CH31 vapor etching of masked and patterned GaAs

CH3I vapor etching of masked and patterned GaAs substrates has been experimentally investigated. For GaAs samples masked with silicon nitride stripes that are wider than 30 tzm, the etch depth increased compared to unmasked samples, the magnitude of which increased with increasing mask width. Etching of bulk substrates of (l l l)Ga and (lll)As GaAs revealed a dependence of etch rate on crystal orientation, with (l l l)Ga > (100)GaAs > (lll)As. Increasing etch temperature reduced the orientation dependence of etch rates. Orientation dependence of etch rates was also observed on non-planar GaAs substrates patterned to expose different orientations on wet-etched groove structures. In this case, etch rate differences between the different orientations were amplified when compared to the bulk substrate results. Finally, it was found that the extent of mask undercutting depended on the direction of mask stripes in a fashion consistent with the orientation reactivity results. Mask stripes on (100)GaAs oriented in the [011] direction were severely undercut whereas stripes oriented in the [011] direction were undercut less. 1. I n t r o d u c t i o n In-situ vapor etching of GaAs can be an effective pre-cleaning treatment for organometallic vapor phase epitaxy (OMVPE). Previously, we demonstrated CH3I vapor etching of (100)GaAs substrates in an OMVPE reactor. Specular surface morphologies were obtained at etching temperatures several hundred degrees lower than were required to obtain specular surfaces with any other vapor etchant [1]. GaAs OMVPE regrowth on CH3I vapor etched GaAs epilayers * Corresponding author. 1 Present address: Department of Chemical Engineering, Tufts University, Medford, Massachusetts 02155, USA. resulted in superior electrical characteristics at the growth interface compared to conventional wet-chemical substrate cleaning [2]. In addition, the etch rate of (100)A1GaAs epilayers was equal to that of GaAs [1]. This result is in contrast to HC1 vapor etching for which the etch rate decreases with increasing AlAs mole fraction [3,4]. Thus, CH3I vapor etching is a promising in-situ process for OMVPE. In this paper, we investigate the application of CH3I vapor etching for patterning GaAs substrates. Development of such an in-situ etch process would be especially advantageous for fabrication of buried heterostructure diode lasers, as previously proposed [3], since ambient exposure of AlGaAs, which leads to oxidation, could be avoided. However, vapor etching of masked sub0022-0248/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved SSDI 0022-0248(95)00062-3 82 C. l~ Krueger et al. /Journal of Crystal Growth 153 (1995) 81-89 strates can be affected by diffusion of etchant species between the mask and substrate surfaces. Both surface diffusion and vapor phase ("volume") diffusion processes could lead to nonuniform etching near masked regions. In addition, an important parameter is crystal orientation, since sidewalls will evolve at the mask/substrate interface as etching proceeds. These sidewalls may possess different crystallographic orientations than the bulk substrate, which may have different reactivities toward the etch. An understanding of these effects is important for evaluating CH3I vapor etching of masked or patterned GaAs. Several aspects of CH3I vapor etching of masked and patterned GaAs substrates are studied. Nonuniform etch profiles are observed on GaAs substrates masked with stripes of silicon nitride. The etch rate is enhanced near the mask/substrate surface boundary, consistent with a model of vapor phase volume diffusion from the region above an inert mask to the reactive GaAs surface. The etch rate of bulk GaAs substrates is found to depend on surface crystallographic orientation. These surface reactivity differences are also observed on nonplanar substrates which are patterned to expose micrometer-sized facets of different orientations on the bulk substrate. In this case, the etch rate differences between the different orientations are amplified when compared to the bulk substrate results. 2. Experimental procedure The substrates used in these experiments were GaAs wafers which were either masked with silicon nitride or patterned with wet-etched groove patterns. To examine the effect of masks on etch uniformity, (100)GaAs substrates misoriented 2 ° to the (110) were covered with silicon nitride stripes of various widths. Silicon nitride, 100 nm thick, was deposited on GaAs substrates by plasma-enhanced chemical vapor deposition (PECVD) from Sill 4 and NH 3. The stripe pattern was then transferred to the nitride by conventional photolithography procedures. Two patterns were used for these investigations. The first had sets of 9 parallel stripes of increasing width centered over a 200 ~m period. The stripes were oriented parallel to the [011] direction, which is shown to minimize mask undercutting effects. The widths of the stripes were 5, 10, 20, 30, 40, 50, 60, 80, and 100/~m. The second pattern had stripes of the same widths but each stripe width was repeated three times before the next width and the stripes were centered over a 250 p.m period. After vapor etching, the etch profiles were measured by a mechanical surface profilometer after removing the remaining silicon nitride in a hydrofluoric acid solution. Bulk etch rates of (111)Ga and ( l l l )A s substrates are measured on half-masked wafers. The mask was 100 nm PECVD silicon nitride which was then removed from half of the wafer using standard photolithographic procedures. Following vapor etching, the etched depth was measured with a surface profilometer after removing the remaining silicon nitride in a hydrofluoric acid solution. Reported etch rates are determined by dividing the etched depth by the etch time. For studying the effect of the crystal orientation on vapor etching on non-planar surfaces, GaAs substrates, (100) misoriented 2 ° to the (110), were etched to expose either ( l l l ) G a or (111)As surfaces by using an established wet-chemical etching procedure [5]. GaAs substrates were patterned with stripes of photoresist, 400 ~m wide separated by 5/~m openings, oriented in the [011] or [011] direction. Grooves were formed by etching the samples in a solution of 5 parts by volume H 2 S O 4 (97 wt%), 1 part H 2 0 , and 1 part H 2 0 2 (30%). "V" grooves were etched with stripes oriented in the [011] direction and "dovetail" grooves for stripes in the [011] direction, as shown in Fig. 1. Although the initial opening in the photoresist was only 5 /~m in width, the sample was etched under the edge of the mask and resulted in 20 /~m openings after 3 min of etching. Additionally, ( l l l ) A s GaAs substrates were patterned with photoresist stripes oriented in the [110] direction. Samples etched in the solution described above resulted in "lambda" grooves [5], shown in Fig. lc. This structure exposes two additional surC.W. Krueger et al. /Journal of Crystal Growth 153 (1995) 81-89 83 (a) "V" GROOVE (b) "DOVETAIL" GROOVE (111) Ga (100) (111) As x 'L j / ( I ° ° )