pubs.acs.org/Langmuir © 2010 American Chemical Society
Development of a Colloidal Lithography Method for Patterning Nonplanar Surfaces Sarang P. Bhawalkar,† Jun Qian,†,§ Michael C. Heiber,† and Li Jia*,† †
Department of Polymer Science, University of Akron, Akron, Ohio 44325-3909, United States, and § School of Chemical Engineering and Technology, Tianjin University, Tianjin 300072, China Received September 2, 2010. Revised Manuscript Received September 23, 2010
A colloidal lithography method has been developed for patterning nonplanar surfaces. Hexagonal noncontiguously packed (HNCP) colloidal particles 127 nm-2.7 μm in diameter were first formed at the air-water interface and then adsorbed onto a substrate coated with a layer of polymer adhesive ∼17 nm thick. The adhesive layer plays the critical role of securing the order of the particles against the destructive lateral capillary force generated by a thin film of water after the initial transfer of the particles from the air-water interface. The soft lithography method is robust and very simple to carry out. It is applicable to a variety of surface curvatures and for both inorganic and organic colloidal particles.
The emerging frontiers of many major scientific and technological areas, including optics,1 imaging and sensing,2 and bioengineering,3,4 require the fabrication of microscopic and nanoscopic patterns on nonplanar surfaces. The workhorse of surface patterning, projection photolithography, has been developed to achieve defect-free, registered, arbitrary patterns on flat surfaces to meet the needs of the microelectronics industry but is severely handicapped for nonplanar surfaces because of the necessity of focusing the beam on the plane of projection.5 New lithographic methods capable of patterning various nonplanar surfaces are needed to fuel the advance of the emerging scientific and technological frontiers. Contact printing and imprinting methods6 can cope with certain curved surfaces7-10 but appear to be restricted to those having a constant magnitude of curvature11 and a large radius of curvature relative to the arc length at least in one dimension. Stress-driven anisotropic buckling has also been employed to generate patterns on curved surfaces.12,13 In this case, the curvature itself is a critical parameter that governs the resulting pattern14 and therefore imposes a restriction on this method. *To whom correspondence should be addressed. E-mail:
[email protected]. (1) Jahns, J.; Brenner, K.-H. Microoptics: From Technology to Applications; Springer-Verlag: New York, 2004. (2) Jeong, K.-H.; Kim, J.; Lee, L. P. Science 2006, 312, 557–561. (3) Bettinger, C. J.; Langer, R.; Borenstein, J. T. Angew. Chem., Int. Ed. 2009, 48, 5406–5415. (4) Hwang, N. S.; Varghese, S.; Elisseeff, J. Adv. Drug Delivery Rev. 2008, 60, 199–214. (5) Moreau, W. M. Semiconductor Lithography: Principles, Practices and Materials; Plenum: New York, 1988. (6) Gates, B. D.; Xu, Q.; Stewart, M.; Ryan, D.; Willson, C. G.; Whitesides, G. M. Chem. Rev. 2005, 105, 1171–1196. (7) (a) Paul, K. E.; Prentiss, M.; Whitesides, G. M. Adv. Funct. Mater. 2003, 13, 259–263. (b) Rogers, J. A.; Jackman, R. J.; Whitesides, G. M. Appl. Phys. Lett. 1997, 70, 7–9. (c) Jackman, R. J.; Wilbur, J. L.; Whitesides, G. M. Science 1995, 269, 664–665. (8) Hong, S.-H.; Han, K.-S.; Byeon, K.-J.; Lee, H.; Choi, K.-W. Jpn. J. Appl. Phys. 2008, 47, 3699–3701. (9) Ruchhoeft, P.; Colburn, M.; Choi, B.; Nounu, H.; Johnson, S.; Bailey, T.; Damle, S.; Stewart, M.; Ekerdt, J.; Sreenivasan, S. V.; Wolfe, J. C.; Willson, C. G. J. Vac. Sci. Technol., B 1999, 17, 2965–2969. (10) Bao, L.-R.; Cheng, X.; Huang, X. D.; Guo, L. J.; Pang, S. W.; Yee, A. F. J. Vac. Sci. Technol., B 2002, 20, 2881–2886. (11) Casey, J. Exploring Curvature; Vieweg: Braunschweig, Germany, 1996; Chapter 10. (12) Bowden, N.; Brittain, S.; Evans, A. G.; Hutchinson, J. W.; Whitesides, G. M. Nature 1998, 393, 146–149. (13) Chan, E. P.; Crosby, A. J. Adv. Mater. 2006, 18, 3238–3242.
