Threading dislocation propagation in AlGaN/GaN based HEMT structures grown on Si (111) by plasma assisted molecular beam epitaxy Jose ´ M. Ma ´ nuel a , Francisco M. Morales a,n , Rafael Garcı ´a a , Rolf Aidam b , Lutz Kirste b , Oliver Ambacher b a Departamento de Ciencia de los Materiales e IM y QI, Universidad de Ca ´diz, 11510 Puerto Real, Ca ´diz, Spain b Fraunhofer Institute for Applied Solid State Physics, Tullastrasse 72, 79108 Freiburg, Germany article info Article history: Received 25 June 2012 Accepted 23 July 2012 Communicated by H. Asahi Available online 3 August 2012 Keywords: A1. High resolution X-ray diffraction A1. Threading dislocation A1. Transmission electron microscopy A3. Molecular beam epitaxy B1. GaN on Si substrate B3. High electron mobility transistors abstract A transmission electron microscopy (TEM) study was carried out on a series of AlGaN/GaN high- electron mobility transistor (HEMT) structures grown by plasma assisted molecular beam epitaxy (PA-MBE) on Si (111). Threading dislocation (TD) behavior and density were investigated for three heterostructures using an AlN/GaN superlattice and/or differently strained GaN layers. Threading dislocation densities (TDDs) were measured by TEM (at different depths) and high resolution x-ray diffraction (HRXRD) allowing one of the most complete and few studies so far, presenting separated values on edge, screw and mixed type TDs quantities. & 2012 Elsevier B.V. All rights reserved. 1. Introduction High power devices have evolved during the last thirty years from structures based on III-As to III-N semiconductors, due to both improvement in group-III nitrides growth processes [1] and the fact that III-N based devices proved to be able to manage up to ten times more power density than their III-As analogues [2]. Among these high power devices, AlGaN/GaN high electron mobility transistors (HEMT) have been reaching a more and more considerable stage of development and commercialization during the last two decades [3]. However, unlike in the case of arsenides, III-N compounds present an important bottleneck for their higher industrial expan- sion: at the present moment, good-quality large enough group-III nitride wafers or bulk crystals are still not available. For this reason, devices implemented with these materials are grown on non-native substrates. Without a doubt, a-Al 2 O 3 (0001) (sapphire) has been the most extensively used, since the demon- stration of candela class LEDs [4]. Nevertheless, other materials offer better results acting as a substrate which is explained by two reasons involving both top layer and substrate: (i) the lattice parameter mismatch between III-N compound and substrate; and (ii) the substrate thermal expansion and conductivity. The last mentioned issue is extremely important, as self-heating is the main problem encountered in high power devices (and therefore in AlGaN/GaN HEMTs) [5] and therefore the power density (which implies temperature increment) in the device is dissipated pre- cisely through the substrate [6]. Lahr  eche et al. elaborated a complete summary of the different substrates (such as SiC, Si, ZnO, LiAlO 2 , etc.), and their relevant properties, used for growing hexagonal (wurtzite) GaN and AlN [7]. Attending to some physical properties, 4H-SiC (0001) substrates lead to the best electronic responses [8], since it presents a  3.5% (  1%) mismatch with GaN (AlN), and the highest thermal conductivity (3.8 W/cm). Sapphire, on the con- trary, has a poor thermal behavior (with a thermal conductivity about six times lower than SiC), this being the reason for the small thermal resistance occurring due to some malfunctions in HEMTs on sapphire, such as the fact that drain current does not reduce with the increase of the source–drain voltage [2]. Despite these reasons, high cost of SiC constitutes a clear disadvantage when considering it as a practical substrate in an industrial context. Hence, Si (111) substrates represent a very attractive alter- native; since it partially solves the power dissipation problem (it has a thermal conductivity value of 1.5 W/cm). Moreover, the possibility of using high quality, low cost, largely available Si and its well established integrated microelectronic technology com- pensates widely the inferior but improvable electronic results obtained compared to the ones that are achieved with SiC; actually, Si is nowadays the most reasonable choice for industrial purposes. Contents lists available at SciVerse ScienceDirect journal homepage: www.elsevier.com/locate/jcrysgro Journal of Crystal Growth 0022-0248/$ - see front matter & 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jcrysgro.2012.07.037 n Corresponding author. Tel.: þ34 956012742. E-mail address: fmiguel.morales@uca.es (F.M. Morales). Journal of Crystal Growth 357 (2012) 35–41