Thursday Morning Sessions (June 27) TMS Logo

About the 1996 Electronic Materials Conference: Thursday Morning Sessions (June 27)



June 26-28, 1996 · 38TH ELECTRONIC MATERIALS CONFERENCE · Santa Barbara, California

Session J: SiC Epitaxy

Session Chairman: Michael Spencer, Room 1124, Materials Science Center, Howard University, School of Engineering, 2300 6th Street NW, Washington, DC 20059

Co-Chairman: David J. Larkin, NASA Lewis Research Center, Mail Stop 77-1, 2100 Brookpark Rd., Cleveland, OH 44135

8:20AM, J1 *Invited

"Polarity-Dependent Step Bunching, Impurity Doping, and Interface Properties of [[alpha]]-SiC:" T. KIMOTO, A. Itoh, O. Takemura, S. Kobayashi, H. Matsunami, Department of Electronic Science and Engineering, Kyoto University, Yoshidahonmachi, Sakyo, Kyoto 606-01, Japan

Silicon carbide (SiC) is a promising wide bandgap material for high-power and high temperature devices. Although SiC device technology is showing rapid progress, most studies have employed only SiC(0001) Si faces and very little attention has been given to SiC( ) C faces. Since SiC is a compound semiconductor like GaAs, phenomena inherent to the substrate polarity should exist, which may affect device performance. In this paper, the authors report systematic investigation on the surface polarity dependence of step structure, impurity doping, metal/SiC interface and oxide/SiC interface properties.

6H- and 4H-SiC epilayers were grown on 3~5deg. off-oriented 6H- and 4H-SiC{0001} substrates by atmospheric-pressure chemical vapor deposition in a SiH4-C3H8H2 system (step-controlled epitaxy). N2 and trimethylaluminum (TMA) or B2H6 were used as n- and p-type dopants, respectively. Typical growth temperature and growth rate were 1500deg.C and 2.5mm/h. Low-temperature photoluminescence and deep level transient spectroscopy measurements revealed that epilayers have high quality, independent of substrate polarity.

In the Nomarski-microscope observation, the densities of surface pits and scratches were much lower on C faces. This is especially true in 4H-SiC growth: No triangular pits and/or macrosteps were observed on C faces. A high-resolution transmission electron microscope (TEM) analysis revealed that bunched steps with 4 Si-C bilayer height (unit cell of 4H-SiC) are mainly formed on 4H-SiC(0001) Si faces. In contrast, epilayer surfaces on C faces were atomically flat with dominating single Si-C bilayer-height steps.

The donor concentration of unintentionally doped epilayers grown on Si faces was drastically reduced under C rich conditions, because the high C coverage on a growing surface prevents the incorporation of N atoms which substitute at the C site. The lowest donor concentration obtained was 5x1013cm-3. However, the donor concentration on C faces was 5x1015cm-3, almost independent of C/Si ratio. On Si faces, the doping efficiency of Al and B increased with increasing the C/Si ratio, whereas such dependence was not observed on C faces. These results indicate that the surface coverage on C faces is not sensitive to the growth ambient.

The Schottky barrier height [[emptyset]][[Beta]]) in metal (Ti, Ni, Au)/4H-SiC structure was determined by internal photoemission, which provides the most reliable values. ØB strongly depends on the metal work function (ØM) with an index of interface behavior S(=(dØB/d(ØM) of 0.7, and ØB was about 0.2V higher on C faces.

We also found the polarity dependence of metal-oxide semiconductor (MOS) interface properties. MOS diodes with Al/SiO2/n-type SiC epilayer structure were fabricated by dry oxidation at 1150deg.C. High-frequency capacitance-voltage (C-V) curves of MOS diodes on Si faces almost coincided with the theoretical curve. On C faces, however, the curve showed a large positive flatband-voltage shift (+3.9V) and high interface state density. This polarity dependence may originate from the difference of bond configuration at the SiO2/SiC interface.

