by Aiwen Wu and Wailup Chow, Entegris Inc., Billerica, MA USA
Executive overview
In photochemical purification applications, bubble elimination is as important as particle removal. This paper describes a technique developed to rapidly eliminate microbubbles, which affect yield, during filter startup in a two-stage dispense system. The experimental results suggest that by providing a constant pressure to the fluid after wetting the filter, it is possible to effectively eliminate microbubbles in the fluid.
In semiconductor manufacturing processes, bubbles (often referred to as microbubbles) can be contaminants that reduce manufacturing yield. In photolithography processes, a point-of-use (POU) filter is used on a clean track system to ensure lower wafer defect levels by providing particle and bubble-free photochemicals. However, filter changeout often results in significant chemical consumption and tool downtime due to the purging of air from the system, hence the need for a method that can eliminate the formation of bubbles.
Microbubble formation
Bubbles, like particles, can cause point defects on semiconductor wafer surfaces. During processes, such as photolithography, bubbles can cause more severe problems than particles. A bubble can act as a lens that focuses the light, thereby magnifying the error relative to the size of the bubble. Bubbles also could fall on a wafer during the spin-on process. Figure 1 shows an example of a bubble-related wafer defect. Particles can be effectively removed from liquids by filtration. However, elimination of microbubbles by filters is particularly challenging.
Figure 1. Wafer defects caused by a bubble coming out of solution during the spin-on process.
Microporous membrane filters with pore size of 0.2μm or smaller are typically used to remove particles from photochemicals. These filters remove hard particles very efficiently, but because, bubbles are deformable they may be extruded through pores. Therefore, the bubbles may attach to the filter, especially if the filter is hydrophobic, and could be released into the fluid under certain conditions. Many fabs either pre-wet filters with organic solvents and/or use photochemicals to purge filters for a long time during filter startup, which results in significant chemical consumption and tool downtime. While filter material selection, filter design optimization, dispense pump design and pump recipe optimization help purge air introduced during system maintenance or filter change out, further improvements in filter startup is necessary.
Microbubbles can be formed by filters when gas bubbles grow in the liquid. Pressure in small gas bubbles is very high therefore, it is unlikely for the gas bubbles to form spontaneously. The bubbles grow on a surface where their radius of curvature can be relatively large, so their pressure can be lower. Bubbles also may form when the pressure in the fluid drops and gas comes out of the solution. If the fluid is not saturated with gas, bubbles will not form. As well, if there is low pressure drop in the system, bubbles are less likely to form.
One approach for improving microbubble performance is to minimize the nucleation sites for bubbles by filling the voids on filter membrane surface with fluid. This may be achieved by two methods:
– Filling the voids with fluid may be done by displacing air in the void with liquid. If a vacuum is pulled on the filter and then the filter is filled with fluid, liquid may go into the crevices on filter membrane surface [1];
– It is also possible to pressurize the fluid and push it into the crevice. As a result, the filter startup process, with a pump, will have an effect on the outgassing performance of a filter.
This paper discusses the bubble behavior in liquids and describes the technique developed to rapidly eliminate microbubbles during filter startup in a two-stage dispense system. Laboratory experiments were conducted with an Entegris IntelliGen Mini two-stage dispense pump and photochemicals. An optical particle counter was installed on the dispense line to detect the microbubble numbers during filter startup. We studied the effect on the microbubble cleanup speed by pressurizing the fluid after it was introduced in the filter and pulled with a vacuum, before filling the filter with more fluid. The experimental results are presented and discussed in this paper.
Behavior of bubbles in liquids
Gas solubility in liquids — Henry’s law. When a gas is in contact with the surface of a liquid, a certain amount of the gas will dissolve into the liquid. The solubility of a gas in a liquid depends on temperature, partial pressure of the gas over the liquid, the nature of the liquid and the nature of the gas. Gas solubility is always limited by the equilibrium between the gas and the saturated solution of the gas. The dissolved gas will always follow Henry’s law, which states that the amount of a gas that dissolves in a liquid is proportional to the partial pressure of the gas over the liquid. A formula for Henry’s law is expressed as:
at constant temperature
Where p gas = partial pressure of the gas over the liquid; c = concentration of dissolved gas in a liquid; and k = Henry’s law constant depending on the nature of the gas and the liquid, which has units such as L*atm/mol, atm/(mol fraction) or Pa*m3/mol.
