Additional pipework opens up transistor applications for SiC

Modified physical vapor transport (MPVT), an alternative method for bulk wafer growth, could be the answer. This approach has produced aluminum-doped substrates that are suitable for insulated-gate bipolar transistors (IGBTs) - devices that could be used in electrical power converters and power circuits for the control of higher-power electrical motors.

The MPVT process could also increase the yield of 4H-SiC substrates through polytype control, and may lead to heavily phosphorus-doped substrates that reduce substrate resistance, thereby enabling the production of low-power-loss Schottky diodes.

Bulk single crystals of SiC that provide substrates for industrial device fabrication are usually grown by a seeded-sublimation technique, referred to as either physical vapor transport (PVT) or the modified Lely technique. The process involves sublimating SiC powder at high temperatures (T>2000°C), followed by recrystallization on a slightly cooler single-crystal SiC seed (figure 1a).

It is a method that was used by current market leader Cree in the early 1990s to produce the first commercial 1 inch substrates. Today the same approach produces industry-standard 2 and 3 inch wafers (see Allen et al. 2003), and it will also be used to fabricate the 4 inch wafers that will enter the market in the near future. The majority of SiC substrates manufactured require n-type doping. Nitrogen gas is the most common n-type donor and, because it does not react with graphite, it can be supplied to the SiC crystal-growth interface through a slightly porous graphite crucible.

However, one drawback of this quasi-closed-graphite-crucible approach is a lack of direct control of the gas-phase composition, which depends on parameters such as the crucible temperature, and the temperature gradient in the growth system. This is a major disadvantage: to produce high-quality crystals with a low defect concentration requires a well-defined supply of the dopant species.

Although control of dopant feeding is well developed for nitrogen, this is not the case for reactive elements such as phosphorus (n-dopant) and aluminum (p-dopant). The growth of aluminum-doped material by PVT involves adding aluminum to the source material (figure 1b). During the initial crystal-growth seeding process, growth defects are formed, caused by the far higher partial pressure of aluminum compared with the silicon- and carbon-containing gas species. As the growth continues, an undesirable dramatic fall in aluminum concentration occurs due to source depletion, leading to lower dopant concentration. We measured a 50-fold variation between the head and tail of the crystal.

At Erlangen, we have demonstrated a modified growth set-up that uses an additional gas pipe to fine-tune the gas-phase composition (figure 1c, see Wellmann et al. 2005). This approach, funded under WE2107/3 from the Deutsche Forschungsgemeinschaft, has produced the first SiC wafers suitable for high-power IGBT device applications: aluminum-doped p-type substrates with a specific resistivity of only 0.1 Ωcm.

The MPVT set-up combines all of the benefits of conventional PVT, such as inexpensive source materials and a well-developed industrial-growth process, with the advantages of chemical vapor deposition (CVD), such as control of the gas-phase composition. The approach is also unaffected by the additional gas flux, with even the addition of propane or silane producing no change to a CVD process.

To grow aluminum-doped substrates by MPVT, an aluminum-helium vapor is added to the growth cell via a pipe (figure 1c). This allows the aluminum concentration to be set within a "growth window" that enables a high concentration of p-type doping, but prevents extended defect formation resulting from too high an aluminum concentration in front of the growth interface.

Continuous dopant supply throughout the entire growth run is guaranteed by permanent maintenance of the aluminum-helium vapor flux. In our case, aluminum-helium vapor was produced by using the PVT growth cell to indirectly heat an aluminum-containing reservoir to 1000°C. Helium gas passing by the cell is enriched with aluminum vapor and the mixture is transported into the growth cell. One inherent advantage of using aluminum vapor instead of trimethyl aluminum is the absence of hydrogen species, which can passivate acceptors.

MPVT has produced aluminum-doped SiC wafers with an aluminum concentration of 1.3 x 1020 cm-3 and a room-temperature hole conductivity concentration of 2 x 1019 cm-3. This difference in density arises from incomplete thermal activation as opposed to electrical activation of the acceptors. Variation in charge-carrier concentration across the wafer is less than 10%, thereby providing further proof that the gas inlet does not alter PVT growth conditions. The material s specific resistivity is 0.1-0.2 Ω cm, a value low enough for us to enter a regime where devices can be produced without exhibiting a large voltage drop across the substrate.

Encouragingly, initial structural characterization studies suggest that defect densities in aluminum p-type doped SiC are comparable to nitrogen n-type doped SiC. Increased concentrations of p-type doping did not produce an adverse contribution to the overall dislocation density, although there were some indications of variation in the dominant dislocation type, i.e. threading versus screw dislocation. To our surprise, in our p-type material we saw none of the basal plane dislocations prevalent in n-type SiC. This may indicate that the stacking faults existing in n-type SiC may be less pronounced or even absent in aluminum p-type doped SiC, a question we are addressing in our current research.

Recent ion-implantation studies have shown that phosphorus exhibits a chemical solubility in SiC that is 10 times as great as that of nitrogen, the standard donor species (see Schmid et al. 2004). However, at typical SiC bulk-crystal-growth temperatures of above 2000°C there is no phosphorus dopant source compatible with the SiC sublimation process that occurs in a closed-graphite crucible. The MPVT growth set-up, in contrast, opens up the possibility of using the standard phosphorous dopant phosphine.

In trials we used 3-10% phosphine diluted in helium for in situ phosphorus doping. The highest dopant density achieved so far is 1.3 x 1018 cm-3, demonstrating that phosphorus doping of SiC is compatible with MPVT. At present there is no indication of a kinetically driven incorporation limit, suggesting that much higher doping levels are achievable. This regime would make phosphorus a strong candidate for doping commercial SiC bulk crystals, with the higher doping level reducing substrate resistance and leading to devices with lower power loss.

Another potential use for MPVT is the control of SiC double-layer stacking through adjustments in the carbon:silicon ratio (figure 3). Two stacking sequences are possible along the c-axis: cubic stacking in a silicon-rich environment and hexagonal stacking in a carbon-dominated atmosphere. Growth with more silane and/or propane present may control the deposited SiC polytype, because 4H-SiC and 6H-SiC differ in their sequence of cubic and hexagonal SiC double layer stacking. 4H-SiC has a periodic repetition of a hexagonal and a cubic SiC double layer, while 6H-SiC is built up of one hexagonal and two cubic SiC double layers. We expect the addition of propane to form the more hexagonal-like 4H-SiC polytype, which could, consequently, increase the substrate yield during 4H-SiC production.

With control of the polytype a possibility with MPVT, alongside well-controlled aluminum doping of substrates for IGBT applications, it may not be too long before today s SiC manufacturers consider adding further gas flow to existing crystal-growth reactors.