16662 DOI: 10.1021/la1035147
Lithography using colloidal self-assemblies has been extensively investigated because it is cost-effective, particularly for the fabrication of patterns over large areas despite the limitation of being able to produce only relatively simple or symmetric patterns.15 For most practical purposes, the individual surface features in a pattern must be separated from each other by a finite distance. This has been achieved in colloidal lithography by utilizing the interstices of hexagonal close-packed (HCP) colloidal spheres,16 trimming HCP arrays by shrinking17 or etching,18 stretching HCP arrays imbedded in a rubbery film,19 and shearinduced ordering during spin coating.20 We have recently demonstrated that hexagonal noncontiguously packed (HNCP) colloidal crystals trapped at the air-water interface can be directly transferred onto solid substrates to give HNCP and distorted HNCP patterns.21,22 The method is conceptually simple. The formation of the interfacial HNCP crystal is a result of the minimization of electrostatic repulsion between the same charged particles in a confined area.23 Transfer of the colloidal crystal from the air-water interface can be accomplished by generating electrostatic and van der Waals attractions between the particles and the solid substrate. These attractive forces are also expected to overcome the lateral capillary force acting on the particles when a thin film of water, which inevitably accompanies the particles onto the substrate, dries. In principle, the method is ideal for patterning topographically uneven surfaces because the initial colloidal array is formed at the fully pliable air-water interface.24,25 (14) Yin, J.; Cao, Z.; Li, C.; Sheinman, I.; Chen, X. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 19132–19135. (15) Velev, O. D.; Gupta, S. Adv. Mater. 2009, 21, 1897–1905. (16) Hulteen, J. C.; Treichel, D. A.; Smith, M. T.; Duval, M. L.; Jensen, T. R.; Van Duyne, R. P. J. Phys. Chem. B 1999, 103, 3854–3863. (17) Zhang, G.; Wang, D.; Gu, Z.-Z.; Hartmann, J.; M€ohwald, H. Chem. Mater. 2005, 17, 5268–5274. (18) Plettl, A.; Enderle, F.; Saitner, M.; Manzke, A.; Pfahler, C.; Wiedemann, S.; Ziemann, P. Adv. Funct. Mater. 2009, 19, 3279–3284. (19) Yan, X.; Yao, J.; Lu, G.; Li, X.; Han, K.; Yang, B. J. Am. Chem. Soc. 2005, 127, 7688–7689. (20) Min, W.-L.; Jiang, P.; Jiang, B. Nanotechnology 2008, 19, 475604/1–475604/7 and references therein . (21) Ray, M. A.; Jia, L. Adv. Mater. 2007, 19, 2020–2022. (22) Ray, M. A.; Shewmon, N.; Bhawalkar, S.; Jia, L.; Yang, Y.; Daniels, E. S. Langmuir 2009, 25, 7265–7270. (23) Pieranski, P. Phys. Rev. Lett. 1980, 45, 569–572.
Published on Web 10/15/2010
Langmuir 2010, 26(22), 16662–16666
Bhawalkar et al.
When reducing the conceptually straightforward process to practice, preservation of the pattern against the lateral capillary force in the drying stage poses a sizable challenge. In our previous work,20 we had to implement post-transfer treatments to mitigate the capillary force or preferably improve the adhesion between the particles and the substrate. The maneuvers, however, would be very cumbersome or impossible to carry out on a nonplanar surface. In addition, the tactics are best suited to organic particles that can be readily deformed but not hard inorganic particles, which are nonetheless important because they offer a variety of useful properties such as magnetism, quantum effects, and dry-etch resistance. In the present work, we utilize a layer of polymer adhesive to improve the adhesion between the particles and the substrate and are consequently able to avoid any posttransfer steps. The present method, which is suitable for both inorganic and organic particles, was first developed using flat silicon wafers and then applied to curved surfaces. In choosing the material for the adhesive layer, we took the following into consideration. First, it must have a glass-transition temperature (Tg) below room temperature. Furthermore, it should possess two surface characteristics that we previously established: a sufficiently high water-contact angle (advancing contact angle >70°) and surface charges opposite to the charges on the particle surface. A copolymer containing 80 mol % n-butyl acrylate and 20 mol % 2-(N,N-diethylamino)ethyl acrylate was synthesized to meet the above requirements. The weight-average and numberaverage molecular weights of the polymer were 33 080 and 18 690 Da, respectively. The Tg of the bulk material was -64 °C. Before the polymer adhesive was coated onto the silicon substrate, the hydrophilic surface of the silicon wafer was treated with a mixture of [2-(4-chlorosulfonylphenyl)-ethyl]trichlorosilane (SPTCS) and [( p-t-butylphenyl)ethyl]trichlorosilane (BPTCS) in a 1:9 mol ratio to retain the charged character of the substrate and improve the wettability of the hydrophobic polymer adhesive. The polymer adhesive in an isopropanol solution was then dip coated on the silanized silicon wafer. The dip-coating procedure that gives a 17 ( 2 nm-thick adhesive layer on flat