IOP PUBLISHING

JOURNAL OF MICROMECHANICS AND MICROENGINEERING

doi:10.1088/0960-1317/17/5/006

J. Micromech. Microeng. 17 (2007) 883–890

Fabrication of precision 3D microstructures by use of a combination of ultraprecision diamond turning and reactive ion etching process Yang Chen, Lei Li and Allen Y Yi Department of Industrial, Welding and Systems Engineering, The Ohio State University, 210 Baker Systems Building, 1971 Neil Ave, Columbus, OH 43210, USA E-mail: [email protected]

Received 21 January 2007, in final form 6 March 2007 Published 3 April 2007 Online at stacks.iop.org/JMM/17/883 Abstract A novel 3D microstructure fabrication process using a combination of an ultraprecision machining process and reactive ion etching technique is described. Compared to the lithographic process, the conventional microstructuring technique, this new process can produce continuous 3D features on wafer substrates instead of the multi-level profiles created using the lithography-based method. Using ultraprecision single point diamond turning and slow tool servo control, true 3D microstructures were fabricated first on a thin SU-8 photoresist layer, and then the patterned 3D structures on the photoresist layer were transferred to the silicon substrate by precisely controlling the etch depth using the timed reactive ion etching technique. Specifically, a diffractive optical element and a sine wave grating were fabricated using this method on silicon and fused silica wafer substrates. The machined wafers were measured by using an optical profilometer, a scanning electron microscope and an atomic force microscope. All measured profiles showed good agreement with the original design. Finally, a diffractive optical pattern that was fabricated using this technique was successfully transferred to a glass optical surface by use of the compression molding process as a demonstration of possible industrial applications. (Some figures in this article are in colour only in the electronic version)

1. Introduction Accurate 3D microstructures on silicon or other wafer substrates are very important for optical, electronic and many other MEMS devices. Traditional cleanroom lithography and the etching process are essentially a 2D method, although complicated procedures were designed to create some 3D microstructures. These processes however are mainly used to create planar features on a silicon wafer substrate using the bulk silicon machining technique [1]. In the past several decades, special techniques were developed for 3D microstructure fabrication. For example, Beuret et al used the multi-lithography step to obtain an inclined structure 0960-1317/07/050883+08$30.00

© 2007 IOP Publishing Ltd

through two shifted masks inclined rotating exposures [2]. Ghodssi’s group developed a method of gray-scale lithography for fabricating 3D silicon MEMS structures used for diffractive optics [3–5]. Tostu et al used a maskless method as part of the multi-layered exposure process to create 100 µm diameter spherical and aspherical microlens arrays on the silicon substrate [6]. Bourouina et al created a new technique using the microloading effect in reactive ion etching (RIE) for micromachining 3D structures with only one mask [7]. Zhang et al used a method based on the micro-stereolithography process to build 3D microstructures on polymer materials by the laser-induced solidifying technique [8]. These methods, however, almost always involved multi-exposure and precision

Printed in the UK

883

Y Chen et al

alignment to create the multi-level profile in the photoresist to simulate the continuous profile of a 3D structure. Therefore, it is difficult to achieve high quality in a true 3D microstructured surface. Moreover, the complexity of these processes often becomes the source of high manufacturing errors, thus resulting in high fabrication cost and a prolonged cycle time. It is also very difficult to make changes to these manufacturing methods since each step is directly associated with the up and down steam processes. In ultraprecision single point diamond turning, real 3D freeform surface features with both geometric tolerance and surface finish required for image optics can be fabricated. In the past, ultraprecision diamond turning was tested for directly fabricating microstructures [9]. More recently, Yi et al developed a simple slow tool servo (STS) process for a microlens array and diffractive optical elements (DOEs) [10, 11]. The slow tool servo process is an ultraprecision fabrication method based on the feedback control of mechanical slides. By use of these methods, true 3D continuous microscale surfaces with optical surface finish can be fabricated without the expensive masks and complex multi-level photolithography setup, which can greatly reduce manufacturing cost and shorten production cycle time. Moreover, this technique can also be used to create microstructures on almost any surface profile other than a flat one that is required for lithography. There are important industrial applications that can benefit from this rather important and unique feature. In this paper, a new method that combines the ultraprecision machining process with the cleanroom RIE for 3D microstructure fabrication was investigated. Using a commercially available single point diamond turning machine and the STS process if necessary, true 3D microstructures were first created on a thin photoresist layer (10–20 µm). These 3D microstructures were then transferred to a silicon or fused silica substrate using the anisotropic RIE method. In a demonstration of their possible applications in industry, using a micromachined silicon wafer as a mold, 3D microstructures were duplicated onto a low transition temperature optical glass using the precision compression molding technique, an emerging high volume precision optical manufacturing process.

