The advance of GaAs/InGaP HBTs gathers momentum
The vertical dimensions of the transistor layers determine the DC and RF properties of HBTs, so the epitaxial process used to form the device layers plays a crucial role. Devices such as PHEMTs are also grown epitaxially, but because these are laid out laterally the gate dimensions defined by lithography are a crucial factor in performance. The vertical nature of the HBT structure means that device performance is less constrained by the lithography steps during processing. When an HBT manufacturing facility receives a supply of epiwafers, the vertical transistor structures are already in place. The manufacturing process involves defining and connecting individual transistors into a circuit (figure 1). The manufacturer can benefit from the relatively relaxed design rules on defining and connecting the individual transistors, compared with those needed to define the gate features in FET-type devices.
The move to InGaP emitters The first GaAs-based HBTs to be commercialized had AlGaAs emitter layers, but the majority of development and new HBT production programs are now focusing on devices with an InGaP emitter layer (figure 2). In the future it looks like the GaAs/InGaP HBT will supersede the GaAs/AlGaAs HBT in many applications. There are a number of reasons why using an InGaP emitter might improve the performance of HBTs. InGaP provides a larger difference in valence band energies between the base and the emitter. Secondly, the high etch selectivity between InGaP and GaAs makes device processing easier. Finally, oxygen incorporation into AlGaAs emitters is thought to give rise to recombination centers, which reduces a device s reliability. InGaP is less reactive to oxygen so higher current gain and reliability can be achieved.
While the side-by-side comparison of InGaP and AlGaAs emitter structures may well show some differences in factors like reliability, it is not clear that users exercise the reliability already available from an AlGaAs structure. The improvements of going to InGaP for an application such as a cell-phone PA may well be insignificant. As Kopin s Dave Davito pointed out, "Whether the device lasts 106 or 5 x 106 hours is really irrelevant as nobody is ever going to see that. Our customers tell us they need to offer an InGaP HBT in their product portfolio because their competitors have one too. That is at least partially driving the shift to InGaP."
The MBE-grown GaAs/AlGaAs HBTs initially developed by TRW use the highly mobile element Be as the p-type base dopant. This dopant needs to be kept to such a level that diffusion into surrounding layers does not occur and affect device performance. A high p-type doping in the base layer is necessary to achieve good RF characteristics and the industry seems to have standardized on a carbon-doping level of 4 x 1019 atoms/cm3 for InGaP HBTs. Be doping levels are typically 1 x 1019 atoms/cm3. The flexibility that the additional RF performance from higher base doping brings to designers will be another factor driving the uptake of InGaP HBTs.
Challenges to epi In order to exploit the benefits of HBTs fully, manufacturers need epiwafers with good uniformity of layer thickness, doping and crystal quality, both across the wafer and from wafer to wafer. Without this, transistor parameters such as turn-on voltage, base resistivity and breakdown voltage will not be reproducible irrespective of the processing steps taken to define and connect the HBTs.
The move to InGaP has highlighted a number of problems that growers have had to tackle. Arsenic is easily incorporated into InGaP to form InGaAsP which has a lower bandgap energy and etch selectivity. A lattice-mismatched InGaAsP layer may be unintentionally formed at the GaAs/InGaP interface, introducing defects at the interface. The electrical properties of InGaP change with its crystal structure and composition, so good control of the In and Ga composition is needed to prevent defects being introduced. In0.51Ga0.49P is required for lattice matching to GaAs.
Another critical step is the growth of a highly n-type InGaAs non-alloy emitter contact layer. This layer is lattice mismatched to the GaAs emitter contact below and care must be taken to reduce surface roughness for high In content layers. Carefully optimizing the growth parameters results in surface roughness of about 1 nm for In0.6Ga0.4As layers.
Hydrogen in MOCVD can cause a number of problems in the base layer. The gas passivates the carbon dopant and causes an initial instability in the current gain of the device. This transient instability is often referred to as "the burn-in effect".
MOCVD gets a head start A search through the marketing literature of the world s epiwafer vendors will reveal that the growth of InGaP HBT epiwafers is largely the territory of MOCVD. The reasons for this are varied and are founded on historical and political grounds as much as economic and technical ones. When AlGaAs HBTs first became commercially available MBE was the dominant technique used to grow them, but the pendulum has swung to MOCVD with the advent of InGaP emitter HBTs.
Phosphorus had an obstacle to overcome before it could be routinely and safely used as a solid source in MBE. The safest form of phosphorus, red phosphorus, does not give stable results from a normal MBE effusion cell. White phosphorus is the more reactive relation of red, and while it is possible to get more stable effusion from this material its reactivity makes it difficult to handle safely.
