Journal of ELECTRONIC MATERIALS, Vol. 39, No. 1, 2010 DOI: 10.1007/s11664-009-0953-6 Ó 2009 TMS Heavily Aluminum-Doped Epitaxial Layers for Ohmic Contact Formation to p-type 4H-SiC Produced by Low-Temperature Homoepitaxial Growth B. KRISHNAN, 1 S.P. KOTAMRAJU, 1 G. MELNYCHUK, 1,2 H. DAS, 1 J. N. MERRETT, 3 and Y. KOSHKA 1,4 1. Electrical and Computer Engineering, Mississippi State University, Box 9571, Mississippi State, MS 39762, USA. 2. BarSiC Semiconductors, LLC, Starkville, MS 39759, USA. 3. Air Force Research Laboratory, Wright-Patterson Air Force Base, Dayton, OH 45433, USA. 4. e-mail: ykoshka@ece.msstate.edu In this work, heavily aluminum (Al)-doped layers for ohmic contact formation to p-type SiC were produced by utilizing the high efficiency of Al incorporation during the epitaxial growth at low temperature, previously demonstrated by the authors group. The low-temperature halo-carbon epitaxial growth technique with in situ trimethylaluminum (TMA) doping was used. Nearly featureless epilayer morphology with an Al atomic concentration exceeding 3 9 10 20 cm 3 was obtained after growth at 1300 C with a growth rate of 1.5 lm/h. Nickel transfer length method (TLM) contacts with a thin adhesion layer of titanium (Ti) were formed. Even prior to contact annealing, the asdeposited metal contacts were almost completely ohmic, with a specific contact resistance of 2 9 10 2 X cm 2. The specific contact resistance was reduced to 6 9 10 5 X cm 2 by employing a conventional rapid thermal anneal (RTA) at 750 C. Resistivity of the epitaxial layers better than 0.01 X cm was measured for an Al atomic concentration of 2.7 9 10 20 cm 3. Key words: Low-temperature epitaxial growth, halo-carbon, ohmic contact, aluminum doping INTRODUCTION Silicon carbide (SiC) devices are being developed and commercialized for high-power and high-temperature applications. Epitaxial growth conducted at conventional, high temperatures above 1500 C remains the main technique to provide low- as well as heavily doped SiC layers for device applications. Selective etching and ion implantation are the main techniques used for device topology formation, selective doping, etc. Recently, epitaxial growth of high-quality SiC at low temperatures (i.e., growth temperatures below 1300 C) using the low-temperature halo-carbon homoepitaxial method was demonstrated, 1 which (Received March 26, 2009; accepted September 9, 2009; published online October 1, 2009) offered a simplified approach to selective doping and self-aligned fabrication. 2 Al doping by epitaxial growth and ion implantation has been extensively investigated in the past. 3 7 Possibilities to achieve very high values of conductivity even when employing such a relatively deep dopant as Al in SiC have been explored. Flashlamp annealing was demonstrated as a promising way of achieving a high degree of activation in highly aluminum-implanted 6H-SiC wafers. 3 For high concentrations of Al implants (>5 9 10 20 cm 3 ), so-called metallic conduction was observed and values of the free hole concentrations exceeded 1 9 10 20 cm 3 for Al implant concentrations of 1.5 9 10 20 cm 3. 3 In addition to other benefits of conducting epitaxial growth at lower temperature, it has been demonstrated that a reduction of the epitaxial 34
Heavily Aluminum-Doped Epitaxial Layers for Ohmic Contact Formation to p-type 4H-SiC Produced by Low-Temperature Homoepitaxial Growth 35 growth temperature may be beneficial for facilitating incorporation of certain dopants during epitaxial growth on Si-face. 4 However, the previously available epitaxial growth processes did not allow reduction of the growth temperature below 1450 C to 1500 C without deterioration of the epilayer morphology. In the previously reported work by the authors group, a possibility of achieving high values of doping by the low-temperature growth method was demonstrated. 