A MEMS nanoplotter with high-density parallel dip-pen nanolithography probe arrays

INSTITUTE OF PHYSICS PUBLISHING Nanotechnology 13 (2002) 212 217 NANOTECHNOLOGY PII: S0957-4484(02)29674-9 A MEMS nanoplotter with high-density parallel dip-pen nanolithography probe arrays Ming Zhang 1, David Bullen 1, Sung-Wook Chung 2, Seunghun Hong 2,3,KeeSRyu 1, Zhifang Fan 1, Chad A Mirkin 2,4 and Chang Liu 1,4 1 Micro Actuators, Sensors, and Systems Group, Microelectronics Laboratory, University of Illinois at Urbana-Champaign, 208 N Wright Street, IL 61801, USA 2 Department of Chemistry, Institute for Nanotechnology, Northwestern University, 2145 Sheridan Road, Evanston, IL 21208, USA E-mail: changliu@uiuc.edu and camirkin@chem.nwu.edu Received 15 October 2001, in final form 19 February 2002 Published 25 March 2002 Online at stacks.iop.org/nano/13/212 Abstract We report on the development of a nanoplotter that consists of an array of microfabricated probes for parallel dip-pen nanolithography. Two types of device have been developed by using microelectromechanical systems micromachining technology. The first consists of 32 silicon nitride cantilevers separated by 100 µm, while the second consists of eight boron-doped silicon tips separated by 310 µm. The former offers writing and imaging capabilities, but is challenged with respect to tip sharpness. The latter offers smaller linewidths and increased imaging capabilities at the expense of probe density. Parallel generation of nanoscopic monolayer patterns with a minimum linewidth of 60 nm has been demonstrated using an eight-pen microfabricated probe array. (Some figures in this article are in colour only in the electronic version) 1. Introduction High-throughput lithography and surface patterning with extremely fine linewidths (e.g. of the order of 10 100 nm) are very important for the future growth of the microelectronics industry and nanotechnology [1]. The emerging field of nanotechnology requires patterning and functionalization of surfaces with a resolution that is comparable with the scale of the molecules and cells that need to be manipulated and modified. Next generation integrated circuit technology will inevitably call for efficient and low-cost generation of features with a sub-100-nm linewidth. Dip-pen nanolithography (DPN) is a new method of scanning probe nanolithography [2]. It functions by depositing nanoscale patterns on surfaces using the diffusion of a chemical species from a scanning probe tip to the surface. 3 Current address: Department of Physics, Center for Materials Research and Technology, Florida State University, Tallahassee, FL 32306, USA. 4 Authors to whom any correspondence should be addressed. Figure 1. A schematic diagram of the DPN process where the ink molecules transfer from a probe tip to the surface. There is strong evidence to suggest that in many cases molecular diffusion is achieved through a water meniscus that naturally forms between tip and sample under ambient conditions [1 5], as shown in figure 1. Currently, several groups are investigating and modelling the molecular transport 0957-4484/02/020212+06$30.00 2002 IOP Publishing Ltd Printed in the UK 212

A MEMS nanoplotter with high-density parallel DPN probe arrays Figure 2. A schematic diagram of a multipen, parallel DPN writing system implemented in a conventional scanning probe microscope. mechanism. Once on the surface, the molecules chemically anchor themselves to the substrate, forming robust patterns. Features in the range of 10 nm to many micrometres can be fabricated with commercially available silicon nitride tips. One of the factors that influences the linewidth of DPN writing is the translation speed of the tip. Smaller linewidths are achieved with faster tip speeds. Other factors that influence the linewidth include tip sharpness and the diffusion constants of the molecules used as inks [2 5]. DPN technology has been implemented using a commercial scanning probe microscope (SPM) instrument. In a conventional experiment, the DPN probe chip is mounted on an SPM scanner tube in a manner similar to commercially available SPM tips. Precise horizontal and vertical movements of the probes are attained by using the internal laser signal feedback control system of the SPM. Initial DPN experiments involved a single