Deposition of Thin Polymeric Films from Liquid Carbon Dioxide Using

Brian J. Novick,† Joseph M. DeSimone,†,‡ and Ruben G. Carbonell*,†. Department of Chemical Engineering, North Carolina State University, Riddi...
0 downloads 0 Views 220KB Size
Ind. Eng. Chem. Res. 2004, 43, 515-524

515

Deposition of Thin Polymeric Films from Liquid Carbon Dioxide Using a High-Pressure Free-Meniscus Coating Process Brian J. Novick,† Joseph M. DeSimone,†,‡ and Ruben G. Carbonell*,† Department of Chemical Engineering, North Carolina State University, Riddick Hall, Raleigh, North Carolina 27695-7905, and Venable and Kenan Laboratories, Department of Chemistry, University of North Carolina, Chapel Hill, North Carolina 27599-3290

Free-meniscus coating processes can be used to deposit a wide variety of coatings. However, the physical properties of the coating solutions often lead to the deposition of nonuniform films. Recently, it has been recognized that compressed carbon dioxide can be used as an environmentally benign solvent for industrial processes. We investigate the use of liquid carbon dioxide as the solvent in free-meniscus coating processes because its physical properties are much different from standard coating solvents. The surface tension and viscosity of liquid carbon dioxide are an order of magnitude smaller than those of typical solvents. Additionally, the density of liquid carbon dioxide is strongly dependent on temperature and pressure. The Tallmadge fourforce inertial theory is used to demonstrate that these unique physical properties will result in the formation of thinner films at the same withdrawal velocities as those used with conventional solvents. We then demonstrate experimentally that process variables can be controlled in a highpressure coating chamber to deposit films of a perfluoropolyether lubricant with controlled thickness in the range of 25-350 Å on a silicon surface with a native oxide. It is shown that the withdrawal velocity, polymer solution concentration, and evaporation rate can be used to control the coating process to produce submicron films on a surface. Introduction The deposition of ultrathin films required for the advancement of the magnetic storage and microelectronics industries poses a technological challenge that might be overcome by developing novel coating techniques. The ability to deposit ultrathin lubricant films is one of the applications that have become important in both of these industries. Currently, information on a hard disk is stored on magnetic platters with three primary coated layers: the magnetic layer used to store the data, a hard protective overcoat to prevent damage to the magnetic layer, and the lubricant layer to reduce friction/stiction (Figure 1).1 Increasing the drive capacity necessitates bringing the read/write head closer to the magnetic layer. This means that the total thickness of future protective coatings, including the lubricant, will need to be less than 50 Å thick.2-7 The lubricant is normally a perfluoropolyether (PFPE)-based polymer that is applied by dip-coating, a free-meniscus process, from solvents such as FC113, tetrahydrofuran, toluene, methyl ethyl ketone, and methyl isobutyl ketone.8-12 As the storage density of magnetic hard drives continues to increase and the distance between the read/write head and the storage platter continues to decrease, this lubricating film becomes more important. Novel deposition techniques and new lubricants are required to meet future storage densities.13 In addition, the ability to protect microelectromechanical (MEM) systems with thin lubricant films has become important to the microelectronics industry because the number of microscopic moving components * To whom correspondence should be addressed. Tel.: (919) 515-5118. Fax: (919) 515-5831. E-mail: [email protected]. † North Carolina State University. ‡ University of North Carolina.

Figure 1. Layers in a magnetic hard disk.

on these devices has increased.14 It has recently been shown that MEM devices, which contain moving parts, suffer from the same tribological limitations as hard disks.13-22 Lubricant coatings similar to those used in the magnetic storage industry that have been introduced commercially include perfluorodecanoic acid and phenylsiloxane.13-22 Other fluorinated compounds, including the PFPEs used in the hard disk industry, are also being investigated.15 Deposition of the thin protective film within the small features present in a MEM device is even more problematic than deposition of thin films onto a magnetic hard disk platter. Not only must a very thin film be deposited, but it must be done without causing collapse of the fine surface features. Deposition techniques that are under investigation include low surface energy plasmas, chemical vapor deposition, physical vapor deposition, and self-assembled monolayers.16,17,20,21 Carbon dioxide in the supercritical state has already been used to dry MEM devices after they have been produced.23,24 However, liquid CO2 has not been used to apply thin protective films to MEM devices or magnetic hard disks. In this paper, we demonstrate that the physical properties of liquid carbon dioxide may be used advantageously as an alternative coating solvent for the formation of thin films by free-meniscus coating (FMC).

