SU-8 Overhanging Structures Using a Photoresist Sacrificial Layer and Embedded Aluminum Mask

Transcription

1 SU-8 Overhanging Structures Using a Photoresist Sacrificial Layer and Embedded Aluminum Mask By: See-Ho Tsang, Dan Sameoto, Sae-Won Lee A MAJOR PROJECT REPORT SUBMITTED IN PARTIAL FULLFILLMENT FOR ENGINEERING SCIENCE 851 In the School of Engineering Science See-Ho Tsang, Dan Sameoto, Sae-Won Lee 2005 SIMON FRASER UNIVERSITY September 13, 2005 All rights reserved. This work may not be reproduced in whole or in part, by photocopy or other means, without the permission of the authors. i

2 Abstract Our lab group has successfully developed a method of creating overhanging structures in SU-8 on a silicon substrate. Two methods of producing moving parts in SU-8 are described; both making use of thick sacrificial layers. The first process uses SU-8 for the structural material and Shipley 1827 positive photoresist as a sacrificial layer. The second method uses SU-8 as both a sacrificial and structural layer, with a layer of metal deposited between the two to allow for separation. The photomasks for each process were designed in Cadence with an array of test structures for determining the effectiveness of each process. These masks were printed on transparencies with a resolution of approximately 30 microns. The results of both methods are presented with several process variations being done with each method. These results include the first demonstration to the authors knowledge of an out of plane pop-up structure made of SU- 8. ii

3 Table of Contents Abstract... ii Table of Contents... iii List of Figures and Tables... iv 1 Introduction Motivation SU-8 Photopolymer Photoresist Sacrificial Layer Process Description Embedded Aluminum Mask Process Description Preliminary Testing and Results Photoresist Spinner Characterization Hot Plate Characterization Photoresist and SU-8 Pouring Method Coating of SU Edge Bead Trials SU-8 Exposure Tests and Profilometer Measurements Preliminary Tests for Aluminum as a Imbedded Mask Mask Design Feature Description Mask fabrication Alignment and Exposure Results of the Photoresist Sacrificial Layer Process Observations and Analysis of Photoresist Spinning Photoresist Patterning and SU-8 Spin Coat SU-8 and Photoresist Development Characterization of Photoresist Features Test Structure Results Stress Observations and Survivability Failure Modes Results of the Embedded Aluminum Mask Process Application of SU-8 to Silicon Substrate Observations and Analysis of Sputtering Process Aluminum Patterning Process SU-8 Post Exposure Bake and Development Characterization of Aluminum Features SU-8 Anchor Layer after Aluminum Patterning Exposed SU-8 Structural Layer Prior to Development SU-8 Structures after SU-8 Development Failure Modes Recommendations and Future Work Ramping hot plate Non expired SU-8 and SU Higher Quality Masks Photoresist Bake time/su-8 Development tests Conclusion Appendix A Sputter log References iii

4 List of Figures and Tables Figure 1: Thick photoresist applied on a wafer... 3 Figure 2: Exposure of photoresist layer... 3 Figure 3: Developed photoresist sacrificial layer... 3 Figure 4: SU-8 10 is spun on the wafer... 4 Figure 5: SU-8 layer is exposed... 4 Figure 6: Final SU-8 overhanging structure... 4 Figure 7: SU-8 10 applied to a silicon wafer... 5 Figure 8: Aluminum is sputtered over SU-8 sacrificial layer... 5 Figure 9: Photoresist is applied over aluminum... 5 Figure 10: Exposure of photoresist... 5 Figure 11: Developed photoresist layer... 6 Figure 12: Etched aluminum mask layer... 6 Figure 13: Second applied layer of SU Figure 14: Exposure of SU Figure 15: Final SU-8 structure... 7 Figure 16: Spin coater... 7 Figure 17: Hot plate ovens Figure 18: Exposure scheme to find optimal exposure time Figure 19: Profilometer measurement of 2 minute exposed area (underdeveloped) Figure 20: Profilometer measurement of 4 minute exposed area (underdeveloped) Figure 21: Profilometer measurement of 6 minute exposed area (underdeveloped) Figure 22: Profilometer measurement of 8 minute exposed area (underdeveloped) Figure 23: profilometer measurement for a properly developed area Figure 24: Revised exposure scheme for SU-8 development Figure 25: Profilometer measurement of 10 seconds exposed area (underdeveloped) Figure 26: Profilometer measurement of 20 seconds exposed area (underdeveloped) Figure 27: Profilometer measurement of 30 seconds exposed area (underdeveloped) Figure 28: Profilometer measurement of 40 seconds exposed area (underdeveloped) Figure 29: Aluminum Etchant Test Figure 30: Close-Up View of Etched Area Figure 31: SU-8 Stripped of Aluminum using Tape Test Figure 32: SU-8 Development of Sputtered Wafer Figure 33: Anchor Mask Cell Figure 34: Surface Feature Mask Cell Figure 35: Anchor Dark Field Mask Figure 36: Structural Layer Dark Field Mask Figure 37: Alignment Markers on Structural Mask Figure 38: Anchor Mask Alignment Markers Figure 39: Overlay of Alignment Markers Figure 40: Transparency mask held by glass overplate Figure 41: Example of Poor Mask Feature Definition of Test Structures (40µm on Left, 30µm on Right) iv

5 Figure 42: Example of Poor Mask Feature Definition of Test Structures (100 µm Size) 20 Figure 43: Example of Spot Defects Across Light Field of Mask Figure 44: Example of Blurred of Features On Comb Drive Figure 45: Shipley 1827 photoresist immediately after being spun onto a silicon wafer. 22 Figure 46: Shipley 1827 photoresist showing tweezer marks caused by wafer handling 22 Figure 47: Shipley 1827 coated wafer showing fringes in coating. Each fringe is ~1/4 wavelength of light, and thick fringes indicate small thickness variations over the wafer area Figure 48: Shipley 1827 coated wafer reflecting objects. Light is reflected by the silicon, but is barely distorted by the photoresist layer, indicating uniform surface coating.23 Figure 49: Wafer #20 immediately after SU-8 layer spin on Figure 50: Close up of wafer #20 after SU-8 spin on Figure 51: Wafer #20 after soft-bake of SU Figure 52: Close-up of wafer #20 after soft-bake showing some surface distortion. This is caused by the SU-8 surface being affected by the topology changes between the photoresist and holes Figure 53: Wafer #20 in SU-8 developer with ultrasonic agitation Figure 54: Wafer #18 after development and rinse. Purple areas are undeveloped photoresist Figure 55: Wafer #20 showing relatively clear area with most photoresist developed. This section of the wafer was resting on the bottom of the jar in the ultrasonic bath Figure 56: Wafer #2 after SU-8 development showing good photoresist removal Figure 57: Smallest L-shaped feature in photoresist Figure 58: Minimum spacing and feature size in photoresist Figure 59: 40 and 50 micron holes in photoresist Figure 60: 50 and 100 micron hole in photoresist Figure 61: Profilometer reading of medium L-shape feature in photoresist prior to being hard baked Figure 62: Profilometer reading of medium L-shape feature in photoresist after being hard baked Figure 63: SU-8 cantilevers suspended over silicon substrate. The clear areas close to the bottom are where the SU-8 is anchored to the substrate. The dark perpendicular line is photoresist remainder Figure 64: Large L-shaped photoresist holes under cross-linked SU-8 layer showing major cracking problems Figure 65: Small pop up mirror design connected to hard baked photoresist after development step. The photoresist could not be removed in developer and shows cracking, possibly induced by ultrasonic agitation Figure 66: Corner of large anchored SU-8 structure and posts, showing edge resolution and residue from development process Figure 67: Large comb-drive on wafer #20 with large film stress. Cracking in remaining photoresist can clearly be seen Figure 68: Large comb-drive fingers curling in towards each other Figure 69: Large comb-drive on wafer #20 showing partial adhesion to substrate Figure 70: Reflection of large comb-drive on wafer # v

