Langmuir 2004, 20, 227-231
227
Planarizing Surface Topography by Polymer Adhesion to Water-Soluble Templates with Replicated Null Pattern Charles D. Schaper* Department of Electrical Engineering, Stanford University, Stanford, California 94305-9510 Received August 13, 2003. In Final Form: October 12, 2003 Water-soluble polymer templates are replicated from flat master surface patterns by spin-casting a poly(vinyl alcohol) film-forming solution that solidifies at standard ambient conditions in less than 1 min. The fabricated water-soluble templates are coupled to substrates with surface topography by polymer adhesion with an intervening reactive or photocurable liquid layer. After curing, the resulting two-layer solid structure is subjected to water thereby dissolving the soluble template to expose the underlying polymer adhesive layer with flat surface topography. The results demonstrate a reduction of surface topography from several micrometers to less than 100 nm. The chemical interactions involved in bonding the soluble template to the polymer adhesive and then dissolving are measured by Raman spectroscopy to demonstrate that the constituents comprising the water-soluble template are absent from the surface of the planarization material.
Introduction High-resolution printing is an important step in manufacturing electronics, displays, and memory devices, in addition to emerging applications in microfluidics, plastic electronics, microelectromechanical systems (MEMS), photonics, and biological sensors. The fabrication of these systems generally involves aligned printing of structures over existing surface topography. For advanced manufacturing applications, the substrate surface must first undergo planarization before printing can take place, such as with integrated circuit fabrication that uses deep UV photolithography possessing a narrow depth of focus field of less than 500 nm. Consequently, one of the more important and practical patterns to generate is flatness over existing surface topography. In general terms, planarizing surface topography is of significant scientific concern because material surface asperities influence friction, optical properties, and surface area effects in mass transport, nucleation, and catalyzed chemical reactions. Common commercial methods to achieve planar surfaces over topographically structured substrates such as in-process semiconductor wafers include chemical mechanical planarization (CMP).1,2 This technique employs slurries and a rotating pad, applying pressure on the substrate to grind the surface and yield a smooth finish for further fabrication. The approach is essentially one of polishing in which the surface irregularities and undesired residual material such as metals are mechanically or electrochemically removed from the surface. An additive approach toward smoothing surface topography involves the implementation of thick polymer coatings with subsequent etching to form a thin layer.3,4 * E-mail:
[email protected]. Phone: 650-723-2873. Fax: 650-723-8473. (1) Jeng, Y. R.; Tsai, H. J. Improved Model of Wafer/Pad Powder Slurry for CMP. J. Electrochem. Soc. 2003, 150, G348-G354. (2) Padhi, D.; Yahalom, J.; Gandikota, S.; Dixit, G. Planarization of copper thin films by electropolishing in phosphoric acid for ULSI applications. J. Electrochem. Soc. 2003, 150, G10-G14. (3) Dow Corning, Inc., USPTO Patent 5,370,903, Method for the Formation of a Silicon Oxide Film. Tokyo Ohka Kogyo Co. Ltd. (TOK) offers materials for spin-coat planarization as part of their OCD product line. (4) Schuck, M. H.; McKnight, D. J.; Johnson, K. M. Spin-cast planarization of liquid-crystal-on-silicon microdisplays. Opt. Lett. 1997, 22, 1512-1514.