Silicon Substrate Spinning coat 40µm SU8-2025 Photoresist Photoresit Silicon Substrate Ultra-Precision Machining 3D Microstructure

Silicon Substrate O2 Plasma Etching

Silicon Substrate CF4-O2 Mix Plasma Reactive Ion Etching

Silicon Substrate

Figure 1. Process flow for 3D microstructures on silicon substrate.

because of its availability and easiness to work with a diamond turning process. Young’s modulus of SU-8 is about 4.5 GPa. The SU-8 photoresist also has the appropriate mechanical properties for the ultraprecision diamond machining process [12, 13]. The viscosity of the SU-8 2025 photoresist (MICRO CHEM) is about 4500 cSt prior to baking. To prepare for the photoresist layer, 40 µm thick SU-8 was spin coated on the silicon wafer. This thickness is desired for the single point diamond turning fabrication process. The spin coater’s rotational speed was 1000 rpm. After the spin coating was completed, the wafer was baked for 3 min at 65 ◦ C and 9 min at 95 ◦ C on a soft bake oven. To improve the mechanical property of the SU-8 photoresist, the wafer was put in a hard oven at 115 ◦ C for 20 min. The complete details of ultraprecision machining, RIE and compression molding process are described below. Furthermore, as a demonstration, a high volume precision compression molding process was introduced to transfer the pattern from a silicon substrate to an optical glass after the micromachining steps were completed.

2. Experiment To start the experimental process, first the silicon wafer was cleaned using a sulfuric and hydrogen peroxide mix. Then a thick layer of SU-8 photoresist was deposited on the surface. After baking, the selected 3D microstructures were then fabricated by means of ultraprecision machining on the photoresist. The machined structures were then transferred to a silicon wafer surface by timed RIE. The process flow is graphically depicted in figure 1. Although the process was demonstrated on silicon and fused silica wafers in our investigation, the fundamental principle can be easily transferred to almost any material with arbitrary shapes. This approach is especially attractive for super hard materials such as silicon carbide, glassy carbon and other materials that cannot be diamond machined to optical surface quality. In our investigation, SU-8 was selected as the photoresist mainly 884

2.1. Ultraprecision machining 3D microstructures on the photoresist Since it is impossible to machine certain substrates such as glassy carbon or silicon carbide using a single point diamond tool, creating 3D microstructures on the photoresist first, then transferring it to the substrates becomes a necessary step on such an occasion. In micromachining of the photoresist layer, we have encountered several difficult issues. First, because the photoresist layer was spin coated on silicon or fused silica wafers, its thickness could not be precisely controlled. Secondly, air bubbles can occur during the curing stage; thus the accuracy and surface finish of the fabricated structure may be compromised. Thirdly, the photoresist can sometimes be accidentally peeled off from the wafer surface during

Fabrication of precision 3D microstructures

Table 1. Ultraprecision machining process parameters. Process

Tool radius (mm)

Spindle speed (rpm)

Feed rate (mm min−1)

Depth of cut (µm)

