192
Strain-Engineered MOSFETs
have explored the LF noise in bipolar junction transistors (BJTs), especially the
source of 1/f noise in BJTs [46, 47]. LF noise characteristics of SiGe HBTs are also
reported [48]. One of the most important sources of 1/f noise is located within
the thin SiO 2 interfacial layer between the monosilicon and polysilicon emitter
regions. The physical mechanism of this noise is interpreted by tunneling probability fluctuation [47] or number carrier fluctuation [46], and it is still unclear.
The random telegraph noise (RTN) observations, on the other hand, in BJTs,
have mainly been interpreted in terms of noise sources in the base-emitter
space charge region [49] and interfacial oxide in the emitter region [50]. In
the former case, the traps in the base-emitter space charge region cause the
barrier height across the junction to fluctuate, resulting in RTS pulses. In the
latter case, the increase of nonideal base current gives rise to RTS pulses,
with the source being the traps in the spacer oxide at the emitter interface.
Several physical mechanisms exist, such as fluctuating barrier height and
trap-assisted tunneling current across the junction, multiphonon capture by
traps, fluctuating recombination rate in the base-emitter space charge region,
and fluctuating capture cross sections. Overall, the trapping/de-trapping
mechanism in bipolar junction transistors is yet to be fully understood.
Studies on low-frequency noise in SiGe:C HBT are scarce and exact physical
mechanisms are vague. In this work, low-frequency noise behaviour consisting of 1/f noise and RTN in SiGe:C high-speed HBT has been characterised. Bias dependency of 1/f noise in the base contact is demonstrated,
and existence of a nonideal base current component has been shown. RTN is
characterised in terms of amplitude and characteristic times in the low and
high level of base current as a function of bias. Mechanisms behind the bias
dependency of RTN are explained based on the experimental results with
the help of fluctuating barrier height theory.
The SiGe:C HBT was fabricated in an industry standard 0.25 μm BiCMOS
process with 25 lithographic steps. The HBT devices used in the work have
an emitter area AE of 0.42 × 0.84 μm 2 with the typical parameters f T /f max =
75/90 GHz, BV CEO = 2.4 V, and β = 100. The HBT SiGe base layer incorporates
a very low carbon concentration of about 1 × 20 Cm –3 to suppress boron diffusion in the base layer and the surrounding Si regions. Further details on
the device can be found in [51].
Figure 6.37(a) shows the typical output characteristics (I C -V CE ) of the NPN
SiGe:C HBT with an emitter area A E of 0.42 × 0.84 μm 2 . The typical gummel
plot of the HBT is shown in Figure 6.37(b), with the inset showing the current
gain β vs. V BE plot. Apart from this, an f max of 90 GHz and a maximum β of
115 were measured.
6.9.1 Low-Frequency Noise Measurement of SiGe:C HBT
Figure 6.38 shows the base current noise power spectral density S IB at three
different base voltages, which show 1/f 2 dependence with frequency. The
Strain-Engineered MOSFETs
have explored the LF noise in bipolar junction transistors (BJTs), especially the
source of 1/f noise in BJTs [46, 47]. LF noise characteristics of SiGe HBTs are also
reported [48]. One of the most important sources of 1/f noise is located within
the thin SiO 2 interfacial layer between the monosilicon and polysilicon emitter
regions. The physical mechanism of this noise is interpreted by tunneling probability fluctuation [47] or number carrier fluctuation [46], and it is still unclear.
The random telegraph noise (RTN) observations, on the other hand, in BJTs,
have mainly been interpreted in terms of noise sources in the base-emitter
space charge region [49] and interfacial oxide in the emitter region [50]. In
the former case, the traps in the base-emitter space charge region cause the
barrier height across the junction to fluctuate, resulting in RTS pulses. In the
latter case, the increase of nonideal base current gives rise to RTS pulses,
with the source being the traps in the spacer oxide at the emitter interface.
Several physical mechanisms exist, such as fluctuating barrier height and
trap-assisted tunneling current across the junction, multiphonon capture by
traps, fluctuating recombination rate in the base-emitter space charge region,
and fluctuating capture cross sections. Overall, the trapping/de-trapping
mechanism in bipolar junction transistors is yet to be fully understood.
Studies on low-frequency noise in SiGe:C HBT are scarce and exact physical
mechanisms are vague. In this work, low-frequency noise behaviour consisting of 1/f noise and RTN in SiGe:C high-speed HBT has been characterised. Bias dependency of 1/f noise in the base contact is demonstrated,
and existence of a nonideal base current component has been shown. RTN is
characterised in terms of amplitude and characteristic times in the low and
high level of base current as a function of bias. Mechanisms behind the bias
dependency of RTN are explained based on the experimental results with
the help of fluctuating barrier height theory.
The SiGe:C HBT was fabricated in an industry standard 0.25 μm BiCMOS
process with 25 lithographic steps. The HBT devices used in the work have
an emitter area AE of 0.42 × 0.84 μm 2 with the typical parameters f T /f max =
75/90 GHz, BV CEO = 2.4 V, and β = 100. The HBT SiGe base layer incorporates
a very low carbon concentration of about 1 × 20 Cm –3 to suppress boron diffusion in the base layer and the surrounding Si regions. Further details on
the device can be found in [51].
Figure 6.37(a) shows the typical output characteristics (I C -V CE ) of the NPN
SiGe:C HBT with an emitter area A E of 0.42 × 0.84 μm 2 . The typical gummel
plot of the HBT is shown in Figure 6.37(b), with the inset showing the current
gain β vs. V BE plot. Apart from this, an f max of 90 GHz and a maximum β of
115 were measured.
6.9.1 Low-Frequency Noise Measurement of SiGe:C HBT
Figure 6.38 shows the base current noise power spectral density S IB at three
different base voltages, which show 1/f 2 dependence with frequency. The