Issue



Handling and filtration evaluation of a colloidal silica CMP slurry


11/01/2004







Maintaining CMP slurry quality is critical for low defectivity in complex CMP processes [1–4]. Consequently, slurry "health" parameters are key considerations in distribution and dispensing systems. Accelerated aging and handling evaluations were conducted to determine the effects of extensive slurry handling using a vacuum-pressure dispense system and bellows pump for bulk delivery. Filtration tests also were conducted with the extensively handled colloidal silica-based slurry to develop recommendations for point-of-use and global distribution-loop filtration in production environments.

Laboratory characterization of chemical mechanical polishing (CMP) slurries provides valuable information on optimum handling and filtration essential to maintaining slurry quality or health during the slurry consumption cycle [5–7]. Accelerated handling and filtration experiments were conducted at the R&D Laboratory of BOC Edwards' Chemical Management Division to determine the effects of vacuum-pressure dispense system (VPDS) and bellows-pump recirculation on the long-term stability of Levasil 50CK/30%-V1 CMP slurry from H.C. Starck in a bulk delivery environment [8].

Levasil 50CK/30%-V1 is a colloidal silica-based, ready-to-use CMP slurry for metal polishing and post-metal buffing in IC manufacturing as well as surface finishing of hard disks. During accelerated handling evaluations in VPDS, the slurry passed through valves, sudden and gradual expansions/contractions, orifices, bends, and fittings. In these tests, the slurry was also exposed to humidified nitrogen and vacuum, as expected in a real fab environment. The slurry was recirculated using a BOC Edwards P2200 vacuum-pressure dispense slurry delivery system [3] in a 200 ft, 3/4-in. PFA tubing global distribution loop for 162 hr (~2770 turnovers @ 17.1 turnovers/hr). In a second test, the slurry was recirculated in a Nippon Pillar PE-20MAN double-bellows pump for 42 hr (~2520 turnovers @ 60 turnovers/hr). Therefore, the VPDS and pump tests were continued for a comparable number of turnovers. The point-of-use (POU) and global loop filtration experiments were conducted with the extensively handled slurry in the VPDS toward the end of the accelerated handling test.

The main objectives of this study were:

  1. Measure the comparative effects of extensive slurry handling using a VPDS and bellows pump on the typical slurry metrology/health parameters.
  2. Identify the sensitive measurement parameters to monitor and control the blend ratio of different constituents in the slurry, and to detect DI water fab leak.
  3. Determine required minimum flow velocity to maintain abrasive-particle dispersion.
  4. Observe loop shutdown and ease of abrasive-particle redispersion.
  5. Define stirring
    edispersion requirements for the source drum/pail and the daytank.
  6. Develop POU and global-loop slurry filtration recommendations for large-particle management.
  7. Determine cleaning and preventive maintenance requirements for the slurry delivery system.

This study showed that accelerated slurry-handling evaluations can provide useful insights for optimizing CMP slurry metrology and quality management [9–12], and also help identify a compatible slurry blending and distribution system (see Table 1 for test system specs).

Click here to enlarge image

Evaluations were conducted with Myron L. Co.'s Ultrameter Model 6P to measure pH, oxidation-reduction potential (ORP), conductivity, and total dissolved solids (TDS); a Mettler Toledo AG204 analytical balance with Lindberg/Blue M Model G01330A gravity oven to evaporate 2ml slurry sample to dryness (for 2 hr at 105°C) for weight percentage of solids (wt% solids) determination; a volumetric flask (50ml) with analytical balance to measure slurry density; a Brookfield Model LVDVI+ viscometer with Spindle-ULA (S00) at 60rpm to determine viscosity; and Particle Sizing Systems' AccuSizer Model 770A for measuring large particle counts (LPC).

In a typical production-scale fab, slurry will circulate through a bulk slurry-dispense system and fab piping approximately 100 times before being consumed. This number is based on typical fab slurry consumption, storage tank size, and flow rate through a global distribution loop (see Table 2).

Click here to enlarge image

null

Evaluation results

Typical CMP slurry health and quality monitoring and control parameters include wt% solids, viscosity, density, and LPC distributions. These parameters are presented for the VPDS and pump handling tests in Figs. 1–3, respectively. In general, extensive handling of this silica slurry in a VPDS showed insignificant changes in the normal slurry-health parameters during the typical turnovers slurry sees in fab operation (i.e., ~100 turnovers). It is important to note that in the VPDS tests presented, these turnovers were completed in the initial ~6 hr handling. There was no visible settling or thickening of the slurry in VPDS accelerated handling tests, despite having 26 times more turnovers than a typical fab operation.