10.1021/ie030688z CCC: $27.50 © 2004 American Chemical Society Published on Web 12/11/2003

516

Ind. Eng. Chem. Res., Vol. 43, No. 2, 2004

Figure 2. Types of free-meniscus coaters: (a) drainage; (b) continuous; (c) withdrawal; (d) slot; (e) “true” roll.25

We describe a novel high-pressure FMC (hFMC) apparatus that has been used successfully in producing uniform thin films of fluorinated lubricants on model silicon surfaces. The goal here is to demonstrate that carbon dioxide can be used to apply ultrathin lubricant films, but we have not determined the effectiveness of the lubricants on magnetic disk platters or on MEM devices. Withdrawal theory is used to demonstrate that the deposition of films from liquid CO2 has numerous advantages including higher uniformity. We then demonstrate experimentally that process variables can be controlled in a high-pressure coating chamber to deposit films of a PFPE lubricant with controlled thickness in the range of 25-350 Å on a silicon surface with a native oxide. It is shown that the withdrawal velocity, polymer solution concentration, and evaporation rate can be used to control the coating process to produce submicron films on a surface. FMC Processes. FMC processes use an interfacial boundary to determine film properties such as thickness and microarchitecture.25 This phase boundary can be a liquid-gas interface or the interface between two different liquids, such as a polymer melt and a solvent. Some examples of FMC processes such as drainage, withdrawal, slot, “true” roll coating, and continuous coating are shown in Figure 2. This paper concentrates on the drainage and withdrawal processes, also known as “dip coating”. In the drainage version of the dipcoating process, the solution is drained at a constant velocity, leaving a wet film on the substrate from which the solvent evaporates to the atmosphere. The evaporation of the solvent results in the deposition of a thin coating of polymer or other coating material onto the surface. In the withdrawal version of the dip-coating process, the solid is pulled out of the coating solution at a constant rate. In these processes, the surface tension, density, and viscosity of the solvent play a major role in determining the film thickness. FMC techniques are typically used to apply protective coatings, lubricants, electronic coatings, optical coatings, and other surface modifications.25 The major advantages of FMC devices include the ability to coat an entire substrate in one step, the relatively simple and inexpensive equipment, and the ease of reuse of coating solutions. However, FMC processes are difficult to

control because uneven evaporation rates can result in local solution physical properties that adversely affect film uniformity. In addition, the leading and trailing edges of the coated material are subject to nonuniformities caused by uneven wetting or varying surface tension. Liquid holdup on edges and the need for heat treatment to remove residual solvent from the surface are also problematic.25-28 Even though some of the challenges associated with the deposition of thin films by FMC may be overcome by choosing a solvent with appropriate physical properties such as viscosity, vapor pressure, etc., the range of physical properties of typical polar and nonpolar solvents and the ability to control them at normal operating conditions are limited. In the section that follows, we compare the properties of liquid carbon dioxide to those of conventional liquid solvents and we examine how these properties can affect the coating performance. Properties of Liquid CO2 and Its Effect on FMC Applications. No prior work has been done on the development of liquid CO2 based FMC processes. However, the use of compressed carbon dioxide in other states and for other coating applications has been investigated.29 Saturated liquid CO2 has a density that can be tuned from 467 kg/m3 at the critical point to 927 kg/m3 at 0 °C (Figure 3a). Cooling the liquid CO2 below the critical temperature results in a lower operating pressure for coating processes, potentially reducing equipment capital costs as well as operating costs. The density increase at lower temperatures also improves the solubility of solutes in liquid CO2 beyond the solubility exhibited near the critical point. It has been shown that the solubility parameter of liquid CO2 at temperatures below the critical temperature can be as high as that of supercritical CO2 at high temperatures and pressures where the density is equally high.30 Fluorinated polymers, such as the PFPE lubricants, and siloxanes are known to exhibit extremely high solubility in liquid and supercritical CO2 compared to other polymers with comparable molecular weight. As a result, they are particularly well-suited for any coating process that utilizes CO2 as a solvent.30-33 The low viscosity of saturated liquid CO2 (99.394 µPa‚ s at T ) 0 °C to 41.109 µPa‚s at T ) 25 °C) can enhance flow rates and diffusive transport of a coating solute, both of which aid in the formation of thin conformal coatings (Figure 3b). In coating processes, the low solution viscosity results in thinner films because viscous shear forces are considerably reduced when compared to normal solvents. The low viscosity can also help reduce crawling and promote leveling. The diffusion coefficient of solutes in liquid CO2, about 1-2 orders of magnitude higher than that in typical liquid solvents, can decrease concentration gradients in the liquid phase and increase the transport of solutes to the substrate during film formation.34-37 Carbon dioxide is also able to dissolve and diffuse quite readily in a large number of polymeric materials. This large diffusivity through a nearly dry polymer can minimize the amount of residual solvent left on the surface and reduce the time and energy required to drive the solvent from the film. The low surface tension of saturated liquid CO2 (0.004 540 3 N/m at T ) 0 °C to 0 N/m T ) Tc) also allows it to wet essentially all known surfaces (Figure 3c). The low liquid-vapor interfacial energy of liquid CO2 indicates that it is one of the few solvents that will wet solids with extremely low interfacial energies, such