Surface rough turning Surface finish turning Structure rough turning Structure finish turning STS rough machining STS finish machining

3.175 3.175 0.0025 0.0025 0.254 0.254

800–1000 800–1000 800–1000 800–1000 Stationary Stationary

6–10 1–2 5 1 600 600

1–4 1 1 1 1 0.5

the diamond turning process, especially when the remaining photoresist layer becomes very thin, e.g., several microns. To solve these problems, several precautionary steps were implemented. First, part of the photoresist was removed by acetone so that a small portion of the wafer substrate was exposed and could be used as a reference for accurate measurement of the photoresist thickness; this reference section should be as close to the section where the 3D microstructure will be machined as possible. Secondly, the photoresist was diamond turned to the required thickness. In our experiments, the diamond turning process was carried out on the Moore 350 FG Ultraprecision Freeform Generator (Moore Nanotechnology Systems, Keene, New Hampshire). The 350 FG ultraprecision machine has three linear axes that are equipped with linear laser scales capable of resolving 8.6 nm while moving at a maximum speed of 1800 mm min−1. The straightness on all slides is less than 250 nm. The work spindle is capable of reaching 6000 rpm while maintaining axial and radial error motion to less than 25 nm. The work spindle can also maintain angular position accuracy to less than 5 arcsecond in a modulated mode. The photoresist layer was rough machined from 40–45 µm to 15–20 µm, and then was finish machined to 10–15 µm thickness. The machining process parameters are summarized in table 1. After each machining step, the photoresist thickness was measured using an ultraprecision indicator (Federal Gage EHE-2056 with 100 nm repeatability and less than 4 g measuring force) to ensure that the remaining thickness is appropriate for microstructure fabrication. After the surface was diamond turned, a smooth and flat optical surface was obtained. The third step was to rough and finish machine the design pattern using the diamond turning process. For complicated 3D structures that are not axysymmetric, the slow tool servo diamond turning process has to be used. In our study, a diffractive optic element was diamond turned and a sine wave grating surface on photoresists was fabricated by use of the slow tool servo process. To eliminate the air bubbles in the photoresist layer, delivery of the photoresist solution must be carefully controlled during the spin coating stage. In case air bubbles develop, vacuum can be applied to the photoresist to remove them. It was also found in our study that thorough cleaning using acids prevents the photoresist layer from peeling off easily. In addition, a minimal photoresist layer thickness should also be maintained. In the case of SU-8, this thickness is about 3–5 µm. 2.2. Reactive ion etching (RIE) The RIE was performed on the Technics Micro-RIE Series 800-II unit. The control of this system can be easily adjusted

Table 2. RIE etch parameters. Etched material Surface profile O2 flow rate CF4 flow rate Chamber pressure Electrode power Etch time Etch rate ratio (PR/substrate)

