Data Input: Seven datasets, representing CD-SEM metrology of six BARC and one bare wafer layer, are combined into a single data analysis workbook for VR Matching.
Analysis: Determine the optimal BARC film for an 80 nm process from six vendor samples.
Data file: Weir_Process_BARC.XLS
Raw Data Configuration
Seven wafers were exposed. One wafer was processed without an anti-reflection coating and each of the remaining six wafers was coated with a different commercial Bottom Anti-Reflection Coating (BARC). The objective of the analysis was to competitively compare Process Window and performance of each commercial BARC.
Each wafer contained a focus-exposure matrix of nested features as shown in Figure 1. The wafers were process identically with the exception of the BARC layer. One nested site of 80 nm features was measured on each exposure field. The features were measured using a commercial CD-SEM. The objective was to determine the most robust BARC coating by measuring the extent of a +/-10% Process Window of Bottom Critical Dimensions (CDs). Since the sites were able to be viewed by the CD-SEM, the operator also recorded the number of features that were resolved in the line-space pairs by counting the number of profiles standing from 1 to a maximum of 7 peaks.
Figure 1: Exposure Dose and Focus Matrix values shown from the Weir Main interface layout.
Data was assembled into one lot of seven wafers for this analysis. Assembly was simple since the imported CD-SEM data of Vector-Raptor could be easily copied and pasted into a master Weir Workbook. Each BARC coating here is designated using the Wafer ID for the dataset.
A total of 836 measurements for the seven wafers is displayed in Figure 2. The process target size is 80 nm with a +/-10% Process Window. Process Window visualization here is enhanced by right clicking the graphic and setting the bar-code range key to the 72 to 88 nm settings shown in the figure. Any exposure-site that is not dark blue or deep red will be within the Process Window.
Figure 2: Bottom Critical Dimension (BCD) SEM Measurements for all seven wafers Color bar-code has been set to 80 nm +/- 10% range of process window
Figure 3: “Collapsed Line” settings for seven BARC Wafers
Features included nested pairs of seven line-space profiles. This data was also gathered by the SEM operator and manually entered into the Weir data using the FEM Format editor of Weir. Figure 3 displays the number of standing lines in the exposure data with values that range from 1 to 7 representing the number of well defined and standing line/space pairs.
Vector Raptor was able to use the “Collapse” variable as a data-culling tool by setting the features minimum threshold to 3 lines and therefore excluding poor exposures outside of the process definition.
The Process Window response of any two BARC constituents can be easily compared using the VR Matching as shown in Figure 4. This method is much more direct than the complexity of adding in a classic Process Window analysis and visualization of the process response as well as the observation of poor metrology points is superior to classic methods since the analysis is not dependant upon a complex polynomial model interpretation. Finally, although this data has only one point per exposure field, the same plot could have been employed on a mult-field-site analysis of the entire exposure field’s Process Window response thereby incorporating the influence of lens aberrations and edge-field effects.
Figure 4: Wafer without BARC (left) compared to wafer with Expo4073D BARC.
Visualization in Figure 4 used the Color-Bar Key to help in determining the Process Window extent. The user could also have used the Variable Culling tool to set the Process Windows maximum and minimum values thereby eliminating “out-of-specification” metrology points.
The areas of Figure 4 that are not deep red or blue define the process window. Placing the process exposure at the center of the window is as simple as either selecting the exposure site on the graph or using the mouse to box-in the desired process space and viewing the resulting box-enclosed data set. The advantage of using a BARC coating in the process is clearly seen in this data since the Process Window of the right vector plot of Figure 4 is much greater than that of the bare wafer.
Figure 5: Difference Comparison Setup
EXPO4008 BARC (Left) as “Reference” compared with remaining six BARC responses (Right).
Difference Comparison of BARC Process Windows
The Expo4008 BARC was selected, by visual comparison using the comparative display method of the previous section, as a film exhibiting one of the largest Process Window responses. Figure 5 displays the Reference data of Expo4008 in comparison with the remaining six film responses that will comprise the “Matching” lot. Once again the bar-key coding in this figure is set to the limits of the desired Process Window of 80 nm +/- 10% allowing both the Process Window as well as the remaining well-defined, but out-of-specification, line-space pairs to be viewed. Selections were performed by loading the same data lot into both Reference and Matching lot objects. Individual films for each were then selected using the “Wafer ID” checkbox. VR Matching can now be used in the Difference Calculation mode to observe the process variation of each wafer.
There is only one point per field so interpolation during the “Difference” calculation is not needed; the “Point-for-point” option was selected. During the Difference calculation, VR Matching will find the nearest field value on the Reference wafer to any fields on the Matching lot wafers. So if a match-wafer, or in this case BARC coating, is on an exposure-field that is not contained in the reference, the program will search for the nearest reference exposure and use that data.