Ind. Eng. Chem. Res., Vol. 43, No. 2, 2004 517

Figure 4. Important flow regions in FMC processes.

wetting properties can be readily exploited. The low surface tension of CO2 will reduce crawling and wetting, help prevent collapse of materials with fine features, and help reduce solvent retention. The low surface tension of liquid CO2 should also allow it to better penetrate into small pores and rough or patterned surfaces.38 Higher penetration and increased wetting will lead to better adhesion and bonding between the substrate and the coating. The better penetration of CO2 into a porous material should allow the film to be deposited deeper into a porous substrate. Even though the properties of liquid CO2 seem to be ideal in many respects for its use as a solvent for coatings applications, very little work has been done to develop high-pressure coating processes that would take maximum advantage of these unique properties. Some preliminary work has shown a great deal of promise for the use of liquid CO2 in spin-coating CO2-soluble photoresists for use in a novel dry photolithography process.39 However, very few other processes that utilize liquid carbon dioxide instead of supercritical carbon dioxide have been investigated. In the following two sections, we will first present theory to show the specific benefits of liquid carbon dioxide for free-meniscus processes and then demonstrate experimental techniques to deposit films from liquid carbon dioxide. Figure 3. Physical properties of compressed carbon dioxide and typical coating solvents: (a) density of coating solvents; (b) viscosity of coating solvents; (c) surface tension of coating solvents.46 0: carbon dioxide. O: water. 4: ethyl acetate. f: toluene. Upper shaded ]: 1,1,2-trichlorotrifluoroethane. Tilted 5: acetone.

as fluorinated polymers like poly(tetrafluoroethylene) (PTFE; γsv ≈ 19-23 dyn/cm). Of course, this low liquidvapor surface energy also allows it to wet solids with extremely high interfacial energy, such as glass or silicon oxide. All fluids near their critical points exhibit surface tensions of approximately zero, so this property is not unique to CO2. However, the readily accessible critical point of CO2 makes it extremely convenient to design high-pressure coating operations in such a way that these extremely favorable solvent penetration and

Theoretical Prediction of Liquid CO2 Behavior in Dip-Coating Processes Theories to estimate the amount of solution entrained in withdrawal/drainage processes include those developed by Landau and Levich, Deryagin et al., Brinker and Hurd, and others.25-27,36-40 Tallmadge’s four-force inertial theory (FFIT) is an extremely convenient tool to estimate the amount of fluid entrained on a substrate during withdrawal processes.40 The FFIT breaks the withdrawal/drainage process into three regions as the entrained film interfacial boundary passes over the substrate, as indicated in Figure 4. The nonconstant film thickness region (NCFTR) is parabolic in shape and occurs near the leading-edge drying line. If the substrate

518

Ind. Eng. Chem. Res., Vol. 43, No. 2, 2004

is long, then a constant film thickness region (CFTR) develops as a result of the balance among inertial, gravitational, and viscous forces. At the bottom of the substrate, the entrained film contacts the bulk fluid through a meniscus region whose shape is strongly dependent on capillary or surface tension forces. Typically, a coating system is designed so that the CFTR covers most of the surface to ensure a uniform coating. The Tallmadge theory is valid for a Newtonian fluid at Capillary numbers of less than 40. The theory is derived by assuming that the fluid flow in the liquid film is one-dimensional and that it results from viscous, gravitational, capillary, and inertial forces. The pressure gradient in the flow direction is assumed to be governed by capillary forces with a small curvature approximation. Under these conditions, the Navier-Stokes equation along the flow direction (x) takes the approximate form:

FU

∂u ∂2u d3h ) γlv 3 + µ 2 - Fg ∂x dx ∂z

Figure 5. Theoretical CFTR film thickness predicted by the FFIT as a function of the withdrawal velocity for common coating solvents at T ) 25 °C.

(1)

Here u is the velocity at an arbitrary point within the entrained film, U is the drainage rate, r is the fluid density, γlv is the surface tension, and µ is the fluid viscosity. Soroka and Tallmadge were able to obtain an approximate solution to eq 1 for the thickness of the film in the constant thickness region during a continuous withdrawal process.40 Tallmadge found that the dimensionless thickness D was dependent on the Capillary number Ca and the fluid property number Fp

(

Ca ) 1.09D3/2 + D2 + 0.5 exp -

)

5.13Fp2 Ca4/3D

D2 (2)

Here Ca ) Uµ/γlv and Fp ) µ(g/Fγlv3)1/4. The constant region film thickness, H, is computed by using the definition of the dimensionless film thickness

D ) H(Fg/γlv)1/2

(3)

