Measuring Particles in CMP Slurries
Kristi Nicholes and Rakesh Singh BOC Edwards, Chaska, Minn. Don Grant and Mark Litchy CT Associates Inc., Bloomington, Minn. -- Semiconductor International, 7/1/2001
| At a Glance |  | ||
|
| ||
| |||
Control of CMP processes has been hindered by the inability to detect slurry damage before it adversely affects product yield. For years, slurry pH, conductivity and total solids have been monitored during CMP processes.1 More recently, CMP slurry analyzers have been used to characterize slurry particles in terms of concentration and size distribution.2 This study was undertaken to determine which slurry characteristics change when slurry is circulated and which instruments can most sensitively detect the onset of change. Effective slurry monitoring depends greatly on the type of analytical instrument, its sensitivity, repeatability and resolution.
Materials and methods
Six slurry analyzers were evaluated (Table 1), along with a description of the slurry characteristics they report and the principles used to measure them. All were tested off-line, although all (except the LS230 and NICOMP 380/ZLS) are designed for on-line use. The slurry used was Cabot Semi-Sperse 12, an oxide CMP slurry containing silica particles.
The first step was to determine situations that lead to changes in slurry characteristics or negatively affect slurry "health." To assess the damage caused by circulation and drying, parallel tests were conducted in three systems. In two parallel systems, a centrifugal and a diaphragm pump were used to circulate slurry from a humidified tank through a distribution loop. In the third system, circulated with a diaphragm pump, compressed dry air (CDA) was passed through the slurry in the tank using a sparger to accelerate change caused by drying.
The slurry was circulated for seven to 10 days to force measurable changes in characteristics. Typically, eight samples were withdrawn over the test period and analyzed within one hour. The measurements were repeated three to five times for each sample to determine sample variability.
| Table 1. Slurry Health Analyzers Evaluated | ||
| Analyzer (manufacturer) | Parameters analyzed | Particle measuring principle |
| LiQuilaz-S05 (Particle Measuring Systems) | Cumulative particle number; particle concentration at 15 different sizes of ³0.55-10 µm | Single-particle light scattering |
| AccuSizer 780/OL (Particle Sizing Systems) | Cumulative particle number; particle concentration at 15 different sizes of ³0.56-10 µm | Single-particle light scattering and light extinction |
| LS230 (Beckman Coulter) | Cumulative particle volume fraction at sizes of 0.04-5 µm | Laser diffraction and light scattering |
| NICOMP 380/ZLS (Particle Sizing Systems) | Mean particle size; standard deviation of PSD; zeta potential | Dynamic light scattering |
| SlurryChek (Particle Measuring Systems) | Extinction coefficients at 17 different wavelengths of 200-1000 nm (wide cell only) | Multiwavelength light extinction |
| Lab CMP slurry monitor (Colloidal Dynamics) | Attenuation and dynamic mobility spectra at frequencies of 1-18 MGz, pH, conductivity, temperature, zeta potential, median diameter, and geometric standard deviation | Electroacoustics |
Data analysis
Slurry analyzers use a variety of detection principles, so their outputs cannot be compared directly. Hence, all output data were converted to a uniform, dimensionless parameter: relative signal strength (RSS), defined as the ratio of the change in instrument response caused by circulation to the variability of the base slurry measurement. The RSS is calculated as:
| 1. Results obtained with the Lab CMP slurry monitor analyzer during measurement of the sparged slurry in the diaphragm pump system show statistically insignificant changes in pH, conductivity and zeta potential. |
The variability of the base slurry measurement was defined as 3× the measurement standard deviation (3 s). A detailed explanation of RSS can be found elsewhere.3 RSS analysis was performed for all of the parameters listed in Table 1. Parameters listed as a range were analyzed at multiple values. For example, cumulative particle concentrations were analyzed at 15 different particle sizes for the LiQuilaz-S05 analyzer.
Results
Several parameters indicated no measurable change even though the slurry was cycled for more than 8000 turnovers. Figure 1 presents some results obtained with the Lab CMP slurry monitor analyzer during measurement of the sparged slurry in the diaphragm pump system. The error bars represent ±3 standard deviations. Figure 1 demonstrates that, as a slurry is used, changes in pH, conductivity and zeta potential are statistically insignificant. These three parameters, then, make poor choices as slurry health indicators.