probe (pen). However, parallel DPN patterns have been realized using an array of up to eight commercial pens with an inter-pen spacing of 1.4 mm [3, 4]. Significantly, it has been determined that the DPN linewidth is not a strong function of the contact force within a limited range [4]. This important characteristic of DPN eliminates the need to perfectly align the probe arrays to the substrate surfaces and, therefore, the need for complicated multiple pen feedback systems. It is believed that the throughput of DPN can be significantly increased if a large and dense array of DPN pens is used to create surface features in parallel. The major objective of this work is to develop a nanoplotter with an array of microfabricated, closely spaced DPN probes and to demonstrate parallel DPN writing with such an array, see figure 2. 2. Design Two types of DPN probe arrays, denoted type-1 and type-2, have been developed in our work. The type-1 probe arrays are made of thin-film silicon nitride, whereas the type-2 probe arrays are constructed of heavily boron-doped silicon. For each type of DPN probe array, microfabricated tips are integrated into the end of the probe shanks. It is preferred that the designs and fabrication processes for these probes can be extended in the future to accommodate active, individually addressable probes. Major DPN probe array design variables include the force constants of the probes, the length of the probes, the sharpness of each tip, and the tip-to-tip spacing. The desired range of force constant is 0.03 0.3 N m 1, as discovered through earlier studies [2, 3]. If the force constant of a probe is too large (i.e. greater than 0.3 N m 1 ), the probe will scratch the substrate instead of following its topography and will be unable to accommodate misalignment between the array and substrate. On the other hand, if the force constant is too small, two problems are introduced. Firstly, the capillary and van der Waals force may be large enough to make it difficult to lift the probe array off the surface, thereby undermining the controllability of the DPN process. In addition, the probes may become too sensitive to external noise such as platform vibrations. The length of an individual probe is limited in our design to comply with the existing SPM instrument, which uses commercial SPM probes with a length of approximately 50 300 µm. If the probes are more than 1.5 mm long it becomes difficult to integrate them into conventional systems with pre-existing laser-based displacement sensing elements. The sharpness of the tip influences the minimal achievable linewidth with smaller radii of curvature leading to narrower linewidths. The type-1 and type-2 probe arrays differ in several aspects. The type-1 probe array has 32 probes separated by a 100 µm spacing, while the type-2 array has eight probes with a 310 µm spacing. A type-1 probe is 400 µm long and 50 µm wide. A type-2 probe consists of folded, segmented cantilever support beams. The overall length is 1400 µm while the beams are 15 µm wide. The thickness of a type-1 probe array, made of LPCVD thin-film silicon nitride, is of the order of 600 nm. The type-2 probe array is made of single-crystal silicon. It is much thicker, with the thickness of the shank being 10 µm. The thickness of the type-2 probe array is determined by that of the heavily doped etch-stop layer. The practical limit for the etch-stop layer thickness under our developed fabrication process is 5 µm; this ensures robust and high-yield processing. The force constant (k) of a fixed-free cantilever beam with a length, width and thickness of l, w and t is given as k = Ewt3 (1) 4l 3 under the small displacement assumption. The term E denotes the modulus of elasticity of the probe material. The approximate modulus of elasticity is 385 GPa for the silicon nitride material (type-1 probe array) and 190 GPa for silicon (type-2 probe array). Since the force constant of a cantilever beam is proportional to the thickness to the third power and inversely proportional to its length to the third power, a type-2 probe array needs to be longer than its type-1 counterpart in order to achieve the same force constant. Furthermore, in order to comply with the overall restriction on probe length, the cantilever beams are folded to reduce their overall length. The folded beams, on the other hand, increase the overall width of each individual probe and increase the inter-probe spacing. Due to the different fabrication processes used for each type of array, the sharpness of the tips as defined by the radius of curvature (r o ) is different. For a type-1 probe array, the radius of each tip is approximately 700 nm. For a type-2 213

