Study of Epitaxial Growth and Fabrication of InGaP/GaAs HBT’s Ricardo T. Yoshioka , Augusto C. Redolfi, Jacobus W. Swart. Jefferson Bettini and Mauro M. G. de Carvalho DSIF/FEEC and LPD/IFGW - UNICAMP Abstract: An initial study of the growth of InGaP/GaAs layers by means of Chemical Beam Epitaxy (CBE) and the fabrication of Heterojunction Bipolar Transistors, HBT’s, is presented. This study shows the dependency of emitter-base I-V curves with growth conditions of the layers and gives the indication for designing the following experiments. The results represent the first working HBT’s totally fabricated at our laboratory and in Brazil. 1- Introduction Heterojunction bipolar transistors, HBT’s, present superior performance compared to homojunction bipolar transistors, BJT’s, due to the fact that they allow the use of highly doped base regions, with a doping level higher than that of the emitter, leading to high maximum oscillation frequency. This inverse emitter/base doping level ratio is possible because high emitter injection efficiency is obtained mainly due to the emitter/base heterojunction band-gap offset. The heterojunction emitter/base can be obtained with III-V materials of different compositions, such as: AlGaAs/GaAs, AlInAs/InGaAs, InP/InGaAs and InGaP/GaAs. HBT’s of AlGaAs/GaAs are the most popular and mostly studied up to now. HBT’s of InGaP/GaAs have been proposed and studied lately to replace the AlGaAs/GaAs structure because of its superior characteristics [1, 2]. Among these, a higher valence band off-set (~300 meV) [3], the absence of DX-centers in the emitter and a high etching selectivity between InGaP and GaAs are the main advantages. A study of the growth of InGaP and GaAs epitaxial layers by means of CBE has been performed at our laboratory [4]. Following this, new layers has been designed and grown for the fabrication of HBT’s. This paper presents this study, including the fabrication of HBT’s and the correlation between the heterojunction characteristics and the growth conditions. 2- Experimental Procedures Four samples has been grown according to a typical layer specification for HBT as shown in table 1. The growth series are named as #CBE616, #CBE636, #CBE638 and #CBE647. Table 1: Specified layer structure for the HBT fabrication. Layer Material Thickness [nm] Doping [cm-3] Cap GaAs (Si) 250 (n+)4e18 Emitter In0.5Ga0.5P (Si) 100 (n)3e17 Spacer GaAs (Un) 10 Base GaAs (Be) 100 (p+)5e19 Collector GaAs (Si) 500 (n)3e16 Sub-collector GaAs (Si) 500 (n+)4e18 GaAs layers were grown using arsine and triethylgallium (TEG) sources while for InGaP layers, trimethylindium (TMI), TEG and phosphine sources were used. Silicon doping was used for n-type layers, while beryllium doping was used for the p-type base layer. A 10 nm thick undoped spacer layer was inserted between the n type InGaP emitter and the p type GaAs base layers. This spacer layer is designed in order to account for the diffusion of the beryllium from the highly doped base layer and to avoid it from entering the emitter layer. This layer was however omitted in the #CBE647 growth run. Table 2 shows the substrate and the solid silicon and beryllium doping cell temperatures during growth and the corresponding surface quality obtained. Table 2: Temperatures of growth conditions and final surface characteristic Growth TSubstrate [oC] TSi-cell [oC] TBe-cell [oC] Surface + run n /n p+ (view by pyrometer) #CBE616 535 1100 / 1050 790 Mirror like #CBE636 535 1140 / 900 800 Slightly milky #CBE638 535 1140 / 900 800 Slightly milky #CBE647 520 1140 / 900 800 Mirror like The doping profile of sample of run #CBE616 were obtained by means of electrochemical C-V profiling. The doping levels of the n+ and n layers were respectively below and above the specified values of table 1. For this reason, the silicon cell temperature was changed to the values as in table 2 for the other growth runs. The beryllium cell was also increased in order to correct the base doping. In order to avoid unwanted beryllium doping in the other layers by increasing the beryllium cell temperature, the following care measures were taken: increase the time interval with no growth, between base and emitter growth, from 30s to 90s; reduce the standby temperature of the beryllium cell from 500oC to 200oC. Even with these precautions, the changes in doping cell temperatures resulted in slightly milky surfaces. In order to improve the surface quality, the growth temperature was reduced to 520oC in run #CBE647, what resulted in a mirror like surface again. Transistors were fabricated on samples of the above grown layers, using a process sequence similar to the one described before for AlGaAs/GaAs HBT’s[5]. The differences were: 1) the following selective etching solutions were used for GaAs and InGaP respectively: H3PO4:H2O2:H2O (3:1:50) and HCl. 2) The metallizations and sinterings after emitter metallization were also specific in each run, as indicated in table 3. Table 3: Metallization and sintering conditions for each run Sample / Emitter [nm] Base [nm] Collector [nm] Sintering Device (oC/s) #CBE616 CPqD LME/USP CPqD 250/10 / Ni/Ge/Au/Ni/Au Ti/Pt/Au Ni/Ge/Au/Ni/Au 