A moving-zone Marangoni drying process for critical cleaning and wet processing
10/01/1997
Jerald A. Britten, Lawrence Livermore National Laboratory, Livermore, California
A moving-zone Marangoni drying process can be integrated into an aqueous processing track to allow processing, rinsing, and drying of a single surface of a large flat substrate in one consolidated application. This ambient temperature process uses negligible amounts of organic material and surfactant-free deionized water to provide ultraclean drying of critical surfaces.
Aqueous cleaning has been increasingly used due to stricter environmental regulation of organic cleaning agents, the widespread use of aqueous processing in the semiconductor, flat panel display, and optics industries, and the increasing need for cleanliness brought about by decreases in pattern geometries and the push for increased yields. The stringent demands all point to the need for an ultraclean drying process that removes all water and contaminants from critical surfaces. Water either adheres to a clean hydrophilic substrate or beads up on a hydrophobic one. Marangoni drying uses surface tension gradients in a thin aqueous film to induce a film of water to flow off the surface, leaving it completely dry. This drying method is intrinsically cleaner than those using heat, forced air, or high-speed motion of the drying object. The evaporation that occurs in the latter processes can leave behind nonvolatile contaminants. In Marangoni drying, the liquid flows off the surface, carrying with it nonvolatile contaminants and entrained particles.
Liquid flows induced by surface tension gradients are generally known as Marangoni flows, in honor of a 19th-century Italian physicist who is credited with their elucidation [1]. Marangoni flow applied to ultraclean rinsing and drying was first described by Leenaars et al. [2] in the context of drying Si wafers emerging from a quiescent rinse tank by blowing solvent vapors entrained in a nitrogen flow onto the surface of the wafers. Recent work includes commercial applications to a drying system for Si wafer batch drying [3, 4], studies on the efficiency of Marangoni drying in production of extremely clean surfaces [2, 3], and fluid dynamics modeling of Marangoni drying [5].
Integrated wet processing, rinsing, and drying of single surfaces
We have been working on a method to use the Marangoni effects to dry large planar surfaces. We use an active rinsing flow near a passive source of a volatile organic compound (VOC, e.g., alcohol or ketone) that has a low to moderate vapor pressure at ambient temperature, some affinity toward water absorption, and significant capacity for lowering the surface tension of water in very low concentrations. The cleaning apparatus was for rinsing/drying the surfaces of large optics for high-power laser systems, but has application wherever ultraclean drying of water films from a large flat surface is required.
Our goal is to develop a process ideal for large, flat substrates in which the processing (e.g., cleaning, developing, or etching), rinsing, and drying take place simultaneously as the substrate moves relative to the units performing these functions [6]. Thus, the three processing steps are combined into one operation (Fig. 1). This geometry is similar to that of meniscus coating [7–9], in that the surface faces down and the rinsing/drying head travels underneath, contacting only the surface to be processed.
An applicator containing the cleaning, developing, or etching solution is placed in close proximity to the surface of a substrate. Liquid pumps into the inner chamber and overflows into an outer chamber. This applicator moves relative to the substrate to form a processing zone. The process contact time depends on the width of the applicator and the translation speed. A rinsewater applicator then irrigates the surface with pure deionized water, which drains away on both sides of the overflow weir. Along the trailing edge of the rinsewater applicator is a small reservoir containing a few milliliters of a low-molecular-weight alcohol or other suitable VOC (Fig. 1). An opening or slit at the top of this linear reservoir allows evaporated vapors to diffuse from the solvent source. Alternatively, we can use spray nozzles to introduce the liquid onto the surface, or bring the substrate in contact with the surface of a stationary pool of processing liquid that extends above the side of a hydrophobic container.
As the applicator assembly moves under the substrate, both processing fluid and rinsewater attach to the surface, forming menisci. A thin film of rinsewater is entrained on the substrate at the trailing edge of the applicator assembly. Evaporated VOC absorbed into this liquid film lowers its surface tension relative to the strongly curved region of the meniscus, where constant replenishment maintains the surface tension to that of essentially pure water. Flow induced by the strong surface tension gradient pulls the liquid film from the substrate into the falling film of the main overflow stream. Dissolved impurities and microscopic particles also flow back into the bulk liquid. This lateral film flow leaves the substrate completely dry within a few millimeters back from the falling water film. As a result, the substrate is processed, rinsed, and dried in one operation.
Drying effectiveness
We have constructed an apparatus to test the moving-zone Marangoni drying concept. A rinsewater applicator similar to that shown schematically in Fig. 1, 370-mm long, is supported on a crosspiece attached to linear rails. A pump, mounted on the crosspiece, pumps deionized water through a diffuser into the overflow reservoir. A surrounding chamber collects and pumps the overflow back into the overflow reservoir in a closed loop. A section of nylon rope placed in a channel, approximately 10 mm from the edge of the overflow weir, and saturated with solvent constitutes the solvent reservoir. Square glass substrate panels (300 × 300 mm) are held by a vacuum plate in which a 220 × 220 mm square opening provides a view of the drying phenomena through the glass. The plate surface is about 1.5 mm above the top of the applicator and solvent reservoirs, and a speed-controlled DC motor with a cable drive scans the applicator underneath the substrate.