Silicon Sine wave 20 sccm 20 sccm 270 mTorr 200 W 40 min 1

Silicon DOE 30 sccm 20 sccm 270 mTorr 200 W 30 min 1.5

Fused silica Sine wave 20 sccm 20 sccm 270 mTorr 300 W 120 min 4

to change the etching process conditions that include gas flow rate, electrode power, chamber pressure and etch time. To replicate the 3D microstructure on the microstructured photoresist layer using the single point diamond turning process, compensation in etch rate between that of silicon and photoresist can be used to compensate for the etch rate difference in the photoresist and the silicon wafer (or other substrate material, e.g., fused silica). This compensation can be achieved by fabricating a deviated profile on the photoresist or by varying the plasma conditions. For the CF4–O2 gas mixture that was used in the plasma etching process for both silicon and photoresist in our first experiment [14, 15], an even etch rate on both silicon and photoresist was identified to transfer the 3D photoresist microstructure to the silicon substrate by carefully varying the ratio of CF4 and O2, supply power, flow rate and chamber pressure. To achieve a higher production (etch) rate, the oxygen plasma was first used to etch the photoresist 3D structure until the lowest point of the structure reached the silicon substrate. At the end of the first step, a formula that can be used to etch both the photoresist and silicon at a similar rate was selected through trial and error. The final conditions used in this paper are summarized in table 2 for three different etching conditions and materials. Using the conditions in table 2, microstructures fabricated on the photoresist layer by ultraprecision diamond turning were accurately transferred to the silicon substrate. The ratio between the etch rate on the photoresist and silicon wafer is also summarized in table 2. Figure 2 shows the crosssectional view and the line scan image of a silicon blaze after etching with this formula. The measurements were acquired using the Veeco NT 3300 optical profilometer. To test this process on different materials, particularly on materials that cannot be diamond turned at all, e.g., ceramic and glass, a sine wave profile was also fabricated on a fused silica wafer. The parameters for machining fused silica are also listed in table 2. 885

Y Chen et al µm

Glassy Carbon

1.98

Forming Force

0.00

-0.66

(a) Heating 2

Molding

1.5

Mold platen (two shown)

µm

1

Silicon Wafer

0.5

Glass

0 -0.5 0

200

400

600

µm

Cooling

Release

(b)

Figure 2. Fabricated silicon blaze ridge after etching.

2.3. Glass molding experiment The glass molding experiment was performed on a Toshiba GMP 211V machine at the Fraunhofer Institute for Production Technology in Germany. The detailed descriptions of the experiment can be found elsewhere [16, 17]. A low transition temperature glass (Tg = 285 ◦ C, K-PG325 from Sumita, Japan) was used as raw material. The refractive index of this glass is 1.50 679 and the thermal expansion coefficient (TEC) is 17.3×10−6 ◦ C−1. In this paper, our main goal is to study the new micromachining method to fabricate true 3D microstructures on hard and brittle materials, such as silicon, fused silica and glassy carbon. These components can be used directly in micro device fabrication but may also be suitable for use as mold/stamper for high volume production. The silicon wafer mold with microstructures and an optical glass specimen were placed between two optically polished glassy carbon wafers. The glassy carbon wafers were used to prevent sticking between the glass material and the carbide mold platens as a precautionary measure, since the carbide platens were not optically polished, thus have a tendency to stick to the glass specimen. The complete compression molding processes are illustrated graphically in figure 3. The glass and silicon wafers were first heated to a temperature that was slightly higher than the temperature Tg of the glass material. The lower mold was then moved up to the molding position, and a forming force was applied. Slow cooling of the glass sample and mold assembly was started after the glass was thermally saturated at the forming temperature for about 1 min. The molding process is completed with fast cooling to save time. The finished glass optic was manually removed from the mold assembly. The total molding cycle time was about 12 min. Glass samples detached from the silicon wafer mold right after the annealing process, indicating a non-sticking situation between the low Tg glass and the micromachined silicon wafer mold. Since only a limited number of DOEs were molded, no reliable information on the tool wear was available. The silicon wafer mold was observed using an optical microscope under 886

Figure 3. Glass molding process using silicon mold.

high intensity light after each cycle, but no sign of wear was detected. For this glass type and the silicon wafer, the selected molding temperature was fairly low; thus we would not expect substantial wear due to thermal loading. Alcohol was used to clean the samples prior to performing the surface measurements.

3. Results and discussion To quantify the process capability, a 10 mm diameter diffractive optical element was fabricated using a single point diamond turning machine on the SU-8 2025 photoresist. The height of the finished microstructure H (r) is a function of radius r as in equation (1): H (r) = r 2 − 3n

