Effects of Na 2 S and ( NH 4 ) 2 S edge passivation treatments on the dark current-voltage characteristics of GaAs pn diodes

We have investigated the dark current‐voltage characteristics of GaAs pn homojunctions whose surfaces have been passivated with Na2S and (NH4)2S chemical treatments. Reductions in 2kT perimeter recombination currents by a factor of 3.2 were obtained for the two treatments. A shunt leakage, observed at low forward bias for the Na2S treated devices, is virtually eliminated with the (NH4)2S treatment. It is also shown that even the high quality, large area (0.25 cm2) pn diodes used in this study are dominated by 2kT edge currents before passivation.


Effects of Na2S and {NH
We have investigated the dark current-voltage characteristics of GaAs pn homojunctions whose surfaces have been passivated with NuzS and (NH4)2S chemical treatments. Reductions in 2kTpcrimeter recombination currents by a factor of 3.2 were obtained for the two treatments. A shunt leakage, observed at low forward bias for the NazS treated devices, is virtually eliminated with the {NH 4 hS treatment. It is also shown that even the high quality, large area (0.25 cm 1 ) pn diodes used in this study are dominated by 2kTedge currents before passivation.
Due to a large density of surface states the performance of GaAs devices, such as solar cells and heterojunction bipolar transistors (HBT's), can be dominated by recombination at exposed mesa edges. This recombination at the exposed perimeter gives rise to an additi.onal component of forward biased current for a GaAs pn diode. The total current in a forward-biased GaAs pn junction can be represented as (1) The saturation current densities J tll and '/"nB are associated with carrier recombination in the quasi-neutral and bulk space-charge regions, respectively, and A is the area of the pn junction. The current density J 02P is associated with recombination at the exposed mesa edge and P is the perimeter of the pn junction. Even for the large square (A = O.25cm 2 ) GaAs diodes reported in this letter, the total perimeter recombination current is much larger than the bulk spacecharge recombination current. I The elimination or lowering of surface recombination can therefore significantly reduce the 2kTcurrent of a GaAs diode. This reduction in 2kT current would result in an in~ crease in gain at a given collector current for a HBT. Lower~ ing of the dark current due to reduction in 2kT current of a solar cell will increase the one~sun fill factor and hence increase the efficiency of the cell. I The lowering of the 2kT perimeter recombination current in pn GaAs di.odes also allows one to observe the 2kT bulk recombination current; thereby studies could be performed to correlate bulk 2kT recombination. current with defects and impurities, and the possibility of further reduction in 2k T current. A reduction in perimeter surface states may also reduce edge generation and increase the storage time of dynamic GaAs memori.es. 2,3 Recently, photochemicaI 4 and chemica15-~ treatments have been shown to be effective in lowering the surface state density of III-V semiconductors. This reduction of the surface state density unpins the Fermi level at the surface and also reduces the nonradiative recombination at the surface. 6 However, these treatments are far from ideal. They last for only a short time in room air (anywhere from 20 min for the photochemical treatment to 18-48 h for a Na2S chemical treatment). Another problem with the Na 2 S chemical treat-ment is the introduction of a surface conduction. For a GaAs solar cell or emitter base junction ofa HBT, the surface con~ duction is manifested as a shunt leakage at low forward bias.
While the nature of the semiconductor surface after treatment is not known, it appears that the common link between various chemical treatments is sulfur-containing compounds. The most widely reported compound is Na 2 S. It is not known if the other elements ofthe variolls sulfide compounds contribute to the nature of the interface or just provide a vehicle to present sulfur to the surface of the semiconductor. Nottenburg et al. 8 have investigated the effects of the NazS treatment on the dark current-voltage (1-V) characteristics of AIGaAs/GaAs pn heterojunctions. In. this letter we report the effects of Na 2 S and (NH 4 ) 2S treatments on the dark 1-V characteristics of GaAs pn homojunctions.