6 Figure 71: Photomicrograph of comb-drive with internal stress causing curling down Figure 72: Photomicrograph showing the radius of a curled comb-drvie. This radius is directly related to the internal stress distribution Figure 73: Large pop-up mirror held in place by friction. Mirror is placed in position by microprobes Figure 74: Wide angle view of large pop-up mirror showing the size relative to other objects Figure 75: Wafer #28 post sputtering process Figure 76: Wafer #19 post sputtering process Figure 77: Wafer #25 post sputtering process Figure 78: Wafer #11 post sputtering process Figure 79: Higher magnification view of wafer # Figure 80: Wafer #19 patterned with thin photoresist Figure 81: Wafer #25 patterned with thin photoresist Figure 82: Wafer #19 after aluminum etch without temperature ramp up Figure 83: Close-up of high surface tension Figure 84: Wafer #25 after aluminum etch with temperature ramp up and cool down Figure 85: Close-up of smooth surface Figure 86: square test structures under aluminum (30µm on left, 40µm on right) Figure 87: Square test structures under aluminum (50µm on left, 100µm on right) Figure 88: L-shape test structures (50µ on left, 30µm on right) Figure 89: Anchor of a small pop-up structure Figure 90: Comb drive fingers defined prior to SU-8 development Figure 91: Stipples on large square test structure, also showing the smoothing of reticulated area Figure 92: Pop-up structure defined prior to SU-8 development Figure 93: L-shape test structures also showing the smoothing of reticulated area Figure 94: Corner of square test structure peeling away Figure 95: Fixed-free cantilevers Figure 96: Fixed-fixed cantilevers Figure 97: Well defined comb drive fingers Figure 98: Cross-linked su-8 on slide due to long exposure to aluminum etchant Figure 99: SU-8 film after aluminum etch Figure 100: Surface of SU-8 film removed after SU-8 development Figure 101: SU-8 film after introduction of isopropyl alcohol Table 1: Comparison of set and measured operation time of the spin coater... 8 Table 2: Comparison between the dial reading of the spin coater and measured rotational speed...9 Table 3: Guideline for creating a SU-8 layer with 15µm in thickness Table 4: Height variation between exposed and unexposed areas Table 5: Thickness of SU-8 features for given exposure times Table 6: Thickness of SU-8 features for given exposure times Table 7: Wafer Preprocess Description for Embedded Aluminum Mask Process Table 8: SU-8 Soft Bake Times for Embedded Aluminum Mask Process vi

7 1 Introduction Our group has developed two new MEMS processes for the creation of compliant SU-8 structures. These processes are significantly different to each other and other processes that have been described in existing literature. Both use the photoresist SU-8 from Microchem corp. as a structural material. SU-8 is a negative photoresist, meaning that exposure to UV light will cause it to polymerize and remain after being immersed in developer. SU-8 is also interesting because once it has been exposed it becomes very difficult to remove by chemical means. This means that SU-8 can form permanent structures with one exposure and development step. The processes differ in the material that is used as a sacrificial layer. The first process uses a positive photoresist for a sacrificial layer, while the second process uses unexposed SU-8 for this purpose. In order to complete this project, we have characterized the process of spinning on both SU-8 and positive photoresist, experimented with exposure dose and baking times and tested several methods of producing structures. The process steps and experimental results are listed for each wafer that we produced, with the main criteria being the release of structures and the quality and repeatability of the process. Our most successful designs have used a photoresist sacrificial layer, although using SU-8 as a sacrificial material holds the most long-term promise. 2 Motivation Most surface micromachining processes in use today produce very low aspect ratio structures (width of structures is much larger than thickness) [1, 2]. The main reason for this has been the difficulty and expense of producing high-resolution structures in thick layers. With the exception of LIGA [3] and silicon-on-insulator (SOI) micromachining, [4] there have been few ways for MEMS designers to create tall structures for their devices. Tall thin structures are desirable because they are very compliant in-plane, and very stiff out of plane. This preferential stiffness is good for effective mechanical amplifiers [5] and certain actuators [6]. SU-8 is a negative photoresist that is capable of producing thick, high aspect ratio structures with a single UV exposure and development step. It is also extremely compliant and shows good elastic properties with no ductile failure when fully polymerized [7]. These features have made SU-8 an important material for use in micromachining and micro-fluidics. To date, almost all research using SU-8 has been done with respect to creating passive structures that are generally fixed in place on the substrate. More recently, research has been done into creating compliant structures in SU-8 that are separated from the substrate such as check valves, mechanical amplifiers, and gears [8]. The next logical step is to create active devices in SU-8, including actuators and sensors. With the advent of electrically conductive SU-8 [9], this possibility is getting closer. Another reason to develop a surface micromachining process based on SU-8 is that all processing with photoresists can be done at low (< 200 C) temperatures. This will allow the integration of the MEMS process we design with existing CMOS or equivalent circuits, creating a complete system on chip with sensing, actuation and feedback built right in. 1

8 Surface micromachining processes require sacrificial layers to create space between the structural layers and the substrate to make moving parts. If SU-8 is to be used as a structural material, a suitable sacrificial layer must be found where overhanging and suspended SU-8 structures can be produced quickly and easily. Most of the previously reported processes generally use sacrificial materials that are very thin relative to the SU- 8 structures and are therefore vulnerable to stiction and other surface interactions. For our purposes it was desired to have an SU-8 structure that is separated from the substrate by several microns to several 10 s of microns. The positive resist Shipley 1827 has been found to create layers up to 6 microns thick, while SU-8 can theoretically produce layers up to hundreds of microns thick [10]. 3 SU-8 Photopolymer SU-8 was originally developed by IBM [11], but is now mainly available through Microchem [10]. It is transparent in the visible light range, but will absorb UV light. When it is exposed to UV light with wavelengths of less than 400 nm it polymerizes and resists development by organic solvents. It is stored as a liquid, but after it is spun on a wafer, it is heated (baked) at a high enough temperature to drive off solvents and solidify it on the wafer. The next step is to expose it to UV light in the desired pattern. This step starts the polymerization process (generally referred to as cross-linking), but to speed it up, the SU-8 is baked again. After this second bake step, a developer is used to remove all the unexposed SU-8, leaving behind the cross-linked material. Sometimes, yet another baking step is used to enhance the cross-linking after this point, but it is not necessary. Once completely cross-linked, SU-8 is very thermally and chemically stable. Several formulations of SU-8 are available, each with a different viscosity for producing layers with large thickness differences. The type of SU-8 that we used for all our preliminary experiments was SU This variation of SU-8 can easily produce structures ranging from 10 to 30 microns depending on spin speed [10]. We chose this particular variation because it was already available at SFU (courtesy of Dr. Bonnie Gray) and it would produce the thinnest layer for a given spin speed of all the SU-8 variants available to us. A thin layer is desirable for proof of concept processing because the film stresses tend to be less and the exposure dose, baking time and development times are all smaller than for thicker films. Because the lateral resolution of our photomasks was ~30 microns, almost all of our structures would then have aspect ratios (height to width) between 1:3 and 1:1. For the majority of our tests, and the spin speed parameters that we chose, this aspect ratio was approximately 1:2. As a result of this, many devices that were designed to move in plane were instead much more compliant in the out of plane direction. Many authors have attempted to characterize the optimal process steps for SU- 8 development, but the exact characteristics depend highly on individual lab conditions in addition to the substrate materials and layer thickness. This optimization for our own equipment was done as well as possible, but the lack of automatic equipment and accurate feedback for several process steps made this difficult. 2

