Homoepitaxial growth and defect characterization of 4H-SiC epilayers
Invited Talk at the International Conference on Sillicon Carbide and Related Materials 2011, ICSCRM 2011, to be held in Cleveland, USA
The structural quality of 4H silicon carbide (SiC) wafers has been significantly improved with respect to micropipes and polytype inclusions in recent years. Nevertheless, commercial 4H-SiC wafers still exhibit dislocation densities in the range of 104 cm-2. Some of these dislocations are regarded as of minor importance to device performance like Threading Edge Dislocations (TEDs) and Threading Screw Dislocations (TSDs). Contrary to that, Basal Plane Dislocations (BPDs) are suspected to trigger the electrical degradation of bipolar devices with blocking voltages > 2 kV . The commonly accepted strategy to achieve long-term stable bipolar devices is to avoid any BPD within the active region of the bipolar device, i.e. within the homoepitaxial layer. Such BPD free epilayers can be achieved by homoepitaxial growth only if (i) all BPDs originated in the substrate are eliminated in the epilayer and (ii) the generation of further BPDs, e.g. at the interface of substrate and epilayer, is avoided. The first part of this paper deals with the elimination of BPDs during homoepitaxial growth of 4HSiC. In substrates, BPDs are common dislocations with a typical density of 103 cm-2 to 104 cm-2. These substrate BPDs can be eliminated during homoepitaxial growth due to dislocation conversion to TEDs, i.e. the overall dislocation density remains constant . The reason for the BPD conversion is, according to the model by Klapper  and Ohno et al. , the minimization of the dislocation line energy during epitaxial growth. As shown in figure 1, the dislocation line energy W and hence the conversion efficiency during growth are mainly depending on the off-cut angle of the substrate. The validity of this model will be verified experimentally. In the second part of this paper, the possibility of BPD generation due to doping-induced lattice mismatch is discussed. Several epigrowth series were conducted on vicinal, highly nitrogen (N) doped, 250 ?m thick substrates by Chemical Vapor Deposition (CVD) using an Aixtron VP508GFR reactor . In a first series, the layer thickness was varied from 12.5 ?m up to 50 ?m maintaining a constant N doping level of low 1015 cm-3. In a second and third series, the N and aluminum (Al) doping concentrations were varied over several orders of magnitude keeping the layer thickness constant. All samples were characterized with respect to their macroscopic bow. Based on Stoney's formula , the macroscopic bow can be converted to lattice mismatch data. For highly Al doped epilayers grown on highly N doped substrates, the lattice mismatch was determined additionally by HRXRD measurements. The doping dependent mismatch, obtained from bow and HRXRD measurements, is compared to theoretical values calculated based on the model by Jacobson . For Al doped epilayers, the experimental and theoretical values of mismatch fit well. For N doped epilayers, the theoretical mismatch deviates from the experimental value. Furthermore, critical epilayer thickness was calculated according to the models by Matthews and Blakeslee  and People and Bean . The limitations to BPD free, homoepitaxial growth with respect to epilayer thickness in conjunction with epilayer doping will be presented.