Thus, each surface (Si/C face) possesses its inherent properties, and the substrate polarity should be selected, depending on the device structure, to achieve the full potential of SiC.

8:40AM, J2

"Study of 3C-SiC Inclusions in 4H-SiC Epitaxial Films Grown on 4H-SiC Single Crystal Substrates:" W. SI, M. Dudley, H.S. Kong*, J. Sumakeris*, C.H. Carter, Jr.*, Department of Materials Science and Engineering, SUNY at Stony Brook, Nicols Rd., Long Island, NY 11794-2275, *Cree Research, Inc., Durham, NC 27713

During the epitaxial growth of 4H-SiC films on 4H-SiC single crystal substrates, some triangular inclusions often occur in the epilayers. In the present study, Synchrotron White Beam X-ray Topography (SWBXT) and Nomarski Optical Microscopy (NOM) have been used to non-destructively characterize the 4H-SiC epilayers and study the character of those triangular inclusions. SiC substrates are Si-face vicinal (0001) wafers with 30 mm in diameter, and epilayer thicknesses are about 7um. In SWBXT experiments, the grazing incidence Bragg-Laue reflection geometry was employed. In the diffraction pattern, besides the 4H-SiC pattern, there are additional spots corresponding to the two different 3C-SiC structures. There are two kinds of epitaxial relationships between 3C and 4H-SiC: for the 3C(I)-SiC, (0001)h//(111)c, [10 0]h//[211]c; for the 3C(II)-SiC, (0001)h//(111)c, [ ]h//[11]c. In general, the crystalline quality of the 4H-SiC epilayer is good, except for the 3C-SiC inclusions. NOM results showed that the 3C-SiC inclusions usually exhibit as triangular shapes. There is one-to-one correspondence of 3C-SiC inclusions in X-ray topographs and optical micrographs, but NOM reveals no crystallographic information. Possible causes for the formation of 3C-SiC inclusions will be discussed.

Research supported by AFOSR/ARPA and DOE.

9:00AM, J3

"Reduction of Memory Effects in p-Doped 6H and 4H SiC Grown by VPE:" N. NORDELL, A. Schöner, Industrial Microelectronics Center, P.O. Box 1084, S-16421 Kista, Sweden; M.K. Linnarsson, Royal Institute of Technology, Solid State Electronics, P.O. Box E229, S-164 40 Kista, Sweden

The recent progress in SiC device development has raised the demands on stable and reproducible processes for crystal growth. An important aspect is the possibility to control the dopant incorporation over a wide range, to enable growth of highly doped contact layers in the same growth run and the same reactor as low doped layers with low compensation. When the dope gases are switched out of the reactor, a considerable memory effect is observed for impurities like aluminum and boron, probably due to absorption at the reactor walls, and subsequent re-evaporation. The memory effect of nitrogen is less pronounced. The incorporation of dopants could to some extent be controlled by adjusting the C:Si ratio in the process gases, but only within a certain range, limited by the crystal quality. By changing the C:Si ratio when the dopant precursor is switched out of the reactor, the memory effect stays in the 1016 cm-3 range, which is not enough for most applications. We have investigated the use of an in situ HCl etch to reduce the memory effects of aluminum and boron doping, and to increase the dynamics of doping incorporation. In addition p-n junctions were formed to investigate the quality of the HC1 etched interface.

The growth was made in a horizontal reactor for vapor phase epitaxy (VPE), with an inner cell made entirely in graphite. As growth precursors, we used silane and propane, and for doping trimethylaluminum and diborane. The growth rate was fixed to about 2mm/h and the C:Si ration was varied between 1:1 and 4:1. Growth temperature was 1550 - 1620deg.C, and the reactor was operated at a pressure of 800 mbar. The doping transients were investigated with secondary ion mass spectrometry (SIMS). A Cameca ims 4f system was used and Al was obtained with O2+ primary beam and 27Al+ detection, and B was obtained with O2+ primary beam and 11B+ detection. C-I-V and capacitance transient measurements were used for p-n-junction investigations.

