MC² Market & Competitive Convergence

1991 PROCEEDINGS — Institute of Environmental Sciences

Ross West and John Ruddock
Parker Hannifin Corporation
Lebanon, Indiana 46052

BIOGRAPHY

Ross West

Ross West joined Parker Hannifin Corporation, Lebanon, Indiana in 1986. He received a BA degree in biology from Alfred University, Alfred, New York, and obtained his MBA degree in marketing from National University, San Diego California. Prior to his Parker experience, he marketed numerous analytical instruments and membrane-based filtration products.

Since joining Parker, he has been involved in the commercialization of several filtration products. He is currently Marketing Manager for Parker's newly developed microfiltration products, and has been responsible for the beta site evaluations of the new Dual Asymmetric PTFE membrane. He is a member of the Institute of Environmental Sciences, the Parenteral Drug Association, and Semiconductor Equipment and Materials International.

BIOGRAPHY

John M. Ruddock

John Ruddock joined Parker Hannifin Corporation in 1989 at Lebanon, Indiana. He received a BS degree in Mechanical Engineering from Purdue University, West Lafayette, Indiana, and is a registered P.E. in the state of Indiana. Prior to his Parker experience he designed test and production equipment. He also designed analytical instruments, including spectro photometers.

Since joining Parker he has been responsible for the automated equipment needs of membrane and cartridge production. He is also responsible for the design of test equipment to evaluate the performance of Parker's new membrane products. 

ABSTRACT

The requirement for lower particulate levels in liquid semiconductor process chemicals and the need for inert wetted surfaces have driven the development of a new Dual Asymmetric PTFE membrane filter. This report will describe the features of the filter and a test method used to assess its performance in a bulk chemical delivery environment. Using a model chemical delivery system, test filters were challenged with a steady-state particle concentration of approximately 60 particles per liter. The new filter was able to significantly reduce the particle concen tration by a factor of 6.

INTRODUCTION 

It became evident that the increasing demands for lower particle levels in ultrapure liquid chemicals would eventually outpace the capabilities of existing polytetrafluoroethylene (PTFE) membrane filters. For instance, we have seen filters of this type retain less than 10% when challenged with monodispersed polystyrene latex (PSL) spheres equal in size to the claimed rating of the filter. Attempts to improve retention with the existing technology has resulted in the availability of multilayer configurations and matrices with finer pore structures from various manufacturers. However, results of these attempts are limited by the corresponding increase in the pressure drop across the filter. To address this problem a new PTFE membrane was developed and incorporated into a filter cartridge of perfluoroalkoxy (PFA) construction.

In order to determine its suitability for filtering semiconductor process chemicals, its performance was assessed in a model chemical delivery system. The system was able to deliver a steady-state, low-level particulate challenge using a selected referee chemical. The need to correlate the counts to existing data led to the selection of a commonly used particle counter, the HIAC/ROYCO 4300 with the 346 BCL sensor. Particle counts with the use of a popular PTFE filter of conventional construction were compared to the new filter. This report will show the design features of the filter the model chemical delivery system, and the results of the evaluation. 

FILTER DESCRIPTION

The design objective was to improve particle retention with a PTFE matrix while minimizing the pressure loss across the membrane. Existing, commonly used PTFE matrices are made through a stretching process and dubbed "expanded membranes." Their porous structure is essentially unchanged throughout the matrix. Increasing particle retention with this technology is accomplished by controlling the expansion during the membrane manufacturing process. Although the use of this tactic results in smaller pores, there are less open areas, and therefore, there is a corresponding increase in pressure loss across the membrane.

A Dual Asymmetric PTFE matrix was developed to overcome these shortcomings. The structure is characterized by three distinct layers: a thin inner layer of fine pores with layers above and below consisting of relatively larger pores. The fine pores of the inner layer contributes to the improved particle removal efficiency while its thinness minimizes the pressure loss across it. The outer upstream layer acts as a prefilter to maximize filter longevity. The upstream and downstream layers protect the inner layer as well as adding strength and durability to the matrix. See Figure 1.

The scanning electron micrographs indicate a substantial gain was made in reducing the size of the pores. This is confirmed by the results of mono- dispersed PSL sphere retention studies. Flow rate/pressure loss comparisons show the membrane's performance to be in the range of acceptability. See Table 1.

The membrane was incorporated into a standard 25 cm (nominal) filter element constructed of PFA in a pleated configuration, with double 222 o-rings of PFA encapsulated Viton (DuPont trade name).

REFEREE CHEMICAL DESCRIPTION

NMP (N-Methyl-2-Pyrrolidone), was chosen as the referee fluid. It was the belief of several beta sites that it was somewhat difficult to reduce NMP particle levels. In addition, their experience was that filters that perform well with NMP, generally speak mg, performed well with other chemicals. NMP was also selected because of its relative safety, and because it is a commonly used solvent in wafer fabrication.

