Kinetic roughening of GaAs(001) during thermalCl2etching

Abstract

The surface morphology of Cl 2 -etched GaAs͑001͒ is measured as a function of etch time by atomic force microscopy and elastic light scattering. A flat surface is found to become rougher during the etch whereas a textured substrate becomes smoother. We have numerically simulated this behavior. It is found that the evolution of surface roughness at length scales between 50 nm and 5 m can be described with excellent accuracy by a continuum equation for the surface height h(x ជ ,t), which is given by dh/dtϭٌ 2 h Ϫ/2(ٌh) 2 ϪKٌ 4 hϩ, where is a random noise input.Typical processes in nanofabrication involve the patterning of surfaces by various etching techniques, and deposition or regrowth of thin films on these patterned substrates to create three-dimensional nanostructures. In the etching process, atoms on the surface chemically react and form volatile etch products. The chemical reactions and the removal of the etch products from the surface are inherently random processes, which are subject to fluctuations on the atomic scale. Surface roughness at mesoscopic length scales created by these fluctuations sets an ultimate limit to the precision with which nanostructures can be fabricated by chemical etching processes. It is, therefore, of great fundamental and practical importance to gain a quantitative understanding of this process.The reaction of GaAs͑001͒ with molecular chlorine is one of the most widely studied and employed etching processes. In recent years, much progress has been made in understanding the adsorption of halogens on semiconductor surfaces and in identifying the main reactions that are taking place on the surface. 1-4 In contrast, little is known about the evolution of surface roughness during the etch. It was found early on by optical microscopy that the thermal chlorine etch leads to a smooth surface at a substrate temperature of 300°C, but at a lower (150°C) or higher temperature (500°C) the substrate becomes rough 5 . Scanning tunneling microscopy studies of GaAs͑110͒ surfaces show that halogen etching proceeds by formation of one layer deep etch patches. 6 In this work, we report a quantitative description of the etched surface morphology using a continuum etch equation. In this approach, the discrete nature of the atoms on the surface is averaged out and the surface height is taken to be a continuous function h(x ជ ,t) whose time rate of change can be related to various spatial derivatives of h. Possible terms that can be included in a continuum equation have been derived within the framework of kinetic roughening theory. 7-9 Generally, in the limit of long times and large length scales, these equations lead to surfaces with characteristic scaling properties, and most experimental studies to date have concentrated on determining the relevant scaling coefficients for a given system. However, a sufficiently wide range of time and length scales to obtain reliable scaling exponents is often not accessible experimentally. In this work we demonstrate that the comparison of numerical simulations of...

Evolution of surface roughness of AlN and GaN induced by inductively coupled Cl2/Ar plasma etching

“…In the case of Cl 2 etching of GaAs, this value was found to be ϳ10 nm. 26 In our case, the corresponding roughness is у5 nm for both GaN and AlN. Associating with the lowest etch rates at which the nonlinear term in Eq.…”

“…25 In Ref. 26, also concerning thermal Cl 2 etching of GaAs, a rounded dependence was seen on the PSD function, with ␣ starting ϳ1 at high k then reducing to saturation with decreasing spatial frequency. Plasma etching of Si produced an exponent within experimental error of unity.…”

“…25͒ and 240 nm 2 /min. 26 However, since the desorption coefficient is proportional to flux at the surface, 47 it will be influenced by surface bombardment under RIE.…”

“…This value is comparable to the value of 36 000 nm 4 /min reported for Cl 2 etching of GaAs. 26 The scaling exponents obtained from the PSD analysis of the GaN surface data range from ϳ0.10 to 1.10. The uncertainty in ␣ is below Ϯ0.10 in all cases.…”

“…Previous etch studies have employed the theory of kinetic roughening to interpret surface properties [25][26][27][28] includ-ing a nonlinear term describing local etching. 29 Here, the time evolution of the surface h(r,t) is described by the nonlinear differential equation 25 ‫ץ‬h/‫ץ‬tϭ des ٌ 2 hϪ…”