8 In the present work, low-temperature halo-carbon epitaxial growth at 1300 C was utilized to produce heavily Al-doped SiC layers for ohmic contact formation to p-type SiC. Structural and electrical properties of the low-temperature p-type epitaxial layers and ohmic contacts are investigated. Different metals have been investigated in previous reports as candidates for ohmic contacts to heavily Al-doped p-type SiC. Interest in nickel (Ni) is dictated primarily by the possibility of forming ohmic contacts simultaneously to n- and p-type regions of SiC devices. Specific contact resistance as low as 7 9 10 6 X cm 2 was reported for Ni contacts to epitaxial p-type SiC. 9 For the same reasons, Ni was also selected as a metal for p-type ohmic contacts developed in this work. EXPERIMENTAL PROCEDURES Commercial n + 4H-SiC wafers vicinally cut at 8 towards the [1120] direction were used as substrates for epitaxial growth of 4H-SiC. The growth was conducted in a low-pressure hot-wall chemical vapor deposition (CVD) reactor at 1300 C and 140 Torr with hydrogen as the carrier gas, chloromethane (CH 3 Cl) as the carbon precursor, and silane (SiH 4 ) as the silicon precursor. TMA doping in situ during the blanket low-temperature epitaxial growth was utilized to produce heavily Al-doped SiC layers on n + substrates. The TMA bubbler (electronic grade) was maintained at a temperature of 18 C and at an output pressure of 760 Torr. With the standard ultra-high purity CH 3 Cl and SiH 4 precursors used, the residual Al concentration without using TMA was determined by the conditions of the susceptor and the thermal insulation foam and was below 10 14 cm 3. The Si/C ratio in all the growth experiments was maintained at 6.0, which is a typically high value, used in the low-temperature halo-carbon process. 1 The thickness of the grown epilayers was varied between 0.2 lm and 0.9 lm by altering the growth duration. The Al-doped films were grown directly on the substrate surface without employing any buffer layer. Surface morphology of the grown layers was monitored by Nomarski optical microscopy. Etching of Al-doped epitaxial layers by molten potassium hydroxide (KOH) was used to delineate defects in the material and evaluate the generation of Fig. 1. SIMS profile of aluminum concentration in the epitaxial layer as a function of TMA flow. additional dislocations caused by the high level of doping. The etch pits were evaluated by Nomarski optical microscopy and verified by scanning electron microscope (SEM) on selected samples. Qualitative changes in resistivity versus the TMA flow were monitored using I V characteristics measured with two tungsten probe needles in contact with a bare SiC surface. Secondary-ion mass spectroscopy (SIMS) was conducted by Evans Analytical Group to measure the Al atomic concentration. Ni TLM contacts (750 Å thick with a thin 20-Å adhesion layer of Ti) were formed by e-beam deposition and metal lift-off. The contacts were formed on the as-grown surface of a few selected samples having thicknesses of the epitaxial layers from 0.2 lm to 0.9 lm. The contacts were subsequently annealed in a flash-lamp annealing system at 750 C in flowing nitrogen ambient. A Rigaku Ultima III x-ray diffraction (XRD) system was used to identify the phase formation in the annealed contacts. RESULTS AND DISCUSSION Low-temperature epitaxial growth with different TMA flow rates was used to produce Al-doped SiC layers with increasingly high Al doping. Al atomic concentrations exceeding 2.7 9 10 20 cm 3, as determined by SIMS measurements (Fig. 1), were achieved without apparent morphology degradation. An epitaxial growth rate of 1.5 lm/h with goodquality, nearly featureless epilayer morphology was achieved at growth temperatures of about 1300 C. Qualitative estimation of the resistivity of the grown epitaxial layers was obtained from the current voltage characteristics measured with two tungsten probe needles in contact with the bare SiC surface of the epitaxial layers. Figure 2 shows current voltage characteristics measured on epitaxial layers grown with different TMA flow rates. A clear transition from Schottky behavior at lower TMA flows to nearly ohmic behavior for higher TMA flows was observed.