A weak HF solution cleans and etches the glass panel prior to each set of experiments to ensure a hydrophilic surface. Three alcohols — ethanol, isopropanol, and sec-butanol — were tested for their effects on Marangoni drying, particularly the highest scanning speed at which the panel was completely dried. The effectiveness of the drying process can be easily observed through the back side of the glass plate. Under normal conditions, a visible, sharp contact line between the dry surface and the liquid film remains more or less stationary with respect to the moving applicator. This line is either flat or corrugated with water droplets resembling "wine tears" on the side of a glass of wine. If the scan speed is too fast, irregular patches of very thin water films or small water droplets remain behind on the glass. These films can be observed via the moving interference fringes produced from reflected light as they continue to dry.
The ability of the various alcohols to accomplish the drying process varies. The table gives the maximum scanning rate at which the applicator can leave behind a surface visually free of retained water. In general, isopropanol slightly outperforms ethanol, and both are more effective than sec-butanol.
 |
The maximum drying rate is greatly influenced by the overall water flow rate for a specific applicator design, and thus appears to be controlled by the liquid flow in the immediate vicinity of the meniscus. The rate will, therefore, depend on the applicator design as well.
Improved drying rates: Active irrigation vs. passive diffusion
The above observations suggest that transport of the absorbed solvent vapor from the thin film to the bulk liquid is the rate-controlling process in Marangoni drying. This implies that systems using passive introduction of solvent vapors and active irrigation of the surface in the immediate vicinity of the entrained thin film can achieve faster drying rates than systems that actively supply solvent vapors to the meniscus region, but drain the liquid from below and rely on passive diffusion of absorbed vapors from the surface layer to the bulk liquid [3]. Data from Wolke et. al. [3] suggest drying rates that are about 10 times slower than those reported here. (They did, however, report on a batch process in which many wafers are dried simultaneously.)
The ratio of diffusion coefficients of solvent vapors in air to those of dissolved solvent molecules in water is on the order of 104:1 [10], again suggesting that transport in the liquid phase is rate limiting. Furthermore, most organic substances in water are preferentially adsorbed at the surface [11]. Therefore, a system that continually exposes a fresh surface by active irrigation works better than one which relies on diffusion into the bulk liquid from a quiescent surface — active irrigation transports more adsorbed vapors and thus maintains a stronger surface tension gradient at the meniscus. The speed at which drying occurs is directly related to the strength of this surface tension gradient.
Preferential accumulation of adsorbed vapor in the water surface layer is proportional to the length of the hydrocarbon chain. Such accumulation may also explain why 2-butanol is not as effective in Marangoni drying as 2-propanol, even though the decrease in surface tension as a function of the molar concentration of adsorbed alcohol is much more pronounced with 2-butanol than with 2-propanol [11]. Other factors, such as the vapor pressure and gaseous diffusion rates, cannot be discounted, however. A fully coupled, two-dimensional gas and liquid-phase transport model, including the effects of surface layer accumulation, is required to understand fully the complexities of Marangoni drying in the region of an active free-surface flow.
Crude modeling estimates of ethanol evaporation and transport into the water film suggest that on the order of 0.1 g (0.15 ml) of alcohol is used during drying of a 30 × 30 cm square plate. This amount is much smaller than amounts used in other methods such as forced convection or droplet atomization to introduce VOCs to the drying surfaces.
Drying of vertical surfaces
Marangoni drying is also effective when the substrate is positioned vertically and the rinsing/drying head moves from top to bottom of the workpiece. Figure 2 shows a cross-sectional schematic of the apparatus in a vertical configuration. This figure depicts both surfaces being rinsed and dried. In the proof-of-principle experiments described below, only one face was treated.
 Figure 2. Cross-sectional schematic showing Marangoni drying of vertical surfaces. The substrate is moving up in relation to the applicator heads. |
Proper introduction of water flow onto the surface is the main problem to be overcome with this orientation. The irrigation rate must be uniformly high across the surface to maintain a high surface tension gradient over a short distance. However, introduction of water at these high rates often results in splashback above the drying line or flooding of the vapor source. By modifying a commercially available, off-the-shelf air knife (Exair Corp., Cincinnati, OH), we can introduce an adequate water flow rate in the proper geometry. In the optimum configuration, the air knife is held upside down from its intended orientation for air flow purposes. A slot gap of 0.1 mm disperses a flow rate of about 4 liters/min uniformly across the 45-cm active length of the air knife. The water is fed to one end of the knife. A gap of about 1 mm exists between the wetted face of the applicator bar and the surface to be rinsed and dried. A source of vapor (again a section of nylon rope saturated with solvent) is mounted on the top of the knife, with a standoff distance of about 3–4 mm from the face of the optic. The air knife applicator moves downward from the top of the optic so that the water continuously flushes the face of the optic.
