Mechanism of Current Gain increase of
Heterostructure Bipolar Transistors Passivated by
Low-Temperature Deposited SiNx
Abstract — The graded base InGaAs/InP heterostructure bipolar transistors (HBTs) were passivated by the low-temperature plasma deposited SiNx. The current gain was found to increase after the passivation. The study of the Gummel plots shows that the passivation results in the reduction of both the collector current and base current, while the decrease of the base current is more significant. This causes the increase of the current gain. Nitrogen plasma treatment results in the increase of the current gain, while Silane plasma treatment results in a small reduction of the current gain. The influence of the plasmas of the two sources on the HBTs' performance are investigated. Insitu ellipsometer study showed that at the initial state of SiNx deposition, the SiNx deposition rate was zero, and nitrogen plasma has strong effects on the passivation of HBTs and results in the increase of the current gain.
I. INTRODUCTION
SiNx (or SiO2) passivation of InP-based HBT was intensively investigated in recent years. But the passivation typically suffers from the degradation of current gain. In SiNx passivation, NH4 and SiH4 are typically used as sources. These two gases generate hydrogen radicals during SiNx deposition. For the SiNx passivation of InP-based HBTs, the hydrogen radicals in the plasma result in the preferential loss of group-V elements and in an group-III element rich surface. This nonstoichiometric surface typically serves as a leakage channel and results in enhanced surface recombination, these cause the increase of the base current and the decrease of current gain. The degradation of the current gain is also related to the high deposition temperature of SiNx. For example, the degradation of the current gain in HBTs was suppressed to a certain degree when SiNx was deposited at low temperature. But the current gain after passivation is always smaller than that before passivation. When the semiconductor surface is exposed to the plasma, the reaction between the surface and plasma becomes stronger with increasing temperature. The higher deposition temperature causes the even more severe nonstoichiometric problem, while the surface nonstoichiometry can be suppressed by the addition of nitrogen gas in plasma. Here we report on SiNx passivation of gradedbase InGaAs/InP HBTs. N2 gas was used as source instead of NH3 in conventional PE-CVD in order to decrease hydrogen radicals. SiNx was deposited at room temperature in order to decrease the thermal damages. An increase of current gain is obtained by the SiNx passivation. The mechanism is investigated.
II. EXPERIMENT
The HBT layers were grown by LP-MOVPE on semi-insulating (001) InP (Fe) substrates. The layer structure is shown in Table I. The HBTs were
fabricated by conventional wet etching and metal deposition. The HBTs were nonself-aligned. The layout is single emitter and double base contacts. The space between the emitter and base contact is 1 μm. For SiNx deposition, a PlasmaLab System 90 ECRPECVD from Oxford Instruments was used. Details about the facility were given in Ref. [3]. The temperature of the sample holder was maintained at 20 ºC during the deposition. The HBTs were covered with 100-nm-thick SiNx without any pre-treatment. The microwave power was set to 180 W. 90 sccm silane (SiH4) diluted in helium (5%/95%) and 5 sccm nitrogen gases were used as sources. The chamber pressure was kept at 5 mTorr during the deposition process. The refractive index of the SiNx is about 1.9 in this deposition condition. DC characteristics was measured by HP 4515B semiconductor analyzer.
III. RESULTS AND DISCUSSION
Figure 1 shows the common-emitter I-V characteristics of the HBT before and after SiNx passivation. Before the SiNx passivation, the current gain of more than 100 can be derived from this figure. The collector current, IC, increases by more than 20% at the same base current after the passivation, indicating the increase of the current gain. Although we found the reverse current of the base-collector diode increased from 6_10-10 to 3_10-7 A at VBC=1.5 V after passivation, ICEO (the collector current at IB=0 in the common emitter configuration) was not affected, as shown in Fig. 1. Figure 2 shows the Gummel plots of the HBT before and after SiNx passivation. Both the collector current and base current, IB, decrease after
passivation. The decrease of the base current is much larger than that of the collector current. The base current is composed of the following four components: the space charge recombination, the bulk recombination in the neutral base region, the surface recombination current at the space between the emitter and base contact, and the interface recombination under the base contact. For a nonselfaligned HBT, the space between the emitter and base contact is large, thus the excess carrier concentration under the base contact is very small and the interface recombination current can be neglected. While the space charge recombination and the bulk recombination components are mainly generated in the intrinsic base region. Thus the decrease of the base current is due to the decrease of the surface recombination. The decrease of the base current is also found in (NH4)2S passivation of InGaAs/InP HBT. But for SiNx passivation the base current typically increases drastically. The study of Kikawa et al. showed that the SiNx passivation causes the shift of the Fermi level position. A surface leakage channel was formed and results in the increase of the base current. In our case, the decrease of the base current indicates that no such surface channel was formed. Figure 3 shows the current gains which are derived from Fig. 2. As shown in this figure, the current gain increases after the passivation. the increase of the current gain is due to the significant decrease of the base current, as shown in Fig. 2. The large current gain before the passivation at very small collector current level is due to the leakage current of the base and collector junction. To our knowledge, there is first report that the current gain can be improved by just SiNx passivation. We now discuss the mechanism of the increase of the current gain in our case. There are two gases in our SiNx deposition: N2 and SiH4. We investigate the effects of the two kinds of plasma treatments on the
HBT performance. Here, the HBTs were treated by N2 and SiH4 plasmas at room temperature, subsequently. Figure 4 shows the Gummel plots of the HBT before and after treatment by the N2 and SiH4 plasmas. The mirowave power was 60 W and the pressure was 5 mTorr during the treatments. The collector current decreases slightly after N2-plasma treatment. In contrast, the base current decreases drastically when treated by the N2-plasma. The behavior is much like that of the HBT passivated by SiNx. The nitrogen plasma can partly remove the oxide on the surface. While the oxide typically acts as non-recombination centers. Thus the nitrogen plasma treatment causes the decrease of the surface recombination and the decrease of the base current. When the same sample was subsequently treated by the SiH4 plasma, the collector current is the same as that treated by N2-plasma, but the base current increases. The reason may be that the stoichiometric surface is broken down by the hydrogen radicals in the SiH4 plasma. This results in the increase of the surface recombination current and in the increase in the base current. Another possibility is that the stress between the InGaAs and the deposited Si attributed to the leakage current. Figure 5 shows the current gain derived from Fig. 4. The derived current gain shows that the current gain of HBT was improved by the N2-plasma treatment and the current gain of HBT was decreased a little by SiH4-plasma, as shown in Fig. 5. But the current gain is still much larger than that before treatment.
It is very helpful to understand the mechnsm of the SiNx passivation through the investigation of the deposition process. Figure 6 shows the SiNx layer thickness as a function of deposition time measured by the in-situ ellipsometry. The deposition rate is 0 at the beginning 70 s. In our system, the plasma is generated by the N2 gas, while the SiH4 gas does not
IV. CONCLUSION
We developed a new SiNx passivation process. The graded based InGaAs/InP HBTs were successfully passivated. After the passivation, both the collector and base currents in Gummel plots were found to decrease, but the base current decreased more significantly. This resulted in the increase of the current gain. When InGaAs/InP HBT is treated by one of the sources, the nitrogen plasma only, the current gain was also found to increase, while Silane treatment resulted in a little decrease of the current gain. The in-situ ellipsometry study showed that the increase of current gain is related to the effect of N2 plasma treatment at the initial deposition process.
Luiggi Marquez C.I 17677911 CRF
Actualízate gratis al nuevo Internet Explorer 8 y navega más seguro
No hay comentarios:
Publicar un comentario