M Zhang et al Figure 3. Results of computer simulation of the vertical displacement of a single type-2 DPN probe under a constant force of 6 µn applied at the end of the beam. probe array, r o is approximately 100 nm. In fact, one of the major motivations for making a type-2 probe array is to increase the sharpness of the tips and hence enable smaller writing linewidths. Although designed for writing purposes, the probe arrays of both types have resonant frequencies (20 60 khz) and force constants (0.03 0.3 N m 1 ) that allow them to perform contactand tapping-mode atomic force microscopy (AFM) imaging similar to the way a commercially available AFM probe tip is used. To optimize the design of the DPN probe arrays and to meet all of the aforementioned performance criteria, we have used computer-aided finite-element simulations for the mechanical design of the beams with complex geometries, see figure 3. The force constant of the beam shown in figure 3 was calculated to be 0.315 N m 1. It is noteworthy that parallel probe arrays for scanning probe microscopy have been developed in the past [6, 7]. However, such probes are not specifically targeted for DPN applications and do not meet the requirements of DPN. The two types of probe array developed in our work address dimensional and performance requirements that are specified for DPN writing and imaging. 3. Development of the fabrication process The development of the microfabrication process follows two guiding principles. Firstly, the fabrication process must be simple and robust, capable of realizing high yield. Secondly, the developed processes should be compatible for use in future generations of devices where the actuation of individual probes may be called for. The fabrication process of a type-1 device begins with a {100} silicon wafer, see figure 4. Firstly, a layer of silicon dioxide film is thermally grown on the wafer and patterned. The silicon dioxide pattern serves as a mask for the subsequent silicon anisotropic etching step, which is used to form a Figure 4. A schematic diagram of the fabrication process of a type-1 DPN probe array. A {100} n-type silicon wafer serves as the starting wafer. Figure 5. SEM micrograph of a type-1 32 DPN probe array made. The insert shows an enlarged view of a single tip at the end of a beam. pyramidal-shaped tip. Because the etching process is not uniform across the wafer surface, it is practically impossible to form sharp tips with uniform radii of curvature at this stage. In order to increase the yield, the tips are intentionally not sharpened. In fact, tips with remaining flat top (1 µm 2 ) surfaces are produced. The tip is sharpened by thermally growing silicon oxide with subsequent oxide removal using buffered hydrofluoric acid. This process results in tips with uniform and reduced radii [8]. A 600 nm thick silicon nitride thin film is deposited onto the silicon wafer using a low-stress, low-pressure chemical vapour deposition process. The conformal, blanket deposition of a thin film with a finite thickness increases the r o of the tips. The silicon nitride layer on the front side of the wafer is patterned to define the shanks of individual probes. A subsequent anisotropic wet etching using aqueous 95 C ethylene-diamine pyrocatechol (EDP) solution (Transene Company Inc.) from the same side of the wafer frees the probe arrays. A scanning electron microscopy (SEM) micrograph of a type-1 probe array is shown in figure 5. 214

