分享

Customer Case Study

 Rivalry 2017-09-28

Chemical mechanical planarization/polishing (CMP) processing is affected by various factors, including consumables, process parameters, and polisher hardware design. Among all these factors, the wafer carrier is particularly important, because it directly affects the performance of the wafer-polish process (e.g., across-wafer uniformity, ovoval rate stability). This article describes the significant process improvement obtained when using an Applied Materials Mirra CMP system equipped with that company’s 150mm Titan Profiler or 200mm Titan Contour wafer-carrier heads, versus a polisher with a basic carrier. The wafer-edge profiles and polish-rate stability will be compared between the polishers using different wafer-carrier heads. The polisher utilization/capacity and CMP cycle time in production will also be compared.

Wafer-Edge Profiles

When Qorvo’s 150mm Bulk Acoustic Wave (BAW) product development first started, an oxide CMP process was developed using a conventional polisher with a basic wafer-carrier design (hereafter referred to as CMP-a). This carrier has a single pressure zone with a number of vacuum holes in the center area of the carrier plate. These vacuum holes are used not only for wafer handling but also for applying extra pressure to the center area of the wafer backside. The center pressure adjustment is used as a process knob to dial in the wafer center removal. This type of carrier has worked well in production to address cross-wafer uniformity, but the design lacks robust control on wafer-edge removal profiles.

A common factor in this, and all CMP carriers, is a wafer- retaining ring to keep the wafer inside the carrier during polishing. This ring has a fixed position in the carrier, determined by the wafer protrusion required for a stable process. In a normal setup the retaining ring is not flush with the wafer surface, and does not contact the CMP pad during polishing. Due to this design feature on such basic carriers, wafer-edge control is managed by the wafer-protrusion setting. This was proven to have limited control over the wafer-edge removal profiles. Since the retaining ring is designed to be stationary and cannot touch the polish pad, it cannot keep the pad flat around the wafer-edge area during polishing. Therefore, the wafer edge protruding out of the carrier is exposed to the rotating and compressed CMP pad, and is susceptible to the adverse eects of elastic pad rebound. This leads to wafer-edge uniformity/profile fluctuations, as shown in figure 1. The measurement site 1 in the delta radius scan profiles represents wafer edge at 3mm edge exclusion, while site 25 represents the wafer center. The profile fluctuations at the wafer edge are a reflection of pad rebound effects.


Figure 1. Delta radius scan profiles of oxide pilot wafers polished on CMP-a.

Using the same consumables (polish pad, conditioning disk, slurry, oxide pilot type/thickness), process specifications (removal rate, within-wafer non-uniformity) and measurement tools/recipes as data collected on the said CMP-a carrier, a new CMP process has been developed on Applied Materials Mirra polishers (hereafter referred to as CMP-b). These polishers are equipped with 150mm Profiler and 200mm Contour wafer-carrier heads, respectively. Unlike the CMP-a wafer carrier, these two carriers have multiple pressure zones designed to provide a better capability for wafer-profile tuning. The retaining rings in these carriers can also be controlled by a pressure channel in the ring. This allows the ring to stay flush with the wafer surface and to contact/press the polish pad during polishing. When the retaining-ring pressure and the carrier-edge-zone pressure are properly configured, the adverse effects of pad rebound on the wafer edge are greatly reduced or eliminated. As a result, wafer edge profiles are more stable, flat and controllable, as illustrated in figure 2. Some minor fluctuation in the profile traces still exists around the wafer edge relative to the center, but compared to the profiles demonstrated in figure 1, these edge profile fluctuations are minimized and negligible.


Figure 2. Delta radius scan profiles of oxide pilot wafers polished on CMP-b.

Applied Materials 150mm Titan Profilers feature three concentric zones of control.

Polish Rate Stability

An advantage of the carrier head used in the CMP-a polisher is its basic design and ease of operation; however, it has limitations in wafer-center and -edge polish control. As a result, the CMP-a polish process used in our BAW production line requires daily qualification to ensure stability in both the process removal-rate and non-uniformity. This daily process qualification involves small adjustments in polish-pressure settings. For example, the center pressure applied on the wafer backside usually needs to be increased by up to 0.4 psi in order to maintain a wafer-center removal-rate equivalent to that around the wafer center. The primary polish pressure may also need to be slightly increased to keep the overall removal-rate within specifications, especially with increasing wafer count on a polish pad. Both the backside and primary polish-pressure settings have windows for adjustment, and finding an appropriate pressure ratio for the daily process qualification is critical to the stability of the removal-rate and within-wafer non-uniformity. However, this process qualification leads to day-to-day removal-rate changes from the required pressure-setting adjustments. Removal-rates of all daily process qualifications performed over a pad life generally follow a gradually declining trend within the specification window.