Changes in pressure above a liquid will affect the solubility (concentration) of a gas in the liquid. If the temperature stays constant, increasing the pressure (e.g., by compressing the gas) above the liquid will increase the frequency of collisions between the gas molecules and the liquid surface. The result is more gas molecules dissolved into the liquid and an increased concentration of the gas in the liquid. An example of air dissolved in water is shown in Table 1.
Table 1. Henry’s Law example of air dissolved in water at 25°C (77°F).
Bubble formation. Bubbles arise in liquids when the solubility of dissolved gases decreases. Pressure fluctuations, such as those created during flow through an orifice or fluid pumping, can cause the formation of bubbles by several different mechanisms. Three mechanisms have been proposed [2, 3] for the formation of bubbles: homogeneous nucleation, heterogeneous nucleation, and cavitation
Homogeneous nucleation results in the formation of microbubbles everywhere in a liquid when gas molecules form clusters and grow to a defined size. This phenomenon occurs when supersaturated dissolved gas in a liquid suddenly becomes insoluble, for example, by a reduction in pressure. Large supersaturation ratios are required for homogeneous nucleation, which is rare and is not a likely mechanism for bubble formation in photochemicals.
Another mechanism, heterogeneous nucleation, is defined as bubble growth that occurs on hydrophobic surfaces. Hydrophobic surfaces, or particles, act as catalysts for bubble formation when gas solubility in a liquid is reduced.
Lastly, cavitation is characterized by bubble formation at nucleation sites caused by a sudden reduction in pressure of a moving fluid. Both heterogeneous nucleation and cavitation are the likely mechanisms for bubble formation in photochemicals. The formation of cavitation nuclei on the membrane internal surface can be explained by the crevice model developed by Atchley and Prosperetti [4]. However, the pressure balance across a crevice bubble can be described by the Young-Laplace equation:
Where Pg = pressure inside gas bubbles; PL = pressure in the liquid; γ = surface tension of the liquid; Κ = the interface curvature.
Different from a bubble existing on a flat surface, a crevice bubble can be stabilized by having a negative curvature. Crevice bubble curvature could be a function of the geometric characteristics of the crevice and the surface hydrophobicity. A small positive or a large negative curvature may occur with hydrophobic material. According to the equation (above), a bubble having a negative curvature would exhibit a lower internal gas partial pressure than the liquid pressure, thus, having a potential to degas from the bulk liquid into the nuclei. This would stabilize the crevice bubble which can serve as an air pool for formation of the nucleus of cavitation. The tension (negative pressure) required to expand the gas pocket in the crevice into the liquid body depends on the geometric characteristics of the surface.
For a sphere shape gas bubble, the Young-Laplace equation can be expressed as:
Where r = radius of gas bubbles.
Κ > 0, Pg > PL
Κ < 0, Pg < PL
Table 2 shows how the internal pressure of an air bubble in water increases with decreasing bubble size. The bubbles grow on a surface where the radius of bubble curvature can be relatively high so the bubble pressure can be lower. The rougher and the more hydrophobic the surface, the easier it is for bubbles to form [5, 6].
Table 2. Air bubble pressure and size in water.
Bubble rise rate in liquids. Once bubbles form and grow in liquids, they tend to rise to the surface of liquids. Forces of rising bubbles include buoyant force and friction force, which can expressed as the equation:
Where F = forces of rising bubbles; Vr = bubble rising velocity; g = gravitational acceleration; r = radius of air bubble; ρg = gas density; ρf = fluid density; μ = fluid viscosity
The rising speed of a bubble is reached when the combination of the frictional force and the buoyant force together equal the gravitational force. The resulting rising velocity is given by:
Equation 5 (above) is referred to as Stokes’ Law and is valid for Reynolds numbers <1. In water, the Reynolds number is <1 for bubbles smaller than ~125μm in diameter. This equation can be used to calculate the rate at which bubbles rise in a liquid. Table 3 describes the results of such calculations for air bubbles in liquids with various densities and viscosities. The table indicates that the bubbles rise fairly slowly. The smaller the size of a bubble, the slower the rising velocity. Microbubbles are non-visible bubbles (<10μm) that have no appreciable rising velocity and act more like particles. If microbubbles are stable at <10μm in size, they cannot completely be removed by simply waiting for them to rise out of the liquid. Other approaches have to be used to accelerate the bubble removal, such as using high flow rate to flush them out of a filter and/or pressurizing the fluid to accelerate their dissolution.