By applying a growth interruption of 10 minutes with an HCl etch we have reduced the memory effects more than an order of magnitude. For the Al doping, we have hence obtained dynamics in doping shut-off of about five orders of magnitude, from a maximum of >1020 cm-3, down to 2x1015 cm-3. From preliminary capacitance measurements it is concluded that the HCl etch is not deleterious to the interface.

9:20AM, J4

"Hollow-Core Screw Dislocations in 6H-SiC Single Crystals: A Test of Frank's Theory:" W. SI, M. Dudley, R. Glass*, V. Tsvetkov*, C. Carter, Jr.*, Department of Materials Science and Engineering, SUNY at Stony Brook, Nicols Rd., Long Island, NY 11794-2275, *Cree Research, Inc. Durham, NC 27713

Hollow-core screw dislocations, also known as "micropipes", along the [0001] axis in 6H-SiC single crystals have been studied extensively by Syncrotron White Beam X-ray Topography (SWBXT), Scanning Electron Microscopy (SEM), and Nomarski Optical Microscopy (NOM). Correlation between topographic images and SEM micrographs shows that micropipes are screw dislocations with Burgers vector magnitudes typically ranging from 2c to 7c (c is the lattice constant along [0001] axis, c = 15.17Å). The Burgers vector magnitude of the screw dislocations (determined from SWBXT), b, and the diameter of the associated micropipes (measured from SEM), D, are fitted to Frank's prediction for hollow-core screw dislocations: D = ub2/4[[pi]]2[[gamma]], where u is shear modulus, [[gamma]] is specific surface energy. Statistical analysis of the relationship between D and b2 shows that it is approximately linear, following Frank's prediction, and the constant, [[gamma]]/u can be obtained from the slope. The mean value of [[gamma]]/u ranges from 0.011 to 0.016Å. Using such values, the criteria for dislocations to form hollow cores are discussed, and the extrapolation to elementary screw dislocations (1c) is presented and discussed.

9:40AM, J5

"Low-Temperature Interface Modification by Hydrocarbon Radicals in Heteroepitaxy of 3C-SiC on Si Clean Surface:" TOMOAKI HATAYAMA, Norihiro Tanaka, Takashi Fuyuki, Hiroyuki Matsunami, Department of Electronic Science and Engineering, Kyoto University, Yoshidahonmachi, Sakyo, Kyoto, 606-01, Japan

Reproducible growth of single-crystalline 3C-SiC on Si was brought by introducing carbonization of a Si surface. With chemically active hydrocarbon radicals produced by cracked propane (C3H8), a Si surface could be carbonized successfully at as low as 750deg.C. To improve crystallinity of 3C-SiC grown layers, the difference (8%) in thermal expansion coefficients and the lattice mismatch (20%) between 3C-SiC and Si have to be relaxed enough in heterointerface. In this paper, an interface modification by hydrocarbon radicals is discussed in the heteroepitaxy of 3C-SiC on a Si clean surface. Carbonization is also studied by the use of dimethylgermane ((CH3)2GeH2:DMGe) to introduce a large size atom (Ge) at the heterointerface.

A Si(001) wafer was used as a substrate. Before the substrate was set in a gas source molecular beam epitaxy system, it was chemically cleaned with a solution (NH4OH : H2O2: H2O = 1:1:5) and covered with a thin protective oxide layer. The oxide layer was removed at 850deg.C in a high vacuum (~10-8Torr). Which generated a Si(2x1) structure. After surface cleaning, the substrate temperature was settled at a given value, and the Si surface was carbonized using hydrocarbon radicals from C3H8 or DMGe. During the modification, diffraction patterns of reflection high-energy electron diffraction (RHEED) were monitored with a charge-coupled device camera.

