Improving superhydrophobic textile materials
Since the wettability of a solid surface is determined by two parameters, the chemical composition and the geometrical structure of a rough surface, the combination of these two factors are used for the development of superhydrophobic surfaces. In this chapter, the relationships among contact angles, surface tension and surface roughness are reviewed; the physical surface modifications for the design of superhydrophobic surfaces are discussed; the wetting behavior of a rough surface is compared with that of a smooth surface; the relationship between the contact-angle hysteresis and the roll-off angle is analyzed, and the preparation of superhydrophobic surfaces using textile structures is discussed.
Science and technology related to superhydrophobicity have recently attracted considerable attention due to their potential applications in medical devices as well as in industrial materials. The idea of superhydrophobicity was introduced in 1940s by A. Cassie who was interested in enhanced water repellency.1 This amazing water repellency has been used in the textile industry ever since.
A superhydrophobic surface is defined as having a water contact angle greater than 150°.2 The high contact angle is obtained by a combination of surface chemistry and surface roughness, while the roll-off angle depends on the droplet size and the receding contact angle.3 A water droplet easily rolls off a superhydrophobic surface, such as lotus-leaves, washing dirt off in the process and effectively cleaning the surface. This unusual wetting behavior is called self-cleaning or the ‘lotus effect’ although there are two different types of water-repellent plant leaves: the first type is macroscopically smooth leaves such as the lotus, and the second type is hair-covered leaves such as lady’s mantle.4 Water droplets completely run off both plant leaves even after heavy rain. Although this phenomenon is observed on other plant leaves besides the lotus plant leaves, the ability has been termed the lotus effect. For these surfaces, since water droplets roll off easily, they remove dirt off the leaves and effectively keep the surface clean and dry.5
This chapter focuses on the physical surface modification and design of superhydrophobic surfaces, compares the wetting behavior of a rough surface with that of a smooth surface, analyzes the relationship between the contact-angle hysteresis and the sliding angles of water droplets, and discusses the preparation of superhydrophobic surfaces using textile structures. Since the wettability of a solid surface is determined by two parameters, the chemical composition and the geometrical structure of a rough surface, the combination of these two factors are used for the development of superhydrophobic surface.6 On the other hand, in this chapter, a superhydrophobic rough surface is designed using a plain woven and hydroentangled non- woven structure. In addition, the wetting behavior of the superhydrophobic woven fabric is compared with that of other materials having a flat surface or a rough surface made of a woven fabric with monofilament yarns.
Although it is hard to measure the surface tension of a solid directly, it is easy to measure its contact angles (Fig. 15.1). Therefore, the contact angles of polymers and organic layers can be used for the prediction of surface tension and wetting behavior for various liquids.7−9
The relation between the surface tension and contact angle is obtained by the Young equation:10
where γ is the surface tension and SV, SL and LV are the solid–vapor, the solid–liquid and the liquid–vapor interfaces, respectively (Fig. 15.2). According to Young’s equation, the contact angle is a well-defined property that depends on the surface tension coefficients of solid, liquid and gas.
The right-hand side of equation [15.1] and γLV can be obtained from experimental measurements, leaving two unknowns, γSV and γSL. When θe for a test liquid > 20°, it is assumed that γSV ≈ γS and γLV ≈ γL and equation [15.1] can then be reformulated as:11
According to Fowkes, the interfacial tension between solid and liquid is given by the following equation when only dispersion interactions are present:12
The geometric mean of the liquid and solid surface tension is used to calculate the thermodynamic work of Lifshitz-van der Waals (LW) components:13
where d, p, H, ind and m mean London dispersion forces, permanent dipoles, hydrogen bonds, induced dipoles and metallic interaction, respectively. Since the first three components in equation [15.7] are the major factors determining the surface tension of most materials, we can use these group contribution methods to calculate γS and γL:
The Young equation is valid for only wetting of smooth surfaces, but real solids are not perfectly flat. The surface structure affects wettability, for example when a surface of solid is superhydrophobic, liquid droplets are in contact with the upper part of a rough surface and the lower part is filled with air. In this chapter, we study the apparent contact angle of a drop deposited on a textured surface, and finally characterize a hydrophobic surface.
Using the Young equation, equation [15.11] can be reformed as:
Equation [15.12] defines a critical contact angle between zero and π/2, since r ≥ 1 and s ≤ 1. For example, the surface should be completely wet when r approaches one since the contact angle becomes zero, but the contact angle will be close to 90° when r > > 1 or r > 1 and s < < 1.
