Of 32nm technology in the copper metal CMP voids
Experiment 1 used to adjust the level of voids, Table 1 summarizes the process parameters of the first set of experiments. Crystal seed layer thickness and the thickness of the POR POR thickness varied between 1.5 times. Experiments using two different plating infiltration chemical reagents, the fill rate of the lower one while the other is higher. CMP cleaning process is also divided into two types, the first effect is stronger than the second.
P>Table 2 summarizes
CMP voids to adjust the level of the design parameters of experiment 2. Crystal seed layer thickness on the basis of Experiment 1 was fixed as a constant, and use one of the "thick" process. Four wafers (6-9 pieces) without CMP process. Thick layer of copper covered by focused ion beam (FIB) microscope was removed and shows the spiral copper wire. This allows pre-scanning electron microscopy (SEM) to examine the defects, thus avoiding the possible impact of CMP process.
Results and discussion
The results of Experiment 1, using a slower fill rate of plating solution and a more intense cleaning process defects obtained the highest level. The lowest defect levels can use a higher rate of plating solution and the cleaning process has been more moderate. For the filling rate of slow chemical reagents, high-level defects in the thin and thick layer of crystal seeds can be observed. Seed for the average thickness of the crystal layer, the use of chemical reagents slower rate will allow a higher defect rate. In this case, the cleaning process 2 1 compared to the cleaning process can reduce the defect rate. When the thickness of crystal seed layer on the common use of faster rate of chemical reagents, the use of the cleaning process for the defect rate of not so big. Seed crystal in the thick layer of high-speed filling of chemical reagents, the cleaning process on the defect rate to a certain extent, better cleaning process 2. Most defects are hollowmetal type. Electroplating chemical reagents and the combination of CMP process even for the thick layer of crystal seeds can also lower the defect rate.
A set of random samples of each of the 50 on the wafer SEM defect inspection and found empty after CMP is the main defect mechanism. Figure 1 and Figure 2 shows the CMP was observed with SEM the nature and severity of empty. Quickly filled with chemical agents (Figure 1) than the slow-fill plating process (Figure 2) resulted in fewer voids. The same, obviously, the first 1-4 wafers under the SEM has a relatively less empty, because the thinner the thickness of the crystal seed layer to make it easier to fill in the groove metal.
Experiment 2 results show that the high fill rate relative to the slow filling plating reagent reagents performance better. Compared with the first experiment, CMP cleaning process after the impact of the defect level is not very obvious.
A set of random samples of each of the 50 on the wafer SEM defect detection, post-CMP metal hole that is still the main defect mechanism. For low fill rate of chemical reagents (4 and 10-12), basically all the defects have been classified as belonging to the metal empty type. Figures 3 and 4 that after the hollow nature of the CMP and severity. Voids caused by slow chemical reagents more, almost every metal line can be observed.
In experiment 2, the copper cladding removed by FIB (This step is omitted CMP), and using SEM to observe the defect before the CMP process. Corner of the copper coating by FIB to open a window of about 15 microns, so that beam is not parallel to the substrate to the surface of the defect caused the most damage. 7 Figure 5 shows the top view SEM wafer, the spiral lines are using the same light field tools to detect defects in the body. As the wafer is quickly filled with the sample kit, only a very small number of defects observed. Near the upper left corner of the image defect is usually visible on the graph lines of some foreign material caused by corrosion. 9 Figure 6 shows the top view SEM wafer. FIB samples obtained by the vertical line is only a slight change in thickness will not affect the observation of planar defects. In this case, filled with the slow rate of chemical reagents, so the defect in the whole image is very clear.
Bright field and defect detection of defects in the metal of the parameters of the rate changes are sensitive enough, making the light field in the defect detection process optimization as a useful tool. When the device feature size due to the use of crystal seed layer thickness reduced, the defect rate level will increase in the trend growth rate of this defect can be quickly filled by using the reagent to reduce. Copper coating technology rather than through the CMP process to remove FIB shows these defects or the density in terms of both shape and polished by the CMP defects similar. It can be inferred, in this case, CMP process steps are observed on the final post-CMP defect rate levels have little effect. P>