Table 2 presents the parameters for the sparged slurry in the diaphragm pump system that did not change with increasing turnovers of the slurry. The maximum RSS for each parameter is listed. The RSS must be greater than 1 for a change to be detected. An RSS of less than 1 indicates that the change in instrument response was less than the variability of the base slurry measurement. All of the RSSs in Table 2 were less than 1. Similar results were observed for the other distribution systems.
| Table 2. Parameters That Did Not Significantly Change in >8000 Turnovers | |||
| System | Instrument | Signal | Max. RSS |
| Diaphragm sparged | Lab slurry monitor | Conductivity | 0.71 |
| Zeta potential | 0.21 | ||
| pH | 0.25 | ||
| Attenuation coefficient | 0.81 | ||
| NICOMP 380/ZLS | Zeta potential | 0.20 | |
Only parameters related to particle characteristics — primarily particle size distribution — changed when circulating slurry. For each instrument, parameters and operating conditions that gave the overall best RSS in all circulation systems were used in the instrument comparison (Table 3).
| Table 3. Summary of Parameters Used in RSS Analysis | |
| Instrument | Signal |
| LiQuilaz-S05 | Cumulative particle concentration ³0.77 µm |
| AccuSizer 780/OL | Cumulative particle concentration ³0.56 µm |
| SlurryChek | Wavelength 750 nm |
| LS230 | Cumulative volume fraction at 1 µm |
| Lab CMP slurry monitor | Frequency 5.4 MHz |
| NICOMP 380/ZLS | Mean diameter (nm) |
RSS analysis allowed comparison of the instruments' abilities to detect change in circulating slurry (Fig. 2). The x-intercept of each linear regression (RSS=1) represents each instrument's detection sensitivity; that is, the fewest number of turnovers at which the instrument could detect change. The detection sensitivities of each instrument in each of the three systems are listed in Table 4.
| Table 4. Summary of Detection Sensitivities | |||
| Instrument | Detection sensitivities (turnovers to measurable signal) | ||
| Diaphragm sparged | Diaphragm humidified | Centrifugal humidified | |
| LiQuilaz-S05 | 4.0 | 0.77 | 9.7 |
| AccuSizer 780/OL | 17 | 2.1 | 1000 |
| SlurryChek | 52 | 18 | 220 |
| LS230 | 500 | 420 | 1800 |
| NICOMP 380/ZLS | 2800 | 2200 | 3200 |
| Lab CMP slurry monitor | >8000 | 3200 | >8000 |
The sensitivities of the instruments varied by more than four orders of magnitude. The LiQuilaz-S05 and AccuSizer 780/OL were the most sensitive instruments for the diaphragm pump systems, followed by the SlurryChek analyzer. For the centrifugal pump system, the LiQuilaz-S05 analyzer was the most sensitive followed by the SlurryChek and the AccuSizer 780/OL analyzers. However, the slope for the LiQuilaz-S05 data in the centrifugal system was very different from slopes for all other instruments in all systems. If the slope were similar to the others and weighted to the RSS values at high turnovers, then the sensitivities of the LiQuilaz-S05, AccuSizer 780/OL, and SlurryChek instruments would all be similar.