9 4 Photoresist Sacrificial Layer Process Description One method of making an overhanging SU-8 10 microstructure is to use a thick photoresist as a sacrificial layer. The process starts with application of thick photoresist Shipley SC1827 on a RCA cleaned wafer (Figure 1). The photoresist is spun for 30 seconds at 900rpm and produces a layer that is approximately 6 microns thick. Thick photoresist Silicon Figure 1: Thick photoresist applied on a wafer After the photoresist is applied, the wafer is pre-baked at high temperature to drive out the solvents and solidify it. Then the wafer is exposed to UV light with an intensity of 10 mw/cm 2. This UV exposure causes a chemical change within the photoresist that makes it dissolve in a solvent (Figure 2). Thick photoresist Silicon Figure 2: Exposure of photoresist layer The photoresist is developed using MF-322 with visual inspection to determine completion (Figure 3). The development usually takes approximately 3 minutes. Thick photoresist Silicon Figure 3: Developed photoresist sacrificial layer After development, the photoresist is hard-baked at high temperature to increase the resistance to solvents. This step is very important because if it is under baked, it will not survive other process steps, but if it is over baked, it becomes very hard to remove with any solvent. After the hard bake, SU-8 10 is spun on the wafer ( Figure 4). The speed of the spin coater is ramped from 0rpm to 500rpm in 5 seconds, from 500rpm to 2000rpm in another 5 seconds, and the speed is kept at 2000rpm for 30 seconds. 3

10 SU-8 10 Thick photoresist Silicon Figure 4: SU-8 10 is spun on the wafer The wafer is then soft-baked at 65 C on a contact hot plate. This step removes the solvents from the SU-8 and ensures good resolution during the SU-8 exposure step. Most times, the temperature of the oven is slowly ramped up to the desired value and after prebake, the oven is slowly ramped down. The wafer is exposed for 40 seconds to define the SU-8 structural features ( Figure 5). SU-8 10 Thick photoresist Silicon Figure 5: SU-8 layer is exposed After exposure the wafer is post-exposure-baked (PEB) to speed up the cross-linking process. Afterwards, the SU-8 and photoresist are developed with SU-8 developer leaving only the SU-8 structure behind (Figure 6). Because of the long hard bake of the photoresist, removal of the photoresist requires approximately 20 minutes in an ultrasound bath. SU-8 10 Silicon Figure 6: Final SU-8 overhanging structure 5 Embedded Aluminum Mask Process Description Two layers of SU-8 and an embedded mask is created such that both top and bottom layers of SU-8 can be cross-linked with a single exposure. Therefore, the bottom layer of SU-8 acts as an anchor as well as a sacrificial layer. The process begins with a coating of a layer of SU-8 10 on a silicon wafer (Figure 7). The speed of the spin coater is ramped from 0rpm to 500rpm in 5 seconds, from 500rpm to 2000rpm in another 5 seconds, and then maintained at 2000rpm for 30 seconds. 4

11 SU-8 10 Silicon Figure 7: SU-8 10 applied to a silicon wafer After SU-8 is applied, the wafer is soft-baked for one day at 65 C. As mentioned in the previous section, the baking temperature for SU-8 is slowly ramped up and down. Then, aluminum is sputtered on the wafer using the Corona sputterer in the clean room (Figure 8). Aluminum SU-8 10 Silicon Figure 8: Aluminum is sputtered over SU-8 sacrificial layer Shipley SPR2 is then applied over the aluminum for aluminum patterning (Figure 9). The resist is spun for 30 seconds at 4000rpm. Photoresist Aluminum SU-8 10 Silicon Figure 9: Photoresist is applied over aluminum Instead of soft-baking the photoresist, the wafer is left at a room temperature (approximately 17 C) for 20 minutes. Then the photoresist is exposed for 20 seconds for patterning aluminum and left at a room temperature again for 30 minutes (Figure 10). Photoresist Aluminum SU-8 10 Silicon Figure 10: Exposure of photoresist The photoresist is developed with MF-319 for one minute leaving the unexposed areas behind (Figure 11). 5

12 Photoresist Aluminum SU-8 10 Silicon Figure 11: Developed photoresist layer After leaving the wafer at a room temperature for 10 minutes, aluminum etching was performed with aluminum etchant type A solution (Figure 12). Photoresist Aluminum SU-8 10 Silicon Figure 12: Etched aluminum mask layer SU-8 10 is applied again with the same method used for the first layer of SU-8 (Figure 13). SU-8 10 Photoresist Aluminum SU-8 10 Silicon Figure 13: Second applied layer of SU-8 SU-8 is soft-baked at 65 C for 10 minutes with a slow ramping of the temperature. After the soft bake the wafer is exposed for 30 seconds to initiate the SU-8 polymerization for creating the desired structures (Figure 14). SU-8 10 Photoresist Aluminum SU-8 10 Silicon Figure 14: Exposure of SU-8 6

13 PEB was performed at 65 C for approximately 15 minutes with a slow ramping of the temperature. The development of the structure is performed in an ultrasonic bath with SU-8 developer leaving the structures shown in Figure 15.. SU-8 10 Silicon Figure 15: Final SU-8 structure 6 Preliminary Testing and Results 6.1 Photoresist Spinner Characterization The behaviour of the spin coater was characterized to for given inputs. Figure 16 shows the spin coater that was used. Figure 16: Spin coater We required the actual spin duration of the spin coater with given settings. We have gathered the following data. 7

14 Table 1: Comparison of set and measured operation time of the spin coater Time Setting Measured operation time (sec) The measurements were taken with a rotational speed of 3380rpm, and the time setting in Table 1 is within accuracy of ±2.5sec. When we are spinning a thick photoresist or SU-8, it is important to follow the suggested time given in the guideline. By following Table 1, we could determine what input time is required to give us desired spinning time. A dial is used to set the rotational speed of the spinner. However, the value written on the dial did not match to the measured rotational value shown on a display. We performed a similar characterization as above to determine the required dial settings for a given rotational speed of the spinner. The speed characterization of the spinner is shown in Table 2. 8

15 Table 2: Comparison between the dial reading of the spin coater and measured rotational speed Dial reading Measured rotational speed (rpm) The 120-mark on the dial was effectively 0rpm with little rotational motion. Any dial setting below 120 did not produce any spinner movement. 6.2 Hot Plate Characterization Two small hot plate ovens in the clean room are used for baking SU-8. Figure 17 shows the two hot plate ovens. 9

16 Figure 17: Hot plate ovens However, the ovens had poor temperature control and measurement lag. The dials for setting the oven temperature did not have proper marks for related temperatures. Even though the ovens were of the same model, they behaved differently. Fortunately, we were only interested in two temperature levels for baking SU-8: 65 C and 95 C. With careful adjustment of the temperature control dial and temperature measurements we were able to determine dial positions for both temperature levels on both ovens. We determined that 65 C was equivalent to about 250 on the oven dial and 95 C was equivalent to about 340 on the dial. After spun on a wafer, the SU-8 layer was still soft and could flow slowly to one side of the wafer while baking if not level. Therefore, it was important to level the hot plates. We used a level and paper supports inserted beneath the ovens to level the hot plates. 6.3 Photoresist and SU-8 Pouring Method Several trials of pouring and spinning of thick photoresist and SU-8 were performed. We found that when either photoresist or SU-8 was poured on a wafer, many small bubbles were formed. When the wafers were spun with bubbles, the quality of the film after spinning was typically poor, with streaks and holes in the film. Bubble formation was most likely caused by liquid hitting the wafer from a large height. The pouring rate was hard to control because we were pouring the liquids directly from their containers. To solve the problem, we poured the liquids into a measuring spoon from their container which gave us repeatable volumes. It was also far easier to pour the liquid onto the wafers from a small height with these spoons. Most of the small bubbles were eliminated in this way. Bubbles that formed when the liquids were poured into the measuring spoon were easily removed by popping them with a piece of aluminum foil. 10

