A Comparison of Dry Plasma and Wet Chemical Etching of GaSb Photodiodes

Abstract

We report on the performance of GaSb pn junction photodiodes fabricated using electron cyclotron resonance plasma etching using Cl 2 /Ar recipe, a mixed gas recipe consisting of Cl 2 /BCl 3 /CH 4 /Ar/H 2 and wet chemical etching. Diodes fabricated using Cl 2 /BCl 3 /CH 4 /Ar/H 2 recipe show an order of magnitude lower leakage current density and lower ideality factor. The highest value of the zero bias dynamic resistance-area product was obtained for Cl 2 /BCl 3 /CH 4 /Ar/H 2 etched diodes and was equal to 830 cm 2 as compared to 300 cm 2 for Cl 2 /Ar and 330 cm 2 for wet etching. Spectral responsivity of Cl 2 /BCl 3 /CH 4 /Ar/H 2 etched diodes was observed to be three times that of Cl 2 /Ar and wet etched diodes. Overall, the diodes etched using the recently reported Cl 2 /BCl 3 /CH 4 /Ar/H 2 recipe provided the best optical and electrical characteristics. Recently, there is increasing interest in the gallium antimonide GaSb based compound semiconductors due to its wide range of optoelectronic applications in the mid-infrared MIR wavelengths. 1 GaSb is an attractive choice as a substrate material because its lattice parameter matches various ternary and quaternary III-V compound semiconductors whose band gaps cover a wide spectral range from 0.3 to 1.58 eV, i.e., 0.8-4.3 m. High quantum efficiency photodetectors, 2,3 photovoltaic cells, 4,5 and laser diodes with low threshold current 6,7 have been demonstrated using GaSb and related compounds. Fabrication of these devices requires etching to form mesa or line structures. Despite the damage associated with dry etching, it can provide highly anisotropic profiles with good repro-ducibility and uniformity, and therefore is desirable compared to wet chemical etching. In spite of the technological importance of GaSb based alloys, little has been reported on the dry etching of GaSb and related materials. A common choice of plasma chemistries for dry etching of GaSb-based structures consists of chlorine based precursors, such as SiCl 4 , BCl 3 , or Cl 2 because the volatilities of the gallium and an-timony chlorides are generally very high. 8 Also, high-density plasma sources, such as electron cyclotron resonance ECR, are preferred because of the increased density of low energy ions in the plasma. 9 Pearton et al. 10,11 have reported on the etching of GaSb in high density plasma using Cl 2 /Ar and BCl 3 /Ar chemistries down to 30°C. Improvement in the anisotropy was observed at lower temperatures but at the cost of a decrease in the etch rates. They also observed that the surface morphology obtained by BCl 3 /Ar etching was smoother than that of Cl 2 /Ar. 10 Dry etching of GaSb using methane/hydrogen (CH 4 /H 2) and ethane/hydrogen (C 2 H 6 /H 2) chemistry has also been reported. 12,13 C 2 H 6-based plasmas show 50% higher etching rates as compared to the CH 4-based plasmas. 13 Addition of a polymer forming gas such as CH 4 results in the formation of a polymeric thin film on the sidewalls, which minimizes the undercutting and provides passivation. 14,15 Although mixed chlorine and methane etches have been reported for various III-V compound semiconductors, systematic studies of antimonide are very few. 16 Langer et al. have recently reported a mixed-gas etching recipe consisting of chlorine, boron trichloride, methane, hydrogen, and argon (Cl 2 /BCl 3 /CH 4 /Ar/H 2) which shows high etching rate of 0.5 m/min and results into smooth surfaces and sharp sidewalls at room temperature. 17 The effect of etching conditions on the device performance of GaSb and related materials has not been well addressed in the literature. In this paper, we have compared the performance of pn junction photodiodes fabricated using a Cl 2 /Ar recipe roughly based on the results reported by Pearton et al. 10,18 a mixed gas recipe consisting of Cl 2 /BCl 3 /CH 4 /Ar/H 2 , 17 and wet chemical etching using a NaK tartrate based recipe. Experimental The substrates used in this study were 100 n-type GaSb Te doped obtained from Galaxy Compound Semiconductor Inc. The carrier concentration was 5 10 17 cm 3. The substrates were de-greased with hot xylene followed by acetone and methanol rinse XAM cleaning. Then the samples were etched in hydrochloric acid HCl to remove the native oxide layer. Zinc Zn acts as p-type impurity in GaSb. Zn diffusion was carried out at 500°C for 5 h using the leaky box technique. 19 Solid Zn pellets were used as the source. The samples were next subjected to backside etching using a 2% solution of bromine in methanol for 30 s. Back side contacts were formed by electron beam E-beam evaporation of 200 Å of tin and 1000 Å of gold; followed by rapid thermal annealing RTA at 350°C for 5 s. The front side metal contact consisted of 400 Å titanium followed by 800 Å gold evaporated using E-beam. The final step in the fabrication process was the mesa etching. The front metal contacts were protected using a photoresist and the mesa areas were pat-terned. The photoresist was cured by baking at 110°C for 10 min. A Plasmatherm Electron Cyclotron Resonance 357 system with a load-locked chamber was used for the dry etching of GaSb. One sample set was etched using Cl 2 /Ar plasma. The gases were in the ratio of 1:6 at a pressure of 1.5 mTorr. 100 W rf power and 300 W plasma power was used and the temperature was maintained at 30°C. The etch rate obtained with this process was 135 nm/min. A Cl 2 /BCl 3 /CH 4 /Ar/H 2 gas mixture in the ratio 2:1:2:6:12 was used to etch the second set of samples. A chamber pressure of 1.3 mTorr, rf power of 150 W, and plasma power of 400 W was used. The etching was carried out at room temperature. This recipe gave an etch rate of about 560 nm/min. The details of the surface morphology after etching are presented elsewhere. 17 The third sample set was prepared by wet etching. For the wet etching of GaSb, a mixture of HCl:H 2 O 2 :NaK tartrate 66 ml:18 ml:24 g in 1 liter of solution was used. The etching rate with this solution was 120 nm/min. After the dry/wet-etching step, the photoresist was removed by using ac-etone and the samples were cleaned by XAM cleaning. Thereafter, the samples were rinsed in pure methanol to avoid the formation of oxide on the sidewalls, which increases the leakage current associated with the diodes. No surface passivation was carried out and no z