A MEMS nanoplotter with high-density parallel DPN probe arrays Figure 8. LFM images of ODT patterns on a gold surface as written by two adjacent type-1 silicon nitride probes. The writing speed is 0.1 µms 1. The contact force is estimated as 1.3 nn. The humidity and the temperature of the ambient are 35% and 23 C, respectively. The linewidth is 260 nm. Figure 6. A schematic diagram of the fabrication process of a type-2 DPN probe array. The process initiates from a three-layered silicon wafer that has a heavily boron-doped silicon layer (etch stop) sandwiched between a substrate and an epitaxial {100} silicon layer. Figure 9. LFM images of ODT patterns on a gold surface as written by two adjacent type-2 silicon probes. The linewidth is 60 nm. The patterns are generated under a writing speed of 0.5 µms 1 and an estimated contact force of 2.5 nn. Figure 7. SEM micrograph of a type-2 eight DPN probe array. The insert shows a magnified view of a single tip at the end of a beam. The radius of curvature of this tip is estimated to be 100 nm. The fabrication process for a type-2 device begins with a silicon wafer that consists of three layers, see figure 6. The top layer is made of 10 µm thick epitaxially grown silicon. The second layer, called the etch-stop layer, is made of silicon that is heavily doped with boron (doping concentration >10 19 cm 3 ). The epitaxial layer and the heavily doped silicon layer lie on top of a single-crystal silicon substrate with a thickness of 300 µm. Firstly, the wafer is oxidized to form a 500 nm thick layer of silicon dioxide on both sides of the wafer. Double-side alignment marks, which can be used for referencing purposes in later processing steps, are first made on both sides of the wafer. The silicon dioxide layer on the front side of the wafer is then patterned to form the mask for the subsequent anisotropic etching of the top epitaxial silicon, which results in a silicon pyramidal tip lying on top of the heavily doped silicon layer. The wafer is then oxidized to sharpen the pyramidal tip and to protect the wafer during subsequent etching. The bulk substrate is etched from the back side of the wafer using EDP, releasing a diaphragm made of the heavily doped silicon layer. The etch rate on the heavily doped silicon layer is much smaller compared with the etch rate of the substrate bulk [8]. The process can withstand over-timed etching to ensure that the etching of the silicon substrate is time-insensitive and the thickness of resultant probes is uniform. This significantly increases the yield and uniformity of the final devices. A reactive ion etching (RIE) process is performed to define the cantilever beams, and the final structure is imaged by SEM, see figure 7. 4. Experimental results The force constants and the resonant frequencies of the type-1 and type-2 probe arrays have been validated experimentally, and a good match between experimental results and numerical simulation was found. The force constant and resonance frequency for the type-2 probes were determined to be 0.327 N m 1 and 23.03 khz, respectively. The force constant estimated using the finite-element simulation is 0.315 N m 1. The discrepancy is attributed to the deviation of dimensions from the design. The type-1 probe has a measured force 215

M Zhang et al 110 nm Figure 10. A three-dimensional LFM image of ODT pattern generated by a type-2 silicon DPN probe array. The image is processed with Z-axis renormalization and 2D FFT filtering for enhanced contrast. Figure 11. A schematic diagram illustrating the potential application of DPN in subtractive lithography. Sub-0.1-µm features can be generated on a {100} silicon wafer in a high-speed, parallel fashion. constant of 0.03 N m 1 and a resonance frequency of 35 khz. The force constant estimated under the small displacement assumption is 0.019 N m 1, see equation (1). The estimation is based on an effective length of 375 µm, measured from the pyramid tip to the fixed support. We have successfully integrated both probe arrays into a ThermoMicroscope AutoProbe M5 SPM with a closed loop scanner. Tests have been conducted to study the linewidth and imaging resolution achievable using the parallel DPN arrays. Both types of array were used to perform DPN patterning of 1-octadecanethiol (ODT) onto a gold surface, and the patterns were imaged using the same probes via lateral force microscopy (LFM). Figure 12. Contact-mode topographic AFM images of eight copies of inductor patterns generated using a type-2 parallel DPN probe array. The ODT layer is used as a mask to pattern gold. The linewidths range from 80 to 100 nm. 216