Experimental verification of the Tallmadge inertial theory shows that it can predict the film thickness in the constant thickness region with an accuracy better than 1.5%. Even though the Tallmadge inertial theory was developed for use during continuous withdrawal processes, it has been shown that it can be applied to unsteady withdrawal processes.41,42 Van Rossum et al. have shown that Jeffrey’s equation (4) can be used to estimate the time-dependent shape of the NCFTR so that the time-dependent thickness of the entrained film from the leading edge through the CFTR is

h(td) )

x

µx Fgtd

when h(td) < H

(4)

and

H ) D(Fg/γlv)-1/2

when h(td) > H

(5)

Comparison of CO2 and Typical Coating Solvents Using the FFIT. The FFIT can be used to identify important differences between coating with standard organic solvents and coating with saturated liquid CO2. Comparisons of the predicted constant region film thickness H between four typical coating

Figure 6. Theoretical film profile predicted by the FFIT as a function of the withdrawal velocity for common coating solvents at U ) 39 cm/s, T ) 25 °C, and td ) 128 ms.

solvents and liquid CO2 are depicted in Figure 5 for drain coating of a 5-cm-long plate at 25 °C. The thickness of the constant region as a function of the drainage rate for Capillary numbers of less than 0.1 has been calculated. The FFIT predicts that the film thickness of the entrained CO2 film is about 2.7 µm at a 10 mm/s drainage rate. This is significantly smaller than the theoretical entrained film thickness of other solvents at the same drainage rate. At faster drainage rates up to 600 mm/s, the constant region entrained film thickness is still significantly smaller for liquid CO2 compared to other solvents. The nonconstant thickness region theory in eq 2 also predicts formation of a much thinner film with CO2 than with other solvents (Figure 6). The differences in the behavior between liquid CO2 and other solvents can be understood by looking at the magnitude of the Capillary number and fluid property number in eq 2, which play such an important role in the FFIT theory. The low surface tension of CO2 near the critical point causes the Capillary number and fluid property number of liquid CO2 to be almost an order of magnitude higher than those of typical solvents at T ) 25 °C and Uw ) 25 mm/s (Figures 7 and 8). These changes in both the fluid property number and Capillary number explain the differences between the entrained film thicknesses predicted for carbon dioxide and those for organic solvent based systems. The differences between the dimensionless numbers are important for a second reason. The Capillary number is the ratio of viscous to capillary forces. At temperatures closer to the

Ind. Eng. Chem. Res., Vol. 43, No. 2, 2004 519

Figure 7. Dependence of the Capillary number on the temperature for typical coating solvents at U ) 25 mm/s.

Figure 8. Dependence of the fluid property number on the temperature for typical coating solvents at U ) 25 mm/s.

critical point, the Capillary number in carbon dioxide based systems will be several orders of magnitude higher than that in organic solvent based systems. However, at lower temperatures, the Capillary number in a carbon dioxide solvent system is similar to that of an organic liquid solvent system. The large change in the Capillary number in carbon dioxide based systems with temperature implies a certain measure of control of the coating process with this variable. This cannot be done effectively at these moderate conditions with organic solvents because the temperature does not affect the Capillary number in these cases. It should be noted that organic solvents should exhibit these advantages near their critical points. However, unlike carbon dioxide, the critical points for most organic solvents are at conditions that are not easily achieved. Although the temperature dependence of the viscosity and surface tension of liquid CO2 gives rise to the ability to change the relative importance of the forces that govern the coating process, temperature cannot be used to control the thickness of the coating (Figure 9). Over the temperature range from 0 to 25 °C, the predicted constant region entrained film thickness for CO2 changes by less than (0.9 µm with a minimum film thickness at about 10 °C when Uw is constant at 390 mm/s. This is in contrast to the temperature behavior observed for a typical organic solvent such as toluene in which the entrained film thickness does not reach a minimum. The small changes in entrained film thickness due to temperature changes in the carbon dioxide based systems are advantageous because changes in temperature that

Figure 9. Theoretical dependence of the CFTR thickness on the temperature for liquid CO2 and toluene predicted using the FFIT at Uw ) 390 mm/s.

occur during the entrainment process due to evaporation will have less of an effect on the film thickness. This can lead to more uniform films. It also means that temperature can be used to adjust the important forces governing the deposition without changing the film thickness or the microarchitecture. It is important to recognize that the predicted entrained film thickness results discussed here are only valid for low solute concentrations because they utilize the physical properties of the pure solvent. Additionally, the FFIT predicts the wet film thickness. By knowing the solute concentration and the wet film thickness, it is possible to estimate the dry film thickness to compare coatings deposited by liquid CO2 and organic coating solvents. The dry film thickness can be estimated from the FFIT only if it is assumed that the entrainment and evaporation processes are independent. In this treatment, it is assumed that the entrainment process is followed by evaporation of all of the solvent from the entrained wet film. The evaporation results in the deposition of all of the entrained solute on the substrate. The average dry film thickness, hdry, is calculated by first using the FFIT to determine the CFTR thickness, ho. This thickness can be multiplied by the substrate surface area, As, and the mass solute concentration, Csol, to determine the mass of solute that is deposited, Mdep. The average film thickness is equal to the weight of the solute deposited divided by the dry film density, Fdry, and the substrate surface area.

