Proper Filtration Removes Oversized Particles from CMP Slurry Systems

Single stage recirculation filtrations were performed to remove oversized particles from colloidal silica based Chemical Mechanical Polishing (CMP) slurry. Proper filtration was demonstrated to achieve rapid removal of oversized particles while not altering the percent solids of the CMP slurry. Mathematical models were developed to simulate particle reduction in CMP slurry distribution systems. The models predict particle concentration as a function of the flow rate, the particle removal efficiency of the filter, and filtration time. The models show that flow rate is the most critical parameter in order to achieve rapid removal of oversized particles. The major role of pre-filtration is to capture a portion of the oversized particles and protect the final filtration from premature plugging in order to deliver maximum filter service life. Based on examination of the limited data, the recirculation flow model adequately predicts the particle concentration profile. Ultimately, selection of the optimum filters depends on flow rate, particle removal efficiency of the filter, filtration scheme (single vs. multi-stage filtration), and filter service life. Introduction For microelectronics applications, CMP slurries have a well defined particle size distribution1 and are often composed of fine abrasive particles up to 0.25 μm. A small population of oversized particles greater than 0.5 μm is typically found in a CMP slurry system. Oversized particles can result from particle agglomeration, drying of slurry from wetted surfaces, interactions within the slurry distribution system, and introduction of contaminants from handling. These oversized particles affect the level of defectivity on the surface of the wafer after CMP has been completed. Filtration has been shown to be effective in removing oversized particles from copper CMP slurries2, resulting in a reduction of micro-scratches on the surface of the polished wafer3. Hence, filtration has been incorporated into the CMP process for improved yield management during the manufacturing of integrated circuits. This study investigates the effectiveness of several filtration modes on the removal of oversized particles from silica based copper CMP slurries and the consistency of solids content before and after filtration. In addition, a mathematical model has been developed to simulate particle reduction in a CMP slurry distribution system when the filtration system is being operated under recirculation mode. The filtration results of the reduction in oversized particle concentration were used to evaluate the particle concentration profile predicted by the mathematical model. Experimental A colloidal silica based CMP slurry of approximately 50% solids was obtained from a commercial source. To simulate a low percent solids slurry, the 50% solids slurry was diluted to 1:~160 ratio using DI water filtered through a 0.1 μm membrane filter. The resulting particle count was approximately in the range of 1,000,000 particles/mL @ ≥0.54 μm. The diluted slurry samples were tested and filtered through all polypropylene (PP) depth filters. Filters tested included BetapureTM CMP series filters, and some competitive PP filters with micron ratings ranging from 0.5 to 10 μm. Filtration experiments were performed via single stage, recirculation mode filtrations under constant pressure. Influent and effluent samples at different filtration times were collected for particle sizing and counting, and percent solids determination. The CMP slurry filtration test stand consisted of a pump to deliver fluid from a reservoir to a series of 316 stainless steel filter housings. Pressure gauges were installed before and after each of the filter housings to monitor differential pressure during filtration. Sanitary connections and 316 stainless steel pipe were used for ease of cleaning. Prior to carrying out any filtration experiment, the CMP slurry filtration test stand was flushed and cleaned with DI water filtered through a 0.1 μm membrane filter. The residual DI water was allowed to drain from the filtration test system and a 0.5 μm filter was installed in the last filter housing. Then, filtered DI water was drawn from a 5 gallon reservoir and fed to the 0.5 μm loaded filtration test stand. To make sure that there was no oversized particle contribution from the test system, this operation was run under recirculation mode until the particle count reached an acceptable level, <50 particles/mL @ ≥0.5 μm. Once the test stand was cleaned, it was ready for the CMP filtration experiment. A 5 gallon batch size was used for the CMP slurry filtration. Particle size analysis for particles greater than 0.5 μm was performed via Accusizer Model 780A by Particle Sizing Systems. The background particle count was maintained at <50 particle/ mL @ ≥0.5 μm throughout the sample analysis by flushing with filtered DI water. The counting threshold was set at 2000 particles/ mL @ ≥0.5 μm. Determination of percent solids was performed using Inductively Coupled Spectroscopy (ICP). Using an ICP Si standard of 10,000 ppm, successive dilutions with DI water were used to prepare 100, 10, 1, and 0.1 ppm standards for calibration. The actual silicon concentrations were 100.382, 10.437, 0.944, and 0.159 ppm for the 100, 10, 1, and 0.1 ppm standards, respectively. Based on the ICP responses, 1.0 ppm may be regarded as the limit of quantitation, (LOQ). Figure 1 shows the expected versus actual silicon concentration as determined by ICP. To further reduce the uncertainty in measuring the content of Si, the influent and effluent samples were diluted with DI water to the upper range of the calibration curve, i.e., between 80 and 90 ppm for ICP analysis. The Si determination was performed in triplicate and the average values were reported. The resulting Si readings were converted to SiO2 and reported as percent solids.