Recent progress in atomic layer epitaxy of III–V compounds

101 RECENT PROGRESS IN ATOMIC LAYER EPITAXY OF rn-V COMPOUNDS S. M. Bedair North Carolina State University, Dept. of Electrical and Computer Engineering, Raleigh, North Carolina 27695-7911 ABSTRACT The potential applications of Atomic Layer Epitaxy of HI-V compounds will be outlined. These include the growth of special structures and devices such as ordered alloys, ultra-thin quantum wells, non-alloyed contacts, planar doped FET's and HBT's. Also, the main challenges facing ALE will be outlined along with possible solutions. These include reactor design, control of carbon doping and the growth of ternary alloys. A general assessment of the ALE technology will be provided. CURRENT CHALLENGES FACING THE ALE TECHNIQUE The ALE technique has suffered from several shortcomings that we believe have slowed down its potential applications and the interest of many researchers. The first problem is the very low growth rates where in some cases growth rates as slow as 0.02 fim/hour were reported. Recently improvement in the growth rate has been achieved and a growth rate of about 0.2 Mm/h was reported, which we still believe to be discouragingly slow"'. The main reason for such a low growth rate is the commonly used approach based on exposure/purging each of the reactants with a vent/run manifold configuration. The finite gas residence time in the reactor and valve switching times will always lead to growth of only a small fraction of a micron per hour. The second problem facing ALE is gas phase reactions which limit ALE growth to low temperatures. The premature decomposition of trimethylgallium (TMG), for example, in the gas phase results in the growth of more than one monolayer per ALE cycle. The process in this case is not controlled by surface reactions. Another problem facing ALE is the synthesis of ternary alloys such as AlGaAs and InGaAs. This problem is due to the narrow temperature range of ALE in most systems and thus the lack of compatible group III precursors that will provide a self-limiting process at the same temperature. Some of the above problems have been eliminated by optimization of ALE reactor design. The approach adopted in our laboratory relies on rotating the substrate between the different source gas streams that are continuously flowing through a specially designed vertical reactor. The growth rate will depend on the substrate rotation speed. Growth rates in the range of 0.4 to 0.7 Mum/h can be achieved with this approach. Such growth rates are comparable with those reported by MBE. For high temperature growth and to achieve compatible growth of ternary alloys the thickness of the thermal boundary layer must be minimized. The approach taken in our laboratory is to mechanically shear off the gaseous boundary layer when the substrate rotates between the reactive gas streams. One ALE reactor used is a modified Emcore 3200 system operating at 30 torr. The reactor chamber is partitioned into six compartments to further separate the reactive gases and to assist in shearing off the boundary layer. The chamber is divided by 0.01" molybdenum sheets (baffles). The detailed design is discussed elsewherer2]. ALE growth of GaAs using TMG and AsH 3, with and without the baffles was studied. We have observed that monolayer per cycle growth is achieved only when the baffles are Mat. Res. Soc. Symp. Proc. Vol. 222. 01991 Materials Research Society 102 1.8600 C A 1.6- U.. S S 650 C 1.4 1• .= 0.8 E 0.60.40.2 0i 0 Figure 1 0.02 0.04 0.06 0.08 0.1 0.12 TMGa Integrated Flux (urnole) 0.14 0.16 Growth of GaAs as a Function of TMG Integrated Flux. used. Also, thickness uniformity across the wafer is only observed under the ALE growth conditions. The range of self-limiting growth is limited, due to incomplete removal of the boundary layer. Another system used in our laboratory operates at atmospheric pressure and uses a graphite susceptor. This system allows better removal of the boundary layer and the self-limiting process was observed over a broader range of growth conditions. Figure 1 shows that the self-limiting conditions are met for growth at 600 and 6500(2 over a wide range of TMG integrated flux. The two systems were used to achieve device quality GaAs, AlGaAs, InGaP and other films. Also several devices were built based on these ALE materials such as 6-doped FETs, heterojunction bipolar transistors, resonant tunneling diodes and others. In the following discussions we will outline some of these results. ALE GROWTH OF GaInP Details of the growth process of Ga 0.5In0.5P grown on GaAs substrates were previously reported"'. The growth relies on the sequential exposure to TMG, PH3, TEI and PH3 to grow Ga-P-In-P, ... layers. Double crystal x-ray diffraction was performed and confirms that the ALE grown InGal.,P films are lattice matched to GaAs with x = 0.5 for films grown on either (100) or (100) 2' off oriented substrates. Cross-sectional TEM was also used to study the crystal structure