17 6.4 Coating of SU-8 10 A guideline for achieving a SU-8 layer with thickness of 15µm is shown in Table 3. The corresponding value for speed adjustment dial is found using Table 2 and is shown in the third column of Table 3. Table 3: Guideline for creating a SU-8 layer with 15µm in thickness Order Actual values to be used Input for the spin coater rpm in 5sec in 5sec rpm in 5sec in 5sec rpm for 30sec 300 for 30sec 6.5 Edge Bead Trials When SU-8 is spun on a wafer, an edge bead was visible. We tried several methods to remove the edge bead. The first method was to spin the wafer very slowly for 20 seconds to spread the SU-8 out. However, using this method created several uncovered areas around the edge. The second method attempted to remove the bead with acetone directly squirted from a bottle while spinning the wafer. However, this attempt resulted in an uneven removal of the edge bead. Another method attempted was using an acetone soaked wipe to dissolve the edge bead. This method worked, but it was difficult to perform reliably. The best solution was to spray SU-8 developer onto the edge from a glass syringe while spinning the wafer. The amount of applied SU-8 developer can be carefully controlled and the developer is not splashed when shot on the edge of the wafer. 6.6 SU-8 Exposure Tests and Profilometer Measurements To determine the optimum exposure time for a 15µm SU-8 layer with a10 mw/cm 2 exposure dose, we exposed a wafer with several different exposure times on separate areas as shown in Figure

18 8 min 6 min 4 min 2 min Figure 18: Exposure scheme to find optimal exposure time The post exposure bake was completed by placing the wafer on a hotplate held at 50 C. This temperature was ramped up slowly for 5 minutes and the wafer was removed after a temperature of 60 C was reached. The latter portion of the post exposure bake was performed for 3 minutes beginning at 90 C where the temperature of the oven was 105 C when the wafer was removed. After development, it was discovered that most of the wafer was covered by cross-linked SU-8 in different thicknesses. Only areas exposed for 2 minutes had their unexposed parts developed. Figure 19 to Figure 22 and Table 1 summarize the profilometer measurements of the overexposed SU-8 in exposure experiments. 12

19 Figure 19: Profilometer measurement of 2 minute exposed area (underdeveloped) Figure 20: Profilometer measurement of 4 minute exposed area (underdeveloped) Figure 21: Profilometer measurement of 6 minute exposed area (underdeveloped) Figure 22: Profilometer measurement of 8 minute exposed area (underdeveloped) Table 4: Height variation between exposed and unexposed areas Exposure Time Height variation between exposed and unexposed areas 2 minutes 6.179µm 4 minutes 6.530µm 6 minutes 5.796µm 8 minutes 6.428µm Figure 23 shows the profilometer measurement of SU-8 exposed for 2 minutes that revealed the surface of the silicon wafer. The height of the SU-8 layer was found to be 18.22µm. 13

20 Figure 23: profilometer measurement for a properly developed area A new wafer for exposure trials was started and exposed for the amounts shown in Figure sec 30 sec 20 sec 10 sec Figure 24: Revised exposure scheme for SU-8 development After baking and development, the wafer showed very good definition for almost every section. Very little or no SU-8 was left on unexposed areas. Microscope inspection revealed that the area that was exposed for 40 seconds did not have as good definition as other sections. The areas that were exposed for 20 and 30 seconds had very good definitions and features. Cracks were spotted on SU-8 surfaces of the area exposed for 10 seconds. 14

21 Figure 25 to Figure 28 are profilometer measurements of SU-8 features with different exposure times. Figure 25: Profilometer measurement of 10 seconds exposed area (underdeveloped) Figure 26: Profilometer measurement of 20 seconds exposed area (underdeveloped) Figure 27: Profilometer measurement of 30 seconds exposed area (underdeveloped) Figure 28: Profilometer measurement of 40 seconds exposed area (underdeveloped) The thickness measurement of the SU-8 features with different exposure times are summarized in Table 5. Table 5: Thickness of SU-8 features for given exposure times Exposure Time Thickness of SU-8 features 10 seconds 17.05µm 20 seconds 17.00µm 30 seconds 16.73µm 40 seconds 17.06µm Exposure for 30 seconds seemed to be enough without losing any definition. For the process, we decided to use either 30 or 40 seconds exposure for SU-8 layers. 15

22 6.7 Preliminary Tests for Aluminum as a Imbedded Mask The preliminary testing of the aluminum sputtering and etching was performed on a glass slide to reduce the consumption of wafers. Figure 29 illustrates a glass slide that was sputtered with aluminum, etched and developed using SU-9 developer. As shown in Figure 30, the SU-8 is not cross-linked and is removed easily after an aluminum etch. Therefore, the preliminary tests show that patterning a sputtered aluminum layer over uncured SU-8 is possible. Aluminum etch and SU-8 Development Aluminum etch only Figure 29: Aluminum Etchant Test Figure 30: Close-Up View of Etched Area In addition to tests using a glass slide, tests are also performed on wafers to quantify the quality of adhesion between the SU-8 and aluminum. Scotch tape is applied to the surface of the aluminum and is torn away. If the aluminum does not peel off, then good adhesion is achieved. However, when this test was performed on the SU-8 sputtered wafers, the aluminum always stripped off and remained stuck on the tape. Figure 31 shows an example of the aluminum removed by a strip of Scotch tape. Although the aluminum can be stripped off, using aluminum as an embedded mask is still possible as long as the SU-8 does not cross-link due to UV exposure in the sputterer. A development test using the same wafer is also completed. The results of the development test are shown in Figure 32. After the development test, aluminum covering part of the developed area was removed to see if any SU-8 was developed. It appeared that only top layer of SU-8 was partially developed. We suspect that the SU-8 did not get developed well because the wafer was left undeveloped for too long. 16

23 Figure 31: SU-8 Stripped of Aluminum using Tape Test Figure 32: SU-8 Development of Sputtered Wafer As shown by the previous figure, the SU-8 area that was cleared using the tape is easily removed using developer. This removal was completed in fewer than 40 seconds of agitation verifying that SU-8 remains uncross-linked after an aluminum sputter process. 7 Mask Design Figure 33: Anchor Mask Cell Figure 34: Surface Feature Mask Cell 7.1 Feature Description The features on each cell of the mask design were specifically placed there to test different aspects of the process capabilities. The lower right hand corner of both mask layers includes designs that test the resolution of the process and masks. Arrays of small squares with dimensions ranging between 30 and 100 microns are used. These features are produced in the anchor layer as holes, and in the SU-8 layer as posts. The anchor 17