Mdep ) hoAsCsol

(6)

hdry ) Mdep/FdryAs

(7)

Evaporation processes need to be considered to get a more complete picture of the differences between FMC with standard solvents and hFMC with liquid CO2. Most liquid coating devices utilize the difference in solvent concentration in the gas phase (usually air) at the liquid-gas interface and the concentration of solvent in the bulk gas to drive evaporation.25,36-38 In this case, the flux of solvent at the interface Js is due to convection so that it can be written in terms of a mass-transfer coefficient km and the difference between the concentration of the solvent at the interface c/s and that in the bulk solution cs∞.

Js ) km(c/s - cs∞)

(8)

520

Ind. Eng. Chem. Res., Vol. 43, No. 2, 2004

The mass-transfer coefficient is dependent on the Reynolds and Schmidt numbers for the solvent in the gas phase

Sh )

kmL ) Sh(Re,Sc) D

(9)

In most FMC processes, mass transfer is driven by natural convection and is not well controlled. In a hFMC apparatus, the vapor consists of pure solvent and there is no concentration gradient to drive mass transfer. Two mechanisms can be used to drive evaporation: First, a second component, such as helium, could be added to the gas phase to create a concentration gradient. However, this component can diffuse into the coating and create defects during film formation. This process also requires careful control of the flow rate of inert gas. Second, a pressure gradient can be used to drive evaporation if the chamber pressure is set at a value just below the vapor pressure of the solvent in the coating solution. This system provides more control because the system pressure can be adjusted quicker and more reliably than a concentration gradient. It also does not require addition and monitoring of an additional component. As a first-order approximation, pressure-gradientdriven evaporation of a species can be described by the Hertz-Knudsen equation.43

Js )

Rv 1 (P/ - P) Na x2πmkT s

(10)

The evaporation rate in this case is proportional to the difference between the vapor pressure of the evaporating species at the interface P/s and the system pressure P. However, the coefficient in front of the pressuregradient driving force is not a function of hydrodynamics or mass-transfer parameters; it is a function only of the thermodynamic properties of the solvent. Of course, heat must be transferred to the interface to drive evaporation in both cases. The flux of carbon dioxide as a function of the evaporation driving force, ∆P, in our experiments has been measured at T ) 25 °C and is presented in the Experimental Section. The authors are not aware of prior use of pressure-gradient-driven evaporation in dip-coating processes, but in principle it offers some advantages in control over diffusion-limited evaporation. The theoretical discussion above argues for the utility of carrying out hFMC applications in CO2, especially for coating processes involving fluoropolymers that are readily soluble in carbon dioxide. In the section that follows, we describe a novel experimental apparatus that has been used to carry out hFMC of a PFPE onto a silicon wafer using liquid carbon dioxide as a solvent. Experimental Section Materials. The PFPE used was Fomblin Z-DOL 2000 with a number-average molar mass of 2000 g/mol (NMR) and a structure given by HOCH2O[CF2CF2O]x[CF2O]yCF2CH2OH. Fomblin Z-DOL 2000 has a specific gravity of 1.81 g/mL, a viscosity of 154 cP, a surface tension of 24 dyn/ cm, a vapor pressure of 2 × 10-5 Torr, and a polydispersity of 1.5. In this case, x/y is from 0.5 to 2. The material was used as received from the manufacturer (Ausimont USA, Inc., Thorofare, NJ). This PFPE lubri-

Figure 10. Schematic of the hFMC: VS1, high-pressure solution vessel; VS2, high-pressure coating vessel; V1, pneumatic control valves; V2, isolation valves; V3, flow rate control valve; V4, pneumatic flow control valve; P2, high-pressure circulation pump; P2, ISCO syringe pump; S1, carbon dioxide source tank; F1, feedback control loop; PT, pressure and temperature transducers.

