Log in / Register
Home arrow Engineering arrow III-nitride semiconductors and their modern devices

Nearly lattice-matched InAlN/(Al)GaN distributed Bragg reflectors

Growth properties

As already mentioned, InAlN with an In content of ~18% is lattice-matched to GaN while exhibiting a refractive index contrast to GaN of 7-8% in the blue- green wavelength range (Carlin and Ilegems, 2003). This makes such an alloy of great interest for the realization of III-nitride-based high-quality DBRs. Other approaches usually rely either on the use of AlGaN/GaN (Waldrip et al., 2001;

Natali et al., 2003) or AlN/GaN (Ng et al., 2000) bilayers. The main drawback of such systems is the large in-plane lattice parameter mismatch between the different layers composing the DBR, causing a large build-up of the strain with an increasing number of pairs, which eventually results in cracks and misfit dislocation generation. A strain-compensation technique based on superlattice interlayers has been proposed to improve the quality of AlN/GaN DBRs (Huang et al., 2006), and it has even led to the demonstration of blue cw electrically driven vertical cavity surface-emitting lasers (VCSELs) on sapphire (Lu et al., 2010). However, such a method is not suited for devices grown on low-dislocation-density FS-GaN substrates since the accumulated strain with this bilayer system would inevitably generate threading dislocations, without resulting in a significant gain over growth carried out on a sapphire substrate.

InAlN-based DBRs with a stopband centered either in the UV (Feltin et al., 2006) or in the visible (Butte et al., 2005) range have been demonstrated on c-plane sapphire. Reflectivity spectra measured on such DBRs are depicted in Fig. 6.9. For the realization of UV DBRs, AlGaN is used as the second layer together with InAlN in order to avoid GaN absorption below 370 nm. Such UV DBRs have been successfully used to demonstrate RT polariton lasing in nitride- based microcavities (Christopoulos et al., 2007; Christmann et al., 2008), and DBRs with a stopband in the visible region have been used for the realization of electrically driven resonant-cavity light-emitting diodes (Dorsaz et al., 20056) and optically pumped VCSELs (Feltin et al., 2007a).

The transfer of such DBRs on FS-GaN substrates would allow a great increase in the device performances, due to the much lower threading dislocation density that would directly benefit the device internal efficiency, and improved thermal management. However, the growth of such DBRs on FS-GaN is not straightforward. Indeed, having fewer threading dislocations also means a reduced residual strain accommodation. In this context, a careful growth optimization of such

Reflectivity spectra measured on three different InAlN-based DBRs (see text for details). The GaN absorption edge is shown as a reference. (Reprinted, with permission, from Butte et al. (2009).)

Fig. 6.9. Reflectivity spectra measured on three different InAlN-based DBRs (see text for details). The GaN absorption edge is shown as a reference. (Reprinted, with permission, from Butte et al. (2009).)

DBRs on FS-GaN has to be performed. It has been found that the indium-to- aluminum precursor gas flux ratio has to be large in order to ensure flat interfaces and a smooth DBR top surface as well as an optimum peak reflectivity (Cosendey et al., 2011). This has been ascribed to indium surface segregation—a well-known effect occurring in indium-containing ternary alloys of the III-nitride and the III- arsenide family, which is due to the much larger bond length involving indium atoms compared to aluminum or gallium ones (Massies et al., 1987). A possible explanation for the way indium surface segregation affects the quality of InAlN-based DBRs is the following. In the presence of a few InAlN atomic layers with an indium content much lower than the LM value, at the beginning of each InAlN layer the lattice-matched condition in such sublayers is no longer fulfilled, so that a large tensile strain builds up. Then, depending on the indium surface segregation coefficient, the total thickness of those indium-poor sublayers might eventually reach the critical thickness where strain is no longer accommodated by elastic deformation but causes the generation of misfit dislocations. Subsequently, it has been assumed that increasing the indium-to-aluminum precursor gas ratio reduces the indium surface segregation coefficient by increasing the amount of indium adatoms on the sample surface during growth. By doing so, the nominal indium content satisfying the lattice-matched condition is reached before the critical thickness.

Another point that has to be stressed once more concerning the MOVPE growth of InAlN-based DBRs is the large-growth temperature difference between GaN (about 1050°C) and InAlN (about 850°C). This means that when considering the most basic approach for growing such DBRs the surface of each InAlN layer is subject to a temperature ramp up to 1050°C. This is above the thermal stability of lattice-matched InAlN, which is about 960°C (Gadanecz et al., 2007). Indeed, such an annealing step greatly affects the InAlN surface morphology (Brice et al., 2010). Figure 6.10 shows two AFM images, the first one for an as-grown 50-nm thick InAlN layer LM to GaN (Fig. 6.10(a)) and the second one for a similar 50-nm thick InAlN layer that has been annealed in situ at 1065° C for two minutes (Fig. 6.10(b)). Both layers have been grown on a FS-GaN substrate under the same conditions than for InAlN layers inserted in the DBRs exhibiting the best optical properties. The AFM images show that after annealing the surface morphology switches from small hillocks to a mesh of small cracks. To illustrate the impact of the temperature ramp, Fig. 6.10(c) shows a Z-contrast STEM image of an InAlN/GaN DBR. InAlN layers appear darker than GaN ones due to their lighter atomic weight. But, close to the InAlN/GaN interfaces, InAlN layers are even darker, suggesting that indium desorption indeed occurs during the temperature ramp. To circumvent this drawback, Sadler et al. (20096) proposed two solutions for the growth of GaN layers used in DBRs. They can be grown at low temperature, but it results in a finger-like GaN surface morphology, which is detrimental to the reflectivity of the DBRs. The second approach relies on the low-temperature growth of a thin GaN sublayer followed by a temperature ramp and the subsequent growth of GaN at high temperature.

x 2 pm AFM images of the surface of 50-nm thick InAlN layers (a) as grown and (b) annealed in situ at 1065°C. (c) Z-contrast STEM image viewed in cross-section of a DBR

Fig. 6.10. 2 x 2 pm2 AFM images of the surface of 50-nm thick InAlN layers (a) as grown and (b) annealed in situ at 1065°C. (c) Z-contrast STEM image viewed in cross-section of a DBR.

This latter method seems much more promising, since it should not alter the GaN surface morphology while preventing indium desorption at each InAlN/GaN interface.

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science