24 layer is done in photoresist directly, or in photoresist then transferred to aluminum for the sputtered wafers. The goal was to compare the features in photoresist before and after the application of the SU-8 layer over top. Because SU-8 contains solvents, it was not known whether the photoresist features would be distorted during the SU-8 spin on step. Two large comb drive type structures are included on the lower left hand side of the structural mask. They have small anchors so that the actual comb fingers are suspended above the substrate. The goal of these was to see if the development process would damage large compliant structures and if there was a very good process developed, a voltage could be applied across them to determine if any actuation would occur. The right hand side of the wafer is composed of fixed-free and fixed-fixed cantilevers of varying lengths and widths. This test would determine the yield of various designs and determine whether stiction would be a significant issue. These cantilevers could also be used to determine mechanical strength of the SU-8 if the displacements and forces could be measured. The three rectangles in the center suspended by serpentine springs are pop up mirror designs designed by See-Ho Tsang. The original design was to be produced in an SOI process, but a rapid prototype was completed on this chip design to determine feasibility. The scale of each of these mirrors is different to determine the effect of minimum feature size on device reliability. The large square surrounded by posts in the upper left is to determine how large arrays of posts survive, and what type of defects can occur in large films of SU-8 with sharp corners. The last test structures to note are located in the center of the cell. They are designed to be channels of SU-8 open at either end. One is completely enclosed in the middle, while the other has etch holes to determine if the removal of the sacrificial layers can be sped up significantly. 7.2 Mask fabrication The layout masks were created using Cadence and exported as CIF files in order to maintain a one-to-one magnification factor. These files are converted into postscript using an exporter program written by the IMMR. The postscript converter is located on the Unix network at: ~param/bin/psmask. Using the psmask generated postscript files, the masks are printed at Alliance Printing using a 2500 dpi printer on film transparency which translates to 25µm resolution. Initially, we intended to transfer the transparency masks to glass emulsion masks for exposure. However, due to the cost of the materials required for the transfer, the transparencies were used directly instead. Figure 35 and Figure 36 show the respective light and dark field masks that were printed. 18

25 Figure 35: Anchor Dark Field Mask Figure 36: Structural Layer Dark Field Mask 7.3 Alignment and Exposure For a two layer process, alignment markers are needed to ensure that the anchor layer and structural layer are aligned during exposure. Identical alignment markers separated 6cm apart at the centerline of the wafer are used along with the split-field view on the aligner to match rotational, vertical and horizontal translations. Figure 37 and Figure 38 are the respective markers on the structural and anchor masks. Figure 39 illustrates the overlay of the two mask layers once alignment is reached. Figure 37: Alignment Markers on Structural Mask Figure 38: Anchor Mask Alignment Markers Figure 39: Overlay of Alignment Markers Since a direct exposure of the flexible transparency masks is required, a method for loading the masks into the aligner was developed. The following list illustrates the basic steps for loading a transparency mask into the aligner. 1. Ensure that the power, mask-load indicator light and suction is turned on. 2. Use visual-align to raise the aligner arm. 3. Place transparency mask onto mask suction gasket. 4. Place a clean a glass overplate over the transparency mask and hold down with pressure. 5. Turn off the mask-load. 19

26 6. Place wafer into wafer chuck. 7. Hold down the glass overplate and load wafer (holding down the overplate is required to keep wafer from pushing through the mask vacuum seal). 8. Lower the aligner arm and align wafer. 9. Raise the aligner arm using visual-align. 10. Hold down the glass overplate and put wafer into contact. 11. Lower the aligner arm and expose. Figure 40: Transparency mask held by glass overplate After examining the transparency mask under photomicrograph, several defects are seen. These defects include small dark spots located throughout the film as well as the inability of the printing process to perfectly reproduce the 30µm features from the post script file. Figure 41 and Figure 42, illustrates the poor definition and blurring of the mask test structures. Figure 43 shows an example of the spot defects that are located throughout the light field areas of the mask. Figure 44 shows blurred features which reduce the structure resolution. Figure 41: Example of Poor Mask Feature Definition of Test Structures (40µm on Left, 30µm on Right) Figure 42: Example of Poor Mask Feature Definition of Test Structures (100 µm Size) 20

27 Figure 43: Example of Spot Defects Across Light Field of Mask Figure 44: Example of Blurred of Features On Comb Drive Due to these defects in the mask, we expect that the photolithographic process will reproduce these problems in the photoresist, aluminum, and as well as SU-8 structures. Therefore, defects that arise due to these problems can be accounted for. 8 Results of the Photoresist Sacrificial Layer Process We were able to produce overhanging structures made of SU-8 using the photoresist sacrificial layer, but several issues need to be resolved. They involve optimizing the hard bake time for the photoresist for best feature size and development rate, proper post exposure bake times and temperatures for the SU-8 to minimize film stress and adjusting the method of development to minimize damage to compliant structures. Four wafers were used with different process parameters. The process steps for each of these wafers are listed below in table 6. Table 6: Thickness of SU-8 features for given exposure times Wafer # Soft-bake 100C, 2 hrs 100C 1 hr 100C 1hr 100C 1hr Exposure time 60 sec 80 sec 80 sec 80 sec Development 5 min 3 min 3 min 3 min Hard-bake 3 days 120C, 2hrs 100C, 1 hr 100C, 1 hr SU-8 soft-bake SU-8 exposure time SU-8 PEB 65C, 20 min, no ramping 65C 10 min, ramp up and down from room temp. 65C, 20 min with ramp up and down 40 sec 40 sec Not done 40 sec 65C, 80 min 95C, 6 min ramp down to 65C 10 min Ramp up and down from Not done 65C, 20 min with ramp up and down 65C, 10 min Ramp up and down from 21

28 SU-8 development Other notes 50C then removed Ultrasonic bath > 20 min Photoresist remains in large areas after SU- 8 development room temp Ultrasonic bath 3 min Photoresist remains in large areas after SU- 8 development Manual agitation 3 min Photoresist exposed to UV for different amounts after hard-bake. No change in development speed noticed room temp Manual agitation 4 min then 11 min without agitation Photoresist easily removed, but manual agitation is highly damaging to SU-8 structures 8.1 Observations and Analysis of Photoresist Spinning After much practice, we were able to get consistent covering of the wafers with Shipley 1827 positive photoresist. Examples for the photoresist spins are shown in Figure 45 through Figure 48. Figure 45: Shipley 1827 photoresist immediately after being spun onto a silicon wafer Figure 46: Shipley 1827 photoresist showing tweezer marks caused by wafer handling 22

29 Figure 47: Shipley 1827 coated wafer showing fringes in coating. Each fringe is ~1/4 wavelength of light, and thick fringes indicate small thickness variations over the wafer area. Figure 48: Shipley 1827 coated wafer reflecting objects. Light is reflected by the silicon, but is barely distorted by the photoresist layer, indicating uniform surface coating Photoresist Patterning and SU-8 Spin Coat Photoresist patterning was done with the anchor mask. All areas where the photoresist was exposed to UV light were developed using MF-322 developer. There was a strong correlation between exposure dose and development time. The larger the dose, the less development time was required for complete photoresist removal. After development, the wafers were rinsed in DI water and examined to determine feature quality before SU was spun on. Figure 49 through to Figure 52 show wafer #20 with SU-8 spun onto the patterned photoresist. Figure 49: Wafer #20 immediately after SU-8 layer spin on. Figure 50: Close up of wafer #20 after SU-8 spin on 23

30 Figure 51: Wafer #20 after soft-bake of SU-8 Figure 52: Close-up of wafer #20 after soft-bake showing some surface distortion. This is caused by the SU-8 surface being affected by the topology changes between the photoresist and holes. 8.2 SU-8 and Photoresist Development It was known from previous experiments that SU-8 developer would remove unexposed photoresist. Because of this feature, it was decided that the final processing step would use SU-8 developer to remove both the unexposed SU-8 and remaining photoresist. The major problem with this method is that if the photoresist is hard-baked too much or too little, major problems will occur with respect to feature resolution or photoresist removal. Figure 53 through to Figure 56show several wafers during and after the development process. Figure 53: Wafer #20 in SU-8 developer with ultrasonic agitation Figure 54: Wafer #18 after development and rinse. Purple areas are undeveloped photoresist 24