Recently, an alternate method of planarization has been introduced that employs mechanical forces to press an optically flat transparent surface on a photopolymerizable liquid organic material.5 The material is cured with ultraviolet light after contact, and then the flat quartz template is lifted to reveal a smoothed polymer surface on the substrate. This approach is one of material addition, employing a process similar to that of imprinting patterns using photocurable polymer systems.6 Areas that use this method include the planarization of organic materials such as benzocyclobutene (BCB) and low-k materials as dielectrics for silicon and compound semiconductor applications. These materials are difficult to planarize using CMP because of the hardness requirements. Additional applications include MEMS and wafer level packaging employing high aspect ratios that are difficult to fill using thick coating methods. This method of contact planarization employing an optical flat and photopolymerization, however, has potential difficulties with the generation of defects during physical retraction of the template from the substrate, requiring a twisting motion for successful removal. The optical flats may require frequent cleaning to avoid defect propagation since they are reused. Moreover, performing planarization over nonflat surfaces, such as cylinders or spheres, is complicated with the use of a hard quartz template. In previous work, a new class has been reported by this laboratory of high-resolution pattern formation and materials transfer printing strategies, collectively referred to as molecular transfer lithography (MxL).7-10 The (5) Brewer Science, Inc., offers a full-wafer contact planarization material and tool that presses an optical flat to the polymer surface, cross-links the material by photopolymerization, and then retracts the quartz template. (6) Resnick, D. J.; Dauksher, W. J.; Mancini, D.; Nordquist, K. J.; Ainley, E.; Gehoski, K.; Baker, J. H.; Bailey, T. C.; Choi, B. J.; Johnson, S. C.; Sreenivasan, S. V.; Ekerdt, J. G.; Willson, C. G. High-Resolution Templates for Step and Flash Imprint Lithography. Proc. SPIE Emerging Lithogr. Technol. VI 2002, 4688, 205-213. (7) Schaper, C. D. MxL: Pseudo-maskless, high throughput, nanolithography. Proc. SPIE Emerging Lithogr. Technol. 2003, 5037, 538549. (8) Schaper, C. D. Water-soluble polymer templates for highresolution pattern formation and materials transfer printing. J. Microlith. Microfab. Microsys., accepted. (9) Schaper, C. D. Molecular transfer lithography for pseudomaskless, high-throughput, aligned nanolithography. J. Vac. Sci. Technol., B, accepted.
10.1021/la035482h CCC: $27.50 © 2004 American Chemical Society Published on Web 11/19/2003
228
Langmuir, Vol. 20, No. 1, 2004
Schaper
Scheme 1. Replication and Transfer of a Null Patterna
Scheme 2. Methods to Planarize a Substrate by Polymer Adhesion to a Flat Water-Soluble Templatea
a (A) The procedure begins with spin-casting a PVA filmforming solution onto a flat master pattern, the binding of a solid PVA sheet, and the removal from the master null pattern to form a water-soluble template with replicated flatness. (B) The transfer procedure bonds the template to the substrate via an intervening adhesion layer. Water is then used to dissolve the template, leaving the polymer adhesive with a flat surface topography.
a (A) The polymer adhesive is coated onto the surface of the substrate, and while in liquid form, a water-soluble template is brought into contact with the adhesive and bonded, using a pressing or rolling action by an external source to achieve the desired level of topographical smoothing. (B) The transfer procedure begins with coating liquid adhesive on the surface of the flat water-soluble template, and then the adhesive surface is brought into contact with the substrate and bonded through UV, thermal, or reactive adhesion schemes. Combinations of the two approaches are also possible. Water is used to dissolve the template and complete the process.
approach replicates surface patterns as water-soluble polymer templates by spin-casting a poly(vinyl alcohol) (PVA) film-forming solution. The solidification process takes place at standard ambient conditions and can complete in less than 1 min. Lateral resolution of replicated surface features below 100 nm has been demonstrated. In this paper, the MxL method is applied to generate a null pattern on a substrate by polymer adhesion to a watersoluble template with replicated flatness. This approach to planarization avoids the generation and propagation of defects since a new template is used for each implementation. Moreover, the procedure utilizes chemical dissolution of the template after bonding, thereby eliminating the need for physical release of the template and removing the possibility of generating defects by such an action. The use of polymer adhesion to the water-soluble template also enables new types of planarization materials, since such materials are allowed to bond to the template, which is not possible to do with alternative strategies. The fabrication process of the water-soluble template is depicted in Scheme 1A. It begins with the replication of flat surface topography from a master surface pattern as a water-soluble polymer template by spin-casting the PVA film-forming solution. The solution undergoes a roomtemperature process to form a thin film with a thickness and drying period dependent upon the rotation speed, process time, and ambient conditions. The ability of spincasting a low-viscosity fluid to fill the voids and recesses of the master surface pattern is exploited to achieve highresolution replicas, including flatness. A silicon wafer may be used as a master surface pattern to generate the replicated null pattern. After replication with detachment, the water-soluble polymer templates are bonded to the substrate through an intervening polymer adhesion layer. The null pattern is transferred after water dissolution of the template as depicted in Scheme 1B. The chemical association of the template to the intervening adhesion layer may be secured using several mechanisms including photopolymerization or multiplecomponent reaction systems such as hardening by interaction with the moisture in the environment, both demonstrated in the paper. In addition, two methods of coating the adhesion layer are examined. In Scheme 2A, the standard method is shown of coating the substrate and then bonding the adhesive material to the replicated flat water-soluble polymer template. An alternative (10) Schaper, C. D. Patterned transfer of metallic thin film nanostructures by water-soluble polymer templates. Nano Lett. 2003, 3, 1305-1309.
