Recombination-current suppression in GaAs p-n junctions grown on AlGaAs buffer layers by molecular-beam epitaxy

n+pp+GaAs and n+pP+ GaAs/GaAs/Al0.3Ga0.7As mesa diodes have been fabricated from films grown by molecular‐beam epitaxy. The diodes made from films employing an AlGaAs buffer layer show marked improvements (a factor of 5 reduction) in recombination current densities. Deep level transient spectroscopy measurements moreover indicate that deep level concentrations are reduced by the AlGaAs buffer.

n+pp+GaAs and n+pP + GaAs/GaAs/A1o.3 Gao. 7 As mesa diodes have been fabricated from tums grown by molecular-beam epitaxy. The diodes made from films employing an AIGaAs buffer layer show marked improvements (a factor of 5 reduction) in recombination current densities. Deep level transient spectroscopy measurements moreover indicate that deep level concentrations are reduced by the AIGaAs buffer.
Suppression of recombination currents is an important factor in the design of high-performance AIGaAs/GaAs solar cells and bipolar transistors. [,2 Nonradiative recombination is typically controlled by deep levels introduced by impurities and/or defects in the semiconductor bulk and surfaces. In this communication we demonstrate that relatively thick OaAs films grown by molecular-beam epitaxy (MBE) atop a O.69-,um Al o . 3 0a n . 7 As buffer layer show a substantially lower deep level concentration than do films grown atop a GaAs buffer layer. p-n junctions fabricated in such films display a corresponding reduction in recombination current. Gale et al. 3 have previously employed AIGaAs buffer layers in the fabrication of GaAs solar cells by metalorganic chemical vapor deposition (MOCVD). These cells showed higher open-circuit voltages and conversion efficiencies than cells made with GaAs buffers. This improvement was attributed to the greater minority-carrier confinement achieved by the GaAsl AIGaAs potential barrier. Our work indicates that reduced recombination current, due to impurity reduction, may also have contributed to the improved solar-cell characteristics. Similar results have been obtai.ned by Beneking et al.,4,5 who observed reduced impurity concentrations for GaAs films grown by MOCVD and liquidphase epitaxy on top of an indium-doped strained layer several micrometers thick. Our report of corresponding benefits for MBE-grown films usi.ng relatively thin AIGaAs layers suggests that the technique is suitable for the routine growth of high-quality MBE-grown AIGaAs/GaAs films for bipolar applications.
The MBE films used in this work were grown in a Perkin-EImer 400 MBE system. The starting substrates were Zn-doped (1.5 X 10 19 em -3) (100) horizontal Bridgeman material with an etch pit density ofless than 500 cm-2 , The GaAs layers were grown at a substrate temperature of 605 "C and the AIGaAs layer at a substrate temperature 625°C. There were a total of five films grown for this work which we have labeled F I-F 5. The first three samples, F 1, F2, and F3, were grown on three consecutive days and the substrates were cleaved from the same wafer. There was a total of 28 !till of material grown in the MBE system prior to the growth of samples F I-F 3. The first film, F 1, had the Bedoped p-type base grown directly on the p+ substrate. The second film, F 2, had a 0.44-,um p + buffer layer below the p.
type base while the third film, F3, had a O.9f.tmp+ buffer layer. (Film structures F I-F 3 are shown in Fig. 1.) Ohmic contacts were made to the n + emitters by alloying Au-Ge, and mesa diodes were defined by photolithography and sub· sequent wet etching in HZS04:H202:H20 (1:8:40). The areas of the diodes ranged from 4X 10-4 cm 2 to 1.6x 10-3 cm 2 • The MBE system was opened to repair a broken weld on its Ga oven before growing the second set of films F 4 and F 5, Films F 4 and F 5 were grown on two consecutive days with substrates cleaved from a second wafer. There was a total of 12,urn of material grown in the MBE system prior to the growth of samples F 4 and F 5. (Because of the shorter burn in time of the ovens, one would suspect samplesF 4 and F 5 to be of inferior quality when compared to samples F 1 through F 3.) The first sample grown, F 4, had a heterojunc-tionpP + barrier while the second sample grown, F5, had an isotype pp + barrier as shown in Fig. 2.
The mesa diodes for all five films were characterized by dark current-voltage (1-V) measurements and by deep level transient spectroscopy (DL TS), The dark 1-V characteristics were measured using a Hewlett Packard 4145A semiconductor parameter analyzer. The n = 1 and n = 2 saturation current densities were extracted from the measured J-V characteristics by curve fitting. The n = 2 current component is of particular interest because it is directly related to both recombination in the space-charge region and to surface recombination around the junction perimeter. 6 ,7 Thus, the n = 2 current can provide an indication of the MBE film quality. Table I shows the average n = 2 saturation current densities J 02 for tUrns F 1-F 5. The first three isotype barrier films F I-F 3 have J 02 values that are five to six times lower than the J 02 value for film F 5. On the other hand, J D2 for the heterojunction barrier film (F 4) falls within the range ofJ 02 values for filmsF I-F 3. One would expect the impurity concentrations in film F 4 to be greater than those in film F 5 because film F 5 was grown on the day following the growth offilmF 4. However, F 4's lower J 02 value indicates that film F 4 is actually of significantly higher quality thanF 5. It is our conclusion that the introduction of an AIGaAs buffer layer has caused the film quality of F 4 to be comparable to that of F I-F 3.
