Wafer environment nanoparticle contamination control and defect reduction in front-end-of-line (FEOL) cleaning processes
Isamu Funahashi, Takuya Nagafuchi, Bipin Parekh, Mykrolis CorporationIntroduction
The complexity of semiconductor device manufacturing processes is increasing as scaling of ICs continues with shrinking feature sizes. With this complexity, contamination control of nanosize particles is increasingly becoming more important during fabrication. The ITRS guidelines in Table 1 show the stringent purity requirements for liquid chemicals to be used in the manufacture of next-generation semiconductors.1
As semiconductor device design rules become more constrained, controlling yield-reducing contaminants in the wet-etch and cleaning process chemicals becomes more critical. An optimally designed filtration process can prevent wafer recontamination from the chemical delivery and cleaning systems in the fab. The filters should be designed to efficiently remove nanoparticles from the cleaning chemicals and also provide high liquid flows at low differential pressures to meet the needs of expanding wafer size from 200 mm to 300 mm.
Importance of stringent particle filtration: Wafer defects caused by undetected sub-0.10-μm particles
Currently, the sensitivity of liquid particle counters for detecting particles in chemicals is 0.065 μm. Any particle smaller than 0.065 μm not detected in the chemical could wind up on the wafer surface and adhere to it. Such a particle can be detected by enhancement with an oxide/nitride deposited film.
Figure 1. Sub-0.065-µm substances on an IC device surface
Figure 1 shows a scanning electron microscope (SEM) image of surface contamination detected on an IC device during its surface evaluation tests. This image shows magnified residual particles on the wafer surface. After the device was cleaned and dried, an oxide/nitride film was deposited on it. The SEM picture clearly shows that the deposited film enhances the presence of smaller size particles, which were probably undetected in the chemical. Based on the film thickness measurements, this device maker estimated the size of residual particles to be less than 0.05 μm. This observation clearly indicates the need for more stringent filtration.
Such surface analysis is appropriate for measuring particle contamination on wafer surfaces. To characterize particle removal efficacy of chemical filters, we have developed several test methods that measure membrane pore size (by bubble point) and determine particle retention using an appropriate particle challenge, in conjunction with either an in-line particle counter or by analytical methods. In the following, we will briefly explain the principle of these test methods and present experimental results that demonstrate the removal of 0.034-μm particles by Teflon® 0.03-μm filters. Additional references are cited to understand the details of the methods.2
Filter pore size and nanoparticle retention test methods (SEM method)
There are several available methods of measuring membrane pore sizes, such as SEM imaging to view it directly or the membrane bubble point test. The particle removal efficiency can be determined by challenging the filters with test particles of the appropriate size and counting the particles before and after filtration using a particle counter or an appropriate analytical detection method.
Table 1. Yield enhancement/wafer environment contamination control (2004 ITRS update)
Figure 2 shows a SEM image of two different membrane surfaces. The left image shows a film-stretched membrane, and the right shows a track-etch membrane. The track-etch membrane has a fairly well-defined circular pore structure that facilitates measurement of the pore size. The PTFE membrane, made by the stretch method, has the pore structure formed between the stretched fibers. Since its pore structure has an undefined, tortuous shape, SEM imaging cannot be used to precisely quantify its pore size.
Figure 2. SEM images of PTFE (left) and track-etched (right) membranes.
Pore size rating by bubble point
Bubble point testing is widely used by filter makers since it can be done without damaging the membrane. The test method is as simple as wetting the membrane with a solvent, such as an alcohol, that completely fills all the pores. The wetted membrane is then pressurized with an inert gas such as nitrogen, gradually raising the pressure. The gas pressure level at which the solvent from the largest pore is completely pushed out is termed the bubble point (rated pore size) of the membrane. The bubble point is directly proportional to the solvent surface tension, and inversely proportional to membrane pore size.3 The mathematical expression of the bubble point is as follows:
P = Bubble point pressure
δ= Surface tension
θ= Liquid-solid contact angle
d = Pore diameter
K = Shape correction factor
The bubble point test method is extensively relied upon in the development of membrane filters.
Particle retention with a monolayer-coverage PSL-bead challenge test
This test involves challenging the membrane with a high concentration of polystyrene latex (PSL) bead solution (monodispersed particle size distribution), containing an added surfactant to cover the equivalent of one layer of PSL beads on the membrane. The light absorption technique is used to measure particle concentration in feed and filtrate. The particle retention measured by this method is attributed to the presence of large diameter pores. (Special fluorescent beads can be used for this challenge test.) This method is used as a quality tool to check the retention performance and not necessarily for retention claim.
Silica challenge test
The silica bead challenge test is performed by flowing a solution of monodispersed silica particles, such as colloidal silica, through the membrane. The silica concentration of the feed and filtrate is measured by the Graphite Furnace Atomic Absorption photometry (GFAA) or ICP-MS. The advantage of using this test is that it is relatively easily done with a smaller membrane area. The test is run in a breakthrough mode and the retention claim is based on at least 99 percent silica removal by the membrane.4 This method has been successfully applied to characterize the retention of the point-of-use (POU) DI filter Durapore® Z, measuring DI water nonsieving retention down to 0.02 μm. It has also been used to characterize the sieving retention of tighter membranes where silica removal can be correlated to the 0.05-μm or 0.034-μm PSL retention as measured with a particle counter.