Figure 1. Wt% solids variation during handling of Levasil 50CK/30%-V1 in a) a VPDS and b) a bellows-pump recirculation loop.
Click here to enlarge image

null


Figure 2. Viscosity and density variation during handling of Levasil 50CK/30%-V1 in a) a VPDS and b) a bellows-pump loop.
Click here to enlarge image

Wt% solids variation in Fig. 1a shows a slight increase during the 162 hr test, due to the limited quantity slurry exposure to repeated vacuum and possible evaporation without any slurry usage and replenishment. The increase is within experimental error for the initial 24 hr (~400 turnovers) continuous handling in VPDS. Similarly, the viscosity and density in Fig. 2a show consistent behavior during this test. Also, a very small increase in the cumulative LPC was observed (Fig. 3a) in the initial 24 hours of the VPDS test. This LPC behavior suggests that Levasil 50CK/30%-V1 slurry should not generate large particle agglomerates during VPDS handling turnovers expected in normal fab operation. When slurry is handled over an extended period without usage, it is possible to generate large particle agglomerates (i.e., an increase in LPC) and changes in the slurry's mean particle size.


Figure 3. LPC distribution during handling of Levasil 50CK/30%-V1 in a) VPDS and b) bellows-pump loop.
Click here to enlarge image

Similar to the VPDS test, the slurry recirculation in a bellows-pump loop showed insignificant changes in pH, ORP, conductivity, TDS, wt% solids, viscosity, and density (Figs. 1b and 2b). The pump test had a duration of 42 hr, compared to 162 hr in VPDS. The results did not show any significant change in the wt% solids, or in the slurry's settling or thickening. Similar to other silica slurries with high wt% solids abrasive content, this slurry does show sensitivity to shear and agglomeration due to repeated pump handling over an extended period of time. The increase in the cumulative LPC for the pump test (Fig. 3b) was much higher when compared with the VPDS system (Fig. 3a) for similar turnovers.


Levasil 50CK/30%-V1 and DI water blend tests indicate that conductivity, TDS, and wt% solids (or density) are good indicators to monitor and control the blend ratio in applications using diluted slurry [8]. Conductivity and TDS may not be used to monitor the blend ratio, however, as they are not controlled in the supply slurry of different lots. Also, pH, ORP, and viscosity do not vary significantly with blend ratio and would not be good indicators of blend change. Wt% solids and/or density may be used to monitor and control the slurry and DI water blend ratio (Fig. 4). These tests also suggest that wt% solids, density, conductivity, and TDS would be good indicators to detect a DI water fab leak, but viscosity, pH, and ORP would not.


Figure 4. Wt% solids and viscosity for Levasil 50CK/.30%-V1 and DI water blends.
Click here to enlarge image

The slurry was recirculated at different flow velocities in the simulated global loop. As with other colloidal silica slurries, Levasil 50CK/30%-V1 has very slow settling characteristics [8]. A minimum flow velocity of 0.5 ft/sec in 3/4-in. tubing (~0.5gpm, ~2 liters/min) was sufficient to maintain total wt% solids variation within ±0.2%. If the slurry must recirculate over extended periods in a complex loop, it may be desirable to maintain a minimum flow velocity of 1.0 ft/sec (~1.0gpm, ~4 liters/min) and to implement periodic recirculation when the slurry is not used or replenished to limit total slurry turnovers in the global distribution loop before it is consumed.

This slurry may only need slow, intermittent mixing using a drum mixer or 1–2 turnovers using a pump for the supply-slurry drum initial homogenization. High-speed, extended mixing, and exposure to air should be avoided. Distribution recirculation should be sufficient to maintain slurry homogeneity, while continuous or intermittent slow mixing in the daytank is optional. No settling of the abrasive particles was visually observed in the global loop after a 15 hr simulated fab shutdown. The wt% solids measurements confirmed that there was no variation in the solids content, beyond experimental uncertainty, during restart after the shutdown.


Figure 5. Planargard CM11 distribution-loop filter cumulative LPC results (slurry flow = 1.5gpm).
Click here to enlarge image

Global distribution loop and POU filtration characterization experiments (Figs. 5 and 6, respectively) suggest that the Mykrolis Solaris-03 POU filter (3µm nominal rating) can be used to reduce the cumulative LPC in this slurry. The data show that the Solaris-03 filter reduced by ~70% the cumulative LPC ≥1.01µm with a POU flow rate of 200ml/min. As expected, in-line global recalculation loop filtration tests showed minimal large particle removal with a relatively open Planargard CM11 graded-density depth media filter. CM11 was used in our evaluation, considering the high wt% solids content and the slurry's extensively handled condition, and to get extended filter lifetime in a limited-volume test. More retentive filtration with Planargard CMP9 or CMP7 filters in the distribution loop can provide tighter retention and further reduction in large particle counts.