(1)

where n denotes the nearest integers less than or equal to r2/3. Surface measurements were performed to evaluate the difference between the design and machined profile on the photoresist. Both the Veeco Dektak Series 3 stylus profiler and the Veeco NT3300 optical profilometer were used to measure the photoresist profile. Figure 4 shows both the diffractive optical element design and the measurement on the photoresist after diamond turning and the scanning electron microscope (SEM) image of the diffractive optical element on the photoresist. The 3D structure profile is considerably better than the result with gray-scale lithography or other multi-level lithographic-based method. As illustrated in figure 4, the single point diamond turning process can produce continuous surface features as compared with many other techniques [1–8]. The 3D microstructures on the photoresist were successfully transferred to a silicon substrate by a carefully formulated RIE process. After the etching step was completed, a continuous 3D profile was created. The Veeco NT3300 optical profilometer was used to measure the contour of the patterned photoresist and etched silicon substrate. As further proof of the effectiveness of the process in creating true 3D microstructure, using the same fabrication technique, a sine

Fabrication of precision 3D microstructures

µm

3

Height (µm)

3.32

2 0.00

1 2.5mm

1.9mm -3.72 (a)

0 0

1000

2000

3000

µm

Radius (µm)

3.13

(a)

0.00

2.5mm

1.9mm

-2.84

(b)

3 2

Photoresist

µm

1 0 -1 -2 (b)

Figure 4. (a) Profile of a diffractive optical element after diamond turning compared with the design. The dashed line is the design curve and the solid line is the diamond turned surface measured by a Veeco Dektak Series 3 stylus profiler system. (b) SEM image of the diffractive optical element on the SU-8 2025 photoresist after diamond turning.

wave grating was created. Because the etch rate of the photoresist using the selected plasma etch recipe was the same as the silicon substrate for the plasma conditions that were used, the height of the sine wave grating after etching was the same as the photoresist after diamond machining. The contours of both photoresist and machined silicon substrate measured using the Veeco optical profilometer are shown in figure 5. The profile of a selected peak from the sine wave grating on photoresist as well as on silicon wafer is plotted together in figure 5(c) for comparison. The 3D analysis plot of the first two ridges of the etched diffractive optical element is shown in figure 6, measured on the Veeco NT3300 optical profilometer. The comparison between the photoresist profile and the etched profile on the silicon wafer is also shown in figure 6(b). Under the second set of etching conditions listed in table 2 (O2 flow rate 30 sccm, CF4 flow rate 20 sccm), the etch rate of the SU-8 photoresist was about 1.5 times faster than that of silicon. Thus, the error caused by the etch rate difference can be precisely compensated by offsetting the design curve. To obtain the proposed profile (the dashed line), the photoresist

-3 0.3

After etching

0.6

0.9 mm

1.2

1.5

(c)

Figure 5. (a) 3D image of a sine wave grating on SU-8 photoresist. (b) 3D profile of the sine wave grating after etching measured by the Veeco optical profilometer. (c) Comparison of the profile of one period from the sine wave grating before and after etching.

was fabricated as the modified profile (the solid line) using the diamond turning process. After the etching process was completed, the final results were measured using the Veeco optical profilometer and plotted as the center line. The small columnar structures at the center in figures 4(a) and 6 are fabrication errors caused by inaccuracy in centering the diamond cutter during the single point diamond turning process. The final results showed good agreement with the design profile. The surface roughness was also studied by using an atomic force microscope (AFM). A surface roughness of Ra = 88 nm or an arithmetic average was obtained after etching as shown in figure 7. This surface roughness increase was mainly due to the aggressiveness in the reactive ion etching process. The same measurement on a diamond machined photoresist indicated that the surface roughness was less than Ra = 10 nm prior to etching. Although the microstructures fabricated in this study showed sub-mm scale resolution in the lateral direction, there may exist certain optical applications that would require a higher precision. We believe that both the diamond turning and etching processes need to be improved 887

Y Chen et al µm

µm

1.5

0.92mm

1.23

0.00

0.00

4.9mm 3.7mm

-1.5

-1.68

1.2mm

(a)

(a)

Etched Profile on silicon substrate

Curve fit result of photoresist profile

6

Diamond turned photoresist profile

4 Height (µm)

Design Profile

Height (µm)

3

2

2 0 -2 Design curve

-4 -6

1

0

0.2

0.4

0.6

0.8

Radius (mm) 0

0

1 2 Radius (mm)

+ Measured profile after etching • Measured profile on photoresist

3

(b)

Figure 6. (a) First two ridges of an etched diffractive optical element (b) Design, diamond machined and etched profile on silicon substrate.