Mixed-bead and single-bead PSL particle challenge tests
Mykrolis has been using the PSL particle challenge method to determine filter particle retention in a dynamic mode with a continuous particle challenge. The details of the test methods are covered in the Applications Note MA041.5 This method involves continuously injecting PSL beads into the filter feed water at a predetermined concentration and measuring the PSL particle concentration in the filtrate with a laser particle counter. For the 0.05-μm pore size membranes, we use a mixed particle size solution based on the SEMATECH method. For the new 0.03-μm rated membrane, a single particle size challenge feed is used.
Measurement of 0.034-μm particles using an in-line particle counter
Recently, a new Liquid Particle Counter (LPC) UDI30 that can measure 0.03-μm size particles has become commercially available for in-line particle measurement in DI water. Currently, this special LPC is being used to evaluate the particle removal performance of filters with a sub-0.05-μm particle retention rating. There are two kinds of PSL bead challenge tests: the first uses PSL beads (0.034 um) in pure water; the second is designed to measure the sieving effect (to retain particles based only on membrane pore size), excluding the membrane thickness effect. This method involves letting the filter surface adsorb a surfactant chemical. We expect the second method to have more potential in the future since more wet-clean processes will be using chemical compounds containing surfactants.
Currently, commercial particle counters are available to measure 0.034-μm particles in DI water and 0.065-μm particles in chemicals.
Fluorescence-based fast-retention testing
This is a batch-type particle retention method for testing membrane coupons, especially the charged membranes. The test procedure is as follows:
1. A suspension of fluorescent PSL particles is filtered through a test membrane.
2. Feed and permeate stream particle concentrations are analyzed by fluorescent spectroscopy.
3. Retention is calculated as:
LRV = Log10 ([feed] / [permeate])
The method was adapted to a 96-well platform (high-throughput) device.
The fluorescence method can be used to map the effects of membrane surface charge, pore size, particle size, solution pH, solution ionic strength and flow rate on particle retention. This method uses a very high particle-challenge load (5 to 6 orders of magnitude higher than the PSL retention test), dictated primarily by the detection limit of the instrument. 6
Attempts to provide high flow with tight pore size nanomembranes
Table 2. Front end of line (FEOL) cleaning process and filtration requirements
As the membrane pore size decreases, from 0.10 μm to 0.05 μm to 0.03 μm, the pressure loss per unit flow rate of a device is expected to increase. Filter manufacturers are improving the membrane structure and device design to pack in more surface area to deliver the higher flows and nanoparticle retention required for 300-mm wafer processing. Table 2 lists key process needs and filter attributes for the next generation of nanofilters.
Nanosize particle reduction in wafer environments is critical for improving yield. Continued scaling of ICs and a trend toward larger diameter wafers impose conflicting demands on particle filters to reduce nanoparticles and maintain high flow. Filter manufacturers continue to respond to these challenges by developing improved membranes and filters, and novel techniques to characterize filter performance. The current limitation on in-line particle detection in chemicals (0.065 μm) has led to advancements in techniques for particle retention and pore size measurement of these nanofilters. Field evaluation in IC manufacturing processes has demonstrated the performance capabilities of the novel filters. 7 III
Isamu Funahashi is a manager of global product support and applications technology development in Nihon Mykrolis.
Takuya Nagafuchi is a senior application engineer in Nihon Mykrolis.
Bipin Parekh is a senior consulting engineer in the technology group at Mykrolis Corporation, where he has served since 1980.
1. SEMATECH, “Front End Processes” in International Technology Roadmap for Semiconductors, 2003 Edition. See page 22: “Table 70a Surface Preparation Technology Requirements-Near-term.”
2. Funahashi, Isamu, Takuya Nagafuchi and Bipin Parekh. “QuickChange® ATM Chemical Filters Enable Rapid Nano-Particle Removal via Low Pressure Drop and High Chemical Flow Rates,” Mykrolis Applications Note AN1042ENUS, 2004.
3. Emory, Scott F., “Principle of Integrity-Testing Hydrophilic Microporous Membrane Filters, Part I and II,” Pharmaceutical Technology, September and October 1989.
4. Parekh, Bipin, Karim Vakhshoori and Joseph Zahka, “Filtration of High-Purity DI Water for Semiconductor Manufacturing,” Ultrapure Water, May/June 1993, pp. 53-59.
5. Wargo, C., V. Kinney, and J. Zahka. “Hard Particle Retention of PTFE Membranes Used for Chemical Filtration,” Millipore Microelectronics Applications Note MA041.
6. Thom, Volkmar. Millipore Technical Report.
7. See reference 2.
Process benefits of nanoparticle reductions using next-generation filters
Rapid bath clean-up rate
Figure 4. Particle reduction vs. bath turnover in pure water.
Figure 4 shows a comparison of the particle reduction in the recirculation bath as a function of bath turnover for three types (pore sizes) of filters using pure DI water. In this test, the recirculation flow rate was set to 25 L/min. The results show that the tightest filter, 0.03 μm, had the fastest particle removal rate as it has a better combination of particle retention and bath turnover than other filters.
Defect reduction in SC-1 and SC-2 baths
Figure 5. Wafer defects with 0.05-μm and 0.10-μm filters (Asian device maker).
Figure 5 shows a comparison of a 0.10-μm filter with a 0.05-μm filter in both SC-1 and SC-2 baths in tests conducted by an Asian device maker.