Table 3. Bubble rising velocity in liquids with different densities and viscosities.
Experimental
Setup. In order to investigate the effect on the microbubble cleanup speed by pressurizing the fluid after it was introduced in the filter, laboratory experiments were conducted with an Entegris IntelliGen Mini dispense pump and an ArF BARC or I-line TARC. These materials were chosen because ArF BARC is a solvent-based photochemical and I-line TARC is an aqueous-based photochemical. Both are good representatives of commonly used chemicals in the photolithography spin-on processes. The Mini pump uses two-stage dispense technology to operate filtration and dispense stages completely independently. This unique pump design allows easy control of the filtration and dispense for this filter priming study.
A re-circulating chemical test-stand was assembled using a chemical reservoir, dispense pump, filter manifold, test filter and an optical particle counter (OPC). Figure 2 contains a diagram of the experimental setup. The OPC used in this study was a Rion KS-41. This counter is capable of detecting particles down to 0.15μm, and it was installed on the outlet line of the dispense system, monitoring the entire downstream of the test filters. The effluent was recycled to the reservoir.
Figure 2. Experimental setup for measuring microbubble level by optical particle counter with a dispense pump.
After the test filters were wetted under several different conditions, the filters were primed with the photochemical, and the dispense recipe was continually performed until particle counts leveled off. The particle counter cannot distinguish between bubbles and particles. Since each new test filter was installed after the particle counts had reached very low background with a filter in place, the particle levels shown by the counter indicated the level of microbubbles in the dispense line during the testing. While optical particle counters are not designed to count bubbles, the results can be used in semi-quantitative manner to see differences in filter bubble cleanup performance.
Procedure. For each photochemical we used for this study, the first test was conducted using the “best known” filter startup method — the best known startup procedure that existed before the study was conducted. The results from this test were used as the baseline to compare filter startup performance with various other filter startup methods.
The best known method was defined as the following: introduce fluid into the filter after installation and then start the priming instantly with a programmed recipe. Then, the effect of pressurizing the fluid, after filter installation and membrane wetting on filter startup speed, was investigated. In this test, we applied a constant pressure to the fluid after filter wetting to push fluid into the crevices that could typically cause nucleation sites for microbubbles, and accelerate the dissolution of existing microbubbles in the filter. Also, the time of pressurizing the fluid was varied to look into the effect of pressurizing time on filter startup speed.
The experimental procedure is summarized as follows:
– Perform dispense using the programmed recipe with a filter in place, and monitor the particle counts in the outlet line of the pump until particle counts reach the steady state. This is referred to as the particle background of the system.
– Stop the system and the particle counter, and replace the filter with a new filter.
– Execute “vent” operation seven times for BARC (11 times for TARC) to introduce the fluid into the filter upstream.
– Execute “purge to vent” operation seven times for both BARC and TARC to introduce the fluid into the filter downstream.
– Test #1: Start the continuous dispense instantly using programmed recipe and start the particle counter.
– Test #2: Apply a constant pressure to the fluid for a pre-determined time, then start the continuous dispense and particle counter.
– Stop the test when the particle counts return to the particle background for the system.
The pump recipe parameters used in this study, as shown in Table 4, were kept constant during the continuous dispense for each testing chemical.
Table 4. Mini dispense pump recipe parameters.
Results
The particle data reported in the following graphs shows the particle counts >0.15μm coming out the filter for various filter startup methods, and particle data was collected in 1.5 minute intervals. This testing was performed using Impact 2 V2 0.05μm ultrahigh molecular weight polyethylene (UPE) filters for TARC and Impact 2 V2 0.01μm UPE filters for BARC. The particle counts were plotted vs. dispense time for BARC and TARC.
