Modeling rhenium metallization of a silicon-rich (0 0 1) 6H-SiC surface
Introduction
As technology pushes toward space and harsh environment applications, the need for high-power, high-temperature, radiation resistant, and high mechanical performance materials has become a motivation in materials research for electronic applications and for mechanical performance of nano-microelectromechanical systems. High thermal conductivity and excellent durability makes silicon carbide an excellent choice for these applications. Since SiC has a bandgap which depends on its polytype, it is considered a versatile and competitive material [1], [2], [3], [4], [5], [6], [7] for applications where other conventional semiconductors fail. Its high electron mobility allows for increased current within the semiconductor and high-energy pulsed applications [8]. Polytypes range from purely wurtzite (cubic) to hexagonal with over 200 variations [9]. The bandgaps range from 2.3 to 3.3 eV with common types being 3C-SiC (2.3 eV) and 6H-SiC (2.9 eV). SiC has a melting temperature of approximately 2800 °C, thus capable of being operated in the temperature range of 600–1000 °C [3], [4].
Many strides are being made in the areas of growth, doping and contamination reduction of SiC [10]. With advances in these fields, the major limiting factor in SiC device technology is reproducible metalization. Resistivities as low as 1×10−6 Ω/cm2 have been achieved, however several problems remain unsolved [11]. Particularly at elevated temperatures, metal atom inter-diffusion, formation of silicates and carbides, degradation of the interface due to lattice mismatch and surface reconstruction are a few areas of research requiring more study. The most notable problem in formation of stable ohmic contacts is Schottky barrier control. Understanding what affects barrier height, how to manipulate it, and finding an appropriate metal that minimizes the barrier height while remaining harsh environment resistant provides invaluable information for advancement in SiC device technology [12].
The encouraging experimental findings as described in the dissertation by G.Y. McDaniel [13] led us to perform the present computation. His results showed that rhenium is an excellent metal for electrical contacts to SiC. Initially, he did an extensive study of the influence of thin layers of Ni on SiC to gain an understanding of contact formation. Variables affecting the work function of the SiC surface were varied and the results tabulated. Calibration during this phase of his research was unable to verify a reproducible procedure for application to the Ni/SiC contact. The Re/6H-SiC contact was demonstrated on three surfaces. Each was characterized with Dektak stylus profilometry, XRD, AES and I–V electrical measurements in the as-deposited condition. They were then exposed to a 120 min anneal at 1000 °C in vacuum at less than 1×10−6 Torr total pressure. Each sample was again characterized and the results compared to the pre-anneal condition.
The first surface was stoichiometric clean SiC. Contacts on this surface proved to be wide ranging in resistivity depending on whether or not the contact was annealed. The pre-annealed average resistance was 3700 Ω and the average post-annealed resistance was 50 Ω. It is important to note that one sample included in this average had contact regions that were annealed at roughly 200 °C lower than other samples due to poor mounting. Removing this sample reduces the post-anneal average from 50 to only 6 Ω. The contacts were rectifying prior to annealing. After annealing, most were still slightly rectifying, however their I–V curves indicated they were transitioning to ohmic character. There was no detectable reaction in either XRD or AES data, however, the film was shown to relax into the (1 0 3) crystallographic plane for Re with heat treatment.
The second surface was graphitized SiC. Contacts on this surface were successful, however, their properties were undesirable for this study. In all cases, the depositions resulted in dull gray or flaky contacts and the Re did not wet with the SiC. The resistivity was less wide ranging than the stoichiometric condition. The pre-annealed average was 72 Ω and the post-annealed average was 21 Ω. All contacts were rectifying prior to annealing. After annealing again, a few contacts were near ohmic, but most still exhibited some recitifying character. There was no detectable reaction from XRD data, however, the film was shown to relax into the (0 0 2) crystallographic plane for Re with heat treatment. The AES data indicated an unusual phenomenon with this surface chemistry. While reaction was not evident, ordering of the Re atoms into the (0 0 2) crystallization created a “self-cleaning” of the film. Films that were unintentionally contaminated with carbon throughout showed strong signs of graphite clumping which was forced to the surface of the contact after heat treatment.
The third surface was a silicon layer deposited on a stoichiometric SiC surface. This contact was the most promising of all conditions. Films were highly ordered in both the as-deposited and the post-annealed conditions. XRD confirmed preferential crystallization along the Re (1 0 1) plane. AES confirmed migration of Re into the interface and possible reaction with the Si film at elevated temperatures. I–V curves prior to annealing were mostly rectifying. While annealing every contact was found to be 100% ohmic with R-squared values of 0.99+ to 1.00 when subjected to a linear regression fit. The pre-anneal average resistance was 212 Ω while the post-anneal average was 3 Ω. This rhenium study was not optimized for low resistivity. Rather, it focussed on material aspects, for example, the substrates used in the experimental effort did not have a heavily doped contact region.
It is this third surface condition that we chose to simulate using CASTEP software by Accelrys Inc. The modeling procedure and results of this computation follows.
Section snippets
Details of the calculation
The theoretical basis of CASTEP is the density functional theory (DFT) in the local density approximation (LDA) or gradient-corrected LDA version as developed by Perdew and Wang [14] (CGA). The default setting in CASTEP is CGA. The electron–ion interaction is described using a pseudopotential concept. Potentials have been generated using the optimization scheme of Lin et al. [15] Pseudopotentials might be either local or nonlocal, the latter in a separable Kleinman–Bylander form [16]. The
Computational results
As mentioned above, the experimental metallization of a silicon-rich 6H-SiC surface with nickel led to a surface with poor current carrying characteristics and which clustered when annealed at elevated temperatures, about 1000 °C. Computationally, nickel atoms deposited onto a silicon-rich cleaved (0 0 1) 6H-SiC surface bonded to the silicon atoms and to the underlying layer of carbon atoms, Fig. 1. The partial density-of-states is shown in Fig. 2. In this plot, the Fermi level is located at 0 eV.
Conclusions
Calculations show that the bond energy of rhenium to a silicon-rich surface is about twice that of nickel to the same surface. This helps explain the better stability of rhenium over nickel when deposited onto a silicon-rich surface of 6H-SiC annealed at 1000 °C for 120 min. This selected bonding of rhenium atoms to only silicon atoms was determined computationally before being notified about the experimental results. It is always encouraging to find two separate activities discover the same
Reference (23)
- et al.
J Appl Phys
(1994) Compound Semiconductors
(1995)- Massalski TB. Binary Alloy Phase Diagrams...
- et al.
J Appl Phys
(1989) - et al.
J Elect Mat
(1995) Surface science, an introduction
(1992)Electronic properties of materials
(1993)- et al.
IEEE Trans Mag
(1999) - et al.
IEEE J Circ Dev
(1992) - Photoemission work currently being studied at Wright–Patterson AFB,...