MODEL DELIVERY SYSTEM DESCRIPTION

The overall goal of the model chemical delivery system was to provide as low a particulate challenge as possible in order to evaluate filter performance.

To achieve our low challenge goals the stand was placed in a Class 100 clean room and the materials of construction of all wetted parts were limited to either PFA or PTFE.

The plumbing, as shown in Figures  2  &  3 , provides the necessary interconnections to the test filters while minimizing dead legs which would lead to particle migration. The components and system design were based on an operational ultrapure chemical delivery system. Refer to Table 2.

Drum "A", the NMP shipping container, sat outside the clean room. Special fluorocarbon adapters were fabricated to screw into the drum ports to make the connection to the system's PFA tubing. Pump "A" transferred the chemical to drum "8" inside the clean room. Drum "B", with ultrapure style connections, served as the recirculation tank (Day Tank) during flushing and the supply during testing. Pump "B" served as the system supply pump and directed the fluid through a PTFE direct-read, rotameter-style flowmeter. Teed into this line was a 20 liter pressurized container, also with ultrapure style connects, which was used as an accumulator to help reduce the pressure pulses generated by the pump. The next items were the System Filters "A" and "B" which were plumbed in series to allow the test filter circuit to be connected to "High", "Medium" or "Low" challenges as desired. 

The test circuit consisted of a manifold fed by the desired particle challenge which was connected to the "Control Filter", "Dual Asymmetric Filter #1", and "Dual Asymmetric Filter #2". The on/off control valves for each of the filters were located down stream. A Return Filter provided one last level of filtration before the test fluid returns to the recirculation Drum "B".

A single in-line particle counter was used for the test. Each filter housing had small, short-coupled valves to allow the particle counter line to be linked to upstream or downstream of any filter while providing a close coupling to reduce particle generation.

A flow controller, in series, provided the 100 mL/min required flow rate for the counter. The raw data from the counter was printed out while simultaneously fed to floppy disk files. Charting software was then used to read the data and plot the results.

The test stand frame and component location is shown in Figure 2. The frame was constructed of welded heavy wall square steel tubing for rigidity and was painted with epoxy paint to prevent particle shedding in the clean room. The filter housings were located for ease of access and minimum disturbance with all valves being mounted to the polypropylene sheeted deck and rear wall.

TEST PROCEDURE

The System Filters and the Return Filter were IPA wetted and installed in their respective housings. The three test filter housings were left empty.

Approximately 120 liters of the referee fluid was transferred to the Day Tank (Drum "B"). The NMP was then recirculated from the Day Tank through the delivery system at 12 L (3 gal.) per minute which provided a pump discharge pressure of 2 bar (30 psi).

The particle counter was connected to the low challenge point with 100 mL/min flowing through the particle counter and back to the chemical supply drum.

At this point, the particle count from the Day Tank was on the order of 80,000 particles per liter ³ 0.5µm while the low challenge was approximately 1000 particles per liter ³ 0.5um. Testing continued at 6 to 8 hours/day for about three weeks until the low challenge goal of 60 particles/liter ³ 0.5µm was achieved. Please note that since 100 mL/min was returning to the chemical supply drum from the particle counter, recirculation was stopped about every 3 days to refill the Day Tank. The chemical supply drum was placed in an area of low vibration and poor access to prevent accidental bumping and subsequent particle generation.

The Control Filter, Dual Asymmetric Filter #1, and Dual Asymmetric Test Filter #2 were then IPA wetted and installed in their respective housings. The Return Filter discharge valve was closed to cause a deadhead condition with the supply pump maintaining a 2 bar (30 psi) supply pressure. In the deadhead test condition the flow rate through the Test Filter and particle counter was 100 mL/minute.

Data was taken for 1 to 1½ hours to ensure that the particle challenge level was stable and at the target level. The counter was connected to the Control Filter and the two Dual Asymmetric Filters in turn, with data taken for 6 hours for each. Between measurements, the counter was reconnected down stream of the System Filters to confirm that the challenge level remained unchanged.

PARTICLE MEASUREMENTS

The NMP particle measurements were taken using a HIAC/ROYCO Model 346 BCL detector coupled to a Model 4300 HIAC/ROYCO Microprocessor. The flow rate through the detector was controlled to 100 mL/minute. All measurements were taken at the filter core via the threaded drainage ports provided in the filter housings. 

Counts were totaled in the 0.5µm and larger channels every minute and recorded as counts per 100 mL. The counts were then averaged at 15 minute intervals and converted to particles per liter. The counts were reported as particles ³ 0.5µm. Extreme care was taken to minimize false positive counts caused by bubbles. This was accomplished by thoroughly flushing the system between hookups and maintaining system pressure. In all cases the samples were taken directly at the filter outlet. Inlet counts were those counts taken directly down stream of the System Filters (2 expanded PTFE elements in series, rated at 0.1µm). To ensure against the possibility of noise interfering with particle counter performance the flow was interrupted between measurements and the counter was monitored for spurious signals. 

