News

New Hot Topic in Cu CMP: Edge-Over-Erosion (EOE)

Dishing and erosion are the two most costly surface anomalies that come about with copper CMP. Dishing occurs when the copper recedes below or protrudes above the level of the adjacent dielectric. Erosion is a localized thinning of the dielectric.
Recently, we have observed the new type of defects associated with the reduction of feature size and integration of low-k materials through edge-over-erosion(EOE) phenomenon. EOE is a defect that is where the erosion is higher at the edge of the recessed region in a feature when compared to the center (figure 1). EOE is increasingly severe when the feature size is small (0.12-0.25µm) and pattern density is high (>35%).

Figure 1:profile scan which shows edge-over-erosion(EOE) phenomenon

The investigation of EOE brings forth several experimental challenges. One of the first challenges is the requirement of a very advanced atomic force profiler.
Secondly, software, which can level and extract the important metrics such as dishing, erosion, and others, is required as well. Our experience shows that in order to systematically investigate the EOE, a number of profilers that scan up to 80 to 100 in number, have to be analyzed per wafer. Without the aid of some sort of analysis software, the analysis can be very tedious and time-consuming. Lastly, the design of test patterns (which is sensitive to EOE after Cu CMP) is required. Our previous experience shows that EOE is most pronounced for fine array line features.
Since the minimum feature size of our current SKW 6-5(which is developed for 90nm technology node) is 0.12µm, we have chosen this patterned wafer as the first test vehicle. With the aids of an advanced atomic force profiler (XE-300P), advanced pattern wafer (SKW 6-5 Cu/TEOS), and in-house developed software ("ChampiAn"), we have investigated EOE phenomenon in Cu CMP. Enclosed is a document summarizing our experimental findings. Based upon these data, we conclude EOE is a function of metal line width, pattern density and array size. In addition, we also conclude that SKW 6-5 is an excellent test pattern wafer for the investigation of EOE.