Roughness makes a significant contribution to the wetting behavior of a surface.16−18 When the surface is roughened the minimization of liquid surface free energy results in two possible contact angles, the Wenzel apparent contact angle or the Cassie–Baxter apparent contact angle.19,20 Figure 15.4 shows the apparent contact angle on a rough surface.
In Wenzel’s approach the liquid fills the grooves on the rough surface (Fig. 15.5a). According to Wenzel, the liquid contact angle at a rough surface can be described as:
Here, r is the ratio of the total wet area of a rough surface to the apparent surface area in contact with the water droplet (r > 1). If the Young contact angle is smaller than a critical contact angle θc, the liquid is sucked into contact with the rough surface. According to equation [15.13], for large r, the rough surface is dry when the contact angle on a flat surface exceeds 90°.
The Cassie and Baxter model is an extended form of the Wenzel model to include porous surfaces. In this model a liquid sits on a composite surface made of a solid and air. Therefore, the liquid does not fill the grooves of a rough solid. In their paper published in 1944, Cassie and Baxter suggested that:
where f1 is the surface area of the liquid in contact with the solid divided by the projected area, and f2 is the surface area of the liquid in contact with air trapped in the pores of the rough surface divided by the projected area. When there is no trapped air, f1 is identical to the value of r in the Wenzel model. Recognizing this, equation [15.14] has recently been rewritten as follows:
where f is the fraction of the projected area of the solid surface in contact with the liquid and rf is defined by analogy with the Wenzel model.21 It is important to note that rf in equation [15.15] is not the roughness ratio of the total surface, but only of that in contact with the liquid. In this form of the Cassie–Baxter equation, the contributions of surface roughness and of trapped air are much clearer than in the other forms of the equation.
Recently, many authors used another approach for the Cassie–Baxter equation to describe contact angles of droplets on heterogeneous rough surfaces that have composite interfaces.22 In the modified Cassie and Baxter model the liquid forms a composite surface made of solid, liquid and air; and the liquid does not fill the grooves on the rough surface (Fig. 15.5b).23 When the top of a rough surface is completely flat, the following equation describes the apparent contact angle on a rough surface:24
where θrCB is the apparent contact angle at a heterogeneous rough surface composed of two different materials and θ1 and θ2 are the droplet contact angles on the two surfaces. A unit area of the surface has a unit surface area fraction 1with a contact angle θ1 and an area fraction 2 with a contact angle θ2. When this rough surface consists of only two materials 2 = 1 – 1. If the liquid does not completely wet the surface, 2 represents the trapped air with θ2 = 180°. Equation [15.14] can be modified as:25
where S is the ratio of the rough surface area in contact with a liquid drop to the total surface covered by a liquid drop. Smaller S increases θrCB and makes the surface more hydrophobic. For an apparent contact angle of water on Teflon™ to be greater than 150° (superhydrophobic), the fraction of the surface in contact with water must be less than 26%.26
Although most studies of the lotus effect have been carried out on inorganic materials, textile materials having a rough surface can also be superhydrophobic by the lotus effect.27,28 In this case, the protruding fibers of woven or knitted fabric, flock fibers of flock fabric, or surface fibers of non-woven fabric can be regarded as grooves of rough surfaces.29 Figure 15.6 shows the upper-sectional view of a roughness pattern when the grooves are thought of as square pillars sticking up from a fabric substrate.30,31
The pillar cross-sectional area is a2, the distance between two pillars is d, and the height of pillar is h. In an analysis of the superhydrophobic effect, Patankar has provided two equations to describe the surface based on the Wenzel and the Cassie-Baxter models:32,33
As a numerical example, if a smooth surface is hydrophobic (θe = 120°), the width of a pillar is 10 μm and the height of the pillar is 1 mm, the distance between two pillars, d, has to be 0.014 mm < d < 0.24 mm for a superhydrophobic surface.
