Measuring Film Thickness after Initial Polish Step

Copper CMP is often performed in two steps. In the first step, most of the soft copper is cleared from the wafer using a copper-selective process. The remaining copper and harder barrier film (tantalum, for example) is then removed using an unselective process. It is important to optimize film-thickness uniformity during the initial step; doing so minimizes the amount of polishing needed in the second step, which, in turn, reduces final dishing and erosion.

Figure 4 shows a typical example of how optoacoustic metrology can characterize film thickness between these two polishing steps. Ideally, no copper would remain after the first step. In the example shown, the initial polishing of a blanket wafer did not completely remove the copper, as is evident in the photo inset as well as the optoacoustic scan data from across the diameter of the wafer. The remaining metal would have to be cleared in the second polishing step. Intermediate optoacoustic measurements such as this can be used to monitor the effectiveness of the initial polish and optimize the second-step process conditions based on those measurements.

Figure 4: Diameter profile of the remaining metal after the initial step of a two-step polish sequence measured with an optoacoustic tool. Inset photo shows the wafer with this diameter profile.

The optoacoustic technique can provide accurate film-thickness measurements between polish steps. While reflectometry or ellipsometry can be used to detect very thin residual layers at the completion of the CMP process, they cannot be used on the thicker opaque metal layers remaining at intermediate stages. Electrical measurements could be taken on blanket wafers, but the presence of two surface metals (copper and tantalum) with vastly different resistivities would present a challenge. Also, as discussed earlier, profilometry provides useful data in some situations but has a limited ability to show very gradual thickness variations in long scans across the full wafer diameter. Therefore, optoacoustics provides an effective complement to these other techniques for CMP process optimization.

Measuring Arrays of Metal Lines

The uniformity of polish rates for different feature sizes and densities is critical to the performance of copper CMP. This uniformity is usually measured on arrays of copper lines of various sizes, from the smallest interconnects to the largest. Both profilometry and optoacoustics are capable of measuring the dishing and erosion of these structures that are caused by overpolishing. For process development applications, where speed is not an issue, both techniques can be used in a complementary fashion. Profilometry can be used to characterize step heights and wafer planarity over small regions with very high spatial resolution, while optoacoustic metrology can provide average film-thickness measurements across the wafer.

To demonstrate the capability of optoacoustics to measure a variety of feature sizes on patterned wafers, the researchers in this study performed optoacoustic measurements on a range of structures with varying feature width and density using the Sematech 854 wafer pattern. Results were checked with profilometry. As shown in Figure 5, optoacoustics was used to directly measure the thickness of the metal film or metal/oxide array, while profilometry measured the step height between the field oxide and the copper surface. It was assumed that the initial etching of the oxide (to open the windows for metal deposition) stopped cleanly at the etch-stop layer, so that the difference between the field oxide thickness, measured with a reflectometer, and the profilometer step height would equal the optoacoustic measurement (i.e., R – P = OA).

Figure 5: Thickness dimensions measured by optoacoustic (OA), reflectometer (R), and profilometer (P) instruments in experiments on arrays of metal lines.

Comparisons of post-CMP measurements for two copper linewidths are shown in Figures 6 and 7. Figure 6 plots the measurements across a 1.25-mm-long array of 0.25-µm-wide copper lines with 0.25-µm-wide oxide spaces between them, and Figure 7 plots the measurements of a single 100-µm-wide copper line. To determine the 0.25-µm line thickness, the optoacoustic measurement was averaged over all the lines within the 25-µm probe laser spot. The results were analyzed using a model that averages the acoustic properties of the copper/oxide combination. As the figures indicate, optoacoustic results for both line sizes correlated well with the corresponding profilometry data.

Figure 6: Optoacoustic and profilometry thickness measurements for a 1.25-mm-long array of 0.25-µm-wide copper lines and oxide spaces.

Figure 7: Optoacoustic and profilometry thickness measurements for a 100-µm-wide copper line.

Figure 8 shows an extension of this correlation for a range of other features on the Sematech 854 pattern. The wafer samples contained linewidths from 0.25 to 100 µm and included a range of pitches yielding different densities. (The 854 pattern has a minimum linewidth of 0.18 µm, but linewidths <0.25 µm could not yet be printed by the lithography equipment on-line during these experiments.) Eleven representative structures with densities from 25 to 90% were measured with both profilometry and optoacoustics on each of two samples—one polished normally and the other overpolished significantly. The resulting data are shown on the figure as two points for each type of structure. As the line through these points indicates, the correlation between the two measurement methods was very close, with a standard deviation of only 112 Å for the range of features, which demonstrates that optoacoustics can characterize CMP performance on structures having the feature sizes and densities commonly found on product wafers.

