A New Method For Simulation Of Etching And Deposition Processes

Accurate simulation of et.ching and deposit.ion processes requires three-dimensional models and algorithms for wafer topography evaluation. \;t’e present a new approach for topography simulation which is based on morphological filter operations for advancing the etch front. The primary advantage of cell-removal methods is the t.opologica1 st,ability during simulat,ion. These algorithms can easily handle arbitrary structures and do not have the looping problem encountered in surfaceadvancement methods [l] [2] [3]. We use an array of rectangular cells for geometry representation, where each cell is characterized as etched or unet.ched. The material is defined through an identifier assigned to each cell. Material boundaries need not be explicitly represented. To advance the etch front we perform adaptive spatial filter operations along the surface boundary. During etching, cells within the filter are etched away, while cells out,side st,ay unchanged. These filter operations are based on Minkowski algebra which allows one to simulate topography processes by use of the fundamental morphological operations of erosion and dilation, as they are termed in image processing [4]. In general, filters are ellipsoids, in case of isotropic movement of surface points filters are spheres, although t.here is no rest.riction on the filter shape. Thus, the simulation of etching with preferred etch directions such as in crystalline etching is also possible. The main axes of each filt.er are relat,ed t,o the simulation time step and to the local etch rates. The etch front at a given t,ime step is obt.ained by t.he envelope of the filtered cells. With our method we avoid the inherent inaccuracy of t.he original cell algorithm [5], which in two dimensions produces an octagon instead of a circle during uniform etching from a single point,, as shown in Figure 1. Filter operations at material boundaries are performed using composite filters. In general! interfaces lead to an abrupt change in etch rates. For this reason: filter operations are always performed selectively on a given material, which means t,hat. only cells of the same material as the actual cell material are etched away during one filter operation. For a given t,ime step, all cells which have to be filtered are stored dynamically in a linked list toget,her with rate informat,ion. At t,he beginning of a simulation st.ep, the linked list consists only of surface cells. Etching and deposit.ion st.eps process the list,. Filters which extend over a material boundary will add cells to that list and are subsequently processed by addit.iona1 filt,er Operations. The et.ch rates for these filter operat,ions depend on etch rates of both sides of the interface and on how far a filter reaches into the other material. On the following page we present simulat,ion results both in two and three dimensions. Figure 2 shows a deposition from a hemispherical vapor source. The growth rate of the evaporated film at each point is strongly dependent on the surface topology. As a result of shadowing, the growth rate varies in tirne. Figure 3 shows anisotropic etching with a mask. The t.opmost layer is a mask that etches slowly with respect to the underlying layers. Through t.he lateral etch rate of the second layer etching under the mask takes place. Figure 4 shows the correct movement of the etch front during etching at several layers starting from a planar geometry. As one can see, we are not restricted t.o a layered geometry. Three dimensional applications can be seen in Figure 5 and Figure 6. Figure 5 shows isotropic deposition to simu1at.e a chemical vapor deposition process and Figure 6 shows an anisotropic etching process with a strong directional rate t.o simulate t,rench etching. REFERENCES