When the volume of a liquid drop placed on the surface is steadily increased until the contact line advances, the contact line begins to move. The contact angle observed when it just begins to move is the advancing contact angle (θA). On the other hand, when the liquid droplet is retracted steadily until the contact line recedes, the contact line begins to move again. The contact angle observed when the contact line is just set in motion by this process is defined as the receding contact angle (θR).34 Alternatively, if the surface with a drop on it is slightly tilted, the drop remains with different contact angles at the front and the back of the drop. Barthlott and Neihuis suggested that the receding contact angle of a water droplet easily reaches the advancing contact angle on a self-cleaning rough surface when the surface is slightly tilted.35 Thus, the drop easily rolls off this surface, washing dirt off and cleaning the surface in the process as shown in the right side of Fig. 15.7.
The sliding angle of droplets, α, on smooth surface can be described as:36
where R is the radius of the contact circle, m is the mass of the droplet, g is the gravitational acceleration and k is a proportionality constant. Roura and Fort demonstrated the work due to the external forces on a drop in Fig. 15.8.37
Figure 15.8 shows that the advancing contact angle is always greater than the receding contact angle on a tilted surface. This condition was described by Furmidge as:
When the surface is tilted, the sliding angle increases until the drop begins to move and equation [15.21] can be expressed in terms of working energy. As mentioned before, it is assumed that the receding contact angle and the advancing contact angle seem to be close to their minimum and maximum values, respectively, when the drop begins to roll off (α = αc):
Equation [15.22] shows an energy balance when the solid surface is progressively inclined and the drop begins to move at αc. In this situation, gravity can supply the necessary energy to develop the back wetted surface, and thereby the energy used to create a unit area of this surface is – 2RγLV (cosθA(max) – cosθR(min)). Although the contact angle changes continuously along the contact line, equation [15.22] can be approximately calculated. The constant k in equation [15.20] is related to contact-angle hysteresis and the interfacial surface tension between water and vapor.
where ρ is the density of water and R′ is the radius of droplet. By multiplying gsinαc to both sides we can describe the relationship between the radius of the contact circle and the radius of droplet sliding on a smooth surface as:
where k′ is constant. Referring to equation [15.22], k′ is related to the radius of the contact circle, the interfacial surface tension between water and vapor, and the contact-angle hysteresis.
Contact-angle hysteresis, ΔθH, is defined as the difference between advancing contact angle, θA, and receding contact angle, θR, that is ΔθH = θA – θR. The gain factor, which is often used to understand the relationship between contactangle hystereses and roll-off angle, is considered as the rate of variation of the contact-angle hysteresis at any operating point.38 The Wenzel gain factor graph shows gain factors equal to the roughness factors for the region close to θe = 90°, and the gain factors dramatically increase on either side of this region. The Cassie–Baxter gain factors increase from zero to maximum value of one. The Wenzel equation gives a change in the Wenzel contact angle, ΔθHw ΔθHW, caused by a change in the contact angle on the smooth surface, ΔθH, as:
where GeW is the Wenzel gain factor, and is ∂θrW/∂θe. The gain factor is very useful since it separates the idea of the equilibrium contact angle increase occurring by surface topography from the observed contact angle. Using the Wenzel equation we can obtain the Wenzel gain factor as follows:
When a contact angle θe is close to 90° the Wenzel gain factor is approximately unity. Since the effect of roughness is proportional to the radian contact angle changes, the Wenzel gain factor rapidly increases as the roughness factor increases.
According to McHale, the gain factor, GeCB, is an attenuation of any contact- angle hysteresis, while hysteresis increases on a Wenzel-type surface. As a numerical example, if the average contact angle on the smooth surface is 70° and the contact-angle hysteresis is 50°, then a roughness factor of 1.9 gives a Wenzel contact angle of 50° and a gain factor of 4.3 so that the hysteresis on the rough surface will be increased to 172°. However, when θe is 120° and S = 1% after roughening, a Cassie–Baxter surface gives an apparent contact angle of 162° and a gain factor of 0.03, reducing the contact-angle hysteresis on the rough surface to 9.5°.
To make a superhydrophobic surface, we first need to make the surface hydrophobic and create the appropriate roughness. As mentioned in the introduction, since cos θrW = r cos θe, θrW goes towards 180° when θe > 90° and the surface has proper roughness. In other words, a hydrophobic surface becomes more hydrophobic when roughness, r, increases. Figure 15.9 shows a cross-sectional view of a model of a plain woven fabric made from monofilament yarns. The surface area of a single round monofilament yarn in the unit fabric can be calculated based on Fig. 15.9.