silicon wafers was adopted throughout this work (Supporting Information). The surface of the adhesive layer had an advancing water-contact angle (θa) of 89 ( 4° and a receding contact angle (θr) of 47 ( 7°. Monodisperse silica particles with average diameters (D) ranging from 127 to 678 nm were used in this work. For the colloidal particles to be stably trapped at the air-water interface and selfassemble into the HNCP lattice, the surfaces of the particles must also have a sufficiently high water-contact angle and be sufficiently charged. To achieve these surface characteristics, the silica particles were treated with a mixture of SPTCS and BPTCS in a 1:4 mol ratio. This empirically adopted formula proved satisfactory for interfacial trapping and self-assembly. To form the HNCP pattern, the silanized silica particles were suspended in isopropanol and spread at the air-water interface in a Langmuir trough. The available area was compressed by the movable barriers of the trough. For an evenly spread particle film, the rise of surface pressure signaled the onset of the self-assembly process. The HNCP particles at the air-water interface were readily adsorbed onto the substrate when the substrate (typically 1 1 cm2 in area) was brought into contact with the interface parallel to the water surface (Scheme 1). No additional post-transfer (24) Previous studies of Pickering emulsions provide a glimpse of the potential: (a) Velev, O. D.; Furusawa, K.; Nagayama, K. Langmuir 1996, 12, 2374–2384. (b) Velev, O. D.; Furusawa, K.; Nagayama, K.; Langmuir 1996, 12, 2385-2391. (c) Velev, O. D.; Nagayama, K. Langmuir 1997, 13, 1856–1859. (25) Dinsmore, A. D.; Hsu, Ming F.; Nikolaides, M. G.; Marquez, M.; Bausch, A. R; Weitz, D. A. Science 2002, 298, 1006–1009.
Langmuir 2010, 26(22), 16662–16666
Letter Scheme 1. Stacks on the Substrate and Experimental Configuration for Transferring Interfacial HNCP Particle Arrays
step was necessary. The very thin layer of adhesive proved sufficient to secure the HNCP lattice of silica particles of all sizes used in this work (Supporting Information). Because the lateral capillary force increases with the decrease in the interparticle distance,26 the minimal achievable interparticle distance can be expected to be a finite value, beyond which adhesion between the particles and the solid surface mediated by the adhesive layer would yield to the capillary force. The limit was probed by a set of experiments using the 247 nm particles. Assuming that all particles spread at the air-water interface are irreversibly trapped and given that the number of particles spread at the air-water interface and the interfacial area are experimentally defined quantities, the interparticle distance in the HNCP lattice can be calculated (Supporting Information). The particle films with successive calculated interparticle distances (Lcalc) were transferred onto the substrates and imaged by scanning electron microscopy (SEM) as shown in Figure 1. The HNCP lattices were analyzed by fast Fourier transform to give the experimentally determined average interparticle distance (Lexp). General agreement between Lexp and Lcalc is evident27 (Figure 2) and therefore validates the above method for predicting the interparticle distance. The attempt to transfer the HNCP particles with Lcalc = 275 nm or 1.1D resulted in a loss of order (Figure 1b). The possibility that the loss of order is due to flocculation while the particles were at the air-water interface can be ruled out by the observation of the HCP lattice (Figure 1a) when the interfacial film was further compressed. For Lcalc = 300 nm or 1.2D, the HNCP lattice with Lexp = 328 ( 35 nm or ∼1.3D was successfully preserved (Figure 1c). The minimum achievable Lexp is evidently between 1.1D and 1.3D for the present colloidal lithography method. The above set of experiments also interrogated the maximum achievable Lexp, which is obviously determined by the range of electrostatic repulsion between the same charged particles at the air-water interface. The attempt to form the HNCP lattice with Lcalc = 750 nm or 3D did not result in any ordered phase (Figure 1i), but the HNCP lattice with Lexp = 745 ( 41 nm or ∼3D (Lcalc = 625 nm or 2.5D) was able to form (Figure 1h). Note that the present formulas for particle and substrate surface modification are both pragmatically adopted but not optimized. The window of achievable Lexp might expand somewhat at both ends if the critical parameters were optimal. The degree of order for the particle arrays whose order was preserved (i.e., patterns with Lcalc = 1.2-2.5D) was evaluated using the bond orientational order parameter28 * + 6 1X Ψ6 ¼ jexpði6θj Þj 6 j¼1 (26) Kralchevsky, P. A.; Denkov, N. D. Curr. Opin. Colloid Interface Sci. 2001, 6, 383–401. (27) The discrepancy is likely attributable to the inadvertent loss of particles in the preparative steps; for example, the particles stuck to the wall of the container when they were redispersed prior to spreading. (28) Marcus, A. H.; Rice, S. A. Phys. Rev. E 1997, 55, 637–656.