In contrast, the carrier heads used in CMP-b polishers employ a more complex hardware design, but have superior performance in both profile control and process stability. The CMP-b polish process also remains stable with similar removal-rates and non-uniformities throughout the pad’s life, without the need for adjustments in process parameters over the pad’s life and after consumable changes. Such stability also removes the need for a daily qualification on the CMP-b process. Process qualification on the CMP-b setup has been reduced to a semiweekly interval, mainly to check process performances on all four carrier heads and obtain more accurate removal-rate readings from each of these individual heads. The removal-rate qualification data from the two types of polishers over a one-year period is plotted in figure 3. Long-term stability in the CMP-b process removal-rate data is demonstrated by the narrower variation as compared to CMP-a. Pad lifetime is a major factor in the wider variability in removal-rate data in the CMP-a process (i.e., higher rate with a newer pad, decreasing throughout the pad’s life), as evidenced by the downward trend in the CMP-a data over time.


Figure 3. Comparison of CMP-a and CMP-b removal-rate qualification data.

Step Height Contour Plots

In order to compensate for the larger removal-rate variations in the CMP-a process, a look-ahead (L/A) wafer is polished from each production lot before committing the rest of the wafers in the lot. The polished L/A wafer provides a more accurate on-product polish rate for process-time calculation, ensuring accuracy of the polish process relative to the desired step-height target. Daily process qualification coupled with the L/A methodology works well with the CMP-a process, delivering satisfactory process performance/product quality with the fixed polish- time approach. Shown in figure 4 is the step-height contour plot of a typical product wafer polished on CMP-a, which meets the step-height specification on all measurement sites. As previously mentioned, the CMP-b process has superior process stability with more advanced wafer-carrier heads as compared to CMP-a. This not only makes daily process qualification unnecessary on CMP-b, it also eliminates the need for an L/A polish step on production material. A production lot therefore is polished with a calculated process time based on the semiweekly process-qualification removal-rate. Figure 5 is the step-height contour plot of a typical product wafer polished on CMP-b, which also meets the step-height specification at all of the measurement sites. The comparison of step-height statistics in figures 4 and 5 demonstrates that the CMP-b process performance is superior on production wafers, with significantly lower across-wafer step-height variation (e.g., standard deviation and high/low variance).


Figure 4. Step-height contour plot of a production wafer polished on CMP-a.


Figure 5. Step-height contour plot of a production wafer polished on CMP-b.

 


The Applied Materials Titan Contour 200mm head provides five concentric zones of control.

Conclusion

In summary, both CMP-a and CMP-b polish tools and processes have been used in volume BAW production, meeting the same process-specification requirements. Due to the hardware limitations of the carrier heads in CMP-a, the process is required to maintain a daily qualification schedule, which consumes approximately 7 hours per week per polisher. Comparatively, the CMP-b process requires only a semiweekly qualification, which consumes only 2 hours per week per polisher. Thus CMP-b tool utilization in production is higher than that of CMP-a by about 5 hours per week per polisher. CMP-a also requires an L/A wafer-polish step on every production lot, adding 30 minutes per lot to overall process time. Historically the L/A wafer-polish methodology has proven adequate for polishing product wafers to the desired target; however it significantly reduces polisher capacity/utilization in production. Because the CMP-b process does not require the L/A wafer polish step, the cycle time is reduced by about 40% over CMP-a.

This improvement in cycle time is key to high-volume manufacturing, allowing increased wafer throughput, faster wafer turns, and cost avoidance in additional tooling to maintain equivalent capacity. CMP-b has many significant advantages over CMP-a to support the continuous ramp of BAW production, given the improvements in process quality as well as polisher utilization and capacity.

For additional information, contact Yanghua.He@Qorvo.com

ABOUT THE AUTHORS
Yanghua He is senior process development engineer, CMP for Qorvo, Inc.; Michael Lube serves as Qorvo’s CMP process section manager. Both are based in Richardson, Texas.

(Sidebar to Qorvo Article): Higher Throughput and Yield with Applied Mirra CMP Contour Upgrades

Applied Materials Mirra CMP systems, first introduced in 1995, provide production-proven, high-performance 150mm and 200mm planarization solutions for silicon, shallow trench isolation (STI), oxide, polysilicon, tungsten, and copper damascene applications. The integrated post-CMP Mesa cleaner, also available for 150mm and 200mm, e™ffectively removes slurry, preventing residue formation and minimizing particles and water marks.

Advanced polishing technologies include the Applied Materials Titan Contour (200mm) and Titan Profiler (150mm) head products. Both are available as upgrade options for existing Mirra CMP tools, and help customers achieve higher throughput and yield.

The Titan Contour 200mm head is based on Applied’s production-proven 300mm Contour heads, and features five independent, concentric pressure zones plus a retaining-ring zone to allow full control of wafer-profile tuning. The 200mm Contour o™ffers best-in-class performance on 200mm processes, including challenging Cu, STI, and ILD polishes.

Scaled from the 200mm Titan Contour design, the 150mm Titan Profiler features three independent, concentric pressure zones and a retaining-ring zone to allow control of wafer-edge profile tuning. TheTitan 150mm Profiler o™ers best-in-class performance on 150mm processes, including challenging GaN, thick oxide and SOI polishes.

For additional information, contact Jamie_Leighton@amat.com

    本站是提供个人知识管理的网络存储空间,所有内容均由用户发布,不代表本站观点。请注意甄别内容中的联系方式、诱导购买等信息,谨防诈骗。如发现有害或侵权内容,请点击一键举报。
    转藏 分享 献花(0

    0条评论

    发表

    请遵守用户 评论公约