method is shown in Scheme 2B involving the coating of the soluble template with the adhesive component, and then the liquid material system is pressed into contact with the topographically varying substrate. This method is interesting because in some situations it is difficult to coat the substrate directly, and this approach provides a solution. It is also useful for localized planarization of the substrate. If necessary, the water-soluble template can be slid across the liquid surface to achieve proper placement without damaging the template prior to chemical bond activation. Combinations of the processes in panels A and B of Scheme 2 are also possible to fully exploit the ability of spin-coating to reduce topographical variations and enhance planarization. Experimental Section Materials. The material selected to conduct replication is a PVA film-forming solution, which is a common material that is available commercially in various formats including blends. The liquid film-forming solution was purchased from a distributor, Fiberlay, Inc. (Seattle, WA), as their P.V.A. product marketed as a mold release agent for polyester and epoxy systems, involving a blend of materials including ethenol homopolymer, ethyl alcohol, butyl alcohol, and water. The solid replication material was obtained in both roll form and as a water-soluble adhesive tape from Shercon, Inc. The roll purchased was 13 in. wide by 36 feet long, although wider rolls are available. These solid PVA materials were purchased with a plastic backing adhered to the film. The overall thickness was 100 µm. In addition to dissolvable solid materials, other nondissolvable materials used for the solid sheet shown in Scheme 1A included silicone rubber. Tap water was used to dissolve the replication materials. Several adhesive materials were used for bonding the dissolvable template to the substrate to carry out the transfer steps. These materials included UV15 (Master Bond, Inc.), a 50:50 blend of UV15 and Zipcone UA (Gelest), MB300 (Master Bond, Inc.), Optical Adhesive 60 from Norland, Inc., and Type SK-9 optical cement from Summers, Inc., obtained from a distributor, Edmunds Industrial Optics. A 100 mm polished silicon wafer (MEMC Electronic Materials, Inc.) was used out of the box as the master surface pattern and included a native oxide film. Null Pattern Replication. Spin-casting to achieve film formation was performed by first pouring the PVA material on the master pattern. The system was ramped to 800 rpm for 3 s, followed by a ramping to the final rotational speed. Various final spin speeds were utilized, as low as 1300 rpm and as high as 6000 rpm. A constant spin time of 12 s was selected. A nominal value for spin speeds was selected as 3000 rpm. The lower spin speeds yielded a thin liquid film allowing for a solid sheet to be added, whereas the higher spin speeds at roughly 5000 rpm yielded a dry film at the conclusion of the spin process.
Planarizing Surface Topography by Polymer Adhesion Alternatively, by spinning at the lower rpm speed and waiting for at least 3 min, the liquid evaporated to yield a solid material. After spin-casting, the preform solid sheet of PVA was placed on the surface of the liquid spin-coated surface by rolling onto the surface of the master pattern starting from one end and moving to the other end. A cylindrical rod of 25 mm diameter aided in this process. Alternatively, the spin-casted film, when dried, can be detached by use of an adhesive film to bind a surface onto the solid spin-cast replication by a simple rolling motion. Successful interaction with the dried film was also achieved by using a clean sheet placed into contact with the solidified replication film to achieve enough binding interaction to enable subsequent detachment from the master pattern. This was achieved with both clean sheets of Riston, Mylar, and PVA. After binding a solid sheet to the spin-casted film, the dissolvable template was removed from the master pattern by peeling. A thin layer of liquid generally remained on the surface after detachment using the slower spin speeds. If a contact alignment system was to be used, the master pattern with the solid preform was placed on a prealigner pin system where marks were made on the polymer sheet prior to removal from the master pattern. The master pattern was held in place either by using a vacuum or manually during the removal process. Planarization. For smoothing surface topography, a material with adhesive properties, which may have additional properties such as plasma etch resistance, was applied to either the target substrate or the water-soluble polymer template by spin-coating. The spin speeds were selected as 2000-6000 rpm depending upon the application. The water-soluble template was placed on the liquid adhesive manually by using a cylindrical rod to roll the replica onto the target substrate coated with the intervening liquid adhesive. The replica may be adjusted after contact and prior to curing to adjust the relative positioning. If a contact aligner was employed for contacting the template and substrate, the template was used in place of the quartz photomask. The operations of the aligner were performed in a normal manner. After