36 Krishnan, Kotamraju, Melnychuk, Das, Merrett, and Koshka 1.50 9 10 6 cm 2 for an epilayer doping of 3.8 9 10 20 cm 3 (Fig. 4). Preliminary analysis indicated that the dislocation concentration exhibited an exponential dependence on the TMA flow. It appears that the trend can be reasonably described by a conventional model based on stress generation with doping. However this study is beyond the scope of the present paper and will be reported elsewhere. Fig. 2. Current voltage characteristics of low-temperature p + epitaxial layers grown with different TLM flow rates. The corresponding Al concentration is indicated. The I V characteristics were measured using two tungsten probe needles in contact with bare SiC surface. The epilayer morphology as observed under the optical microscope for all TMA flows up to 0.29 sccm remained virtually as good as that in the TMA-free growth. However, still higher TMA flows resulted in noticeable roughening of the epilayer surface. Figure 3 shows optical microscope images obtained from an epilayer with doping of about 2.7 9 10 20 cm 3 without surface morphology degradation and from a higher-doped epilayer of 3.8 9 10 20 cm 3, where surface roughening took place. Dislocations in the epitaxial layers were investigated using the molten KOH defect delineation technique. KOH etching revealed progressively higher dislocation densities with increasing TMA flow (i.e., doping of the epitaxial layers). The etch pit densities, revealed in the epilayers after 2 min of etching in a molten KOH solution maintained at 450 C, increased from about 2.79 9 10 3 cm 2 for an epilayer doping of 2.4 9 10 19 cm 3 to about Ohmic Contacts Ni contacts with a 20-Å-thick adhesion layer of Ti were formed on the surfaces of the Al-doped epilayer. Transfer length method (TLM) measurements were carried out in order to establish the specific contact resistance of the ohmic contact and the sheet resistivity of the epitaxial layers. On a sample with an Al atomic concentration of 2.7 9 10 20 cm 3, as measured by SIMS, the as-deposited metal contacts (without contact annealing) were found to be nearly ohmic (dotted line in Fig. 5). The value of the contact resistivity in the nonannealed contacts was about 2 9 10 2 X cm 2. Standard RTA process was applied to obtain lower values of specific contact resistance. I V characteristics after contact annealing at 750 Cis compared in Fig. 5 with I V measured before contact annealing. I V characteristics from two different spacing of the Ni TLM contacts are shown. Results of the XRD measurements of the samples with metal contacts are shown in Fig. 6. While only a Ni-related XRD signal was present in the as-deposited contact, the Ni 2 Si phase appeared and dominated after the contacts were annealed. 10 The best value of the specific contact resistivity for this group of samples was found to be in the range of 6 9 10 5 X cm 2. The results from the TLM measurements before and after annealing are summarized in Table I. The value of the sheet resistance measured in the heavily doped p-type epitaxial layers before contact Fig. 3. Optical microscopic image of the epilayer surface for (a) no morphology degradation in highly doped epilayer ([Al] = 2.7 9 10 20 cm 3 ) and (b) degraded surface of even higher-doped epilayer ([Al] = 3.8 9 10 20 cm 3 ).
