1.1.2 Etching and Deposition

  To transfer the resist patterns onto the semiconductor wafer, several etching techniques are used. With the decreasing feature size the trend is going towards dry etching processes, but there are still applications for the classical wet etching processes. The wet etching process has a high reliability and is used for total or partial removal of whole layers. Due to the so-called ``under-etch'' effect, which can lead to a full detaching of the mask from the underlying material, wet etching is not suited to transfer patterns with sub-micron feature size. The process modeling equivalent is an isotropic etching model with uniform etchrate distribution everywhere. Dry etching is based on reactive ion-etch processes (RIE) and plasma-etch processes (PE). For special applications reactive ion-beam etching (RIBE) is also used, which is favored to be the method of the future. But there is still a lack of understanding of the process mechanisms and of the coupling between the variety of physical and chemical parameters for etching processes. Therefore, it is very difficult to perform predictive etching simulation. There are two ways to perform dry etching processes. One is concerned with the etching reactor itself, dealing with its geometry and chemical gas flow using methods of fluid dynamics [Ula89]. Another and more popular way is to simulate the microscopic surface movements, as done by several process simulators [Old80] [McV90a] [Str93]. Accurate physical models for dry etching processes are complex, due to the chemical and physical processes occurring within the plasma. The real etching process is a combination and variation of interaction between neutrals, radicals, and high energetic ions. Effects like chemical-enhanced sputtering, where a certain mixture of process gases amplifies the chemical etch rate for silicon, or damage-induced etching, where the ion bombardment damage caused at the wafer surface increases the chemical reaction between the target and the reactive neutrals, must be captured by simulation models. Also sidewall passivation and hence the burrowling effect are important areas of modeling interests [Zhe94]. Thereby two species of chemical neutrals are considered, one attacks the substrate material and the other passivates the underetched sidewalls. Reliable models should consider reflections of particles at the topography, sticking coefficients and flux distributions for the species.

Nevertheless, there are some three-dimensional effects like the proximity effect which gave motivation for the development of three-dimensional etching and deposition simulators. The basic surface movement algorithm plays the key role by increasing the dimension. The classification of the applied algorithms is the same as for lithography simulation (see 1.1.1). By the string algorithm the wafer surface is represented by two-dimensional surface elements which are moved according to the local etch rates. Special data management is necessary to avoid loops and overlapping regions. The cell-removal algorithm requires the discretization of the whole simulation domain by tiny cells. By applying the local etch rates onto the cells, it is decided whether to remove the cell or not. For implementation into an etching simulator this algorithm is quite simple and absolutely stable. The applicability is limited by the large memory consumption. Physical etching models are similar to those known from string-algorithm based simulators. The decision whether to favor the cell or string algorithm is more complicated as it seems to be at first glance, especially when the simulation of the whole fabrication process is considered. Most of the available three-dimensional process simulators are using polygonal geometry representations [Oda88] [Ush90] [Lei95]. Efforts have been made to convert the structures between a cellular geometry and a polygonal geometry representation [Mle95] to convey the geometry information between these simulators.

Thin film deposition is used for fabrication of various materials like polycrystalline silicon, nitride, silicon dioxide, and silicides. The common deposition methods employed are chemical vapor deposition (CVD) and physical vapor deposition (PVD). From the modeling point of view the same algorithms are used as for the etching processes accept some modifications to capture deposition specific material properties. Re-emission of deposition particles is modeled by uniform sticking coefficients and a cosine distribution for the particle flux [McV90b] [Egu93]. There are also popular algorithms based on ballistic aggregation concepts [Smy90]. Thereby statistically distributed particles and their trajectories are calculated during the deposition process and added up at the surface, which provides additional information about the shape of the films as well as about the quality of the deposited films. The computational effort is high compared to other algorithms and therefore not applicable for three-dimensional problems.