SiGe HBT Technology:The SiGe Heterojunction Bipolar Transistor.

The SiGe Heterojunction Bipolar Transistor

The SiGe HBT is by far the most mature Si-based bandgap-engineered electronic device [5]. The first SiGe HBT was reported in 1987 [6,7], and is in commercial production today worldwide at upwards of 25 different companies. Evolution in transistor performance has been exceptionally rapid (Figure 4.2). Significant steps along the path to manufacturing included the first demonstration of high-frequency (75 GHz) operation of a SiGe HBT in a non-self-aligned structure early 1990 [8]. This result garnered significant attention worldwide since the performance of the SiGe HBT was roughly twice what a state- of-the-art Si BJT(Bipolar Junction Transistor) could achieve. The first fully integrated, self-aligned SiGe HBT technology was demonstrated later in 1990 [9], the first fully integrated 0.5 m SiGe BiCMOS technology (SiGe HBT + Si CMOS) in 1992 [10], and SiGe HBTs with frequency response above 100 GHz in 1993 and 1994 [11,12]. A number of companies around the world have demonstrated robust SiGe HBT technologies (upwards of 25 companies in 2005) [13–25]. Recent work has focused on practical applications of SiGe HBT circuits for a wide variety of mixed-signal circuit applications [26–28].

Because the intent in SiGe technology is to combine bandgap engineering with conventional Si fabrica- tion techniques, most SiGe HBT technologies appear very similar in structure to conventional Si bipolar technologies. A typical device cross section is shown in Figure 4.3. This SiGe HBT has a planar, self-aligned structure with a polysilicon emitter contact, silicided extrinsic base, and deep- and shallow-trench isolation. A 5–7 level, chemical-mechanical-polishing (CMP) planarized, W-stud, AlCu (or full Cu metallization followed by a final thick Al layer) CMOS-like metalization scheme is used. The extrinsic resistive and capacitive parasitics are intentionally minimized to improve the maximum oscillation frequency (fmax) of

the transistor. Observe that the Ge is introduced only into the thin base region of the transistor, and is deposited with a thickness and Ge content that ensures the film is thermodynamically stable. The in situ boron-doped, graded SiGe base is deposited across the entire wafer using growth techniques such as UHV/ CVD. In areas that are not covered by oxide, the SiGe film consisting of an intrinsic-Si/strained boron- doped SiGe/intrinsic-Si stack is deposited as a perfect single-crystal layer on the Si substrate. Over the oxide, the deposited layer is polycrystalline (poly), and will serve either as the extrinsic base contact of the SiGe HBT, the poly-on-oxide resistor, or the gate electrode of the Si CMOS devices. The metallurgical base and single-crystal emitter widths range from 30 to 90 nm and 25 to 35 nm, respectively. A masked phosphorus implant is used to tailor the intrinsic collector profile for optimum frequency response at high current densities as well as define multiple breakdown voltage devices on the same wafer. A conventional deep-trench/shallow-trench bipolar isolation scheme is used. This approach ensures that the SiGe HBT is compatible with commonly used (low-cost) bipolar/CMOS fabrication processes. A typical doping profile measured by secondary ion mass spectroscopy (SIMS) of the resultant SiGe HBT is shown in Figure 4.4. The smaller base bandgap of the SiGe HBT can be exploited in three major ways, and is best illustrated by examining an energy band diagram comparing a SiGe HBT with a Si BJT (Figure 4.5). First, note the reduction in base bandgap at the emitter–base junction. The reduction in the potential barrier at the emitter–base junction in a SiGe HBT will exponentially increase the collector current density, and hence current gain (b = JC/JB) for a given bias voltage compared with a comparably designed Si BJT. Compared with a Si BJT of identical doping profile, this enhancement in current gain is given by

where h = NCNV(SiGe)/NCNV (Si) is the ratio of the density-of-states product between SiGe and Si, and g = Dnb(SiGe)/Dnb(Si) accounts for the differences between the electron mobilities in the base between Si and SiGe. The position dependence of the band offset with respect to Si is conveniently expressed as a bandgap grading term DEg,Ge (grade) = (DEg,Ge (Wb) – DEg,Ge (0)). As can be seen in Figure 4.6, which compares the measured Gummel characteristics for two identically constructed SiGe HBTs and Si BJTs, these theoretical expectations are clearly borne out in practice.

Second, if the Ge content is graded across the base region of the transistor, the conduction band edge becomes position-dependent (refer to Figure 4.5), inducing an electric field in the base which accelerates the injected electrons. The base transit time is thereby shortened and the frequency response of the transistor is improved according to

Figure 4.7 compares the measured unity gain cutoff frequency (fT) of a first-generation SiGe HBT and a comparably constructed Si BJT, and shows that an improvement in peak fT of 1.7´ can be obtained with relatively modest Ge profile grading (0 to 7.5% in this case). Aggressive vertical profile scaling has resulted in dramatic improvements in device performance, with peak cutoff frequencies currently in excess of 300 GHz (Figure 4.8).

The final advantage of using a graded Ge profile in a SiGe HBT is the improvement in the output conductance of the transistor, an important analog design metric. For a graded-base SiGe HBT the Early

In essence, the position dependence of the bandgap in the graded-base SiGe HBT weights the base profile toward the collector region, making it harder to deplete the base with collector–base bias, hence yielding a larger Early voltage. A transistor with a high Early voltage has a very flat common-emitter output characteristic, and hence low output conductance. For the device shown in Figure 4.6, the Early voltage increases from 18 V in the Si BJT to 53 V in the SiGe HBT, a 3´ improvement.