spin-coating the photocurable material on the substrate, solidification was achieved by exposing the materials to light for 60 s and then waiting an additional 30 s. These times were not optimized but were functional. The material was spin-coated at a number of speeds from 1800 to 6500 rpm, for various times from 20 to 60 s. All were found to work acceptably. The materials that bond due to a moisture mechanism, such as cyanoacrylates, were integrated by spin-coating the material onto the substrate or template at 1800-6500 rpm for 12 s. The template was then placed on the material with manual pressure. After adhesion with the substrate, the bonded system was placed in water and the water-soluble material was allowed to dissolve. It took about 10 min to dissolve in a quiescent state since it was necessary to break through an adhesive on the outer surface from the plastic backing. Without the adhesive present, the dissolution time was shorter, about 1 min. To quicken the dissolution time, the samples were subjected to a combination of a gentle water stream impacting the surface, followed by an air jet to push the watersoluble material off the surface. Characterization. Raman spectroscopy was performed using equipment from JY Horiba (Dilor) with a LabRam spectrometer at a wavelength of 633 nm. Scanning electron micrographs were taken using equipment from Hitachi. Atomic force microscopy (AFM) scanning measurements were taken with a Digital Instruments Nanoscope E operating in contact mode with a Nanoprobe SPM tip.
Results and Discussion Several materials and methods of applying the polymer adhesive were evaluated in terms of filling and planarizing trenches using the MxL procedure of replicating the surface of a (null) master surface pattern, bonding the template to the substrate via an intervening polymer adhesive material, and dissolving the template with water. The formation of the replicated null template was performed using a master surface pattern comprised of a silicon wafer with native oxide film. The direct binding of the template to the master surface pattern was not
Langmuir, Vol. 20, No. 1, 2004 229
Figure 1. Filling and planarizing trenches using the MxL procedure of replicating the surface of a null master surface pattern, bonding to the substrate via an intervening adhesive material, and dissolving with water, depicting material fills for several situations: (A) spinning photocurable cement on the template and then bonding over a range of feature sizes; (B) planarizing over large area topography by spinning photocurable cement on the template and then bonding; (C) spinning mid-viscosity, 20 cP, reactive material on the template and bonding on the substrate; (D) spin-coating on the substrate with mid-viscosity contact adhesive material; (E) spinning lowviscosity, 2-3 cP, adhesive material on the substrate and bonding the water-soluble template to planarize.
observed for any of the hydrophilic agents comprising the film-forming PVA coating solution. A sufficiently high spin speed was also used to achieve rapid solidification and to minimize the appearance of a thin residual liquid layer on the master surface pattern after detachment. To reduce the possibility of defect propagation in the event that such bonding should occur, water may be applied to the master pattern prior to replication to dissolve any material potentially left behind from the previous replication cycle. Test wafers with a surface topography comprised of a series of trenches in silicon dioxide over silicon were used to evaluate the ability to fill and planarize dimensions of various shapes and length scales. A planarization result is shown in Figure 1A for the method that involved the spinning of photocurable cement of viscosity 80-100 cP onto the flat template and then bringing that into contact with the substrate, followed by photocuring. The result demonstrates planarization over a range of feature sizes from several micrometers down to 140 nm. The ability to planarize isolated and densely distributed features to the same level relative to a common base is an important issue to consider for integrated circuit applications. To improve uniformity and flatness despite differences in topography and therefore the need to generate local volumetric differences in adhesive materials to properly planarize, the MxL technique employs single- or dualsided coating, spatial sliding to enhance filling, and the application of force prior to solidification/dissolution. Nonetheless, in Figure 1A, a slight differential of thickness
230
Langmuir, Vol. 20, No. 1, 2004
Schaper
Figure 3. Raman spectra measured at the surface of the nullreplication material, the planarization material, and the smoothed surface after bonding and water dissolution. The absence of peaks at a shift less than 2000 cm-1 demonstrates that the replication material is not present in the processed sample as the original form. The relative increase in the CH peak intensity around a Raman shift of 2947 cm-1 is noted. A pure form of poly(vinyl alcohol) demonstrates a peak at 2907 cm-1 representative of the replication material, and the CN bond exhibits a peak at 2251 cm-1 representative of the planarizing and fill material.