A MEMS nanoplotter with high-density parallel DPN probe arrays 5. Conclusions Figure 13. A perspective contact-mode topographic image of DPN-generated gold inductor pattern. The linewidth is 500 nm. With the type-1 probe array, we have achieved an average linewidth of 260 nm with a writing speed of 0.1 µm s 1 and a contact force of 1.3 nn (humidity 35%, temperature 23 C) (figure 8). However, with the type-2 probe array, structures with a minimum linewidth of 60 nm have been realized with a linear writing speed of 0.5 µms 1 and a contact force of 2.5 nn (figure 9). Note that the pattern displayed on the right side of figure 9 is legible even though portions of the gold substrate on which it was written contain surface contaminants. This demonstrates that the probe array is tolerant of adverse surface conditions during DPN writing and imaging operations. Patterns produced from ODT can be transferred to underlying layers. For example, the sub-0.1-µm ODT patterns can act as masks for etching the gold layer beneath them. In single tip experiments, it has also been shown that these gold patterns can be further transferred to layers underneath the gold, such as the silicon substrate [8]. The process for pattern transfer is illustrated in figure 11. Further experiments have shown that eight identical copies of an inductor pattern can be made using a type-2 DPN probe array, see figure 12. These patterns are first generated on a gold surface with ODT via DPN. The patterns are then etched for 20 min with the aqueous solution of 1.0 M KOH, 0.1 M Na 2 S 2 O 3, 0.01 M K 3 Fe(CN) 6, 0.001 M K 4 Fe(CN) 6 while stirring with a magnetic stirring bar, forming elevated gold nanostructures, see figure 12. Each individual inductor pattern is 6 6 µm 2 in size. The linewidths of the patterns range from 80 to 100 nm. Patterns with greater linewidths (500 nm) can be intentionally realized by decreasing the writing speed. For example, a linewidth of 500 nm was obtained with a writing speed of 0.1 µms 1 at 40% humidity and 23.3 C, see figure 13. We have successfully developed arrays of high-density probes with integrated tips for DPN patterning. We have used the newly developed probes to achieve parallel writing of DPN patterns by transferring ODT to gold-coated surfaces. Furthermore, we have identified several key design criteria for making parallel arrays with closely spaced tips suitable for DPN. Specifically, this work shows that one must balance the desire for tip sharpness with writing density and imaging capabilities. A type-1 probe array made of silicon nitride is thin and can realize high probe density. However, the sharpness of each tip is limited by the conformal blanket deposition of the silicon nitride thin film. A type-2 probe array, on the other hand, exhibits smaller tip radii of curvature, but the inter-probe spacing is greater than that of type-1 probes because of the necessary folding of support beams. In the future, we plan to develop new designs and robust fabricating processes to realize DPN probes that provide both sharp tips (comparable to current type-2 probes) and high integration density (comparable to current type-1 probes). Acknowledgments This work was performed under the Advanced Lithography Program of the Defense Advanced Research Projects Agency (DARPA) (grant DAAD19-00-1-0414). The authors also wish to acknowledge financial support from the Air Force Office of Scientific Research and the NSF-supported Nanoscale Science and Engineering Center at Northwestern University. Researchers at the Micro and Nanotechnology Laboratory of the University of Illinois thank Richard Blaney and John Hughes, along with other staff members. References [1] Service R F 2001 Science 293 785 6 [2] Piner R, Zhu J, Xu F, Hong S and Mirkin C A 1999 Science 283 661 3 [3] Jang J, Hong S, Ratner M A, Schatz G and Mirkin C A 2001 J. Chem. Phys. 115 2721 [4] Hong S, Zhu J and Mirkin C A 1999 Science 286 523 5 [5] Hong S and Mirkin C A 2000 Science 288 1808 11 [6] Minne S, Yaralioglu G, Manalis S, Adams F, Zesch J, Atalar A and Quate C Appl. Phys. Lett. 72 2340 2 [7] Lutwyche M, Andreoli C, Binning G, Brugger J, Drechsler U, Haberle W, Rohrer H, Rothuizen H, Vettiger P, Yaralioglu G and Quate C 1999 Sensors Actuators 73 89 94 [8] Liu C and Gamble R 1998 Sensors Actuators A 71 233 7 [9] Weinberger D and Mirkin C A 2000 Adv. Mater. 12 1600 3 [10] Xia Y, Zhao X, Kim E and Whitesides G 1995 Chem. Mater. 7 2332 7 217