cant was chosen for the coating formulation because it is currently used in the production of magnetic media such as hard drives and mass storage media. The subtrates consisted of 125-mm-diameter type 〈100〉 or type 〈111〉 silicon wafers obtained from Silicon Valley Microelectronics. The polished wafers were 600-650 µm thick and grown by the Czochralski growth method with a resistivity of 1-10 Ω‚cm. The wafers had either a 2700-Å thermal oxide layer or a 20-Å native oxide layer. The substrates were prepared as discussed below in the Methods and Procedures section of this paper. The carbon dioxide source consisted of a 17.73-kg size B(AI) SFC/SFE high-purity tank with no helium headspace. The carbon dioxide was obtained from Air Products and Chemicals, Inc., Allentown, PA, with a purity certified at greater than 99.9999%. Experimental Apparatus. A novel hFMC apparatus has been assembled to probe the effect of critical process variables including the polymer concentration, evaporation driving force, and withdrawal velocity on the final dry film thickness. A schematic of the device is shown in Figure 10. The apparatus consists of six key components: two operational vessels (VS1 and VS2), a circulation pump (P1), a vent valve (V4), two control valves (V1), a metering valve (V3), and an inlet pump (P2). Operational vessel VS1 is used for mixing the solution and as a storage vessel during the drainage process. It is made from type 316 stainless steel and contains 1/8and 1/4-in. gas- and liquid-side communication ports that connect to operational vessel VS2 for the drainage process. The vessel is also equipped with two sapphire view ports for visual inspection of the coating solution. These view ports are sealed with PTFE O rings. The vessel is cylindrical on the inside with a diameter of 17.8 mm and a volume of approximately 30 mL. The coating vessel (Figure 10, VS2) is similar to the solution vessel

Ind. Eng. Chem. Res., Vol. 43, No. 2, 2004 521

with two exceptions. First, the volume of the coating vessel is about 20% smaller than the volume of the solution vessel. This ensures that there will be enough empty volume remaining in the solution vessel during drainage to remove all solution from the coating vessel. It also ensures that there will be enough solution to cover the substrate completely because the entire coating vessel will be full before the drainage process begins. Second, the coating vessel contains a small device that can be used to hold the substrate either at the top of the vessel or at the bottom of the vessel. Both vessels were manufactured at the North Carolina State University Instrumentation Shop. The holder requires that a small portion of the substrate (2 × 14 mm2) remain unexposed to the coating solution. The high-pressure circulation pump (P1) is a gear pump manufactured by Micropump, Inc. It is used to mix the carbon dioxide with the solute in the solution vessel and to move the solution between the coating vessel and the solution vessel. The pump is capable of withstanding pressures of up to 5000 psi and can produce a maximum flow rate of liquid CO2 at 25 °C of about 100 mL/min. The vent valve is used to adjust the pressure in the coating vessel by controlling the vent rate of carbon dioxide. The valve is a special pneumatic metering valve (PMV) manufactured by Badgemeter, Inc. Control of the vent rate is accomplished through a computer-regulated feedback loop by combining the PMV with a high-accuracy pressure transducer ((0.1 psi) and a custom process control algorithm. The control valves (V1), manufactured by High-Pressure Equipment Co., are used to isolate the coating vessel from the solution vessel through the gas- and liquid-side communication ports. These high-pressure pneumatic solenoid valves are normally closed, and they are controlled by a computer throughout the coating process. The pneumatic valves (as opposed to electric valves) prevent excess heating of the fluid as it passes through the gas- and liquid-side communication tubing. The metering valve is used to adjust the drainage rate of fluid flowing from the coating vessel to the solution vessel. Two different metering valves were used depending on the required flow rate. Both valves are union bonnet structured metering valves, purchased from Swagelok, that require approximately 12 turns to achieve a fully opened state. The valve flow coefficient, or the number of gallons of water that will flow through the valve at 60 °F/min at a pressure drop of 1 psi, can be adjusted from between 0.005 and 0.04. In the following experiments, gravity drainage was used to control the withdrawal rate because the drainage rate is more precise and there are no mechanical vibrations. However, the Micropump can also be used to control the drainage rate if a flow-dampening filter is used. The inlet pump is a type 500D ISCO syringe pump equipped with high-pressure microporous fiberglass particle filters purchased from Matheson Tri-Gas to remove 100% of particles > 0.02 µm in size. Additional filters, manufactured by R&D Separations, are used to trap hydrocarbons and water. The syringe pump is connected through separate lines to both operational vessels and is used to pressurize the carbon dioxide from gas tank pressures of 700-860 psi to the system operating pressure. The complete system provides control of the drainage velocity to (0.02 mm/s and control of the evaporation driving force to within 0.1 psi.

