Enhanced CMP Slurry Filter Offering Low CoO

Each slurry formulation requires specific filtration technologies to ensure proper purity at the point of dispense onto the wafer.

Chintan Patel, Entegris Inc., Billerica, Mass. -- Semiconductor International, 10/1/2007

Chemical mechanical planarization (CMP)¹ for IC fabrication was researched in the 1980s² as an expansion of glass polishing³ used in 1927. CMP was first used for polishing tungsten interconnects and silicon dioxide interlayer dielectrics (ILDs), but has enabled other key process developments, including shallow trench isolation (STI), pre-metal dielectrics (PMDs), polysilicon, copper interconnects and porous low-k ILDs.⁴ In the past few decades, CMP has evolved as a true enabling technology for IC chip fabrication processes, and its importance has been widely recognized by IC chip designers and fabricators.

As the technology node size decreases, the demand for CMP steps increases. CMP slurry formulation and filtration play a critical role in ensuring the quality of CMP process output. Focusing on the fab environment, CMP slurry filtration at the point-of-use (POU) has been proven to be beneficial in reducing wafer defects associated with large particles.⁵ Costs associated with filtration contribute toward the overall CMP process consumable cost. With the increasing number of CMP steps, the consumable costs associated with the CMP process also increase. As fabs become more capital-sensitive, lowering the consumable costs becomes essential. This study is focused toward achieving lower cost-of-ownership (CoO) relative to CMP filtration without sacrificing the effective particle removal performance.

There are various options available for using POU filtration. From the CMP process point of view, to achieve the target planarization performance with minimal defects, it is important to control the slurry particle distribution and slurry dispense volume on the wafer during the CMP process. Slurry particle distribution is maintained with the POU filtration approach. By prematurely removing the filter, the end user can experience a higher cost associated with filter replacement and tool downtime. An optimum filter lifetime detection method is recommended through which the filter is used for maximum length of time without a negative impact on wafer defect and polishing performance. To ensure precise slurry dispense, end users are implementing newer devices at the POU branch, such as flow controllers, pressure transducers and dispense pumps. Considering these devices consume pressure, a new POU filter capable of achieving particle removal performance at a lower pressure for a longer period of time is introduced.

Experimental procedure

Filter resistance to liquid permeation is measured in terms of pressure drop at a specific flow rate. In this study, filter resistance or pressure drop was measured with two chemicals, first using deionized (DI) water and second using actual silica-based CMP slurry.

Particle removal efficiency is described in terms of retention. The definition of retention is the percentage of particles retained by the filter at the particular particle size. Generally, we display this value as a retention curve, but for this study retention is displayed in terms of a single point at a particular particle size. Retention data was first generated using polystyrene latex (PSL) beads in a single-pass filtration mode. These are standard beads that can be accurately and precisely counted and sized. PSL beads are also most commonly used as calibration standards for calibrating liquid particle counters. Particle counts before and after filtration were measured using a particle counter. New sets of filters were used to measure retention with actual silica-based slurry. Slurry large particle measurement was performed using an automatic particle sizer.

A special test to measure the relative lifetime of a filter was performed to establish the relative lifetime on an actual CMP slurry. The silica-based slurry is recirculated through multiple filters installed in parallel. The pressure upstream and flow downstream of the filters are monitored, and filter life is measured by monitoring filter pressure drop at a constant outlet flow. The slurry is modified with a special method to create an artificial gel to accelerate the filter plugging. Slurry samples from downstream of the filter are collected at a regular interval. These samples were analyzed for mean particle counts. Slurry mean particle were measured using a mean particle analyzer.

Results and discussion

Non-woven polypropylene depth media is widely used for industrial filtration applications, and is very well-suited for the CMP slurry filtration application. CMP slurry filtration is not absolute; rather, it is a size-separation process.⁶ For CMP slurry filtration, the key is to remove the large particles from the high population of small or working particles at an acceptable pressure drop. For such an application, non-woven polypropylene depth media is an ideal choice. The filter designs discussed in this article contain non-woven polypropylene media fabricated using a melt blown process.⁷

1. PSL beads particle removal efficiency against DI water flow resistance.

Typically depth media is rated on the removal efficiency of the industry standard test dust, such as JIS (Japanese Industrial Standard) test dust, ACFTD (air cleaner fine test dust) or the ISO standard test dust. The filter retention value or particle removal efficiency generated using these test dust standards are typically of a very high value. Dust particle distribution may not be an ideal representation of the CMP slurry particle distribution. For this study, particle removal efficiency results were generated based on using PSL beads.