Figure 2. Surface topography after undergoing the MxL smoothing procedure: (A) a full area scan using the atomic force microscope demonstrating a range of displacement of 80 nm showing consistency with the underlying substrate topography. Diagonal line scans are shown in panel B to note the average flatness parallel and perpendicular to underlying channels of the substrate.
between the isolated and dense features is noted. To improve further on this bias, further process optimization and engineering of the planarization system may be useful. In Figure 1B, the method of planarization is shown over large feature sizes with lateral dimensions greater than 50 µm. In Figure 1C, the planarization results are shown for a mid-viscosity reactive adhesive material of viscosity 45 cP that was spun on the template and then bonded to the substrate. Figure 1D depicts a planarization result for spin-coating the reactive adhesive of mid-viscosity directly on the substrate and then bonding the template to the surface. In Figure 1E, a low-viscosity adhesive material of 2-3 cP was spun onto the substrate and planarized using the water-soluble template with a replicated null pattern, to generate a small ranging overfill. Additional materials were also tested for their ability to planarize. These included a photocurable optical adhesive (Norland 60) and an epoxy (UV 15) that did planarize the surface well but detached from the underlying silicon dioxide topography due to significant shrinkage during the curing process. The surface roughness of the planarized film was measured using AFM scanning of an area of 16.7 µm × 16.7 µm of the planarized surface topography for features similar to that of Figure 1D,E, performed using the process shown in Scheme 2B. The result is depicted in Figure 2A. There is an oscillation of the surface topography related
to the position of the trenches in the original pattern. However, the reduction in variation went from about 2.5 µm to less than 100 nm, adequate for deep UV photolithography applications. In Figure 2B, line scans are depicted of the horizontal and vertical measurement planes. The full scan root-mean-square surface roughness was measured at 16.0 nm for the clean silicon wafer, 17.4 nm for the original silicon dioxide surface, and 19.2 nm for the planarized surface, as averaged over several samples. Although the average roughness is slightly greater than that of the reference samples, the lowfrequency oscillations of the planarized surface are seen in coordination with the underlying topography. The nature of the chemical interactions occurring during the planarization process was also investigated to assess transfer of the null-replication material to the surface where the organic liquid adhesive contacts and solidifies in adhering to the water-soluble polymer template. In Figure 3, the Raman spectrum is presented for three materials on glass: the solidified planarizing polymer material, the water-soluble template before dissolution and bonding, and the solidified planarizing polymer material after the planarization process. The absence of peaks at a shift less than 1750 cm-1 demonstrates that the replication material is not present in the processed sample in the original form. No relative increase in the CH peak intensity around a Raman shift of 2947 cm-1 is noted. A pure form of poly(vinyl alcohol) demonstrates a peak at 2907 cm-1 representative of the replication material, and the CN bond exhibits a peak at 2251 cm-1 representative of the planarizing and fill material. This result suggests no elemental material transfer related to the bonding of the adhesive to the PVA template. The remaining constituents of the replicated flat template were removed after water dissolution. Conclusions A new process for planarizing surface topography has been demonstrated. The technique replicates flatness from
Planarizing Surface Topography by Polymer Adhesion
a master surface and transfers that null pattern to a substrate by bonding to an intervening polymer adhesion layer, followed by template disassociation by dissolving with water. The technique enables new materials for consideration as planarizing materials, namely, adhesives designed to bond materials together. Because the release of the single-use soluble template from the substrate is accomplished chemically through water dissolution, there is little chance for defect generation or defect propagation as compared to alternative methods. The conformable nature of the replicated flat soluble template is also useful to planarize roughness of nonflat substrates. Planarizing
Langmuir, Vol. 20, No. 1, 2004 231
surface topography by polymer adhesion to replicated water-soluble templates, using the MxL method, represents a high-precision method for development of microstructured and nanostructured material devices requiring flat surfaces as part of the final product or to enable further processing. Acknowledgment. DARPA Advanced Lithography and NSF ECS sponsored this research. LA035482H