For this rough surface, r is defined using flux integral. As shown in Fig. 15.9, the distance from the center of a weft yarn to the center of an adjacent weft yarn is 4R; in the same manner, the distance from the center of a warp yarn to the center of an adjacent warp yarn is 4R; and the distance from the center of a weft yarn to the center of an adjacent warp yarn is 2R. Hence, according to Pythagoras’s theorem, the vector from the center of one weft yarn to the center of an adjacent weft yarn makes a 30° angle to the plane of the fabric. Using flux integral, the area of one yarn in the unit fabric is calculated as:
where R is the radius of yarn, A is the area, i, j and k are the vectors to x, y, and z axis directions, respectively, and u and v are the notations for the variables of integration. Then, we determine the true fabric surface area as follows:
where Afabricapparent is the apparent area of the unit fabric shown in Fig. 15.9. Finally, the roughness, r, is just the ratio of these areas:
Next, we look at a plain woven fabric made of multifilament yarns. Clearly, a multifilament yarn will have even higher values of r, since the space between the fibers will increase the real surface area while the apparent surface area remains the same. In this case, equation [15.32] becomes:
When N = 1, Rf = R. Otherwise, Rf < R but NRf > R. For example, a plain woven fabric could have R ≈ 150 μm, N > 70, and Rf ≈ 10 μm. Substituting these values into equation [15.38] gives r > 20.48. Since r > 20.48 for the multifilament fabric, we again expect that the surface is rough enough to be superhydrophobic when θe ≥ 92.5° for this plain woven fabric.
According to Cassie and Baxter, the Wenzel model is a special case of the Cassie-Baxter equation where f2 = 0 in equation [15.14]. For a material with a smooth surface water contact angle of 93°, for example, the Wenzel surface roughness, r, must be greater than 16.5 for the apparent contact angle to exceed 150°. However, according to Marmur, the minimization of the free energy requires that, for a hydrophobic surface with f = 1, θe = 180° in equation [15.15]. Since the only material known with θe = 180° is air or vacuum, f cannot be equal to one. In other words, the Wenzel model is invalid for hydrophobic surfaces. In order to develop superhydrophobic surfaces, we need to use a different approach, namely the Cassie–Baxter model.
We begin with an analysis of f, the fraction of the projected area in contact with the water droplet. For parallel cylinders viewed normal to the cylinders’ axes there are two cases: (a) the cylinders are packed tightly together, or (b) they are separated by some distance, as shown in Fig. 15.10. In case (a), the distance from the center on one cylinder to the center of the next is 2R, where R is the radius of the cylinder. In case (b), by analogy, the center-to-center distance is considered to be 2(R + d). In case (a), Marmur showed that f = sinα where α is the angle between the top of the cylinder and the liquid contact line and α = π – θe. In case (b), f = R sinα/(R + d); and in both cases, rf = α/sinα. Therefore, using simple differentials we obtain d(f rf)/df = (cosα)− 1 and d2 (f rf)/df2 > 0 in both (a) and (b). According to Marmur, under these conditions, there is a minimum surface free energy on each surface such that α = π – θe. Substitution of f and rf with the case (b) into equation [15.15] results in:
We consider two cases for solving equation [15.40]. First, we solve it for a monofilament plain woven fabric shown in Fig. 15.9. Again, according to Pythagoras’s theorem, the angle that a weft yarn (or warp yarn) makes with the fabric plane in traveling from one warp yarn (or weft yarn) to another is 30°. Therefore, the distance from the top of a weft yarn (or warp yarn) to the top of an adjacent warp yarn (or weft yarn) is . From this geometric consideration, the center-to-center distance of adjacent cylinders in equation [15.40], 2(R + d),is equal to ,and f = .Substituting these values into equation [15.40] along with the measured contact angles from the flat nylon films, we find 118° ≤ θrCB ≤ 134° for the fluoroamine grafted monofilament woven fabric. These values are in good agreement with the measured values shown in Table 15.1.
In the second case, we again extend the analysis to a multifilament fabric. We begin by determining the apparent contact angle of the liquid with the multifilament yarns. For example, if the fiber spacing is approximately equal to the fiber diameter, that is 2d ≈ 2Rf, where Rf is the fiber radius, substituting Rf for R and d into equation [15.40] carries out θrCB ranging from 123° to 138° for the fluoroamine-grafted single fiber. Then, using these values as the effective contact angles for the yarns in the woven structure and re-solving equation [15.40] with R being the yarn radius and such that as before. We obtained 146° ≤ θrCB ≤ 158° for the fluoro- amine-grafted fabric. As seen in Table 15.1, the measured values are slightly larger than our predicted values. This is probably due to the real value of d being larger than the values chosen in this analysis.
