next up previous contents
Next: 4.4.2 Device Reliability Up: 4.4 Analysis of HBT Previous: 4.4 Analysis of HBT

Subsections


4.4.1 Importance of the InGaP Ledge

It is well known that GaAs-HBTs with an InGaP ledge have an improved reliability [203]. The emitter material covers the complete p-doped base layer, thus forming the so-called ledge. The impact of the ledge thickness $d$ and negative surface charges, which exist at the ledge/nitride interface, on the device performance is investigated using MINIMOS-NT. The surface charges have large impact on the Fermi-level pinning at the InGaP/SiN interface. A schematic drawing of the simulated device structure is shown in Fig. 4.14. In order to save computational effort, the simulation domain covers only one half of the real symmetric device structure.

Figure 4.28: Hole current density [A/cm$^2$]: Leakage path near the Si$_3$N$_4$ interface occurring in the presence of negative charges
\resizebox{1.0\halflength}{!}{
\psfrag{Base}{\hspace*{-5mm}\large{\bf Base}}
\ps...
...arge{\bf Si$_3$N$_4$}}
\includegraphics[width=1.1\halflength]{figs/colSiN.epsi}}

4.4.1.1 Degradation Mechanisms in Devices without Ledge

In case of devices where no ledge is present (see Fig. 4.28), the simulation results suppose that during stress some of the electrons flowing in the emitter are injected in the insulator and get trapped there. The negative charge at the semiconductor/insulator interface can lead to a hole leakage path in the vicinity of the interface, and therefore, to undesirably high base currents.

4.4.1.2 InGaP Ledge Thickness

In Fig. 4.29 measured and simulated collector and base currents of one-finger InGaP/GaAs HBTs with different ledge thickness operating under forward Gummel plot conditions with V$_\mathrm {BC}$ = 0 V are shown. Measurement refers to a device with 40 nm ledge thickness. Surface charges at any of the device interfaces are not yet considered in the simulation. Note the strong increase in the base current at low bias with increasing ledge thickness. As can be seen from Fig. 4.29 simulated and measured base currents differ significantly in the case of 40 nm ledge thickness. Only simulation with a ledge thickness of less than 20 nm delivers a good match. The reason is that insulator surface Fermi-level pinning is not taken into account if surface charges are not considered in the simulation. Therefore, a non-physical electron current path occurs in the upper part of the ledge, as shown in Fig. 4.30. However, this leakage path can be overcome by means of electrically isolated base contacts. The corresponding electron distribution in the ledge using vertical cross-sections at x = 1.6 $\mu$m, 2.0 $\mu$m, and 2.4 $\mu$m is shown in Fig. 4.31. The hole distribution in the middle of the ledge (x = 2.0 $\mu$m) is also included. These concentrations shall be compared to the ones in the case of surface charges in the next subsection.

Figure 4.29: Dependence of I $_{\mathrm {B}}$ on the InGaP ledge thickness compared to measurement
\resizebox{\halflength}{!}{
\includegraphics[width=\halflength]{figs/Nsis1.eps}}

Figure 4.30: Electron current density [A/cm$^2$] at V $_{\mathrm {BE}}$=1.2V: Simulation without surface charges
\resizebox{\halflength}{!}{
\includegraphics[width=\halflength]{figs/0e12.epsi}}

Figure 4.31: Electron and hole distribution in the ledge: Simulation without surface charges
\resizebox{\halflength}{!}{
\includegraphics[width=\halflength]{figs/np0e12.eps}}

4.4.1.3 Negative Surface Charges

As can be seen from Fig. 4.32, where symbols represent experimental data for the collector current I$_\mathrm {C}$ and the base current I$_\mathrm {B}$ and simulation refers to a device with 40 nm ledge, the base current decreases if more negative surface charges are introduced. The upper part of the ledge is depleted as well [204] and the leakage is reduced (Fig. 4.33). The corresponding electron distribution in the ledge at x = 1.6 $\mu$m, 2.0 $\mu$m, and 2.4 $\mu$m, and the hole distribution at x = 2.0 $\mu$m are presented in Fig. 4.34. Note that even in this case the ledge is not completely depleted. However, the electron concentrations near the InGaP/SiN interface are significantly lower in comparison to the ones shown in Fig. 4.31. Thus, with a surface charge density of $10^{12}$ cm$^{-2}$ the measured base current can be simulated very well. Note that in the case of negative surface charges the hole concentration in the ledge increases and at higher values gives the opportunity a hole current path to occur.

Figure 4.32: Dependence of I $_{\mathrm {B}}$ on the negative charge density at the ledge/nitride interface with d = 40 nm: A charge density of $10^{12}$ cm$^{-2}$ is sufficient to get good fit to the measurements
\resizebox{\halflength}{!}{
\includegraphics[width=\halflength]{figs/Nsis2.eps}}

Figure 4.33: Electron current density [A/cm$^2$] at V $_{\mathrm {BE}}$=1.2V: Simulation with a surface charge density of $10^{12}$ cm$^{-2}$
\resizebox{\halflength}{!}{
\includegraphics[width=\halflength]{figs/1e12.epsi}}

Figure 4.34: Electron and hole distribution in the ledge: Simulation with a surface charge density of $10^{12}$ cm$^{-2}$
\resizebox{\halflength}{!}{
\includegraphics[width=\halflength]{figs/np1e12.eps}}


next up previous contents
Next: 4.4.2 Device Reliability Up: 4.4 Analysis of HBT Previous: 4.4 Analysis of HBT
Vassil Palankovski
2001-02-28