(b)

Figure 8. (a) Image of a sine wave grating after etching on fused silica substrate measured by the Veeco optical profilometer. (b) Profiles of one period sine wave grating before and after etching as compared to the design profile.

µm 1.57

0.00

-1.25 (a)

X: 20 µm/division Z: 4 µm/division

Figure 7. AFM image of an etched diffractive optical element surface acquired using the Veeco NanoMan Dimension 3000 SPM.

in order to achieve higher lateral resolution. In addition, using a different masking layer other than the SU-8 photoresist may also provide a better result in lateral resolution as the etching of the SU-8 photoresist layer using the selected plasma might not have been a perfect anisotropic process. 888

Height(µm)

2 1.5 1 0.5 0

0

0.5

1 1.5 Radius(mm) (b)

2

Figure 9. (a) First two periods of a molded glass diffractive optical element. (b) Comparison between the silicon mold and the molded glass diffractive optical element.

To illustrate the 3D microfabrication capability on nondiamond turnable materials, a thin layer of the SU-8

Fabrication of precision 3D microstructures

photoresist was spin coated on a fused silica (amorphous) wafer. The same sine wave pattern that was machined on the silicon wafer was also diamond machined on the photoresist using the slow tool servo process. After the diamond machining process, the fused silica wafer was etched to replicate the pattern on the wafer substrate. Figure 8 shows the result of the fabricated fused silica wafer. Unlike the etching process on silicon, the etch rate for fused silica was much lower than that of the SU-8 photoresist. Under this etching condition, the etch rate of the SU-8 photoresist was determined to be four times that of fused silica. Based on this result, a modified sine wave grating profile with a depth four times that of design can be implemented in single point diamond turning since the total depth of etching was linearly proportional to etch time as indicated in figure 8. The molded glass diffractive optical element and the 3D microstructure silicon mold were also measured using the Veeco optical profilometer and the results are plotted in figure 9. An excellent geometry accuracy and surface finish match was clearly demonstrated.

(c) Improve the diamond machining process to obtain a higher structure accuracy and surface finish. (d) Improve etching uniformity and identify a better etch selectivity. (e) Identify other etching methods such as the ion beam milling process to improve the finished surface roughness. (f) Use a fast tool servo to increase fabrication rate.

Acknowledgments This material is based upon work supported by National Science Foundation under grant nos CMMI 0547311 and EEC-0425626. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. The authors also would like to thank MicroMD at the Ohio State University for assistance in cleanroom operation. The glass molding experiments were conducted at the Fraunhofer Institute for Production Technology in Aachen, Germany.

4. Conclusions References Ultraprecision single point diamond turning, in combination with the slow tool servo process if necessary, can be readily used to create an accurate 3D microstructure with sub-micron features on the SU-8 photoresist. This is possible due to the unique mechanical properties of the photoresist. The machined 3D microstructures on the photoresist layer can be etched into the substrate surface using the timed reactive ion etching process. Unlike the traditional silicon etching technique, accurately controlled etch selectivity is necessary for transferring the fabricated pattern from photoresist onto the silicon substrate. To do this, a carefully formulated CF4–O2 gas mixture plasma was identified experimentally to achieve preselected etching rates on both the SU-8 photoresist and the silicon wafer. The advantage of this method is the capability of manufacturing true 3D microstructures on almost any materials and any geometrical substrate without using a photo mask. This method is essentially a maskless process in which any patterns can be easily transferred from design to an ultraprecision machine, then to the final substrates. It effectively eliminates multiple masks required for multiple exposures in some lithographic methods, thus resulting in lower manufacturing cost and a shorter production cycle. This process also has less accumulation errors in manufacturing compared to the previous processes where multiple steps were involved. The fabricated microcomponents can be used directly in MEMS devices or as mold inserts in other high volume micromanufacturing processes. One example of the latter applications has been showcased in the compression molding of glass diffractive optical elements. To improve the quality of the 3D microstructures, we would continue our investigation further in the following areas. (a) Precise control and measurement of the photoresist thickness for the single point diamond machining process. (b) Control the quality of the photoresist to improve the uniformity in photoresist and obtain higher surface finish.