In general, the two optical particle counters — the LiQuilaz-S05 and the AccuSizer 780/OL — were the most sensitive, with the LiQuilaz-S05 analyzer being somewhat more sensitive. The difference between the sensitivities of the two instruments is believed to be a result of the dilution systems used to supply sample to the sensors. The LiQuilaz-S05 sensor used a laboratory dilution system while the AccuSizer sensor used a patented autodilution system. The laboratory system achieved lower measurement variability than the autodilution system, as shown in Table 5. The AccuSizer 780/OL analyzer's higher variability means that a larger change is required before a signal can be detected by this analyzer. If the AccuSizer 780/OL sensor were to be used with the laboratory dilution system, similar sensitivities would be expected. In fact, because the AccuSizer 780/OL sensor uses light extinction rather than light scattering to size large particles, it is expected to provide better sensitivity than the LiQuilaz-S05 sensor for particles greater than ~1.5 µm.4
| Table 5. Comparison of Optical Particle Counter Measurement Variables | |||
| Instrument | Baseline measurement (#/mL) | Measurement variability 3 s (#/mL) | Relative variability (variability/baseline) |
| LiQuilaz-S05 (³0.77 µm) | 34,479 | 1310 | 3.8% |
| AccuSizer 780/OL (³0.56 µm) | 427,776 | 52,831 | 12.4% |
Summary
| 2. Using RSS analysis, slurry health analyzers are compared in systems with a diaphragm pump with humidified nitrogen blanket (a), centrifugal pump with humidified nitrogen blanket (b), and diaphragm pump with air-sparged slurry (c). |
The LiQuilaz-S05 optical particle counter was the most sensitive analyzer tested. The AccuSizer 780/OL was the second most sensitive analyzer in the diaphragm pump systems, and the SlurryChek analyzer had intermediate sensitivity in all three systems. The difference between the sensitivities of the LiQuilaz-S05 and AccuSizer 780/OL analyzers was attributed to the dilution systems used.
CMP particle size analyzers can provide a good basis for monitoring slurry health. With automated and in-line slurry sampling systems, handling variability can be reduced, yielding process-important data. It has been demonstrated in the production environment that changes in slurry particle size monitoring data correspond to changes in the slurry distribution system such as filter swaps. These correlations provide valuable information about the health of the slurry as it reaches the CMP tools for processing.
Kristi Nicholes is a process engineer for BOC Edwards' Chemical Management Division (CMD). Her responsibilities focus on developing new technology in slurry and high-purity blending and distribution, as well as slurry metrology. She holds a B.S. in chemical engineering from the University of Washington.Phone: 1-952-556-2700, x.2631
e-mail: kristi.nicholes@edwards.boc.com
Rakesh K. Singh is a manager of BOC Edwards CMD's Santa Clara R&D laboratory. He is responsible for CMP slurry and ultrapure chemical handling systems. He has more than 20 years of industrial and research experience in process pumps, complex flow characterization, turbulent mixing and CMP slurry characterization. He has a B.E. in mechanical engineering, an M.E. in design of process machines, an Advanced Diploma in management, and a Ph.D. in fluid dynamics.
Phone: 1-408-496-5411, x.2210
e-mail: rakesh.singh@edwards.boc.com
Don Grant, president of CT Associates, performs R&D in contamination control, particle measurement and control, filtration and chemical engineering. He has more than 25 years of experience in analysis and purification of fluids. He has an M.S. in mechanical engineering from the University of Minnesota, and a B.S. in chemical engineering from Case Western Reserve University. He is a recipient of the Maurice Simpson Award from the Institute of Environmental Sciences.
Phone: 1-952-944-4766
e-mail: don@ctassociatesinc.com
Mark R. Litchy is a research engineer for CT Associates, and has several years of experience in particle measurement and control in both high-purity liquid chemicals and gases as well as surface contamination and cleaning. He has an M.S. in mechanical engineering from the University of Minnesota, and a B.A. in physics from St. John's University.
Phone: 1-952-944-4773
e-mail: mark@ctassociatesinc.com
REFERENCES
- J. Bare, T. Lemke, "Parameters for Monitoring CMP Slurry Stability and Contamination," Proc. of SEMICON West Contamination in Liquid Chemical Distribution Systems Workshop, July 1997.
- J. Bare, B. Johl, T. Lemke, "Comparison of Vacuum-Pressure vs. Pump Dispense Engines for CMP Slurry Distribution," Proc. of SEMICON West Contamination in Liquid Chemical Distribution Systems Workshop, July 1998.
- M. Litchy, K. Nicholes, D. Grant, R. Singh, "Perturbation Detection Analysis: A Method for Comparing Instruments That Measure CMP Slurry Health," Proc. of 20th Annual Semiconductor Pure Water and Chemicals Conference, 2001, in press.
- D. Nicoli, et al, "Automatic, High-Resolution Particle Size Analysis by Single-Particle Optical Sensing," American Laboratory, July 1992.