340/5 DN17 (5/50/100/50/100) (50/50/100) (5/50/100/50/100) #CBE636 LPD LPD LPD 150/15 / AuGeNi/Au Cr/Al AuGeNi/Au 300/15 DN19 #CBE638 LPD LPD LPD (1o)150 /15 / AuGeNi/Au Cr/Al AuGeNi/Au 300/15 DN20 (2o)150/10 330/30 LPD #CBE647 LPD LPD 250 /15 / AuGeNi/Au Cr/Al AuGeNi/Au 350 /30 DN21 The metallizations of samples #CBE636, #CBE638, #CBE647 were all performed at our laboratory using a refractory wire coil as resistive heating source. The metallizations for sample #CBE616 were done at CPqD/Telebrás and at LME/USP using E-beam metallization systems. The sintering done after the emitter metallizations aims to increase its adhesion to endure the following wet etching step and to improve its ohmic contact resistance. This is however a critical step because it can also degrade the performance of the device. Sample #CBE616 went through a final alloy step in order to optimize its ohmic contact resistance. This was also done in a RTP system in a two temperature step sequence of 250/10 and 350/10 [oC/s]. Similar alloy step on samples #CBE636, #CBE638 and #CBE647 degraded these devices. For this reason, only characteristics measured before alloy will be presented for these devices. 3- Results and Discussion Device DN17 was the first InGaP/GaAs HBT fabricated at our laboratory. The Gummel-Plot of the transistor is shown in Fig.1. It shows that the ideallity factor for the collector current is quite good, namely 1.2. However, the ideallity factor for the base current is high, 3.9, indicating that the emitter-base junction is of poor quality. The growth runs of #CBE636, #CBE638, #CBE647 were aimed to improve this junction quality, by changing the growth conditions as described above. As mentioned before, the final alloy step of devices DN19, DN20 and DN21 resulted in over-alloy with leaky junctions (due to a problem with the thermocouple of the RTP as identified afterwards). The emitter-base I-V curves of the devices (Fig. 2) were measured before the final alloy step. The ideallity factors can be estimated from these curves and are shown in table 4. These results indicate that the quality of the emitter-base junction increased with the improvement in the epitaxial growth conditions as described above. The improvement of the ideallity factor of devices DN19 and DN20 compared to device DN17 can be attributed to increased time interval used between the growth of the base and emitter layers and reduce the standby temperature of the beryllium cell from 500oC to 200oC. It also indicates that the milky like surface does not affect much the junction. The further improved ideallity factor of device DN21 is probably related to the better InGaP material due to the lower substrate temperature during growth. The effect of using or not the 10 nm spacer layer between emitter and base needs to be investigated further, because in growth #CBE647 two variables were altered at once (temperature and no spacer). A new growth is scheduled with the same parameters and adding the spacer layer. The optimization of the growth conditions and the device processing will be continued. The obtained results are giving the directions of this work and represent the results of the first HBT totally fabricated at our laboratory and in Brazil. Fig.ure 1- The Gummel Plot of DN17 device after alloy process. Figure 2- Emitter-base I-V curve of DN21 device before alloy process. Table 4: Ideallity factor obtained from Gummel-Plot and emitter-base I-V curves. Device Ideallity Factor 1.2 DN17 (#CBE616) 3.9 DN19 (#CBE636) DN20 (#CBE638) DN21 (#CBE647) 3.4 3.0 2.5 Measurement Gummel Plot (Transistor) - collector current Gummel Plot (Transistor) - base current I-V of E/B junction I-V of E/B junction I-V of E/B junction Acknowledgments: This work has been supported by FAPESP, FINEP and CNPq. Additionally the authors would like to acknowledge LME/USP and CPqD/Telebrás for metallization processes and Mrs. K. M. I. Landers, Mr. A. A. Von Zuben and Mr. A. C. Silveira for sample preparation, metallization and C-V profile measurements respectively, at our laboratory. References: [1] M. J. Mondry and Kroemer, “Heterojunction bipolar transistor using (Ga,In)P emitter on a base, grown by MBE”, IEEE Electron Device Lett., vol. EDL-6, pp 175-177, 1985. [2] T. Kobayashi, K.Taira, F. Nakamura, and H. Kawai, “Band lineup for a GaInP/GaAs heterojunction measured by a high-gain npn heterojunction bipolar transistor grown by metalorganic chemical vapor deposition”, J.Appl. Phys., vol 65, pp.4898-4902, 1989. [3] M. A. Rao, E. J. Caine, H. Kroemer, S. I. Loun, and D. I. Babic,: “Determination of valence and conductionband discontinuities at the (Ga, In)P/GaAs heterojunction by C-V profiling”, J.Appl. Phys., vol. 61, no. 2, pp.643-649, 1987. [4] J. Bettini, Master dissertation, “Crescimento Epitaxial de InxGa1-xP sobre GaAs pela Técnica de Epitaxia por Feixe Químico (CBE)”. March 1997. [5] A. C. Redolfi, R. T. Yoshioka and J. W. Swart, “Design of a Test Chip for HBT Process Development and Results of AlGaAs/GaAs HBT Fabrication”, 10th Congress of the Brazilian Microelectronics Society and 1st Ibero American Microelectronics Conference, Canela - RS - Brazil, July 1995, pp.503-514.