As shown in Figure 3 for the 0.01μm filter started up in BARC using the best known method, the particle counts returned to the system background after ~70 minutes of dispense time. However, pressurizing the filter at 15 psi or 25 psi for 15 minutes after filter wetting, significantly improved the filter startup time. Particle counts dropped to <1 particle (≥0.15μm in size) per mL in ~30 minutes dispense time.
Figure 3. Bubble cleanup of 0.01μm UPE filter on a two-stage dispense pump with BARC.
The beneficial effect of pressurizing the filter fluid after filter installation and wetting on startup also was observed for the 0.05μm filter startup in TARC. As shown in Figure 4, the system only took ~50 minutes to return to the background when the filter was pressurized at 25psi for 30 minutes after installation and wetting. However, it took ~230 minutes for particle counts to return to the system background level when the filter was started up by the best known method. Both the pressure and the time of pressurizing the fluid seemed to be a factor in filter startup with a TARC system. Bubble cleanup performance of fluid pressurizing at 25 psi for 15 minutes, and at 15 psi for 30 minutes, was not quite as good as 30 minutes pressurizing time at 25 psi. However, all pressurizing recipes showed significant improvement on filter startup when compared to the best known method.
Figure 4. Bubble cleanup of 0.05μm UPE filter on a two-stage dispense pump with TARC.
Table 5 is the summary of the bubble cleanup data for this study. It shows that with both a solvent-based photochemical and an aqueous-based photochemical, applying a constant pressure to the fluid after wetting the filter significantly reduced the bubble cleanup time when a new filter was installed in the system.
Table 5. Bubble cleanup performance comparison.
Conclusion
In photochemical purification applications, bubble elimination is as important as particle removal. Removal of bubbles by filtration is more complicated, especially at the filter startup stage. The amount of tool downtime for filter change and priming drives reduces the operating efficiency for tracks and scanners. As tracks and scanners get faster, the cost of downtime becomes greater even if the actual amount of filter priming time stays the same. Further reducing filter priming time is needed to meet the high throughput goal in photolithography processes. We have developed a technique using the unique features of two-stage dispense pumps to rapidly eliminate microbbubles during filter changeout. The experimental results suggest that by applying a constant static pressure to the fluid, after wetting the filter, we can effectively push fluid into the crevices that typically cause nucleation sites for microbubbles, with the outcome of preventing microbubble formation and significantly improved filter priming time.
Acknowledgments
The authors would like to thank the photochemical suppliers for providing the photochemicals used in this study. We also would like to thank Joseph Zahka for valuable discussions. IntelliGen Mini is a registered trademark of Entegris.
References
[1] U.S. patent pending, Provisional Application Serial No. 60/674,594.
[2] D.E. Yount, “Skins of Varying Permeability: A Stabilization Mechanism for Gas Cavitation Nuclei,” J. Acoust. Soc. Am. 65, 1429, 1979
[3] Research on microbubble formation and detection conducted for Mykrolis by CT Associates Bloomington, MN.
[4] A.A. Atchley, A. Prosperetti, “The Crevice Model of Bubble Nucleation,” J. Acoust. Soc. Am. 86 (3), 1065-1084, 1989.
[5] R. Cole, “Boiling nucleation,” Adv. in Heat Transfer, 10:85-166, 1974.
[6] S. D. Lubetkin, “Bubble Nucleation and Growth in Controlled Particle, Droplet and Bubble Formation,” ed. D. J. Wedlock, Butterworth Heinemann, 167-168, 1994.
[7] J. O. Osburn, P. L. Markovic, “Calculating Henry’s Law Constant for Gases in Organic Liquids,” Chem. Eng., 76(18): 105-108, 1969.
[8] X. Li, Y.C. Yortsos, “Theory of Multiple Bubble Growth in Porous Media by Solution Diffusion,” Chem. Eng. Sci., 50(8), 1247-1271, 1995.
Biographies
Aiwen Wu received his Ph.D in chemical engineering from the U. of New Hampshire at Durham, NH and is a senior applications development engineer in the liquid filtration business unit at Entegris Inc., 129 Concord Road, Billerica, MA, USA 01821; ph.: 978-436-6820; email: [email protected].
Wailup Chow is applications development engineer in the photochemical dispense business unit at Entegris Inc., Billerica, MA USA.