Methods and Procedures. Both operational vessels were purged with carbon dioxide at 500 psi for several minutes. Extremely pure carbon dioxide, 99.9999% SCF/ SFE grade, was used and passed through the Matheson particle filter to remove dust and other materials that might interfere with the coating process. The solution vessel was then vented back to atmospheric pressure and loaded with a known quantity of Fomblin Z-DOL 2000. Carbon dioxide at low pressure was purged through the solution vessel during the loading process to prevent ambient air from entering the system. The weight of Z-DOL was measured on a Mettler Toledo AB204 microbalance to within 0.0001 g. The solution vessel was purged again after addition of the polymer at 500 psi for 1 min. The pressure in the solution vessel was increased to the saturation pressure with the SCF/ SFE-grade CO2 after the purge. The pumping was terminated when the vessel was filled with about 30 mL of liquid carbon dioxide at vapor pressure. Although the mixture becomes clear and appears to be well mixed after 30 min, the mixture was stirred for at least 18 h to ensure homogeneity. The type 〈100〉 or 〈111〉 silicon wafer substrates were broken into small coupons that were roughly 15 mm wide and 51 mm long. The coupons were each wiped with an acetone-soaked lint-free clean room wipe to remove any loose particles and contaminants. They were then placed in separate baths of a 50/50 acetone/water mixture. The baths were heated to 40 °C and agitated using an Aquasonic 750D ultrasonic cleaner for at least 30 min. The coupons were then rinsed with acetone, soaked in acetone for 18 h, blown-dry with nitrogen, and cleaned in a Jelight Co. model 42 Suprasil lamp UV/ ozone cleaner for 25 min. This device uses UV radiation to produce ozone and atomic oxygen. These species react with hydrocarbon impurities at the sample surface, hence making sure the substrate is thoroughly cleaned. The thickness of silicon dioxide was measured after cleaning using a Rudolph Technologies AutoEL II ellipsometer. This device operates at a fixed angle of 75° with a single wavelength of 632.8 nm. The coupons were coated with PFPE within a few hours of the final cleaning process. The loading of the coupon into the coating vessel utilized the same purge scheme described above for solution creation. The coating process consisted of several steps. The substrate was first placed in the prepurged coating chamber and the pressure raised to that of saturated carbon dioxide. The pressure of the two primary vessels was then equilibrated by opening the gas-side communication line. The coating chamber was filled with liquid from the solution chamber using the highpressure gear pump. After the solution contacted the substrate for approximately 30 min, the coating chamber pressure was decreased to achieve the experimentally specified ∆P. The gas-side communication valve was then opened, and a short equilibration period was used to ensure that ∆P was stable. The FMC process was then initiated by opening the liquid-side communication valve (V1, Figure 10). This caused liquid to drain into the solution vessel. The maintenance of ∆P during drainage provided the evaporation driving force. After the solution had fully drained from the coating vessel and the PFPE had been coated onto the substrate, the coating vessel was isolated from the rest of the system. The pressure was then lowered to atmospheric pressure at 2.5-5 psi/s using the PMV and the pressure

522

Ind. Eng. Chem. Res., Vol. 43, No. 2, 2004

Figure 11. Dependence of the average dry film thickness on the solution concentration for Fomblin Z-DOL deposited by hFMC with liquid CO2 onto type 〈100〉 and type 〈111〉 silicon wafers at Uw ) 1.5 mm/s and T ) 25 °C: (A) ∆P ) 7; (B) ∆P ) 0; (C) theoretical prediction estimated using the FFIT with liquid CO2 properties.

feedback control loop. The thickness of the dry deposited PFPE film was measured using a Rudolph Technologies AutoEL II ellipsometer at a fixed angle of incidence of 75° and a wavelength of 632.8 nm. A two-layer model that incorporated the previously measured thickness of the silicon dioxide layer was used. The refractive index of the film was assumed to be 1.300. Previous authors have demonstrated the validity of this assumption.11,44 Results and Discussion The dependence of the dry film thickness on the polymer concentration is depicted in Figure 11. An increase in the PFPE concentration from 0 to 3 wt % resulted in an increase in the film thickness from 0 to 300 Å when an evaporation driving force of 7 psi was used. The increase in the film thickness with concentration was linear. However, when an evaporation driving force of 0 psi was used (Figure 11B), the dry film thickness only increases from 17 to 38 Å with a concentration increase from 0 to 3 wt %. In fact, the film thickness increases with a slope that is 10 times smaller than that when the larger evaporation driving force of 7 psi was used. We attribute this to a difference in the primary driving force for deposition. The polymer left on the substrate when a 0 psi driving force is used is mostly due to adsorption of species prior to the entrainment step. This is because in the case of a 0 psi driving force there is almost no evaporation during withdrawal. After the solution is entrained, the solvent in the entrained film does not evaporate. This results in the solute and solvent draining back into the bulk liquid during the withdrawal process. The solute does not concentrate during the drainage step and is therefore not deposited onto the substrate during the withdrawal process. However, prior to the withdrawal step, the solution and solute are in contact with the substrate. Some of the polymer adsorbs to the substrate during this time period. After drainage, the only polymer that remains on the substrate is due to the attractive forces between the substrate and the solute. This causes the concentration dependence of the dry film thickness for the 0 psi driving force case to resemble an adsorption isotherm. Furthermore, the thickness of this adsorbed film without evaporation is roughly equivalent to a few

Figure 12. Dependence of the average dry film thickness on the withdrawal velocity for Fomblin Z-DOL deposited by hFMC with liquid CO2 onto type 〈100〉 and type 〈111〉 silicon wafers at C ) 0.51 wt %, ∆P ) 3 psi, and T ) 25 °C: (A) measured film thickness; (B) theoretical prediction estimated using the FFIT with pure liquid CO2 properties.