of the ALE GaInP films. Ordering was not observed for films grown on the (100) nominal substrates. However, ordering was observed for all samples grown on the 2* misoriented substrates. The ordering is found to be CuPt, where the Ga and In atoms alternate on (111) planes on the column III sublattice. This can be explained based on the atomic arrangement which favours that phosphorus atoms be attached to three In atoms (or Ga and one Ga (or In) rather than two In and two Ga atoms. Thus, a highly strained InP-GaP monolayer superlattice is formed as a result of the ordered arrangement of this structure. Photoluminescence (PL) at liquid helium temperature was also performed on the ordered and 103 8000 5000 angstroms Figure 2 4K PL Spectrum of GaInP for; a) Ordered Structure, b) Disordered Structure. disordered structures and results are shown in Figure 2. The PL shows emission peaks at 1.85 and 2.01 eV for the ordered and disordered structures respectively as shown in the figure. This difference in the value of Eý is confirmed by photoreflectance measurements141 , where ordered structures showed a record low bandgap of 1.76 eV at room temperature. This value of the bandgap is the lowest reported for this ternary and may be a result of achieving such a high degree of ordering by the ALE technique. To assess the quality of the GaInP grown by ALE, a series of GaAs quantum wells (QWs) were grown using approximately 350 A of GaInP on either side for the high bandgap barrier. Since the ordered material forms domains and has a suppressed bandgap, the disordered alloy was used for the barriers. Growth temperature was 4800C to minimize intermixing at the GaAs/GaInP heterointerfaces. Phosphorous based compounds tend to have a surface exchange reaction when exposed solely to arsine at typical growth temperatures. Therefore, to prevent this intermixing the sample was hidden from the column V flux for 15 s while the two gases were switched. This was accomplished in our system by rotating the sample the sample underneath the fixed part of the susceptor, effectively hiding it from exposure. No detrimental effect to the surface morphology was detected upon examination of QWs grown in this manner. PL at 25 K was performed on the samples to detect QW emission. Three QWs with 5.65, 20 and 31 A were grown, based on thicknesses produced by ALE of 2, 7 and II GaAs monolayers, respectively. Figure 3 shows the 4K PL of the 5.65 A quantum well. The full width at half maximum (FWHM) is around 30 meV. The FWHMs measured here are comparable to those of gas source molecular beam epitaxial material"'. 104 5x10 5 5X10 5 _ 65-90 6 4x10 5 4x 10 O' 3x1066 C 3x10 z., •- - 0 2xl0 5.6-eA Q z 2x10 6 106 FWHM 5x 105 30 me Ox 100 5000 Figure 3 6000 7000 WAVELENGTH (A) 8000 9000 4K PL Spectrum of 5.65 A GaAs Quantum Well with 350 A GaInP Barriers. ALE GROWTH OF PLANAR DOPED FET Atomic Layer Epitaxy (ALE) was used for growth of the 5-doped device structures. ALE offers an attractive approach for the synthesis of 5-doped structures, since it is a lowtemperature growth process which minimizes dopant diffusion. ALE also allows an accurate control of epilayer thickness between the dopant plane and the gate over a large area wafer, resulting in uniform device performance. TMG and AsH 3 have been used for the growth of GaAs. Hydrogen selenide (H2Se) was used as the n-type dopant source. A cross-section of the device is shown in Figure 4. The growth starts with an undoped GaAs buffer layer grown at 600°C. The GaAs buffer layer is necessary for good device pinch-off and low output conductance. The ALE grown GaAs buffer is n-type with a background carrier concentration in the low l0'3 /cm 3 to the high 1014 /cm 3, and a room temperature mobility of 4200 cm 2/V.sec. Following this layer, the substrate temperature was lowered to 480'C to optimize the incorporation of selenium into GaAs. At temperatures much above or below 4800C the fraction of selenium atoms which are electrically active tends to drop resulting in low free carrier concentrations. The H2Se (30ppm) was introduced during the arsenic exposure cycle of the growth with minimum arsine flow. The wafer was kept under H2Se for one minute with 20 sccm of H2Se flowing. This procedure resulted in a doping profile with a peak carrier concentration of about 6 x 101S/cm 3 and a full width at half maximum (FWHM) of about 40 A as obtained by C-V measurement. The undoped top GaAs layer was grown at 6001C to achieve a better quality undoped film and thus improve the breakdown voltage characteristics. The final layer was the contacting layer which resulted in uniform and reproducible nonalloyed low resistance ohmic contacts. This contacting layer consists of 10 sheets of Se planar doping separated from each other by 50 A of undoped GaAs. The flow of H2Se was set to its 105 GD S n++ _n-+ 7GaAs Top !Lay~er ( i!)600 A Contact Layer 600 °C GaAs Channel (i) "GaAsChannel ( i) 20 A 480 °C 20 A.48 oC GaAs Buffer Layer 2000 A Planar Doped Channel 600 °C