An alternative technology for dose and focus monitoring
02/01/2007
CD budget constraints necessitate focus and dose monitoring as standard process procedure. Traditional SEM metrology has a few techniques that may allow monitoring focus on standard production wafers, but the lack of sensitivity limits its capability. To overcome this limit, alternative technologies are required. One such technology, referred to by KLA-Tencor as MPX, uses a dual-tone line-end shortening (LES) target, which can separate focus from dose with sufficient sensitivity to each. Experiments show this technology is a viable candidate for a production focus and dose monitor.
Metrology methodology
The dual-tone LES technology target is shown in Fig. 1. It uses grating targets consisting of arrays of lines and spaces at or near the critical dimension of the device layer. The technology being evaluated takes advantage of the LES of array targets in brightfield “islands” or darkfield “windows,” respectively. Relative to CDs, LES will be several times (10×) more sensitive to focus-exposure changes.
 Figure 1. Dual-tone MPX target based on line-end shortening. |
The behavior of both target dimensions (TD1 and TD2) is approximately parabolic through focus and approximately linear through small ranges of exposure dose. The combination of these two properties enables the technique to separate focus from exposure.
The target, though compact enough to fit in the most restricted scribelines, contains metrology dimensions large enough to be measured using a standard optical overlay tool. The coupling of focus-exposure metrology with overlay metrology maximizes the value of the optical overlay tool.
Use of two target dimensions allows simultaneous determination of focus-exposure conditions on a standard production wafer [1]. The technique requires a single calibration step before production monitoring can begin. The calibration step uses F-E matrices (FEMs), and the measured TD1 and TD2 data are fed into an empirical model of the form: TD = k0 + k1E + (k2 + k3E)(F-F0)2. This model is solved for the model coefficients K0-K3 and “best focus” position Z0. The Z0 of the target is typically within <30nm of the standard process best focus, as determined by the SEM results. The model coefficients are saved to a reference table and used during production monitoring to calculate focus-exposure values for each measured location.
Implant layer challenges
The implant layer use case evaluated for this work was selected because of its challenging topography that creates difficulties for traditional SEM metrology. The challenges presented with this layer include multiple edges along the trench sidewall, and the tight focus requirements to clear the sublayer trench at the bottom of the topography. The close proximity of secondary edges from the sublayers provides difficult circumstances for the edge-detection algorithm, resulting in false edge detections for one or both sides of the trench. The consequence of a false edge is that the CD value will be considered out of the targeted limits and thus lots will be put on hold for engineering analysis.
Though typically considered a noncritical layer for overlay and CD, its topography makes it very sensitive to focus. Focus sensitivity becomes critical based on the topography of the structure. For this layer, one is trying to control two parameters: standard CD control at the top of the trench and trench control at the bottom. Between the two areas of interest, this process stack can vary up to 2µm in height. The height difference coupled with the characteristics from each structure make the focus and exposure requirements quite different. The process window (PW) for the CD can be visualized as a more rectangular box, which has a much higher focus window, but a relatively narrow exposure window. This is typical for standard CD control where exposure provides the largest impact on the overall CD dimension. A smaller square PW is required for the trench at the bottom; this feature has a different PW requirement that is very sensitive to focus but not as much to exposure. The tighter focus requirements ensure that the sidewall of the resist does not block the trench. Another interesting characteristic of this scenario is that there is an offset between the centers of focus (COF) of the CD and the trench, which is done to make sure that the sidewalls of the resist at the bottom trench are optimal, so that it is clear.
 Figure 2. MPX dose vs. scanner inputs. |
Both of these concerns compound and drive the need for an alternative technique that could be used to effectively monitor focus and exposure on the implant layer, resulting in improved excursion monitoring.
Production results
In this section, we present summaries created from approximately five months of high-volume production data. The data shown in Fig. 2 and Fig. 3 present the MPX parameters of dose and focus vs. the scanner inputs. Figure 2 shows the MPX dose as compared against the scanner input dose. The input dose is continually being adjusted by the APC feedback based on the standard SEM CD feedback. The main point here is to notice that the MPX dose tracks the scanner input fairly well, which presents an interesting effect: the APC is adjusted to maintain a constant CD value, which in turn should maintain a consistent LES value for the MPX target.
 Figure 3. MPX focus vs. scanner inputs. |
Figure 3 represents the MPX focus vs. the focus input of the scanner. For the most part, the scanner input is a constant value until ~2/3 across the chart, at which point a preventive maintenance (PM) activity was performed. At the exact point of this scanner input change, the MPX focus reacts accordingly. This data appropriately illustrates the MPX capability to detect focus variations on production wafers. The next two figures (Fig. 4 and Fig. 5) show the MPX results vs. the within-field CD SEM metrology that was performed on the same wafers. Since the field sampling for the two different metrology methods is not identical, we should not expect to see a direct high correlation result. One of the driving factors supporting this MPX use case is the fact that the CD metrology has not been stable. If we analyze the data separately, we can begin to see the effects of the focus and dose on the CD values.
 Figure 4. MPX focus vs. CF CD. |
Toward the middle of Fig. 4, in the white circle, the CD values increase in the area about where MPX shows a decrease in focus. Towards the later portion of the chart, the CD returns to the normal condition and the MPX focus also stabilizes back to the higher level. Dose, on the other hand (Fig. 5), is showing a gradual trend. The CD values on the left of the chart are less than those that appear on the right end of the chart. The MPX dose also shows a trend where the dose begins higher and then decreases gradually over time, which makes practical sense since a decrease in dose results in an increase in CD.