RESULTS

Results graphed in Figure 4 show particle counts reported in 15 minute intervals and represent the average of 15, 1-minute total counts from the 0.5µm and larger particle counter channels. The counts are converted to particles per liter. The inlet counts represent the NMP particle levels downstream of the system filters (2 expanded PTFE elements rated at 0.1µm, in series). The inlet counts decreased during the first 2 hours and then remained steady at approximately 60 particles per liter. The control filter (A duplicate of the system filters) immediatelyachieved particle levels equal to the inlet. The graph shows that the Dual Asymmetric elements achieved levels significantly below the inlet levels within two hours of installation and trended downward through out the test interval. 

The mean counts for the last 3 hours of the test are shown in Table 3. The results show the that expand ed PTFE element was unable to reduce the particle concentration below the inlet level. On the other hand, the Dual Asymmetric PTFE elements significantly reduced the particle concentration by a factor of 6.

DISCUSSIONS

Our experience, and that of several of our beta site partners, is that most filters are highly efficient when challenged with dirty liquids, and what distinguishes one filter from the other is their ability to make a clean fluid cleaner. In the latter situation it is possible to see filters adding to the dirt load. This was not the case here, but we did see the chemical processed through 2 expanded PTFE filters and then show no improvement when filtered with a third duplicate filter.

We selected the particle counter based on its wide spread use at our beta sites. However, if we make the adjustment in apparent particle size based on the refractive index of the NMP (1.47), the particles being detected are in the range of 0.7µm. Clearly, the test should be repeated employing detectors sensitive to smaller size particles.

We chose to flow at a rate of 100 mL/minute, the maximum flow rate for the particle counter, in order to count 100% of the filter effluent. This was considered important because of the very low particle counts we experienced and our desire to distinguish filter performance. Although the flow rate for the test was much lower than production chemical delivery flow, it was the experience of the beta site that there was a linkage between test performance and production performance. However, we were not able to find any reference in the literature to any such linkage.

We were careful to design the system to minimize any hydraulic stress to the filters. For instance, the accumulator helped dampened the pump pulses and all valves were operated slowly. This reflected the conditions at the beta site.

Microporous membrane filters have played a major role in semiconductor fabrication. Indeed, they have served the industry well, right up to the present generation of devices. This has been the case due in no small part to the continuous improvement in quality. For instance, filters are now available from various sources where clean room manufacturing is employed. An important aspect of quality that was not addressed in this study is the consistency of filter performance from filter to filter, from lot to lot. Additional testing and further beta site experience is needed to address this aspect of quality, but this is the case for all new products.

CONCLUSIONS

The Dual Asymmetric PTFE filter was able to reduce particle levels in a referee liquid chemical below those achieved with PTFE filters of conventional construction. A consistent and sustainable low level particle challenge can be achieved using a model chemical delivery system. It is possible to distinguish filter performance when ultra-low particle level conditions exist.

ACKNOWLEDGMENTS

The authors thank David Mayette and Gary Husted of MicroAssays Of Vermont who designed the delivery system and monitored the study, and our beta site participants for their encouragement and fortitude.

REFERENCES

1. Donovan, R. P., ed., "Particle Control For Semiconductor Manufacturing," Marcel Dekker, Inc.: New York. NY, 1990.

2. Duke, S. D., "Particle Retention Testing Of 0.05 To 0.5mm Membrane Filters," Proceedings Of The International Technical Conference On Filtration And Separation, American Filtration Society, March, 1988, pp. 525-532.

3. Goldsmith, S. H., Barski, J. P., and Grundelman, G. P., "A Method For Measuring Particle Shedding From Microporous Membrane Filter Cartridges In Liquid Streams," Microcontamination, June/July, 1984, pp. 47-52.

4. Grant, D. C., "Measurement Of Particle Concentrations In Central Chemical Delivery Systems," Journal Of The IES, July/August, 1 990, pp. 32- 37.

5. Gruver, R., Silverman R., and Kehley J., "Correlation Of Particulates In Process Liquids And Wafer Contamination," ES Technical Proceedings, 1990, pp. 312-315.

6. Hashimoto, S., Kaya, M., and Ohmi, T., "Improving And Maintaining Electronics-Grade Chemical Quality Requires Technology Advances," The Ohmi Papers, Keeley, R. W. and Cheyney, T. H., ed., Santa Monica: Canon Communications, Inc., 1990, pp. 66-74.

7. Krygier, V., Latham, M., and Conway, "Automatic Particle Measurement In Liquids Downstream Of Fine Membrane Filters," Microfiltration, April, 1985, pp. 33-39.

8. Seaman, G., and BilImaier A., "Poster 110," Microcontamination Conference, November 16-18,1988.

9. Simonetti, J. A., Schroeder H. G., and Meltzer T. H., "A Review Of Latex Sphere Retention Work: Its Application To Membrane Pore-Size Rating, Ultrapure Water, July/August, 1986, pp. 46-51.