Conclusion


Figure 6. Solaris-03 POU filter cumulative LPC results (slurry flow = 200ml/min).
Click here to enlarge image

Accelerated handling testing of Levasil 50CK/30%-V1 colloidal silica-based CMP slurry in VPDS and bellows-pump loops showed insignificant changes in the slurry-health monitoring parameters. The slurry did not generate large agglomerates during the typical turnovers expected in a fab operation. Similar to other silica slurries, however, this slurry also showed sensitivity to shear and agglomeration when subjected to extended VPDS or bellows-pump handling well beyond what is likely in production fabs. The increase in cumulative LPC for the pump test was much higher than LPC increases produced by the VPDS test.

The slurry POU filtration test with Mykrolis' Solaris-03 filter showed ~70% reduction in cumulative LPC ≥1.01µm. The global loop filtration resulted in minimal LPC reduction with a relatively open Planargard CM11 filter. More retentive filtration can be achieved by using Planargard CMP9 or CMP7 filters in the distribution loop.

The slurry shows very slow settling characteristics and needs minimal mixing to maintain abrasive suspension and redispersion. Density and wt% solids can be used to monitor and control this slurry blend with DI water and to detect DI water fab leak. This study shows the suitability of a VPDS for handling Levasil 50CK/30%-V1 slurry over an extended period while maintaining slurry quality.

Acknowledgments

The authors thank Stephan Kirchmeyer and Gabriele Hey of H.C. Starck for reviewing the original manuscript, and H.C. Starck for providing the slurry samples. Solaris and Planargard are registered trademarks of Mykrolis Corp. Trademarks and their respective owners are: AccuSizer, Particle Sizing Systems Inc.; Levasil, H.C. Starck; Pillar, Nippon Pillar Packing Co. Ltd.; and Ultrameter, Myron L. Co.

References

  1. M. Moinpour, A. Tregub, A. Oehler, K. Cadien, "Advances in Characterization of CMP Consumables," MRS Bulletin, Vol. 27 (10), pp. 766–771, 2002.
  2. R.K. Singh, C. Patel, G. Conner T. Towle, R. Viscomi, et al., "Efficient Filtration of New-generation CMP Slurries: Challenges and Solutions," Semiconductor Manufacturing, Vol 5 (6), pp. 70–84, 2004.
  3. B. Johl, R.K. Singh, "Optimum Process Performance Through Better CMP Slurry Management," Solid State Technology, pp. 63–66, Aug. 2003.
  4. R.K. Singh, "Challenges and Solutions for Current and Next Generation CMP Slurry Filtration," Mykrolis Application Note AN1041ENUS, Aug. 2004.
  5. R.K. Singh, B.R. Roberts, "On Sedimentation and Redispersion of Abrasive Particles in CMP Slurries," Proc. 6th Int. CMP-MIC, pp. 441–448, 2001.
  6. R.K. Singh, B.R. Roberts, "On Extensive Pump Handling of CMP Slurries," Proc. 12th Annual IEEE/SEMI ASMC, April 23–24, 2001.
  7. K. Derivendt, et al., "CMP Defectivity and Slurry Filtration," Proc. MRS Meeting, Apr. 1999.
  8. R.K. Singh, B.R. Roberts, "Handling Characteristics of Levasil 50CK/30%-V1 CMP Slurry in BOC Edwards Vacuum Pressure Dispense Slurry Delivery System and a Bellows Pump Loop," BOC Edwards Technical Report CMD1115/TR/0804, Aug. 2004.
  9. R.K. Singh, B.R. Roberts, "CMP Slurry Metrology: Various Approaches," Proc. 2nd Intl. AVS ICMI Conf., paper CM-TuM12, Feb. 2001.
  10. R.K. Singh, B.R. Roberts, "Behavior of CMP Slurry Properties in Continuous Blending and Distribution Systems," Proc. 17th Intl. VLSI Multilevel Interconnection Conf., pp. 545–547, 2000.
  11. R.K. Singh, B.R. Roberts, R. Viscomi, M. Maxim, M. Diaz, et al., "Behavior of EP-C600Y-75B Copper CMP Slurry Under Extensive Handling," Proc. 8th Intl. CMP-MIC, Feb. 2003.
  12. K. Nicholes, R.K. Singh, D.C. Grant, M.R. Litchy, "Measuring Particles in CMP Slurries," Semiconductor International, pp. 201–206, July 2001.

For more information, contact Rakesh K. Singh at Mykrolis Corp., 129 Concord Rd., Billerica, MA 01821; ph 978/436-6556; e-mail [email protected].

Rakesh K. Singh, Gregg Conner, Mykrolis Corp.,
Billerica, Massachusetts
Benjamin R. Roberts, BOC Edwards, Santa Clara,
California