31 Figure 55: Wafer #20 showing relatively clear area with most photoresist developed. This section of the wafer was resting on the bottom of the jar in the ultrasonic bath Figure 56: Wafer #2 after SU-8 development showing good photoresist removal 8.3 Characterization of Photoresist Features The photoresist features were generally poor, being limited mainly by the quality of our emulsion masks. Figure 57 through to Figure 60 show some of the smallest features produced in photoresist prior to being covered with SU-8. The wafers that had the photoresist hard-baked at high temperatures for long periods of time maintained feature size best and showed no noticeable distortion after the SU-8 spin on. The wafers on which photoresist was hard-baked only for short periods of time at low temperatures or not at all showed much poorer maintenance of feature resolution. In the case where the photoresist was not baked at all in between development and SU-8 application, the photoresist features became so poor that the wafers were stripped because they would have been impossible to align. All features without hard-baking could be seen to smear in the direction of SU-8 spinning (towards the edge of the wafer) so it was very obvious the SU-8 solvent had a significant effect. In addition to this, the SU-8 surface also became much more uneven after soft baking these wafers, adding to alignment difficulties. The hard bake for the positive photoresist therefore requires greater than 1 hour at 100C. 25

32 Figure 57: Smallest L-shaped feature in photoresist Figure 58: Minimum spacing and feature size in photoresist Figure 59: 40 and 50 micron holes in photoresist Figure 60: 50 and 100 micron hole in photoresist Another feature of interest was the reflow of photoresist during the hard bake. During later trials, profilometer measurements were done on a wafer before and after being hard baked. Prior to being hard baked, the thickness of all sections was nearly uniform before sloping down to the substrate in the developed sections. After hard baking, the edges of all photoresist windows exhibit a small lip that rises approximately half a micron above the normal photoresist thickness. Profilometer measurements are shown for the same features before and after hard-bake in Figure 61 and Figure 62. This seems to be due to surface tension effects during the reflow of material and could cause small indentations in the underside of the SU-8 structural layer during later steps. 26

33 Figure 61: Profilometer reading of medium L-shape feature in photoresist prior to being hard baked Figure 62: Profilometer reading of medium L-shape feature in photoresist after being hard baked 8.4 Test Structure Results In general, the test structure feature resolution was very poor, but this is mainly due to the poor mask quality. Most features that were directly anchored to the substrate but not overhanging (such as posts and anchors) had high survivability and yield. Most cantilevers survived, but many had photoresist still remaining underneath, and as a result were not completely released. The structures that fared poorest were the comb-drives and the smallest of the three pop up mirror designs. In all these cases the anchor area and beam widths were too small for the size of the device. The structures would typically either separate completely off the substrate (from poor anchor adhesion) or have their long thin beams break during the development process. Excessive agitation was the main problem when developing by hand, or too much airflow if an N 2 dry was attempted. On most wafers, the largest mirror designs never completely released, nor did the channel designs completely empty out. The longer the hard bake time, the longer it took to remove the photoresist, but for the early trials with the long hard bake times, certain amounts of the photoresist could not be removed at all. It was observed that the visibility through the SU-8 structures is affected by whether they are anchored to the substrate or suspended. Most cantilevered structures appear gray rather than transparent if they were over top of photoresist and this can be seen in Figure 63. It is unknown whether this is caused by some photoresist remaining on the underside of the SU-8 cantilevers or some chemical reaction. The most likely explanation for this phenomenon is that the interface between photoresist and SU-8 is not smooth due to diffusion or different thermal expansion, and that the rough undersides of the resulting SU-8 structures scatter light and appear darker as a result. 27

34 Figure 63: SU-8 cantilevers suspended over silicon substrate. The clear areas close to the bottom are where the SU-8 is anchored to the substrate. The dark perpendicular line is photoresist remainder Figure 64: Large L-shaped photoresist holes under cross-linked SU-8 layer showing major cracking problems Figure 65: Small pop up mirror design connected to hard baked photoresist after development step. The photoresist could not be removed in developer and shows cracking, possibly induced by ultrasonic agitation Figure 66: Corner of large anchored SU-8 structure and posts, showing edge resolution and residue from development process 8.5 Stress Observations and Survivability For the earliest wafer attempted with this process, no ramping was used during soft-bake or post-exposure-bake. The result was a highly stressed film that caused cantilever structures to bend and curl dramatically. The introduction of a slow ramp up and down from room temperature during SU-8 baking steps reduced the incident of this failure mode. The lack of a hot plate with controlled ramp up and down limits the repeatability of each of these trials, but a reasonable means of control is to just place the wafer on the hot plate when it is at room temperature and turn it on for the ramp up. Ramping down is accomplished by turning off the hot plate and opening the door to dissipate heat. It has been observed by other authors that increasing the baking temperature during each baking step can result in better cross-linking of the SU-8 and producing a stronger material. This 28

35 higher temperature baking was not done during these trials because of the much longer baking time it would introduce to each process. Given that many of our structures were too weak to survive the development process, it may be worthwhile to introduce these higher baking temperatures to improve structural strength and yield. Care must be taken with all higher baking temperatures to prevent the creation of large film stress in the SU- 8. The result of this high film stress can be seen in Figure 67 to Figure 72. Figure 67: Large comb-drive on wafer #20 with large film stress. Cracking in remaining photoresist can clearly be seen. Figure 68: Large comb-drive fingers curling in towards each other Figure 69: Large comb-drive on wafer #20 showing partial adhesion to substrate Figure 70: Reflection of large comb-drive on wafer #20 29

36 Figure 71: Photomicrograph of comb-drive with internal stress causing curling down. Figure 72: Photomicrograph showing the radius of a curled comb-drvie. This radius is directly related to the internal stress distribution. 8.6 Failure Modes The possible failure modes while using the photoresist as a sacrificial material include: Poor feature resolution due to mask problems Poor alignment of structures Photoresist dissolving on contact with liquid SU-8 (under baked) Photoresist remaining during SU-8 development (over baked) SU-8 surface rippled and bubbles within film after soft bake (happens when photoresist is under baked) Large film stress in SU-8 (caused by thermal shock) Cracks/fractures in SU-8 (non-optimal exposure and bake times) Damage to large structures from development The feature resolution and alignment issues are directly related to the masks and UV source that we used and are not specific to the photoresist process. The under baking/over baking of the photoresist should be corrected by optimal process parameters and will solve many issues related to alignment and development. The large film stress in SU-8 has already been improved by ramping the hot plate up and down rather than using a step input. This can be corrected further by using an automatic ramping hotplate. The cracks and fractures in SU-8 should be avoided once proper exposure and baking times are determined, but they may also be caused by a problem during the development process. No cracks are visible before development, but large SU-8 areas that are fixed to the substrate typically are full of cracks after being developed and rinsed. The last failure mode comes from the fact that many of our large structures (mirrors and comb-drives) are too large for the strength of their anchors and thin beams and are severely damaged during the development and rinse process. Better designs and process parameters will help, but allowing development to occur without agitation may be the easiest way to avoid this problem. Despite all these problems with some devices, one notable success was the successful manufacture of a pop-up mirror. This mirror survived on wafer #20 30