GaAs Sub. (S.1) Figure 4 Cross Section of GaAs 5-doped MESFET. maximum of 170 sccm and the substrate was held under H2Se for 30 sec at 4800C after being3 exposed to TMG. The resulting peak carrier concentration in each plane was 2 x 10 19/cm as obtained by capacitance-voltage profiling. The lowest specific contact resistance achieved by this structure was 7 x 10- 7 i- cm 2 obtained from a transmission line measurement (TLM). The device fabrication used a standard FET process starting with a mesa etch, followed by source and drain metallization consisting of AuGe (800 A)/Ni(200 A)/Au (2000 A), and finally a gate recess and gate metallization of Ti(500 A)/Au(1500 A). The gate length and gate width of the device were 1.5 Am and 75 Am, respectively. The I-V characteristics of this device are shown in Figure 5. A maximum transconductance of 144 mS/mm was obtained at a current density of 460 mA/mm. The extremely high gate-drain breakdown voltage exceeded 25V. This is one of the highest values reported for a gate to drain spacing of 1.6 Am in spite of the heavily doped n' region between gate and drain161. The I-V characteristics show excellent pinch-off with a very small output conductance which is the result of good quality GaAs buffer layers. The transconductance showed a broad maximum range as a function of gate-to-source voltage, typical of delta-doped FET structures. pnp HETEROJUNCTION BIPOLAR COLLECTOR AND EMITTER TRANSISTOR WITH CARBON-DOPED Carbon acts as a p-type dopant that exhibits very low diffusivity and a very high solubility limit in GaAs. These qualities make carbon doping attractive for the base of npn HBT's as well as emitter and collector of pnp HBT's. ALE has been used for the growth of AlGaAs/GaAs pnp HBT's, with carbon as the p-type dopant. Carbon doping was controlled by adjusting the growth conditions such as growth temperature, group III and group V fluxes 106 VGS (Volts) 35.0 .6 a) 0 :3 0 C 3.5 /div -1.2 C -1.8 -2.4 .00 " .00 Figure 5 -3.0 4.0 _ VrS 0.4/div Source Drain Voltage (V) Drain Current Characteristics of a 1.5 x 75 jm 2 GaAs 5-doped MESFET. and exposure times. P-type carrier concentrations in the range 101-10 2°/cm3 were achieved for the ALE grown GaAs. However for AlGaAs, due the strong Al-C bond, the lowest background p-type doping was in the high 1016/cm 3 range. Table I shows the HBT structure with the corresponding carrier concentrations and thicknesses. SiH4 was used as the n-type dopant for GaAs. Wet chemical etching was used to form the base and the emitter mesas, as well as to provide for device isolation. Ti/Pt/Au was used to contact the emitter and collector. AuGe/Ni/Au was used for the base contact. A common emitter current gain over 100 was obtained with excellent I-V characteristics. Early voltage is a relatively high 30V. Details of the HBT results will be published elsewherer'1 . Doping Thickness LAYER (cm-3) (Am) 0.15 0.075 p-GaAs 2 x 1018 1 x 101T 8 x 1017 5 x 1018 p+-GaAs 1 x 1018 0.5 GaAs Buffer Undoped 0.1 S.I. SUBSTRATE Undoped 500 _ p+-GaAs p-A1. 3Ga.7 As n+-GaAs Table 1 0.12 0.4 pnp AIGaAs/GaAs HBT Structure Grown by ALE. 107 CONCLUSION The ALE growth of III-V compounds has suffered from several shortcomings such as a limited range of growth temperatures, high carbon background and difficulties in the synthesis of ternary alloys. We have demonstrated that these problems can be minimized by special reactor and susceptor designs that allow the substrate to rotate between streams of the precursor gases. Such an approach was used to grow ternary alloys such as AlGaAs, InGaP and InGaAs, and to deposit ordered ternary alloys. ALE was also used for the deposition of highly doped structures for nonalloyed contacts, planar-doped FET's, and pnp HBT's. We believe that ALE is now capable of providing material quality comparable with other techniques, while offering control of the deposition process on the monolayer level. ACKNOWLEDGMENTS I would like to thank, J.R. Hauser, N. El-Masry, B. McDermott, Kim Reid, A. Dip, P. Colter, S. Huessin, M. Hashemi, T. Henderson and B. Bayraktaroglu for achieving these new ALE results. This work is supported by ONR/SDIO, SERI and NSF. REFERENCES 1. K. Mochizuki, M. Ozeki, K. Kodama and N. Ohtsuka, J. Crystal Growth, 93, 557-561 (1988). 2. P. C. Colter, S. A. Hussien, A. Dip, W. M. Duncan and S. M. Bedair, Appl. Phys. Lett. (accepted). 3. B. T. McDermott, N. A. El-Masry, B. L. Jiang, F. Hyuga, W. M. Duncan and S. M. Bedair, J. Crystal Growth, 107, 96-101 (1991). 4. B. T. McDermott, K. G. Reid, N. A. El-Masry, S. M. Bedair, W. M. Duncan, X. Yin and Fred Pollak, Appl. Phys. Lett. 56, 1172-1174 (1990). 5. M. J. Hafich, J. H. Quingley, R. E. Owens, G. Y. Robinson, D. Li and N. Otsuka, Appl. Phys. Letter L4, 2686-2688 (1989). 6. M. Hashemi, B. McDermott, U. R. Mishra, J. Ramdani, A. Morris, J. R. Hauser and S. M. Bedair, to be published in Elect. Dev. Lett. 7. T. Henderson, B. Bayraktavoglu, S. Hussien, A. Dip, P. Colter and S. M. Bedair, Electronic Lett. 27, 692 (1991).