2. Silica slurry particle removal efficiency against slurry flow resistance.

Figure 1 compares Entegris’ Solaris 01 and the new Solaris CL design based on PSL beads retention and flow resistance. A similar comparison is shown in Figure 2, but the retention and flow resistance data is generated using actual silica-based CMP slurry. Both graphs indicate that the Solaris CL design is capable of achieving similar particle removal efficiency at much lower flow resistance. If tighter filtration is required, Solaris CL 0.7 is also available with lower flow resistance than the Solaris 01. Particle removal efficiency trends obtained for three different filter designs were analogous with PSL beads and with actual silica slurry.

3. Silica slurry mean particle distribution over the lifetime of Solaris CL1.0.

During the relative filter lifetime test, slurry samples were collected from downstream of the Solaris CL1.0 at different time intervals. Mean particle distribution of these slurry sample is plotted against the pressure drop at the time of sample collection in Figure 3. As Solaris CL1.0 gets plugged or loaded with particles over time, there is no negative impact on slurry mean particle size. This suggests that filter plugging should not have any negative impact on polishing performance (MRR) because the slurry mean particle distribution remains unchanged.

4. Pressure drop increase in Solaris 01 during relative filter lifetime test with silica slurry.

Figures 4 and 5 display the pressure drop profile of Solaris01 and Solaris CL1.0, respectively, as they are plugged with silica slurry over time. There is a rapid increase in pressure drop values with the Solaris01 design, while the pressure drop increase in Solaris CL1.0 is slower at the beginning and increases gradually. In processes with a 10 psi pressure budget, Solaris01 plugging will have higher impact on downstream flow than Solaris CL1.0 design because the rate of pressure drop increase in Solaris CL 1.0 is slower.

5. Pressure drop increase in Solaris CL1.0 during relative filter lifetime test with silica based slurry.

Summary

For filter lifetime monitoring, the key is to have precise sensitivity to filter loading. At a constant flow, pressure drop monitoring serves as a good indicator of filter life and, at a constant pressure, flow monitoring serves as good indicator. For CMP applications, maintaining precise slurry dispense is a primary goal. Because maintaining constant flow throughout the filter life is more critical, monitoring filter pressure drop should be used. Figure 6 shows appropriate locations for pressure sensors and flow sensors to be installed in the recirculation loop and POU loop for monitoring precise filter lifetime and ensuring accurate process condition. By analyzing historical pressure drop data, accurate filter lifetime can be measured. A filter can be used to its maximum lifetime before it reaches the saturation point with loading and has an impact on the flow rate decrease. By monitoring the pressure drop behavior, estimation on slurry condition can also be obtained.

6. Typical recirculation and POU loop (a). Suggested recirculation and POU loop for precise filter life monitoring (b).

Without pressure drop monitoring, the direct impact of a decrease in system pressure or filter plugging can be seen on the dispense rate. This could result in scrapped wafers. Pressure drop and flow signals can be also utilized as a part of the process control loop for maintaining a constant flow rate during the planarization process. By implementing this recommended loop, end users can adjust their system pressure if needed. Once the POU branch is operating within sufficient pressure, filters can be used for a longer period of time. This can reduce the change-out frequency; hence, minimizing the possible untimed tool down cost. In addition to implementing these steps, if you add a filter product, end users can further enhance filter lifetime, which will enable meeting the lower cost-of-ownership model.

References

1. F. Kaufman et al., J. Electrochem. Soc., 1991, Vol. 138, p. 3460.
2. S. Wolf and R.N. Tauber, Silicon Processing for the VLSI Era, 2000, p. 742.
3. F. Preston, J. Soc. Glass Tech., 1927, Vol. 11, p. 214.
4. R.K. Singh and R. Bajaj, “Advances in Chemical Mechanical Planarization,” MRS Bulletin, 2002, Vol. 27, No. 10, p. 743.
5. G. Vasilopoulos, Z. Lin, B. Johl, S. Joshi. and B. Chatterjee., “Copper CMP Defect Reduction Using POU Filtration,” Proc. of CMP Tech. for ULSI Interconnection, SEMICON West 2000, p. S-1.
6. R.K. Singh et al., “Efficient Filtration of New-Generation CMP Slurries: Challenges and Solutions,” Semiconductor Manufacturing, 2004, Vol 5, No. 6, p. 70.
7. E. Mayer and H.S. Lim, “New Non-woven Microfiltration Membrane Material,” Fluid/Particle Separation Journal, 1989, Vol. 2, No. 1, p. 17.

About the Author

Chintan Patel is responsible for the development and launch of chemical mechanical planarization (CMP) filtration products as a product manager of the Liquid MicroContamination strategic business unit for Entegris Inc. Prior to this position, Patel was a senior application development engineer managing CMP applications. Patel joined Entegris in 1999. Prior to joining Entegris, Patel was with Cabot Corp., where he served as a research engineer, developing catalyst for fuel cell membrane electrode assembly (MEA). He also held a process engineer position in the area of electro-optics with Corning-Lasertron. Patel has a B.S. in chemical engineering from the University of Massachusetts.