[1] Kovacs G T A, Maluf N I and Petersen K E 1998 Bulk micromachining of silicon Proc. IEEE 86 1536–51 [2] Beuret C, Racine G-A, Gobet J, Luthier R and de Rooij N F 1994 Microfabrication of 3D multidirectional inclined structures by UV lithography and electroplating Proc. IEEE Int. Conf. on Micro Electro Mechanical Systems pp 81–5 [3] Waits C M, Modafe A and Ghodssi R 2003 Investigation of gray-scale technology for large area 3D silicon MEMS structures J. Micromech. Microeng. 13 170–7 [4] Morgan B, Waits C M, Krizmanic J and Ghodssi R 2004 Development of a deep silicon phase Fresnel lens using gray-scale lithography and deep reactive ion etching J. Microelectromech. Syst. 13 113–20 [5] Waits C M, Morgan B, Kastantin M and Ghodssi R 2005 Microfabriction of 3D silicon MEMS structures using gray-scale lithography and deep reactive ion etching Sensors Actuators 119 245–53 [6] Totsu K, Fujishiro K, Tanaka S and Esashi M 2005 Gray-scale lithography using mask-less exposure system 13th Int. Conf. on Solid-State Sensors pp 1441–4 [7] Bourounia T, Masuzawa T and Fujita H 2004 The MEMSNAS process: microloading effect for micromachining 3D structures of nearly all shapes J. Microelectromech. Syst. 13 190–9 [8] Zhang X, Jiang X N and Sun C 1999 Micro-stereolithography of polymeric and ceramic microstructures Sensors Actuators 77 149–56 [9] Kawai T, Sawada K and Takeuchi Y 2001 Ultra-precision micro structuring by means of mechanical machining 14th IEEE Int. Conf. on Micro Electro Mechanical Systems (2001 MEMS) [10] Yi A Y and Li L 2005 Design and fabrication of a microlens array by use of a slow tool servo Opt. Lett. 30 1707–9 [11] Li L, Yi A Y and Huang C et al 2006 Fabrication of diffractive optics by use of a slow tool servo diamond turning process Opt. Eng. 45 113401 [12] Hopcroft M, Kramer T, Km G, Takashima K, Higo Y, Moore D and Brugger J 2005 Micromechanical test of SU-8 cantilevers Fatigue Fracture Eng. Mater. Struct. 28 735–42 [13] Yu H, Balogun O, Li B, Murray T W and Zhang X 2006 Fabrication of three-dimensional microstructures based on

889

Y Chen et al

singled-layered SU-8 for lab-on-chip application Sensors Actuators A 127 228–34 [14] White L K and Maa J S 1985 Etch rate enhancement of silicon in CF4–xO2 plasmas Appl. Phys. Lett. 46 1050–2 [15] Rauf S 2004 Model for photoresist trim etch in inductively coupled CF4/O2 plasma J. Vac. Sci. Technol. B 22 202–11

890

[16] Jain A and Yi A Y 2005 Compression molding of aspherical glass lenses—a combined experimental and numerical analysis J. Am. Ceram. Soc. 88 579–86 [17] Jain A and Yi A Y 2005 Numerical modeling of viscoelastic stress relaxation during glass lens forming process. J. Am. Ceram. Soc. 88 530–5