DOI: 10.1021/la1035147
16663
Letter
Bhawalkar et al.
Figure 1. SEM images of 247 nm particles on solid substrates aiming at the Lcalc/D ratio: 1 (a), 1.1 (b), 1.2 (c), 1.4 (d), 1.6 (e), 1.8 (f ), 2.0 (g), 2.5 (h), and 3 (i). Scale bar: 1 μm.
Figure 2. Plot of Lexp vs Lcalc. The diagonal line, which depicts the ideal situation for Lexp = Lcalc, is a guide to the eye.
where θj is the angle between the bond connecting a particle and its jth nearest neighbor and a fixed arbitrary reference axis. The Ψ6 value of an ideal hexagonal lattice is 1. It is immediately noticeable, as shown in Figure 3, that Ψ6 of the particle array with Lcalc = 1.2D (Figure 1c) abnormally falls far from unity. This is mostly because this lattice is severely distorted from hexagonal symmetry and is better characterized as a rhombic lattice. In fact, (29) Distortion from the hexagonal lattice often arises from stochastic factors during the transfer process; for example, the substrate was not perfectly parallel to the water surface. However, a fundamental root cannot be ruled out, particularly when the particle array is highly compressed at the air-water interface. See Aveyard, R.; Clint, J. H.; Nees, D.; Paunov, V. N. Langmuir 2000, 16, 1969–1979.
16664 DOI: 10.1021/la1035147
Figure 3. Plot of bond orientational order parameter vs interparticle distance.
some extent of distortion from perfect hexagonal symmetry is commonly observed for the particle arrays on the solid substrate and was presumably introduced inadvertently during the transfer process in most cases.29 Because Ψ6 for the ideal rhombic lattice (Ψ6,ideal) varies according to the angle R between the primitive vectors, we suggest using the quotient Ψ6/Ψ6,ideal as a measure of the bond orientational order. To calculate Ψ6,ideal = 1/3 |1-2 cos(3R)|,30 the average bond-angle value in the particle arrays (Supporting Information) was adopted as R in the corresponding ideal rhombic lattice. According to the Ψ6/Ψ6,ideal measure, the particle arrays with the highest degree of order were obtained when the interparticle distances were small. (30) Mazars, M. EPL 2008, 84, 55002p1–55002p6.
Langmuir 2010, 26(22), 16662–16666
Bhawalkar et al.
Letter
Figure 4. (a) Distorted HNCP arrays of 370 nm particles coated on a spherical mound formed by the deformation of a 10 μm polystyrene particle. (b) Distorted HNCP arrays of 370 nm particles coated on an irregularly shaped mound formed by the deformation of a cluster of several 2.7 μm polystyrene particles. (c) Top-down view of a 10 μm latex sphere covered by distorted HNCP arrays of 247 nm particles. (d) Side view of a 10 μm latex sphere covered by distorted HNCP arrays of 247 nm particles. Scale bar: 1 μm.