The MBE film quality was further investigated by means ofDLTS. Our DLTS measurements have shown that deep level concentrations are greatly reduced when an AIGaAs buffer layer is used. We suspect that the AiGaAs layer traps impurities that are floating up from the substrate, andlor getters impurities from the subsequent GaAs layers during film growth. This is in agreement with McAfee et al.,!l who observed a highly peaked spatial profile of deep levels located in a 140-A-wide region at the GaAsl Al GaAs interface in MBE double heterostructure lasers. Photoluminescence 9 • 10 and capacitance-voltage ll studies of GaAs quantum weBs have also indicated the presence of impurities at the GaAsl AIGaAs interface.
Typical D LTS spectra for films F I-F 5 are compared in Fig. 3. Figure 3(a), representative of films F I-F3, shows a prominent DLTS peal( at about 320 K, a second peak at about 430 K, and lesser peaks at 360 K and between 100 and 200 K. Figure 3 (b) is typical of film F 4. Again there is a prominent peak at 320 K and lesser peaks to either side. Finally, Fig. 3 (c) shows a representative DLTS scan for film F 5. We reiterate that F 5 was grown after F 4 and is therefore expected to contain a lesser number of impurities. However, in addition to the ever present peak at 320 K, F 5 has signifi- With the exception of the deep level associated with the peak at 320 K, the AIGaAs buffer layer appears to have blocked andlor gettered impurities from the layers grown above it.
The DLTS peak at 320 K requires additional discussion, since it appears in all of the samples and seems unaffected by the AIGaAs layer. The average deep level concentration NT associated with the 320 K peak for films F I-F3 was NT = 4.4 X 1013 cm-3 .ForfilmF4theaveragetrapconcentration was NT = 1.0 X 10 14 em -3; for film F 5 it was NT = 7.8 X 1013 cm-3 • The results suggest that the concentration of this particular deep level is correlated with the amount of MBE material grown prior to each film growth. This observation in turn would imply that the deep level is related to impurities introduced by the system during film growth. Failure of these impurities to be gettered by the AIGaAs layer may indicate that they occupy substitutional lattice sites, or form vacancy complexes that do not readily diffuse to the AIGaAs layer during film growth. They appear to be continually incorporated into the film during the growth process along with the Ga, As, and dopant atoms.
The thermal activation energy for the 320 K peak was measured by DLTS to be E1' -Ev = 0.55 eV. The capture cross section for the majority-carrier holes was found to be up = 6.8X 10-16 cm 2 • The impurity could therefore be iron, 12.13 or possibly a Ga vacancy 14 or vacancy complex. We suspect that this particular deep level does not control the J 02 current. This is suggested by a comparison of the carrier Rancour et al.
lifetimes as determined from DLTS and dark I· V data, For the deep level concentrations encountered here, the measured up would imply hole lifetimes, on the order of 0, I-LO,us, However, the average recombination lifetime .J'i n 7'p' as determined from the measured n = 2 saturation current density (J oz ) was on the order of LO ns, Consequently, 'in would have to be -La ps and Un on the order of 6,8 X 10-10 cm 2 , A capture cross section on the Ofder of 10-9 _10-10 cm 2 is unusually large, The concentration of the O,55-eV level does not correlate with the recombination current density, This further confirms our suspicions that the O's5-eV level does not dominate the recombination current, However, the controlling recombination rate, whatever the mechanism, does appear to have been reduced by the presence of the AIGaAs layer. It is possible that a deep level, as yet undetected by our DLTS measurements, is controlling the n = 2 current, Another factor to consider is surface recombination around the device perimeter, which may make a significant contribution to J 02 ; gross defects (e,g" oval defects) are another possible contributor, Further work is needed to establish the controlling recombination mechanism, In summary, we have shown that p-n junctions, fabricated from MBE GaAs films, show marked reductions in recombination current when an AIGaAs buffer layer is employed, Deep level concentrations are also reduced by the presence of an AIGaAs buffer layer, One deep level, possibly iron or a Ga vacancy/vacancy complex, remains unaffected by the AIGaAs buffer layer, This impurity/defect is either rigidly incorporated into the lattice during MBE film growth, or is highly soluble in AIGaAs, Fortunately, the cited deep level does not appear to significantly affect the n = 2 current density, The use of AIGaAs buffers should be independent of growth technology and could provide significant improvements in AIGaAs/GaAs films grown for bipolar applications, Shallow p+ diffusion into InGaAsP (it = 1.3 pm) has been improved by employing a new spin-on source based on Zn-doped alumina, Thereby the thermal expansion coefficients of diffusion source and semiconductor are better matched together than in case of the more common Zn-doped silica films, Consequently, besides an excellent mechanical stability of the spin-on films over a wide temperature range, the influence of mechanical stress on the diffusion process is effectively reduced, Applying diffusion temperatures around 600 °C surface hole concentrations above 6 X 10 19 cm -3 and extremely low specific p-contact resistances of 2-3 X 10.-6 n cm 2 have been achieved, Presently, optoelectronic devices in the InGaAsP-InP material system are widely used in optical communication system in the wavelength ranges around 1,3 and 1.55 pm, where dispersion and attenuation of optical fibers exhibit their minimum values, For these devices, in particular for light emitting diodes and laser diodes, low resistive contacts are required in order to minimize heat generation and to enable high-speed operation, In case of laser diodes, the maximum temperature for cw operation is decisively influenced by the contact resistance on account of the superlin-