Over the last 15 years, many studies of superhydrophobicity and contact- angle hysteresis have been performed. Recently, Zhu et al. developed a superhydrophobic surface by electrospinning using a hydrophilic material, poly(hydroxybutyrate-co-hydroxyvalerate) (PHBV).39 Gao and McCarthy made artificial superhydrophobic surfaces with conventional polyester and microfiber polyester fabrics rendered hydrophobic by using a simple patented water-repellent silicone coating procedure.40 Lee and Michielsen developed superhydrophobic surfaces via the flocking process and achieved contact angles as high as 178°41 Jopp et al. researched the wetting behavior of water droplets on periodically structured hydrophobic surfaces and the effect of structure geometry.42 Liu et al. studied the creation of stable superhydrophobic surfaces using vertically aligned carbon nanotubes.43 Nakajima et al. prepared superhydrophobic thin films with TiO2 photocatalyst by coating a fluoroalkyl silane onto the original films.44 Zhai et al. demonstrated that the superhydrophobic behavior of the lotus leaf structure can be achieved by creating a semifluorinated silane coated polyelectrolyte multilayer surface.45 Han et al. created a superhydrophobic surface using a block copolymer micelle solution and silica nanoparticles. He also provided a strategy for the fabrication of a wettability-controlled organic–inorganic hybrid. Many other studies of superhydrophobicity have been accomplished in inorganic materials.46 Huang et al. showed a superhydrophobic surface from nanostructure materials can be applied to microfluidic devices by preparing a stable superhydrophobic surface via aligned carbon nanotubes (CNTs) coated with a zinc oxide (ZnO) thin film.47 Shang et al. prepared optically transparent superhydrophobic silica- based films on glass substrates by making a nanoscale rough surface using nanoclusters and nanoparticles.48 Shi et al. obtained a superhydrophobic surface by modifying a silver covered silicon wafer with a self-assembled monolayer of n-dodecanethiol.49 Although most studies of the lotus effect have been carried out on inorganic materials, organic composite materials having a rough surface can also be superhydrophobic by using the lotus effect. Jeong et al. fabricated micro/nanoscale hierarchical structures using a molding technique.50 To modify an organic film surface and create superhydrophobicity, Kim et al. used He plasma and CFx nanoparticle coatings and Fresnais et al. applied CF4 and O2 plasma.51,52
Successful product design and development result in products that can be sold profitably. However, profitability is often difficult to evaluate quickly and directly. The following five specific dimensions related to profit can be used to evaluate the performance of product design and development: product quality, product cost, development time, development cost and market feasibility. Since these factors present huge challenges for product designers, only a few companies are successful in their product development in hightech product markets. In developing new high-tech textile products based on such excellent superhydrophobic materials or the surface modifying technologies introduced in this research, the product designer should search potential markets in which to apply the developed materials or technologies. For new high-tech product designs, the new materials or technologies should fit easily into existing products or processes. Therefore, new product development often uses modifications of generic materials to easily access the potential markets.
Superhydrophobic textile fabrics developed in this research can be used as new platform products. Platform products are generally built near an existing technological system. The technology platform shown in the previous section has already demonstrated its usefulness in mass production and its huge potential for the existing markets as well as new markets. Therefore, product design built on a technology platform is simpler than product design built on technology-push products. In designing a new technology- push product, a designer begins to work with a novel material or a unique technology to look for potential markets where the material or the technology can be applied. The potential markets for the new products consisting of self-cleaning surfaces are shown in Table 15.2.