37 and was strong enough to be deflected into position. Figure 73 and Figure 74 show this structure after being assembled. Figure 73: Large pop-up mirror held in place by friction. Mirror is placed in position by microprobes Figure 74: Wide angle view of large pop-up mirror showing the size relative to other objects 9 Results of the Embedded Aluminum Mask Process The aluminum process as described in section 5 was performed on four separate wafers where each wafer was treated slightly differently to experiment with process parameters and to find optimization points. Table 7 lists the wafer number and corresponding preprocess parameters prior to applying SU-8 and sputtering. Table 7: Wafer Preprocess Description for Embedded Aluminum Mask Process Wafer Number Cleaning Process Description Preprocessed ENSC 495 wafers from This wafer was RCA 28 cleaned omitting the HF dip. We felt that this wafer would provide a basis for testing the aluminum adhesion and feasibility for sputtering onto an SU-8 surface. 19 New wafer. RCA cleaned, complete process. 11 New wafer. RCA cleaned, complete process. 25 New wafer. RCA cleaned, complete process. 9.1 Application of SU-8 to Silicon Substrate The application of SU-8 to the silicon wafer was performed as in section 6.4. Once complete, the wafers are soft baked at different temperatures and lengths of time to optimize the baking such that the sputtered aluminum will adhere. Table 8 lists the times and temperatures for which the wafers are baked. 31

38 Table 8: SU-8 Soft Bake Times for Embedded Aluminum Mask Process Wafer Number Bake Time and Method 28 Baked without ramping for 5 days at 65 C 19 Baked without ramping overnight at 65 C 11 Baked with ramp up of 12min and held at 65 C for 10min and ramped down to room temperature. Same as previous with the addition of a layer of SPR2 prior to 25 sputtering. This layer of photoresist is added to aid in blocking UV and to test whether or not there is a difference in UV absorption. The different bake times and temperatures are examined to find an optimal point where the SU-8 becomes a solid substrate to allow a smooth film of aluminum to adhere. Because of the concern over the thermal stress of the SU-8 film during the sputtering process, over baking of the uncross-linked SU-8 must be avoided. In addition to thermal stress, SU-8 may cross-link during the sputtering process due to any stray UV emitted from the argon plasma. Therefore, testing the feasibility of sputtering to create an aluminum film on SU-8 without cross-linkage was examined. 9.2 Observations and Analysis of Sputtering Process All wafers are sputtered at the lowest power setting in order to reduce the chance of UV emission of the argon plasma. The log of the power settings is attached in the Appendix. The following figures illustrate the result of sputtering the wafers described in the previous section. Figure 75: Wafer #28 post sputtering process Figure 76: Wafer #19 post sputtering process 32

39 Figure 77: Wafer #25 post sputtering process Figure 78: Wafer #11 post sputtering process As shown by the figures above, only wafers #19 and #25 produced smooth and reflective surfaces that can be used for an embedded mask. The main consideration for creating a successful sputter relies on the sturdiness of the uncross-linked SU-8 since any reflow of the material while it is suspended on the planetary will cause the surface to roughen. In addition to the hardness of the SU-8, the other conclusion that was reached is that the SU- 8 must be spun on a clean wafer that is free of oxide. The adhesion of the SU-8 to silicon is far better than that to oxide. Because of this difference in adhesion, the wafer that was not cleaned using a HF dip produced a film that is highly uneven and reticulated. Figure 79 shows a close-up view of this problem. Rough aluminum surface on oxide Smooth aluminum surface on silicon Figure 79: Higher magnification view of wafer #28 As shown above, sections of aluminum that was sputtered on sections of bare silicon appear smoother whereas areas that were sputtered on oxide are rough and uneven. The main issue which caused the rough surface of wafer #11 is mainly due to the thin photoresist layer that exists between the SU-8 and aluminum. Because the photoresist was not hard-baked, due to the concern of thermally stressing the SU-8, the photoresist hardness may not have sufficed to produce a smooth surface for the aluminum sputtering 33

40 process. In the future, hard-baking the entire wafer prior to sputtering may avoid this problem. 9.3 Aluminum Patterning Process In order to pattern the aluminum layer on the SU-8, a coat of SPR2 photoresist was spun onto the aluminum and exposed following the guidelines in section 5. After exposure, both wafers #19 and #25 were baked at room temperature for 30mins. The results of the patterning can be seen in Figure 80 and Figure 81. Initially, we encountered problems attempting to expose the wafer using a flexible mask since the suction between the mask and the loaded wafer did not suffice to stop the wafer chuck from ramming into the mask. Therefore, we used a cut out of the mask taped to the bottom of a glass overplate. This difference in mask alignment can be seen on wafer #19 as shown in Figure 80 and Figure 82 where the mask did not reach the edges of the wafer. However, as our experience with the aligner increased, we were able to develop the method described in section 7.3, which solved the wafer chuck problem. The full sized masks are shown on wafer #25 in Figure 81 and Figure 84. Figure 80: Wafer #19 patterned with thin photoresist Figure 81: Wafer #25 patterned with thin photoresist After development, the features of the photoresist structures appear slightly over developed on both wafers. However, these effects are most likely attributed to the mask defects that have been transferred onto the photoresist. Wafer #19 was not hard-baked due to time constraints and was etched within 10 minutes of inspection. The aluminum etch that we used is an electronics grade Type A activated at 40 C. Initially, the etchant was warmed using a hot plate and wafer #19 was added directly into the etch at 50 C and was completed in less than 30secs. However, the sudden increase in temperature caused the SU-8 under the aluminum to buckle due to thermal shock. Figure 82 shows the thermally shocked wafer and Figure 83 shows the rippled surface due to the high surface tension. Because thermal shock was occurring, we modified the etching step by placing the wafer into the solution at room temperature and 34

41 ramping up to 40 C. Wafer #11 was etched this way taking 7 mins to ramp up to 35 C where etching completed within 5 mins. A ramp down was performed for 35mins to bring the solution back down to room temperature. By ramping the temperature, a smooth un-stressed film was achieved and is shown in Figure 84 with a close up in Figure 85. Figure 82: Wafer #19 after aluminum etch without temperature ramp up Figure 83: Close-up of high surface tension Figure 84: Wafer #25 after aluminum etch with temperature ramp up and cool down Figure 85: Close-up of smooth surface Although the temperature ramping allowed the aluminum etch to complete without thermally shocking the wafer, a longer contact time between the etchant and SU-8 was needed. This increase in time caused problems with cross-linking the SU-8 and is discussed further in section 9.6. Once the aluminum etch is complete, the thin photoresist mask is quickly removed by spraying the wafer lightly with a stream of acetone. The contact time of the acetone must be kept to a minimum to ensure that the SU-8 under the aluminum is not washed away. 35

42 9.4 SU-8 Post Exposure Bake and Development Once the aluminum mask is patterned and the SU-8 structural layers are spun on and baked at room temperature for 30mins, they are exposed. Wafer #19 was post exposure baked by a ramp-and-hold at 65 C with a ramp time of 12mins and hold time of 10mins. The ramp down required 20mins to return to room temperature before the wafer was removed. Similarly, wafer #25 was also baked at 65 C but the temperature knob was set all the way to 600 C during the ramp up and the temperature overshoot thermally shocked the wafer. Once the baking was complete, the development of both wafers was completed. In order to speed up the development time, wafer #19 was developed with SU-8 developer using with the ultrasonic bath for over 30mins. However, the aluminum film remained which led to us to speculate that the SU-8 has somehow cross-linked during the processing. Wafer #25 was developed by hand for 10mins where the top layer of SU-8 was easily removed but SU-8 under the aluminum was also cross-linked. Therefore, an investigation on the reasons for the cross-linking was completed and is discussed in section Characterization of Aluminum Features During the embedded aluminum mask process, the test features are examined at process intervals to inspect for defects and inconsistencies. The times at which the features are examined are: after aluminum patterning, after SU-8 structural definition, and after SU-8 development SU-8 Anchor Layer after Aluminum Patterning The following figures illustrate the anchor layer features of a thermally shocked aluminum wafer under photomicrograph. As shown, the SU-8 under the aluminum layer is highly reticulated and the aluminum around these features is pulled into ridges due to the high surface stresses. 36