Tip: Including all valid data points, both in the process window and external to it, as in Figure 5, is an excellent method for discovering process extension capabilities. Any process capable of moving to smaller feature sizes can easily be evaluated and centered using this method by simply changing the color-bar key limits.
However, when the process limits are well defined we can obtain a precise measurement of the extent of the Process Window for each film by restricting the feature values to these limits. In this dataset we used the variable-culling tool to restrict BCD values to the process-window range of 72 < BCD < 88 nm and re-plotted the graphs of Figure 5 into those of Figure 6. Now the exact shape and extent of the process-exposure space can be comparatively seen. In addition to shape and range the stability of critical feature size within the space can also be evaluated as a measure of the robustness of the process.
Figure 6: Difference Comparison Setup of the exact Process Window
EXPO4008 BARC (Left) as reference compared with remaining six wafers (Right).
Bottom CD values here are restricted to the process window range of 72 < BCD < 88 nm
Since the exposure layout centers the Process Window near the physical center of the wafer, we’ll observe BCD variation of BARC as a function of the exposure radius from the wafer center by selecting the “Radial Plot” drop-down graph. This method implies that the greater the radial distance, the greater the extant of the film’s Process Window. We generate the Difference Calculation by simply pressing the “Difference” Command button to create the plot interface of Figure 7.
In the graphic of Figure 7 the response of each BARC film can be easily viewed by clicking the appropriate tab on the interface. There is one tab set up for each wafer, in this case BARC film, of the data lot. The final three (3) tabs of the interface show the Trend Plots associated with each BARC’s response.
Figure 7: Process Window, Difference Comparison for All BARCs
The six remaining BARC Process Window plots, which are accessed using each of the interface tabs, are shown in Figure 8. The comparison is now restricted to only points whose feature size lies within the desired Process Window so we can see the performance quality and extent of each BARC clearly defined as the number of data points actually on the chart. That is, the greater the number of plotted points and their range from Radius=0, the greater the extent of the Process Window.
Based purely on Process Window size the Expo4073D BARC exhibits the largest window. Looking at the approximate average CD difference value of this film, we can see that it also exhibits a small feature size shift from the bare, or “No BARC”, process wafer. This suggests that exposure times should not have to significantly increase using this BARC. Feature size ranges can be estimated by an examination of the population spread, or envelope, of points along the ordinate of the graph. The Expo4073D range of CD values appears to be approximately the same as those of the No-BARC or bare wafer. This implies that the process window has been expanded but process performance is not improved beyond bare wafer values.
Examination of the “envelope” of the plotted data points yields the comparative performance of each BARC film. A film that exhibits a point-spread envelope centered nearest the zero ordinate value exhibits a closer match performance in exposure response to the reference BARC. An example of a good match to the reference is found in the Expo4061H_vk90 plot.
A different mean value implies an offset in the average feature size from the selected reference BARC. This is commonly observed between any BARC film and a process not using BARC as shown in the graph of the upper left of the figure. The greater the shift in this mean value from the bare wafer exposure, the greater the exposure change needed for the film.
The plot whose population contains the minimum envelope-spread along the BCD ordinate values along with the greatest range along the abscissa, or exposure values will sustain the greatest process and the most robust response of critical feature profile size. This film will not only exhibit a large process window but will also support the tightest histogram-spread of points within that window. For example, although the Expo4037D BARC exhibits the greatest process window size, the EXPO40618_vk90 film supports a window that is only slightly smaller while at the same time exhibiting a much tighter variation in BCD values across the window and a closer average value to that of the bare wafer response.
The last three tabs of the Difference Interface exhibit the statistical trend response of each film relative to the reference BARC using summary statistics displayed on Trend Plots.
Figure 8: Process Window Difference Comparison for each BARC
Film Radial range along the abscissa indicates the extent of exposure size of the process window.
Spread of the population along the ordinate indicates the robustness of the film to process variance.
The Mean Difference +/- Standard Deviation for each film, shown in Figure 9, can be selected from the “Mean” tab. In this instance, the shift in exposure or BCD offset values shows the match similarity of the film to that of the reference. The error-bars show the robustness of the film and process to exposure variation.
Figure 9: Mean-Difference Tab Plot showing BCD Mean Difference from the Reference and Standard Deviation for each film within the defined Process Window.
Here we can see that the closest match to the bare wafer CD values is the Expo4073D film. The closest match to our reference BARC is found on the Expo4061B_vk90 film. An examination of the error bars for each process window provides instant evaluation of the standard deviation of features sizes within the process window. Here the Expo4061B_vk90 BARC film also exhibits the tightest population spread of feature sizes and will result in the most robust process performance.
Figure 10 further clarifies the process robustness of each film by plotting the range of BCD variation from the reference values for each film.
Figure 10: Range of BCD Differences within the Process Window from the reference film for each BARC tested.