|Potential markets||Possible new products|
|Protective equipment||Self-cleaning outerwear, shoes, bags, individual and collective protection gear|
|Residential interiors and exteriors||Self-cleaning outdoor furniture, windows, roofs; Stain-release carpets, paint, wallpaper, etc|
|Automotive interiors and exteriors||Self-cleaning auto glass, body shells, convertible tops|
|Biomedical textiles||Oil and water-release medical gowns, stain-release bed clothes;|
|artificial organs such as blood vessels and scaffold, etc|
The product design cannot be successful if the technology does not offer a clear competitive advantage satisfying customers’ needs. For example, stain-release children’s wear has to maintain soft hand, dimensional stability after 30 launderings, and colorfastness to saliva and perspiration. Self- cleaning convertible tops have to consist of coating materials having good durability after car washing, sunlight and high or low temperature exposure. Self-cleaning roofs have to be well designed to enhance the lotus effect while maintaining the original external appearance. Superhydrophobicoleophobic surgical gowns need materials having appropriate softness, high thermal conductivity and perspiration permeability. In developing artificial organs, it is important to design the new products while considering the biocompatibility for their applications. Therefore, the prototype samples of new biotextile products have to be evaluated through in vitro and in vivo tests.
Wetting behavior of solid materials has recently attracted a great deal of interest from both academia and industry. This interest extends beyond the bio-inspired, lotus leaf property of superhydrophobicity to materials that exhibit similar properties to oils, thus dubbed superoleophobic properties. The applications of such materials are far-reaching and include medical clothing, protective gear and high performance technical textiles. As a surface with a water contact angle exceeding 150° is called a superhydrophobic surface, we define a surface having an oil contact angle over 150° as a superoleophobic surface. Since the wettability of a solid surface is determined by two parameters, the chemical composition and the geometrical structure of a rough surface, the combination of these two factors can also be used for the development of a superoleophobic surface.
In order to design a superoleophobic surface, two predominant rough wetting models are used: the Wenzel model and the Cassie-Baxter model. In the Wenzel model a liquid fills the grooves of a rough surface and completely wets the surface, while in the Cassie–Baxter model, a liquid sits on top of the surface and repels the liquid. As mentioned, to create a Cassie–Baxter surface, the Young contact angle of a liquid, θe, must be greater than 90°. Analysis of the Cassie-Baxter model for assessing feasibility of superoleophobicity yields equation [15.10]. Since the surface tensions of dodecane and most polymeric surfaces are determined by London dispersion forces, this equation can be simplified to:
Substituting γL = 25.4 dyne/cm for dodecane into equation [15.41] suggests γS must be smaller than 6.35 dyne/cm, and a smooth surface having γS ≤ 6.35 dyne/cm can be oleophobic (θe > 90°). Since most solid surfaces typically possess γS ≤ 6.35 dyne/cm, the Cassie–Baxter model does not allow for permanent superoleophobicity under normal circumstances. On a metastable Cassie-Baxter surface, a liquid initially sits on top of the surface due to the minimization of surface energy caused by air pockets inside the grooves of the rough surface. However, the liquid is drawn into contact with the rough surface over time, with the time to absorption being dependent on the surface tension and volume of the liquid, and the surface tension and surface morphology of the solid. Hence, a superoleophobic surface can be produced by designing a metastable Cassie–Baxter surface.
2. Wu, X., Shi, G. Production and characterization of stable superhydrophobic surfaces based on copper hydroxide nanoneedles mimicking the legs of water striders. Journal of Physical Chemistry B. 2006; 110:11247–11252.
10. Pal, S., Weiss, H., Keller, H., Mueller-Plathe, F. Effect of nanostructure on the properties of water at the water-hydrophobic interface: a molecular dynamics simulation. Langmuir. 2005; 21:3699–3709.
11. Balkenede, A.R., van de Boogaard, H.J.A.P., Scholten, M., Willard, N.P. Evaluation of different approaches to assess the surface tension of lowenergy solids by means of contact angle measurements. Langmuir. 1998; 14:5907–5912.
12. Fowkes, F.M. Additivity of intermolecular forces at interfaces. I. determination of the contribution to surface and interfacial tensions of dispersion forces in various liquids. Journal of Physical Chemistry. 1963; 67:2538–2541.
18. Rosario, R., Gust, D., Garcia, A.A., Hayes, M., Taraci, J.L., Clement, T., Dailey, J.W., Picraux, S.T. Lotus effect amplifi es light-induced contact angle switching. Journal of Physical Cemistry B. 2004; 108:12640–12642.
23. Zhao, X., Jiang, P., Xie, S., Feng, J., Gao, Y., Wang, J., Liu, D., Song, L., Liu, L., Dou, X., Luo, X., Zhang, Z., Xiang, Y., Zhou, W., Wang, G. Patterned anodic aluminium oxide fabricated with a Ta mask. Nanotechnology. 2006; 17:35–39.