In the supply of C3H8, it was thermally cracked, which produced methyl (CH3) radicals in the range of 109~1011cm-3. The diffraction pattern of 3C-SiC did not appear immediately after the supply of cracked C3H8: the Si surface became 3C-SiC after the change of Si-related surface reconstructions in the incubation time. The Si(2x1) structure of a clean surface changed to a mixture of Si(2x1) and Si c(4x4) structures. The diffraction pattern of 3C-SiC appeared after the decrease of diffraction intensities of Si-related surface reconstructions. The change of RHEED patterns is believed to be related to the chemical reactions between CH3 radicals and the Si surface. The RHEED pattern of the carbonized layer indicated a mixture of single-crystalline 3C-SiC and its twin crystals. The growth rate in the initial stage of 3C-SiC formation below 770deg.C was regulated by the substrate temperature, indicating surface-reaction limited growth. The activation energy was about 46.9kcal/mol.

In the supply of DMGe, to control precisely the amount of Ge atoms in the heterointerface, it was fed intermittently for 0.2 seconds in one shot. The Si clean surface could be carbonized reproducibly with 20 shots of DMGe at as low as 650deg.C. The change of surface reconstructions was the same sequence as in the case of cracked C3H8. It can be considered that CH3 radicals are generated by thermal decomposition of DMGe on a Si surface, forming the carbonized layer at such a low temperature. The RHEED pattern of the carbonized layer indicated single-crystalline 3C-SiC without Ge-related diffraction spots and any stacking-faults streaks. From the result of Rutherford backscattering measurement, Ge atoms of about 0.1% exist in the carbonized layer. The carbonized layer with the incorporation of Ge atoms will relax the lattice mismatch between 3C-SiC and Si.

10:20AM, J6+

"SiC SOI Structures as Substrates for III-N Growth:" J. DEVRAJAN, A.J. Steckl, University of Cincinnati, Nanoelectronics Laboratory, 899 Rhodes Hall, Department of Electrical and Computer Engineering, P.O. Box 210030, Cincinnati, OH 45221-0030

The silicon carbide (SiC) semiconductor-on-insulator (SOI) structure represents an exciting approach for the development of large area, low cost SiC substrates. The starting point for SiC SOI is a Si SOI wafer consisting of a normal Si substrate, an oxide layer a and thin Si device layer. The Si SOI structure is produced either by thermal bonding and etch back of two Si wafers or by oxygen ion implantation. The SiC SOI structure can be utilized for the fabrication of SiC devices or it can serve as substrate for the thin film growth of other materials. The SiC SOI approach is superior to growth of SiC directly on a Si wafer. This is due to the fact that the Si device layer in the SOI structure can be made quite thin, enabling the complete conversion of Si to SiC and minimizing the stress in the film. Furthermore, the oxide layer is beneficial in reducing parasitic capacitance and absorbing dislocations.

In this paper we report on the growth of (111) SiC SOI structures using carbonization with mixtures of propane (5% in H2) and H2 at temperatures from 1100 to 1300deg.C. Growth on the (111) plane was chosen for two reasons: (a) superior results with growth of SiC directly on (111) Si substrates were previously reported; (b) the (111) SiC surface is a better match for subsequent growth of III-N films. To prevent unnecessary exposure to high temperature, the SiC growth was carried out in an RTCVD reactor which allowed for reaction times ranging from a few seconds to several minutes. The resulting SiC SOI structures have been characterized with regard to structure (XRD), composition (FTIR) and surface morphology (SEM, AFM). The thickness of the SiC layer produced by carbonization was approximately equal to that of the original Si device layer (900-1000 Å). XRD reveals a single peak at 2[[Theta]] = 35.7deg. corresponding to the (111) reflection. The uncorrected peak FWHM was ~0.24deg., compared to 0.3deg. for the peak at 44.4deg. in the equivalent (100) SiC SOI structure.