The diodes used in this study were grown by metalorganic chemical vapor deposition (MOCVD) in a commercial five-wafer reactor. The same reactor and growth procedure has recently produced pin heteroface solar cells which have the highest reported AM1.5 efficiency, indicating very high film quality. 9 The device structure and relevant device parameters are shown in Fig, 1. After film growth, metal patterns were defined using image reversal photolithography and lift-off. Then conventional photolithography and wet chemical etching were used to define mesas. Two size devices were fabricated, one a 0.5 em on a side square and the second a 0.5 cm by 310 ftm rectangle. The chemical treatments started by etching the devices for 10 s in 1:1:500 H2S04:H202:H20, fonowed by a rinse in de-ionized water. The Na 2 S treatment then consisted of soaking the wafer in aIM solution of the sulfide. The device was allowed to soak for up to 10 min to allow for chemical reaction. The devices were then spun dry at 5000 rpm for 60 s, which left a thin polycrystalline film of N a 2 S' 9H z O over the surface of the wafer. If insufficient time (approximately 7 min or less) was allowed for the chemical to react, no surface passivation took place and no reduction of edge currents was observed. Such devices, however, exhibited a hysteresis in their 1-V characteristics indicating a charge trapping at the improperly treated surface.
The (NH4)2S treatment consisted of making a saturated solution of (NH 4 hS from H 2 S and electronic grade NH 4 0H. After etching the sample for 10 s in the 1:1:500 H2S04:H202:H20 solution, the wafer was rinsed in de-ionized water and soaked in the (NH4)2S solution for up to 5 min. The wafer was then rinsed with de-ionized water and blown dry with N 2 • This treatment left no visible film on the wafer.
Displayed in Fig. 2 is a typical diode 1-V characteristic before treatment and after the above described Na 2 S and (NH.fhS chemical treatments. The Na 2 S treatment was removed by rinsing in de-ionized water 5 before performing the (NH 4 ) 2S chemical treatment. [1-V characteristics similar to those shown in Fig. 2 were obtained when the order of the Na 2 S and (NH4)2S chemical treatments was reversed. The (NH4)2S treatment can be removed by spinning on AZ1350J photoresist and then rinsing off the photoresist with acetone. J As is readily visible, the device initially exhibited no k T diffusion current but only a 2k T recombination current (series resistance effects are observed at high bias). After the Nu 2 S treatment, the device exhibits lower 2kTrecombination current and a shunt leakage at low forward bias. As also seen in Fig. 2, after the (NH 4 hS treatment the device exhibited a slightly better reduction in the 2kTrecombination current than after the Na 2 S treatment. In addition, after the (NH 4 ) 2S treatment the device has one order of magnitude less shunt leakage at low forward bias than after the Nu 2 S treatment.
We have also observed a much slower aging of our (NH 4 )zS treated devices than our NazS treated devices. The cause of the aging of the Nu 2 S treatment is probably the humidity of the room air since the Na 2 S treatment can be removed by rinsing with de-ionized water. s Since the (NH 4 hS passivation persists after the de-ionized water rinse, it is not surprising that the (NH 4 ) 2S treatment is less reactive with room air.
The expression for the pre-exponential factor for the recombination current is 102 = AJ02B + PJ 02P ' (2) where A and P are the area and perimeter of the device, respectively. Since I02 is the measured current, it will contain a bulk component and perimeter component. To quantify the reduction in the surface component, the bulk component must be known. However, to investigate extreme cases where surface or bulk currents are dominating, one needs only to examine the scaling of the current, If the devices are bulk dominated, the rati.o of currents for two different size devices will be equal to the ratio of their areas. If the devices are surface domina ted, then the ratio of their currents will be equal to the ratio of their perimeters.
The average observed initial 102 for our large area devices was 1.76 pA and for our small area devices was 0.931 pA. The ratio of these currents is 1.89. A comparison of this to the ratio of the perimeters ( 1.88) and areas ( 16) indicates that the devices are perimeter dominated. Since the initial 2kT current is an perimeter current, the perimeter recombination current density is J 02P = 8.80X 10-13 A/cm. For the treated devices, the average values of 102 were 1.02 and 0.325 pA for the large and small area devices, respectively. This gives a ratio of3.14 for the currents, indicating that the devices are neither perimeter nor bulk dominated. However, the surface component has been reduced to a value comparable with the bulk component. We have used Eq. (2) to estimate the 2kT recombination current densities after the chemical treatments obtaining J 02B = 1.85 X 10-12 A/cm 2 and J OlP = 2.79X 10-13 A/cm. The treatments therefore produced a reduction by a factor of 3.2 in perimeter current.
In conclusion, we have investigated the effects of NazS and (NH 4 ) 2S chemical passivation on the dark 1-V characteristics of GaAs pn homo junction diodes. Comparable reduction in 2k T edge recombination current by a factor of 3.2 was observed for NuzS and (NH 4 hS chemically passivated diodes. The shunt leakage observed at low forward biases for Na 2 S treated diodes was virtually eliminated with the (NH4)2S surface passivation. We have also demonstrated that even for large area pn GaAs homojunction diodes