Figure 8: Correlation of optoacoustic thickness measurements (OA) with profilometry results (R ­ P) for 11 structures of varying linewidth, pitch, and density. Each structure was measured on two wafers, one polished normally and the other overpolished. The standard deviation was 112 Å.

Figure 9 shows an example of the type of dishing/erosion process characterization information that can be obtained with optoacoustic measurement. Four structures with varying linewidths and line spaces were monitored with the in-line tool over a wafer run consisting of two lots. Film loss was determined for the center die of each wafer by measuring film thickness and subtracting the result from the targeted initial trench depth. The average film loss was lowest for the 0.25-µm features, larger for the easily dished 50- and 100-µm lines, and greatest for the 9-µm-wide copper lines with 1-µm spaces. The results shown in the figure are typical for a variety of process conditions.

Figure 9: Film loss at center die on four different line arrays averaged over a wafer run of two lots.

Verification of the optoacoustic measurements was performed by correlating the measurements with electrical test measurements on the line arrays. Like electrical test results, optoacoustic measurements on line arrays are sensitive to the total metal cross-sectional area. A change in either thickness or linewidth is registered as a change in "effective" thickness. Figure 10 shows the correlation between the effective array film thickness for a 0.25-µm 50% line/space array, measured with the optoacoustic tool, and the inverse of the line resistance measured electrically. Standard error of the correlation was only 0.9%. Investigation with CD SEM on these samples showed that for the 0.25-µm-linewidth arrays (the critical linewidth for the lithography system used), the edge die had linewidths ~10% smaller than the center die, which explains why these die appear at ~10% lower effective thickness in the correlation plot.

Figure 10: Correlation of optoacoustic film-thickness measurements with electrical results for 0.25-µm-wide lines in a 50% density array.

The use of optoacoustics can be easily extended to measuring arrays with smaller linewidths than the example discussed here. Because the technique measures the thickness of line arrays by treating the arrays as composite copper/dielectric films, such analyses are valid whenever the instrument's acoustic wavelength (4–10 µm for current-generation tools) is longer than the linewidth. Thus far, optoacoustic tools have been used on linewidths as small as 0.07 µm in proprietary studies.

In-Line Process Control

Once a CMP process has been optimized and implemented, process engineers can use optoacoustics to track wafer-to-wafer thickness variations during production. For example, Figure 11 shows wafer-to-wafer post-CMP copper-thickness measurements performed by optoacoustics, profilometry, and electrical resistance tests on a 100-µm-wide line in the Sematech 854 test pattern. Despite the fact that all of the wafers measured had been subject to the same deposition and CMP processes, the resulting data revealed significant wafer-to-wafer thickness variations of ~10% of target. The figure also shows that the optoacoustic and electrical resistance measurements were in good agreement with each other, with a correlation standard error of only 0.9%. Profilometry results correlated less well with the other two techniques, perhaps because such measurements are much more localized along a particular cross section of the metal lines.

Figure 11: Post-CMP thickness trend data for a run of ~45 wafers, comparing optoacoustic, electrical, and profilometry results measured on a 100-µm-wide copper line. The standard error in correlating optoacoustic to electrical results was 0.9%.

Conclusion

CMP must be optimized to minimize both under- and overpolishing at various process steps and variability in dishing and erosion for various feature dimensions. Achieving good results requires careful process characterization. This study showed that optoacoustics can provide process characterization capabilities that complement those of such established methods as profilometry and electrical testing. It can measure film thickness directly and rapidly, with high spatial resolution, across long scans on either patterned or blanket wafers. The technique enables process engineers to easily characterize within-wafer polish uniformity, film profiles at intermediate polish steps, and CMP planarization length. Because it can be used for both solid pads and arrays of submicron lines and correlates well with profilometry and electrical tests, optoacoustics also can be used to monitor wafer-to-wafer post-CMP thickness trends during production.

Acknowledgments

The authors would like to thank Josh Tower and Chris Moore of Philips Analytical and Tamba Tugbawa of the Massachusetts Institute of Technology, all of whom made valuable contributions to the joint research project. This article is partially based on presentations made at the spring 2001 meeting of the Materials Research Society.