Heavily Aluminum-Doped Epitaxial Layers for Ohmic Contact Formation to p-type 4H-SiC Produced by Low-Temperature Homoepitaxial Growth 37 Fig. 4. Dislocations in p + epitaxial layers exposed by molten KOH etching, observed under a Nomarski optical microscope, for different TMA flow rates. Fig. 5. I V characteristics of Ni TLM contacts: as-deposited and after RTA at 750 C, for two different distances between the contacts (3.5 lm and 10 lm). Fig. 6. XRD spectra from the Ni/Ti/4H-SiC contact, as deposited and annealed at 750 C. annealing was used to estimate the resistivity of the epitaxial layers. For the given epilayer thickness, the resistivity was better than 0.01 X cm in the samples with a total Al concentration close to 2.7 9 10 20 cm 3. When referring to the wide range of values of the free-hole mobility (from 65 cm 2 /V s to 115 cm 2 /V s) reported in the literature for epitaxial and ionimplanted p-type 4H-SiC, the net free-hole concentration in our heavily doped p + epitaxial layers (influenced by the carrier freeze-out) was estimated to be in the range from 5.5 9 10 18 cm 3 to 1 9 10 19 cm 3. It is expected that the mobility in these samples should be on the lower side due to strain and dislocation generation caused by the high level of Al doping. The increase of the sheet resistance after contact annealing (Table I) is yet to be explained. It could be caused by both degradation of materials properties and TLM measurement error caused, for example, by changes in the active epilayer thickness under the contacts caused by Ni silicide formation. CONCLUSIONS Aluminum doping by low-temperature epitaxial growth was shown to be a convenient technique for achieving p-type resistivity less than 0.01 X cm on Si face of 4H-SiC substrates. With Ni metal, which is considered inferior to some other metals for ohmic contacts to p-type SiC, a contact resistivity of 6 9 10 5 cm 3 was achieved with conventional RTA at an annealing temperature of 750 C. These promising results make this approach attractive for ohmic contact formation simultaneously to n- and p-type regions of SiC devices. Furthermore, the possibility of achieving high values of Al doping at an epitaxial growth temperature as low as 1300 C makes this doping technique compatible with selective epitaxial growth, which is suitable for selective doping as an alternative to the presently dominant method of ion implantation. Results for Ni contact formation to heavily Al-doped mesas formed by low-temperature selective epitaxial growth (LTSEG) will be reported elsewhere.
38 Krishnan, Kotamraju, Melnychuk, Das, Merrett, and Koshka Table I. Summary of Ni TLM measurement results as-deposited and after contact annealing TMA Flow (sccm) Thickness (nm) TLM R SH (X/sq) TLM R C (X cm 2 ) As-deposited 0.29 250 392.33 2.01 9 10 2 0.35 290 474.5 2.63 9 10 2 After contact anneal 0.29 250 632.69 6.04 9 10 5 0.35 290 681.15 7.81 9 10 5 ACKNOWLEDGEMENTS The authors are grateful to SemiSouth Laboratories, Inc. for providing contact metal deposition and annealing services. This work was supported by National Science Foundation, Grant Nos. ECS0622184 and IIP-0839748. REFERENCES 1. Y. Koshka, H.D. Lin, G. Melnychuk, and C. Wood, J. Cryst. Growth 294, 260 (2006). 2. B. Krishnan, H. Das, H.-De. Lin, and Y. Koshka, Appl. Phys. Lett. 89, 262103 (2006). 3. H. Wirth, D. Panknin, W. Skorupa, and E. Niemann, Appl. Phys. Lett. 74, 979 (1999). 4. U. Forsberg, Ö. Danielsson, A. Henry, M.K. Linnarsson, and E. Janzén, J. Cryst. Growth 253, 340 (2003). 5. M.A. Fanton, B.E. Weiland, and J.M. Redwing, J. Cryst. Growth 310, 4088 (2008). 6. T. Kimoto, A. Itoh, H. Matsunami, T. Nakata, and M. Watanabe, J. Electron. Mater. 25, 879 (1996). 7. N.S. Sak, A.K. Agarwal, S.-H. Ryu, and J.W. Palmour, J. Appl. Phys. 90, 2796 (2001). 8. H.D. Lin, G. Melnychuk, J.L. Wyatt, and Y. Koshka, Mater. Sci. Forum 556 557, 133 (2007). 9. J. Crofton, J.R. Williams, A.V. Adedeji, J.D. Scofield, S. Dhar, L.C. Feldman, and M.J. Bozack, Mater. Sci. Forum 527 529, 895 (2006). 10. R. Perez, N. Mestres, D. Tournier, P. Godignon, and J. Millan, Diam. Relat. Mater. 14, 1146 (2005).