Technical Insight
Designing GaAs HBTs based on silicon (Silicon Update)
Bob Metzger describes a patent for a GaP/GaAs HBT emitter and Si base/collector structure.
A team of researchers at the National Scientific Corporation (NSC) in Phoenix, Arizona has received a patent (#6 171 920) on the design of a novel HBT that combines a silicon base and collector structure with a GaP/GaAs emitter (see ). One of the major advantages of using a heterojunction at the emitter-base junction of a transistor, in which the emitter bandgap is larger than that of the base bandgap, is the formation of an energy band discontinuity at the valence band (Ev). This effect inhibits hole transport from the base to the emitter, which in turn increases the device s emitter injection efficiency and therefore its DC current gain (). However, for those applications where no further increase in is needed, the improvement in emitter injection efficiency can be traded off against an increase in base doping (which will decrease base resistance), and a decrease in emitter doping (which will decrease emitter-base capacitance). The effect of both these modifications is to improve the RF performance of the device. NSC hopes to combine this enhanced performance with the cost savings gained by implementing the base-collector structure using a conventional silicon bipolar process to create a high performance, cost effective RF device. Problems with interdiffusion From a bandgap engineering perspective, GaP would be an excellent candidate for the emitter as it has a bandgap of 2.24 eV. This is twice that of silicon, with the majority of the bandgap difference observed at Ev. While the lattice mismatch between GaP and silicon is relatively small at 0.34% an important consideration in inhibiting the formation of defects at the emitter-base junction extensive interdiffusion of gallium and phosphorus out of the emitter and into the silicon base, along with silicon diffusion from the base and into the GaP emitter,is observed. Such an effect displaces the electrical junction from the heterojunction and severely degrades device performance. NSC proposes to eliminate this diffusion by inserting a thin GaAs layer between the GaP emitter and silicon emitter. This undoped GaAs layer, which has a lattice mismatch of 3.7%, must be kept thin so it does not exceed critical strain limits and generate defects at the heterointerface. NSC estimates that this layer should be between 5.0 and 20.0 nm. Graham Clark, director of Sales and Marketing at NSC commented on the intended applications for such a device: "Having obtained such an HBT we will have RF power amplifiers operating at frequencies much higher than silicon homojunctions." "Since you can use a very thin highly-doped base region and still maintain extremely good emitter efficiency, in addition to having a highly doped base, you could achieve a transistor with much higher ft and fmax," he explained. "Therefore, this HBT will not compete with conventional homojunction devices, since the operating frequency will be very different." "However, if the HBT device is used at the same frequency as a silicon homojunction, the advantages would be lower noise (due to low base resistance), higher gain (due to higher emitter injection efficiency), and better linearity (due to higher Early voltage and lower output conductance), while operating at the same power."