times the radius of gyration of PFPE in solution (12 Å).45 When an evaporation driving force is used, the polymer concentrates during drainage and deposits onto the substrate as the solvent evaporates. Eventually, all of the solvent evaporates and the remaining solute in the entrained film is deposited. When the coating thickness is thin, then both adsorption and deposition from the entrained film are significant. The effect of the drainage rate on the dry film thickness was also investigated. Figure 12 shows that the dry film thickness increases from 38 to 277 Å when the drainage rate is increased from 0.25 to 5.2 mm/s at a fixed polymer concentration of 0.55 wt % and room temperature. A theoretical average dry film thickness based on the FFIT theory can be estimated using the methods described above. The theoretical prediction and the experimentally measured points do not agree qualitatively. Instead, the experimentally measured curves of dry film thickness versus drainage rate exhibit an increasing slope as the drainage rate increases. This is contrary to the curve predicted using entrainment theory in which the slope of film thickness versus drainage rate continuously decreases as the drainage rate increases. The discrepancy between the theoretical fit and the experimental data may occur for three main reasons: First, a low drainage rate range is investigated. At very low drainage rates, the adsorption processes may dominate the deposition process. As the drainage rate increases, the adsorption process begins to have less significance and the entrainment process becomes more important. The decreasing slope may only be seen at high drainage rates far from the transition between deposition by adsorption and deposition by entrainment. Second, it is likely that some of the differences between the theoretical estimate and the experimentally measured values arise because of evaporation. The average dry film thickness prediction using the FFIT does not take into account that entrainment and evaporation occur simultaneously. In the real system, the solution is concentrating during the deposition process. This causes the solution physical properties to change significantly throughout deposition. At the beginning of the entrainment process, the solution physical properties are close to those of pure liquid CO2 because the solution is dilute. The theoretical predictions use the physical properties of pure liquid CO2.

Ind. Eng. Chem. Res., Vol. 43, No. 2, 2004 523

Figure 13. Dependence of the average dry film thickness on the evaporation driving force for Fomblin Z-DOL deposited by hFMC with liquid CO2 onto type 〈100〉 and type 〈111〉 silicon wafers at Uw ) 1.5 mm/s, C ) 0.61 wt %, and T ) 25 °C.

Figure 14. Evaporation rate for pure carbon dioxide expressed as the difference between the vapor pressure and the system pressure, ∆P.

However, at the end of the experiment, the entrained film will consist mostly of polymer and will therefore have a much higher viscosity, density, and surface tension. Last, evaporation may also be important because it is possible that higher drainage rates actually cause an increase in the evaporation rate (see below). In the liquid CO2 system, the evaporation rate is dependent on the difference between the vapor pressure and the system pressure. A pressure drop is associated with an increase in the flowing fluid velocity. At higher drainage rates, this local pressure drop may cause a significant increase in the evaporation rate. The importance of increasing the evaporation driving force on the average dry film thickness has been investigated to help understand this phenomenon. Experiments were conducted to determine the dependence of the film thickness on the evaporation driving force at constant drainage velocity and polymer concentration (Figure 13). At ∆P between 0 and 4 psi, the average dry film thickness remains close to 25 Å. This is similar to results obtained in the experiments with variable concentration at a fixed evaporation driving force of 0 psi. The evaporation rate of pure carbon dioxide has been measured for these conditions and is depicted in Figure 14. This measurement was performed by measuring the quantity of bulk pure liquid CO2 lost over time at each ∆P. At ∆P between 0 and 4 psi, the evaporation rate is less than 0.03 mg/cm2‚s and the film

thickness only increases slightly as the pressure increases. Under these conditions, most of the solution drains back into the bulk before Z-DOL can deposit onto the substrate and adsorption is the main driving force for deposition. When the evaporation driving force is large, 4-14 psi, then the dry film thickness quickly increases from 46 to 222 Å. The quick increase in the film thickness with increasing ∆P can be attributed to two effects. First, the increase in evaporation causes the entrained film to dry quicker. This ultimately causes more material to be deposited onto the substrate because there is less time for drainage. Second, the evaporation causes the remaining entrained film to become concentrated with polymer. This increases the viscosity of the entrained film, causing slower drainage. This may also be exacerbated by local decreases in temperature and changes in other solution physical properties due to thermal effects during solvent evaporation. These experiments show that only a small increase in the evaporation driving force is needed to cause a significant increase in the average dry film thickness. It is therefore likely that this phenomenon plays a role in causing the average dry film thickness to increase significantly at higher withdrawal velocities as mentioned above. When ∆P is greater than 4 psi, even small changes in the withdrawal velocity can change the apparent evaporation driving force enough to cause a significant increase in the film thickness. Better control of the hFMC process can be achieved by using smaller evaporation driving forces,