Growth of single crystal silicon carbide by halide chemical vapor deposition

Open Access
Fanton, Mark
Graduate Program:
Materials Science and Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
October 05, 2007
Committee Members:
  • Joan Marie Redwing, Committee Chair
  • Joseph R Flemish, Committee Member
  • Suzanne E Mohney, Committee Member
  • Thomas R Shrout, Committee Member
  • chemical vapor deposition
  • silicon carbide
  • epitaxy
Semiconductor grade single crystals of SiC were first produced in the late 1950’s by Lely in Germany and by Hamilton in the United States. Over the last 30 years this process has been improved and scaled to produce single crystal boules that yield 4” diameter wafers. These substrates are of interest for two primary classes of semiconductor devices because of its excellent thermal and high temperature electrical properties compared to competing substrates such as Si, GaAs, or sapphire. For high frequency radar applications semi-insulating SiC wafers are used as substrates for GaN/AlGaN epitaxy. Production of high resistivity material relies on careful control of the concentrations of impurities and point defects in the single crystal. Trace amounts of boron and nitrogen are of particular concern as they act as residual p-type and n-type dopants respectively. Silicon carbide is also an ideal semiconductor for high voltage and high temperature power switching devices such as Schottky diodes, PiN diodes, and MOSFET’s. The most challenging aspect of SiC substrate production and epitaxial layer growth for power switching applications is the elimination of structural defects that lead to device failure. Many of the shortcomings of conventional SiC growth processes can be overcome through the use of a novel halide chemical vapor deposition (HCVD) process for bulk SiC and thick SiC epitaxial layers. The HCVD process provides significant advantages over the conventional physical vapor transport (PVT) process for bulk crystal growth and silane-based CVD for growth of epitaxial layers. However, the HCVD technique requires the balance of several competing processes in order to successfully grow large crystals and high quality epitaxial layers. A detailed study of halide chemical vapor deposition is required to understand the relationships between crystal growth conditions and the resulting structural properties of the SiC deposited. The goal of this thesis is to understand relationships between the major process variables and the growth rate, doping, and defect density of SiC grown by HCVD. Specifically this work addresses the maximum C/Si ratios that can be utilized for single crystal SiC growth by providing a thermodynamic model for determining the boundary between single crystal growth and SiC+C mixed phase growth in the Si-C-Cl-H system. The impact of growth conditions, especially the C/Si ratio, on doping and overall structural quality are assessed within the boundaries determined by the thermodynamic model. SiC epitaxial layers ranging from 50-200µm thick were grown at temperatures near 2000°C on 6H and 4H-SiC substrates at rates up to 250µm/hr. Experimental trends in the growth rate as a function of precursor flow rates and temperature closely match those expected from thermodynamic equilibrium in a closed system. The equilibrium model can be used to predict the trends in growth rate with the changes in precursor flow rates as well as the boundary between deposition of pure SiC and deposition of a mixture of SiC and C. Calculation of the boundary position in terms of the SiCl4 and CH4 concentrations provides an upper limit on the C/Si ratio that can be achieved for any given set of crystal growth conditions. The model can be adjusted for changes in temperature, pressure, and chlorine concentration as well. The boundary between phase pure and mixed phase growth was experimentally shown to be very abrupt, thereby providing a well defined window for Si-rich and C-rich growth conditions. Growth of SiC epitaxial layers by HCVD under both Si-rich and C-rich conditions generally yielded the same trends in dopant incorporation as those observed in conventional silane-based CVD processes. Nitrogen incorporation was highest on the C-face of 4H-SiC substrates but could be reduced to concentrations as low as 1x1015 atoms/cm3 at C/Si ratios greater than 1. Residual B concentrations were slightly higher for epitaxial layers grown on the Si-face of substrates. However, changes in the C/Si ratio had no effect on B incorporation at concentrations on the order of 1x1015 atoms/cm3. No significant trends in structural quality or defect density were evident as the C/Si ratio was varied from 0.72 to 1.81. Structural quality and defect density were more closely related to substrate off-cut and polarity. The highest quality crystals were grown on the C-face of 4° off-axis substrates as measured by HRXRD rocking curves. Growth on on-axis substrates was most successful on the C-face, although the x-ray rocking curves were nearly twice as wide as those on off-axis substrates. Etch pit densities obtained by KOH etching layers grown on Si-face substrates were closely related to the defect density of the substrate not the C/Si ratio. Thick p-type layers with B or Al dopant concentrations on the order of 1019 atoms/cm3 were readily achieved with the HCVD process. Trimethylaluminum and BCl3 were successfully employed as dopant sources. Aluminum incorporation was sensitive to both the substrate surface polarity and the C/Si ratio employed for growth. Dopant concentrations were maximized under C-rich growth conditions on the Si-face of SiC substrates. Boron incorporation was insensitive to both the surface polarity of the substrate and the C/Si used for layer growth even though B appears to favor incorporation on Si lattice sites. Boron acceptors in HCVD grown SiC are not passivated by H to any significant extent based on a comparison of net acceptor concentrations and B doping concentrations. In addition, the lattice parameters epitaxial layers doped with B at concentrations on the order of 1019 atoms/cm3 showed no change as a function of B concentration. This was in contrast to the lattice parameter decrease as expected from a comparison between the size of the Si and B atoms. The HCVD process has demonstrated an order of magnitude higher growth rates than conventional SiC CVD and while providing control over the C/Si ratio. This allows the user to directly influence dopant incorporation and growth morphology. However, this control should also permit several other material properties to be tailored. Future work should focus on determining the range of C/Si ratios required to achieve specific material characteristics. Electronic and structural defects should be the primary focus since they directly impact SiC device design and performance.