In laboratory tests, isopropanol, acetone, and methanol all proved to be effective in establishing the drying effect on the hydrophilic surface of a 40 × 40 cm borosilicate square glass plate in a vertical orientation. Scanning rates (velocity of the rinse/dry head relative to the optic surface) of greater than 4 mm/sec leave the surface completely dry. Due to a lack of smooth motion capability in these manual tests, an upper limit to the scanning rate beyond which water remains behind on the surface has not been established.
Conclusion
The moving-zone Marangoni rinsing/drying process described above uses active irrigation coupled with close placement of an organic vapor source in the immediate vicinity of the solid-liquid-vapor phase contact line. We obtained simultaneous rinsing and drying along this contact line as the surface of the substrate moved relative to the irrigation source. Active flushing of the surface occurs during the drying process and the last layer of water flows off instead of evaporating. Moving-zone Marangoni rinsing/drying is an ambient temperature process where consumption of VOCs is negligible. Passive evaporation and diffusion of these vapors is confined to the immediate vicinity of use, and no elaborate solvent delivery system is necessary. The moving-zone Marangoni drying process works effectively with substrates vertically oriented and horizontally oriented with the working surface facing downward. In the latter orientation, the processing fluids and rinsewater only contact one surface, and aqueous processing (cleaning, developing, or etching) rinsing, and drying can occur simultaneously in a linear track arrangement. Drying rates appear to be limited more by removal of absorbed vapors from the vicinity of the thin film into the bulk liquid than by transport of the vapor through the air gap and into the film.
Acknowledgment
This work was performed under the auspices of the US Department of Energy by the Lawrence Livermore National Laboratory under contract no.W-7405-Eng-48.
References
- L.E. Scriven, C.V. Sternling, "The Marangoni Effects," Nature, Vol. 187, pp. 186–188, 1960.
- A.F.M. Leenaars, J.A.M. Huethorst, J.J. van Oekel, "Marangoni Drying: A New Extremely Clean Drying Process," Langmuir, Vol. 6, pp. 1701–1703, 1990.
- K. Wolke, B. Eitel, M. Schenkl, S. Rümmelin, R. Schild, "Marangoni Wafer Drying Avoids Disadvantages of Spin Drying or Alcohol Rinse," Solid State Technology, Vol. 39, No. 8, pp. 87–90, 1996.
- R. Schild, K. Licke, M. Kozak, "Marangoni Drying: A New Concept for Drying," Proc. 2nd Intl. Symp. on Ultra-Clean Processing of Silicon Surfaces, September 19–21, 1994.
- S.B.G.M. O'Brien, "On Marangoni Drying: Nonlinear Kinematic Waves in a Thin Film," J. Fluid Mech., Vol. 254, pp. 649–670, 1993.
- "Moving Zone Marangoni Drying Process," U.S. Patent No. 5660642, 1997.
- J.A. Britten and I.M. Thomas, "Sol-Gel Metal Oxide and Metal Oxide/Polymer Multilayers Applied by Meniscus Coating," in Laser Damage in Optical Materials-93, 25th Laser Damage Symposium, Boulder, CO, SPIE Vol. 2114, pp. 244–250, October 26–28, 1993.
- J.A. Britten, "A Simple Theory for the Entrained Film Thickness During Meniscus Coating," Chemical Engineering Communications, Vol. 120, pp. 59–71, 1993.
- H. Bok, "Coating Efficiently from Beneath," Photonics Spectra, Vol. 28, pp. 143–146, 1994.
- J. Marra, J.A.M. Huethorst, "Physical Principles of Marangoni Drying," Langmuir, Vol. 7, pp. 2748–2755, 1991.
- H.P. Meissner, A.S. Michaels, "Surface Tensions of Pure Liquids and Liquid Mixtures," Industrial and Engineering Chemistry, Vol. 41, pp. 2782–2787, 1949.
 |
Jerald A. Britten received his BS degree in chemical engineering from Michigan State University in 1979, and his PhD degree in chemical engineering from the University of Colorado in 1984. He is group leader of the Diffractive Optics Group of the Laser Program at Lawrence Livermore National Laboratory, which focuses on the design and manufacture of large-aperture diffractive optics for high-intensity lasers and other applications. Lawrence Livermore National Laboratory, 7000 East Ave., L-439, Livermore, CA 94550; ph 510/423-7653, fax 510/422-5537, e-mail [email protected].