Raw Data Configuration
Data Input: Two FOCAL data sets measured by the scanner tool. Best Focus for both chucks plus the stage-precision data has been imported.
Analysis: Examine across field focus uniformity for the tool prior to FOCAL® optimization. Chuck performance is determined and the source of overall performance degradation is calculated. Next the data taken by FOCAL after correction is analyzed. The performance after correction is shown to have improved center-field performance but also resulted in a greater range in overall full-field focus error. The Before & After data sets are then subtracted to calculate the corrections actually applied to the tool by FOCAL®.
Data files: FODSC_FocalImmersion_1stSet.XLS
FOCAL® Data Layout: Before Correction
Figure 1: FOCAL Data imported by "Family" using VR Matching
An ASML Focal Analysis was performed on an immersion scanner. The metrology is converted from registration data to three distinct sets of focus data. The focus data consists of full field, or “Best Focus”, for each scanner chuck, and a second focus set of focus stage-precision data.
The two sets of “Scanner Chuck” data comprise that of the measured focus uniformity on the pre-alignment chuck (“Chuck 1”) and the field exposure chuck (“Chuck 2’).
VR Matching imports this data as recorded by metrology but the Stage Precision data is condensed to a single field on the wafer for convenience as shown in Figure 1. The data is identified by the software as three families as shown in the “Family” checkbox from the interface in the lower right side of the figure.
We will examine only the Best Focus data in the remainder of this analysis. Statistics shown in Figure 1 include not only the chuck-response but also the stage precision numbers.
Performance Variation between Wafer Stages
Figure 2 illustrates both the method and data gathered for the focus data before correction. The same data set, ”FODSC_FocalImmersion_1stSet.XLS” , was loaded into the Reference and the Matching interfaces. The “Before Correction” “Family” check box was used to restrict the reference data to only that measured on chuck #1 as shown in the top half of Figure 2. Similarly the matching data shown in the bottom half of the figure presented the focus uniformity of the second chuck.
Figure 2: Chuck #1 (top) compared to Chuck #2 (bottom).
Note that the focus ranges and standard deviation on both stages is rather large but very repeatable for both chuck #1 and chuck #2. The average focus value between the two chucks is about -5 nm lower on chuck #2. The source of the large range in focus values becomes apparent if the vector-plotted field graphs are observed. Chuck-to-chuck differences, other than the 5 nm offset, are easily seen. The major variance in this data is caused by the left-most site-column of focus data with some lesser but still strong contribution from that on the right side of the field. The field uniformity without these edge-columns is much more uniform across the majority of the field exposure’s center area. This response suggests that the source of the focus error may in fact be aperture vignetting of the scan-slit but the more likely cause is the limitation of the FOCAL sensor array, which is results in poor metrology at the extremities of the scan slit.
FOCAL however does not have the capability of noticing the field-location based source of variance and simply chose to correct the overall field focus in an attempt to minimize the variance. This resulted in a minimization of the errors for the field edge while moving the corresponding error values into field center as is shown in Figure 3. The actual mechanism for this will be shown in the Difference measurement calculation presented later in this section.
Figure 3: Focus uniformity Before (top) and After (Bottom) correction. Data shown overlays chuck 1 & 2. Histograms represent data exclusive of the left-and-right most focus columns illustrating the superior performance of the tool prior to FOCAL correction across field center.
Performance after focus correction
Figure 3 plots the focus uniformity across each field for both before and after focus correction on the scanner. Each field displays both chuck #1 & #2 data. Therefore, each site on the field displays two vector focus values. Scaling on each plot has been set to 10 nm for each comparison. Scaling was fixed by right-clicking the image to use the plot editor.
The mean focus value on the corrected field has shifted down to values about 4 nm more negative in focus. X and Y focus means are closer than before the correction but still exhibit the significant shift. Unfortunately the range of focus values across the field has significantly risen after correction as has the variance. So while the focus correction improved the difference between X and Y average values the full-field range and variance has degraded. This effect on focus becomes apparent if the two columns of sites located at X=+/- 12 mm from field center are excluded and plotted on histograms as shown in the figure. Quite clearly the before-correction spread of focus has been degraded and will result in greater critical feature variation across the exposure field in the field-center critical areas.
Examination of the after-correction field in the lower half of Figure 3 reveals both a general tilt of the field resulting in over 10 nm of variance from upper left to lower right. Also notice the chuck-to-chuck variation between focus vectors in the upper left quadrant of the exposure.
We can investigate this across-field variation with greater detail by using the left mouse-button to box in the center row of data in each field plot to obtain response corresponding to focus uniformity across the field-center slit location. Generate the graphs of Figure 4 by selecting the XY Plot from the pop-up menu that appears when the button is lifted.
Chuck #1 X & Y focus values are plotted as circles in these plots while the Chuck #2 data are displayed as filled squares. X-focus corresponds to the focus variation seen by the edges of a vertically-oriented feature and is therefore called “dz-V” by the metrology. Similarly “dz-H” will sometimes be called the Y focus value. Variable names are dictated by the user metrology rather than Vector Raptor.
The offset in average focus and field-center-slit improvement in focus range resulting from the focus improvement is easily seen in Figure 4. Also notice that the chuck-#1 to #2 splitting has been removed.
Examining field response on the graphs on the right side of Figure 4 for the row located at +10 mm above field center we now increase the plot scales to almost double those of the field center. This graph clearly shows the degradation in performance resulting from the correction.
Figure 4: Across-slit focus uniformity before (top) & after (bottom) correction
Difference Measurement of the FOCAL Correction Applied
We next select the “Difference” tab in the vector raptor interface. The “before correction” data is used as the reference and subtracted from the ‘after correction” data. No actual data culling is needed except for the removal of the precision focal data family. To calculate the data we selected a the “Point-for-point” subtraction option and the pressed the “Calculate” command button. The results, representing the focus change actually made to the tool, are shown in figures 22 and 23.
The matching report shown in Figure 22 illustrates the mean -5 nm shift imparted by the correction. The overall correction range of values added removed a 12 nm spread of values with only 0.6 nm of difference between X and Y correction values. However the vector plot of figure 23 is more interesting since the source of the focus-error signature seen in the bottom plot of Figure 3 now becomes apparent.
Figure 23 clearly shows the induced field tilt. The barrel distortion layout of the plot was used to correct for the error contributed by the high focus errors seen in the sites located at the field edges, +/- 12 mm from field center. If the -12 mm left-hand site-column data is closely examined, we can see that the correction tracked the general signature of the columns focus error going from high values at the top of the column at the top of the exposure to small values with opposite sign at the bottom. Examine the top plot of Figure 3 in comparison with that of figure 23 to see this signature. However the large values of this column’s distortions were not damped quickly enough as the focus moved to those located nearer the center of the field.
The overall effect of the FOCAL correction was to reduce the focus errors at the extremities of the field while increasing focus non-uniformity, and therefore increasing critical feature dimensional nonuniformity, in the field center.