Etch rates of ≤ 0.25 µm min −1 are obtained for both GaSb and InSb in electron cyclotron resonance CH 4 /H 2 /Ar discharges. The GaSb surfaces remain smooth and stoichiometric over a wide range of plasma conditions (microwave power 400-1000 W, pressure 1.5-10 mTorr, CH 4 to H 2 ratio 0.1-1.5). However, we were unable to prevent preferential loss of Sb from InSb causing rough, non-stoichiometric surfaces under all conditions investigated. It appears that low-profile electromagnet ECR sources are not good choices for CH 4 /H 2 -based dry etching of InSb.

CHLORINE-BASED PLASMA ETCHING OF GaN l*, ~a *

We report results of a study on the inductively coupled plasma (ICP) etching of GaAs, AlGaAs, GaSb, and GaP in pure Ar, CH 4 /H 2 /Ar and CH 4 /H 2 /N 2 plasma chemistries. Etch rates of the semiconductors initially increase with ICP source power, reaching maxima around 1500-2000 Å min −1 with ∼500-700 W ICP power, and then decrease with further increase of ICP power because of the decrease in cathode dc self-bias. Etch rates increase at fixed ICP power with rf chuck power in the range of 100 W to 450 W, while they showed little dependence on the chamber pressure (2 mTorr-20 mTorr) in CH 4 /H 2 /Ar discharges. We found that the dc self-bias on the rf chuck decreased exponentially as ICP power increased, while it increased with rf chuck power and pressure in these plasma chemistries. A simple calculation of ion flux and etch yield based on dc self-bias of the plasmas on the sample chuck was used to measure typical values of etch yield. In both ICP CH 4 /H 2 /Ar and CH 4 /H 2 /N 2 discharges these were close to the pure Ar sputter yield.

GaAs MESFETs may be exposed to H 2 -containing plasmas during various etch and deposition steps. We have found that both inductively coupled plasma (ICP) and electron cyclotron resonance (ECR) H 2 plasmas create severe reductions in MESFET mutual transconductance (g m ) and reverse breakdown voltage (V RB ) through reductions in channel layer doping and surface stoichiometry changes. While changes in channel sheet resistance and diode ideality factor may be minimized by limiting the plasma exposure time, g m and V RB are still reduced by up to a factor of two even for 30 s exposures. The results show that there are no conditions under which there are not substantial changes in device performance, and unless H 2 -free plasma chemistries are used, post-plasma annealing will always be necessary to restore the device characteristics.

Etch rates above 1 µm min −1 have been obtained for BCl 3 /Ar and BCl 3 /N 2 plasma etching of In 0.5 Ga 0.5 P, Al 0.5 Ga 0.5 P and Al 0.5 In 0.5 P under electron cyclotron resonance (ECR) conditions. The etched surface morphologies are a strong function of the ratio of chlorine radicals to total ion density in the discharges, and are smoother in BCl 3 /Ar. Non-optimized etching conditions produce rough surfaces characterized by the presence of chlorine and boron-containing residues and substantial post-oxidation. We believe that the high ion flux available under ECR conditions provides efficient sputtering of the normally involatile InCl 3 , and thus a selvedge layer of this material is prevented from forming. Therefore the etch rates are much higher and the surface morphologies smoother in ECR discharges. The etch rates of Al x Ga 1−x P are much lower than the other two ternary compounds and are independent of composition over the range x = 0.05-0.70 for fixed plasma conditions.