 Figure 5. MPX dose vs. CF CD. |
What this data provides is confidence that MPX can be used to monitor production wafers. Control limits can be established about the MPX focus and dose parameters to assure that the CD remains within the desired range. Through special sampling and analysis, we are able to extract information regarding scan dependency, tilt, uniformity, and general trends in focus and dose.
Monitoring strategy
Adding MPX tests into the standard overlay recipe provides a simple and efficient way to collect additional data from the same load of the wafers. For the same load of a wafer, a user can obtain overlay, focus, and dose information with only minimal additional metrology time. Spansion’s focus and dose sampling for the implant layer is 3 wafers/lot, 5 fields/wafer, and 4 points/field.
The 3-wafers/lot figure is based on the standard lot sampling for overlay. Adding these tests results in a 2-3 min/lot increase, but provides two more process control parameters. Based on experience, this level of lot sampling seems to be sufficient. Four targets/field were selected in order to provide relevant data to monitor tilt and dose uniformity. The five fields are selected to mitigate the spatial effects of cross-wafer focus and dose variations related to varying track conditions.
Spansion’s typical best-known method (BKM) for field sampling is to select fields in an even distribution based on the exposure tool’s scan direction. This BKM enables one to monitor the cross-wafer 3σ statistic as a method to monitor scan-direction offsets, which seems to be one of the largest contributors of focus variation across the wafers. We have seen numerous occurrences of a scan-direction signature in wafer evaluations. With the methods used today, this error goes unnoticed on production wafers.
The main benefit of performing focus and dose monitoring at the overlay step is to reduce wafer moves in the litho cell, which decreases the litho cycle time, a cumulative benefit. The logic works so that focus and dose are dispositioned at the overlay step. If the lot fails, one of the parameters is then forwarded to the CD SEM for further metrology and verification. This does indicate that there will need to be some level of standard production lots processed by the CD SEM to maintain a sufficient baseline; this limited sampling of the standard production may be as low as 10%.
Additional production data
With the implementation of MPX, the photolithography module can realize a reduction in cycle time of >50%. By enabling focus and dose monitoring with MPX on the overlay tool, all litho process feedback can be obtained in a single metrology step, eliminating queue time at the CD tool for most lots. Not only is litho cycle time improved, but increases in CD-SEM availability for development activities are also realized. Based on average Q times (i.e., queuing times: the amount of time required to move the wafers from one tool to the next) and CD metrology times, it is estimated that there is an ~53% decrease in litho-cell time when eliminating the extra Q time and CD metrology time. This data was based on an average time as recorded in a real production environment.
In addition, layers that prove problematic for CD measurements are prone to false reworks and escapes. Incorrect measurements or fliers can obfuscate the metrics used in a CD control chart. In such circumstances, either lots will be in jeopardy of being needlessly reworked or escaping the litho area with genuine process deficiencies, or significant engineering intervention is required to ensure proper dispositioning.
MPX measurements are not sensitive to the same process factors that complicate CD measurements, so the metrics provided with it prove more robust for control. When a lot fails MPX metrics, the lot is simply sent for CD measurement, so the rework decision will be based on a larger data set than CD alone. The technology also reduces the number of lots requiring engineering intervention for dispositioning, reduces rework rate by eliminating false reworks, and reduces escaped lots.
Conclusion
MPX provides an alternative means for monitoring dose and focus on production wafers, with greater focus sensitivity than top-down CD-SEM. When comparing CD and MPX data from wafers with intentional variations in focus and dose, MPX shows superior capability for separating focus from dose. The CD data has poor correlation to the input dose due to the focus variation effects, and the correlation of CD to focus is worse. MPX correlates well to both dose and focus, and its response to each is highly predictable.
Implementing an MPX control strategy saves time in the overall lithography process loop, as the dose and focus information are obtained at the same time as the overlay data. In addition, the data provided by the technology proves more reliable for lot disposition than CD data in the case illustrated in this paper. Based on volume production data, MPX provides metrics both more stable and reliable than CD data for process monitoring. Even if CD data is indispensable for process control, the focus monitoring capability provided by MPX will prove invaluable as focus latitude decreases, and it adds little overhead to the overlay measurement.
Acknowledgments
Spansion would like to thank the contributors from Spansion and KLA-Tencor for their time, effort, and learning.
References
- Alexander Starikov, “Trends in Microscopy-Based Metrology for Dimension Control,” SI Webcast Presentation, June 23rd 2004.
- B. Eichelberger, B. Dinu, H. Pedut, “Simultaneous Dose and Focus Monitoring on Product Wafers,” Proc. of SPIE, Vol. 5038, pp. 247-254, February 2003.
Sean Hannon received his masters in mathematics from the U. of Texas at Austin. He is a senior product development engineer in the mask data preparation group, which processes all reticle data for AMD at Spansion LLC Fab 25, 5204 East Ben White Blvd., Austin, TX 78741; ph 512/602-1880, fax 512/602-5155, e-mail [email protected].
Brad Eichelberger received his bachelors in electronics engineering technology from DeVry Institute of Technology in Columbus, OH, and is a product marketing manager for the Optical Metrology Division at KLA-Tencor, One Technology Drive, Milpitas, CA 95035; ph 717/502-8279, e-mail [email protected].
Harold Kennemer received his bachelors in biological sciences from the U. of North Texas in Denton, TX. He is a senior manufacturing engineer at Spansion LLC, Fab25, Austin, TX.
Chris Nelson received his AS in electronics technology from Texas State Technical College in Harlingen, TX. He is a product marketing manager at KLA-Tencor.
Berta Dinu received her MSc in optics, spectroscopy, and plasma from the U. of Bucharest, Romania. She is a senior system engineer at KLA-Tencor.
This article contains some information originally presented in the Proceedings of the SPIE, Vol. 5752, pp. 1127-1136, 2005.