43 Figure 86: square test structures under aluminum (30µm on left, 40µm on right) Figure 87: Square test structures under aluminum (50µm on left, 100µm on right) Figure 88: L-shape test structures (50µ on left, 30µm on right) Figure 89: Anchor of a small pop-up structure In addition to the high stress film, the aluminum appears to be over-etched since the squares on the 30µm test structure are missing, and are highly distorted on the 40µm structures as shown in Figure Exposed SU-8 Structural Layer Prior to Development Although the reticulation is easily seen in the anchor layer as shown in the previous section, the structural SU-8 layer appears to conform to these ridges which reduce the effect. In addition, the junctions where the anchor layer and SU-8 structural layer come into contact blend together and become invisible because the refractive index of exposed SU-8 is identical to unexposed SU-8. Figure 90 to Figure 93 illustrates this phenomenon. 37

44 Figure 90: Comb drive fingers defined prior to SU-8 development Figure 91: Stipples on large square test structure, also showing the smoothing of reticulated area Figure 92: Pop-up structure defined prior to SU-8 development Figure 93: L-shape test structures also showing the smoothing of reticulated area As shown in Figure 91 and Figure 93, the ripples in the anchor layer are reduced significantly. In the future, a method for smoothing out high stressed areas of SU-8 may be used to attempt to compensate for thermal shock SU-8 Structures after SU-8 Development When the unexposed SU-8 is developed, we expected the aluminum to flake off and release the overhanging SU-8 structures. However, it appeared that the SU-8 under the aluminum became cross-linked and bonded to the aluminum. Although none of the structures created with the embedded aluminum mask were released, the structural layers are clearly defined. An important observation to note is that SU-8 structural layers only attached to anchors, which are completely segregated from the aluminum, did not adhere to the anchors well. As shown in Figure 94, the square test structures peeled away from the SU-8 anchors. 38

45 Figure 94: Corner of square test structure peeling away Figure 95: Fixed-free cantilevers Figure 96: Fixed-fixed cantilevers Figure 97: Well defined comb drive fingers 9.6 Failure Modes The possible failure modes while using the embedded aluminum mask include: SU-8 cross-linking in the sputterer (power too high) SU-8 cross-linking by aluminum etchant Aluminum uneven and reticulated (under baked SU-8 and poor adhesion to substrate) Poor feature resolution due to mask problems Poor alignment of structures Large film stress in SU-8 (caused by thermal shock) Cracks/fractures in SU-8 (non-optimal exposure and bake times) Damage to large structures from development 39

46 The failure modes for the embedded aluminum mask are similar to the ones stated using the photoresist as a sacrificial material. These similarities are not surprising since the SU- 8 remains a constant source for problems to occur. However, unlike the experiments with the photoresist sacrificial layer, a major cause of failure was due to the aluminum etch. Although preliminary tests showed that the etch did not cross-link the SU-8, cross-linking was definitely occurring. The most likely reason for this oversight is that the initial tests did not account for the time that the SU-8 remained in contact with the aluminum etch when a temperature ramp up was performed. In the initial tests, the SU-8 contacted the only etchant for several seconds compared to the several minutes required for the temperature ramp up. Therefore we believe that the extended period of time, in addition to the increased temperature, may be the cause for the SU-8 cross-linking. In addition, due to the short period of exposure to aluminum etchant, the top surface layer of the SU-8 may have cross-linked but remained thin enough to be removed when the bulk of the SU-8 is developed. Tests conditions identical to the wafers were performed on a glass slide to confirm these results. Figure 98 illustrates the unexposed cross-linked SU-8 after contact with the aluminum etch for 7 minutes. In addition to the slide, wafer #11 was also used for an aluminum etch test and is shown in Figure 99 through to Figure 101. Figure 98: Cross-linked su-8 on slide due to long exposure to aluminum etchant Figure 99: SU-8 film after aluminum etch 40

47 Figure 100: Surface of SU-8 film removed after SU-8 development Figure 101: SU-8 film after introduction of isopropyl alcohol When the aluminum etch is performed on wafer #11, the surface of the remaining SU-8 becomes cloudy and dull in colour (Figure 99). When the SU-8 is developed, a skin is shown to peel back as the uncross-linked layer of SU-8 is dissolved (Figure 100). In addition, when isopropyl alcohol is used to rinse off the SU-8 developer, a film corresponding to the wrinkles in the aluminum become apparent (Figure 101). The cause of this phenomenon is unknown. It appears that the major failure of the embedded aluminum mask process is the incompatibility of SU-8 and the aluminum etchant. However, we have shown that metals can definitely be sputtered onto uncross-linked SU-8 without cross-linking which shows promise for further research using other metals as an embedded mask. 10 Recommendations and Future Work 10.1 Ramping hot plate One of the major problems for the both processes is the need for careful temperature control. Currently, the hot plates we are using require 20mins to reach 65 C from room temperature. The heating process also introduces a large overshoot which creates difficulties when attempting to characterize the baking times. In addition, it appears that there is a large delay between the temperature increase and the thermometer response which may also be causing problems when are attempting to control the temperature carefully Non expired SU-8 and SU The SU-8 that we are currently using is expired. Although the functionality did not seem to have deteriorated, using new SU-8 may give us different results. In addition, with the 41

48 introduction of SU , which promotes better wetting on silicon and has faster drying times, we may be able to improve our process Higher Quality Masks The masks that we are currently using are limited to 30µm resolution and the feature edges are very rough in nature. Because of the low resolution, it is far more difficult to create high aspect ratio structures without using a thicker viscosity SU-8. In addition, with higher quality masks, the blurring of the structures will be eliminated Photoresist Bake time/su-8 Development tests According to various authors, by increasing the SU-8 post-exposure-bake, the stiffness of the SU-8 structures can be increased significantly. Therefore, we need to find optimal bake parameters that can provide us with stronger structures. In addition, the optimal time for the SU-8 development and methods should also be examined in order to optimize the time required and yield. 42

49 11 Conclusion Our group has successfully developed a process that allows us to create suspended compliant SU-8 structures. We have created our own photomasks with test structures by which we can test the effectiveness of each process in creating detailed devices for future use. Two distinct process types were attempted using different materials for the sacrificial layer. The greatest success was achieved by using a sacrificial layer of thick positive photoresist. Using this material allowed us to produce cantilever structures and the first reported SU-8 pop-up structure. The process remains to be optimized with respect to photoresist baking time which appears to be critical to final design quality but the concept has been proven to be sound and is worth pursuing further. The second major process design that we attempted uses unexposed SU-8 as the sacrificial layer, which is protected from UV exposure by a thin layer of aluminum. The advantage with this arrangement is the large range of thicknesses that we can produce for the sacrificial layer. We were able to determine a successful recipe for sputtering on a thin smooth film of aluminum onto unexposed SU-8 and were able to pattern this aluminum using aluminum etch. Unfortunately it was discovered that the aluminum etch itself would react with the SU-8, resulting in a cross-linking and severe reticulation of the SU-8 surface. Alternate aluminum etches or lift-off processes may provide a more reliable method of SU-8 patterning. During experiments it was determined that the sputtering process itself did not crosslink the SU-8, so that gives us a wide range of options for depositing metal layers on SU-8 in the future. During the course of this project we have significantly expanded the process capabilities for SU-8 at SFU. The continuation of this research will allow us to design new MEMS devices in a process that will give us far more flexibility than traditional micromachining can offer. It is the intent of the group to continue with this research to develop a reliable and low cost photopolymer micromachining process. 43

50 12 Appendix A 12.1 Sputter log 44

51 45

52 46

53 47