Preliminary results have been obtained from MOCVD growth of GaN layers on (111) SiC SOI with TMGa and NH3 precursors. XRD indicates highly oriented hexagonal GaN with dominant growth orientation along the c-axis. The FWHM of the (0002) peak is ~0.15deg.. At 300deg.K, the PL spectrum of GaN films exhibits a strong near band-edge peak (at [[lambda]]p~371 nm, with an FWHM of ~100-150 meV) and a weak yellow band emission. Under low excitation, the band-edge PL emission from the GaN/SiC SOI structure had an emission intensity ~10x higher than that of equivalent GaN films grown on sapphire.

10:40AM, J7

"Improvement of the Crystallinity of 3C-SiC Films by Lowering the Electron Temperature in the Afterglow Plasma Region:" K. YASUI, N. Ninagawa, T. Akahane, Nagaoka University of Technology, Kamitomioka, Nagaoka-shi, Niigata 940-21, Japan

Crystalline SiC films were grown at low temperature (<500deg.C) by the triode plasma CVD using dimethylchlorosilane diluted with hydrogen as source gas. DC bias effects of the mesh electrode (grid) on the properties of SiC films, i.e. crystallinity, chemical bonding structure and composition, were investigated. The electron temperatures and electron densities of the discharge region above grid and the afterglow region below grid electrode were estimated using the probe measurement.

Under a negative grid bias, the crystallinity of SiC films were remarkable improved and the composition of SiC films became stoichiometrical. On the other hand, carbon-rich amorphous SiCx films were grown under positive grid bias conditions. From the probe measurement, the negative grid bias was found to lower the electron temperature of the afterglow region and to raise the electron temperature of the discharge region. The bombardment of ions onto the film surface was also found to be suppressed under the same bias condition. Under such a condition, high density RF discharge was effectively confined above the grid electrode. Then the active high density hydrogen radicals were generated there, diffused toward the substrate and extracted the weak bonds or excessive methyl groups from the film growing surface. As a result of these processes, the growth of crystalline SiC film was considered to be enhanced.

11:00AM, J8

"Characterization of 3C-SiC Crystals Grown by Thermal Decomposition of Methyltrichlorosilane:" A.J. STECKL, J. Devrajan, Nanoelectronics Laboratory, University of Cincinnati, 899 Rhodes Hall, P.O. Box 210030, Cincinnati, OH 45221-0030; S.N. Gorin, L.M. Ivanova, Baikov Institute of Metallurgy, Russian Academy of Sciences, Moscow, 117334, Russia

Crystalline SiC films were grown at low temperature (<500deg.C) by the triode plasma CVD using dimethylchlorosilane diluted with hydrogen as source gas. DC bias effects of the mesh electrode (grid) on the properties of SiC films, i.e. crystallinity, chemical bonding structure and composition, were investigated. The electron temperatures and electron densities of the discharge region above grid and the afterglow region below grid electrode were estimated using the probe measurement.

Under a negative grid bias, the crystallinity of SiC films were remarkable improved and the composition of SiC films became stoichiometrical. On the other hand, carbon-rich amorphous SiCx films were grown under positive grid bias conditions. From the probe measurement, the negative grid bias was found to lower the electron temperature of the afterglow region and to raise the electron temperature of the discharge region. The bombardment of ions onto the film surface was also found to be suppressed under the same bias condition. Under such a condition, high density RF discharge was effectively confined above the grid electrode. Then the active high density hydrogen radicals were generated there, diffused toward the substrate and extracted the weak bonds or excessive methyl groups from the film growing surface. As a result of these processes, the growth of crystalline SiC film was considered to be enhanced.

11:20AM, J9

LATE NEWS


The information on this page is maintained by TMS Customer Service Center (csc@tms.org).

Search TMS Specialty Meetings Page TMS Meetings Page About TMS TMS OnLine