by photoluminescence, of 624 nm at 290 K with an optical internal quantum efficiency value of 6.5%.

Introduction

Micro-displays for virtual and augmented reality (AR/VR) is a new emerging application for inorganic light emitting diodes (LEDs) [1]. Their high brightness is a key point for such application compared to their organic counterpart [2]. While for large multi-color displays, the pick and place process can be used to merge different types of materials to get native red, green and blue (RGB) pixels on same wafer, such as III-nitride materials for blue and green pixels and phosphide material family for red pixels, it is no more possible to use this technique for AR/VR applications. Indeed, in this case, pixel size has to be reduced below 10 × 10 µm². Color conversion is a possible approach to overcome this drawback by combining blue micro-LEDs and nanophosphors or quantum dots. However, their lifetime is limited and their deposition not easy for such a pixel size. The three primary colors should then be achieved with the same material family in a monolithic approach. In_xGa_1-xN alloy seems to be the best candidate as it can theoretically cover the whole visible range by tuning its InN mole fraction x. However, as it is well known, the so-called green gap prevents this goal at the moment [3]. While blue LEDs provide high quantum efficiency, the efficiency drops dramatically for emission wavelength beyond 500 nm. Indeed, strong material degradation is observed for high InN mole fraction In_xGa_1-xN based quantum wells (QWs). This is mainly due to the low miscibility of InN in GaN [4] and to the high lattice mismatch between GaN buffer layer and InGaN wells. Lymperakis et al. [5] have in addition demonstrated that the maximum InN mole fraction x of an In_xGa_1-xN layer coherently grown on GaN is 25% regardless of the growth conditions. Above this value, additional defects [6] and local alloy fluctuations [7] are usually observed, and even phase separation [8]. It has also recently been pointed out that alloy fluctuations have an important impact on the internal quantum efficiency [3], [9]. Another important point is the presence of an internal electric field in the QWs which leads to the quantum confined Stark effect (QCSE). It permits to obtain redshifted emission but this is at the expense of the quantum efficiency [10]. One of the main solutions could be to reduce the strain in the overall structure to limit the compositional pulling effect [11], [12], [13] and thus increase the In incorporation rate without forming more defects. As the QCSE is mainly due to piezoelectric polarization in InGaN based QWs, reducing the strain will in the same time decrease the QCSE by comparison to fully strained QWs of same In content [14]. To obtain this strain release, one of the best ways is to use an InGaN pseudo-substrate [14], [15]. Many attempts have already been done starting either from a GaN template or a sapphire substrate but no satisfactory results have been demonstrated yet [16], [17], [18], [19], [20], [21], [22], [23]. Soitec has developed an InGaN pseudo-substrate, namely InGaNOS (InGaN On Sapphire), based on its Smart Cut™ technology, at first to reduce the droop in high brightness blue LEDs. This novel substrate offers a thin partly relaxed InGaN seed layer that is epiready for metal organic vapor phase epitaxy (MOVPE) of III-N heterostructures. Its a lattice parameter can vary from 3.190 to 3.205 Ȧ at the moment. It can be used to increase the InN mole fraction of In_xGa_1-xN based multiple QWs (MQWs) in order to get long emission wavelength. Indeed, by using a full InGaN structure grown on InGaNOS, it has been demonstrated that the In incorporation rate is enhanced with the a lattice parameter of the substrate [24]. The whole visible range can then be covered with only thin QW width (2.5 nm) by using the appropriate InGaNOS substrate (i.e. the appropriate a lattice parameter) and/or by adapting the QW growth conditions [24], [25], [26]. However, as the InGaNOS substrate is initially fabricated from a donor structure which is composed of an In_xGa_1-yN layer grown on a c-plane GaN layer on sapphire, it experiences the well-known V shaped defect (V pit) assisted relaxation process [27], [28], especially in the case of a 200 nm thick In_xGa_1-yN layer with an InN mole fraction x = 8% which is needed for the fabrication of the InGaNOS substrate with a 3.205Ȧ a lattice parameter (InGaNOS-3.205). In the case of a pseudomorphic In_xGa_1-xN layer grown on a c-plane GaN layer, the main favourable energy relaxation process is the formation of V pits [27], [28]. The presence of In atoms on the strained In_xGa_1-xN layer surface reduces the surface energy required to create $\left(10\overline{1}1\right)$ planes compared to the (0 0 0 1) surface [27]. When enough compressive strain has been stored by the system, six $\left\{10\overline{1}1\right\}$sidewall facets are created to release strain [26]. It happens usually at a dislocation core where tensile strain is present so that In atoms accumulate in this area and form In-N-In chains [29], which, in addition, act as localization centres to prevent carriers from being wasted in non radiative recombination processes [29]. Furthermore, V pit density and V pit diameter increase with the In_xGa_1-xN layer thickness [28]. A V pit density as high as 2 × 10⁸ cm⁻² is usually observed on an In_0.08Ga_0.92N donor structure surface, which is unfortunately preserved on the InGaNOS-3.205 surface after the InGaNOS fabrication. For efficient red LEDs grown on this substrate [30], it is important to prevent the propagation of these extended defects through the whole structure. There are two possibilities to reduce the V pit density: improve the material quality of the In_0.08Ga_0.92N based donor structure or adapt the overgrown buffer layer to fill the V pits.

In this paper, we will show that it is possible to recover a smooth surface with strongly reduced V pit density and better material quality by using a In_xGa_1-xN/GaN superlattice as a buffer layer grown on InGaNOS-3.205. The improved material quality will, in addition, be confirmed by the internal quantum efficiency value of red emitting In_yGa_1-yN/In_xGa_1-xN multiple quantum wells (MQWs).