To generate the curved surfaces used in this study, 2.7- or 9.4-μm-diameter polystyrene latex particles with surface sulfate groups were randomly deposited on silicon wafers pretreated with the SPTCS/BPTCS mixture. The deposited latex particles and their small clusters were then deformed with diethyl ether to generate various curvatures on the substrate. The method developed for patterning the flat substrate was directly applied to the topographically uneven substrates. The thickness of the adhesive layer was not determined but was likely in the same range as that on the flat surface. Figure 4 shows the SEM images of several representative curved surfaces decorated with distorted HNCP arrays of silica particles. The spherical mound in Figure 4a serves as an example of the curved surfaces, of which the magnitude of curvature is constant and the radius of curvature is large relative to the arc length. Contact printing and imprinting methods can also pattern this type of curved surface9,10,12,13 but will experience difficulties when encountering curved surfaces such as the irregular mound in Figure 4b, where the magnitude of the curvature is not constant. This is not a problem for the present method. The HNCP arrays at the air-water interface can readily adapt to the changing curvature and the overall topography on the substrate while maintaining the order. Note that the 2D HNCP lattice has to distort or stretch when it is coated onto curved surfaces by virtue of the fact that the surface area of the curved surface is larger than its area of projection. The full sphere in Figure 4c represents the type of curve where the arch length is Langmuir 2010, 26(22), 16662–16666
greater than the radius of curvature. As clearly observable from the side view in Figure 4d, the small silica particles cover a large fraction of the large sphere well past the equatorial plane into the undercut area. The large coverage, which may be intuitively surprising, is likely attributable to the stick-slip motion of the three-phase contact line pinned on the backs of the small silica particles immediately after they are deposited on the large 10 μm sphere. The pinned contact line can conceivably drag the interfacial particle film along to cover the area that it could not reach otherwise. The situation is similar to our previous observation that dipping a flat substrate perpendicularly through the airwater interface also resulted in distorted HNCP patterns over the entire surface of the substrate.21 Note that when the difference between the surface areas of the mound and its projection (equal to 3πR2 in the case of a full spherical mound, where R is the radius of the sphere) is large as in the case of the 10 μm spherical mound, the film of the HNCP small particles may break when stretched beyond a certain point, resulting in irregular deposition of the particles in the area around the large sphere. When 3πR2 is not excessive (i.e., R is smaller), the order of the particles around the mound can be retained. At the present time, we do not know the critical 3πR2 value and how it is related to other parameters such as the radius of the small particle and the interparticle distance. Finally, we demonstrate that the present colloidal lithography can be repeated at two or conceivably several length scales to form hierarchical patterns. In Figure 5, the HNCP spherical mounds DOI: 10.1021/la1035147
16665
Letter
Bhawalkar et al.
Figure 5. (a) HNCP lattice of 2.7 μm latex particles deformed after 1 min in diethyl ether. (b) Distorted HNCP arrays of 370 nm particles on the substrate in image a. (c) HNCP lattice of 2.7 μm particles deformed after 3 min in diethyl ether. (d) Distorted HNCP arrays of 370 nm particles on the substrate in image c. Scale bar: 1 μm.
were fabricated by transferring the interfacial HNCP lattice of 2.7 μm polystyrene particles onto the adhesive-coated silicon wafer followed by deformation of the polystyrene particles with diethyl ether. The substrates bearing the hexagonally arranged spherical mounds were then dip-coated again with the adhesive polymer. The interfacial HNCP arrays of small silica particles were then transferred onto the micropatterned surface to achieve the submicrometer pattern. We note in conclusion that the present colloidal lithography method is very general and can in principle be adopted for any particles and substrates as long as an appropriate adhesive layer is applied. Although the size of the substrate in our experiments has typically been 1 1 cm2 , there does not appear to be a fundamental reason that prevents the present method from being applied to much larger substrates. Furthermore, the patterns that it can produce should not be limited to simple hexagonal or distorted hexagonal symmetries. Nonspherical particles and binary particles can potentially be employed to expand pattern diversity and complexity. From a practical viewpoint, the fabrication of optical elements with combined refractive, diffractive, and subwavelength structures, for example, microlens arrays with
16666 DOI: 10.1021/la1035147
antireflective subwavelength gratings,1 is an immediate application with a high probability of succeeding. Other near-term applications likely would emerge from the bioengineering area, where microscopic and nanoscopic surface topographies have been shown to play important roles in cell proliferation and differentiation.3,4 Acknowledgment. We thank The University of Akron for the financial support of this work (startup funds awarded to L.J.). J.Q. thanks the China Scholarship Council (CSC) for awarding him a scholarship and Tianjin University for giving him the opportunity to pursue his studies in the United States as a joint Ph.D. student. L.J. thanks the Office of Research Services and Sponsored Programs of The University of Akron for a faculty research grant. We also thank the National Science Foundation (CHE-9977144) for funding The University of Akron Magnetic Resonance Center’s purchase of the NMR instrument used in this work. Supporting Information Available: Experimental section. This material is available free of charge via the Internet at http://pubs.acs.org.
Langmuir 2010, 26(22), 16662–16666
