Transcript
The Design and Fabrication of a Low-Cost, DMD Based Projection Lithography System A Thesis Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School of The Ohio State University By David Lindsey McCray, B.S. Graduate Program in Industrial and Systems Engineering The Ohio State University 2012 Master’s Examination Committee: Dr. Allen Yi, Advisor Dr. José Castro
Copyright by David Lindsey McCray 2012
Abstract
A low-cost, digital micro-mirror (DMD) based projection lithography machine was designed and constructed. The base of the machine utilizes a surplus Dover Instruments Corp. hard drive tester. The superstructure of the machine is constructed from aluminum plates, which may be bolted together in different positions to allow reconfiguration. The machine offers a positional resolution of 0.1 micron in 2 axes. A vacuum work holding device was also designed, built, and evaluated, along with a commercial alternative. It is determined that these devices may hold a silicon wafer flat within approximately +\- 0.7 micron. The final structure is tested for straightness of axis travel, which results in a maximum deviation of +/- 4 arcseconds. The optical system is also evaluated for its ability to produce an image and replicate a pattern loaded onto the DMD via a PC. A photoresist-covered wafer is exposed using multiple exposure times and patterns. The system is able to replicate the more basic patterns, but needs improvement to shrink feature size and improve definition. It is also found that the optical system needs modification to improve efficiency.
ii
Once the optical system is improved, this machine should be an excellent platform for scanning or step-and-scan lithography, and be able to produce small features covering an entire wafer.
iii
Acknowledgements I would like to thank Dr. Allen Yi for his support and patience throughout my entire time as a research assistant at The Ohio State University, as well as Dr. José Castro for his participation in my committee. I would also like to thank Josh Hassenzahl for his support throughout the manufacturing process, and his generous provision of shop facilities. Additionally, I would like to thank my fellow members of Dr. Yi’s group who assisted me throughout the construction and evaluation of this machine: Dr. Mujun Li, Dr. Lei Li, Hao Zhang, Jingbo Zhou, and Neil Naples. Most importantly, I would like to thank my wife and family, particularly my parents, for their exceptional support during my time in school. This would not have been possible without them.
iv
Vita April 19, 1984.........................................................Born - Lakewood, Ohio June 2002...............................................................Lakewood High School June 2010...............................................................B.S. Mechanical Engineering, The Ohio State University June 2010 to Present.............................................Graduate Research Associate, Industrial and Systems Engineering, The Ohio State University
Publications 1.
B. L. Anderson, J. G. Ho, W. D. Cowan, O. Blum-Spahn, A. Y. Yi, D. J. Rowe, M. R. Flannery, D. L. McCray, P. Chen, D. J. Rabb “Hardware Demonstration of Extremely Compact Optical True Time Delay Device for Wideband Phased Array Radars,” Journal of Lightwave Technology, Volume. 29, No. 9, May (2011).
v
2.
B. L. Anderson, J. G. Ho, W. D. Cowan, O. B. Spahn, A. Y. Yi, M. R. Flannery, D. J. Rowe, D. L. McCray, D. J. Rabb, P. Chen, “Ultra-compact Optical True Time Delay Device for Wideband Phased Array Radars,” SPIE Proceedings, Vol. 7669, Orlando, FL, Apr 26 (2010). Fields of Study
Major Field: Industrial and Systems Engineering Areas of Interest: Manufacturing and Machine Design
vi
Table Of Contents Abstract ............................................................................................................................................ ii Acknowledgements......................................................................................................................... iv Vita ................................................................................................................................................... v Table Of Contents .......................................................................................................................... vii List of Tables ................................................................................................................................... ix List of Figures ................................................................................................................................... x Chapter 1:
Introduction ............................................................................................................. 1
1.1
Project Overview .............................................................................................................. 2
1.2
A Basic History of Lithography ......................................................................................... 3
1.3
DMD ................................................................................................................................. 8
Chapter 2:
Mechanical Systems ............................................................................................... 10
2.1
The Dover Stage ............................................................................................................. 11
2.2
The Superstructure ........................................................................................................ 13
2.3
Mechanical Aspects of the Optical System .................................................................... 18
2.3.1
General Lens Mounting .......................................................................................... 18
2.3.2
Condenser Subassembly ........................................................................................ 19 vii
2.3.3
Projection Lens....................................................................................................... 21
2.3.4
UV Lamp ................................................................................................................. 22
2.3.5
DMD Holder ........................................................................................................... 23
Chapter 3:
Optical System ....................................................................................................... 28
Chapter 4:
Work-Holding ......................................................................................................... 33
4.1
Work-Holding Devices.................................................................................................... 34
4.2
Evaluation and Performance of Work Holding Devices ................................................. 35
Chapter 5:
Machine Performance and Evaluation................................................................... 50
5.1
Measuring Axis Straightness .......................................................................................... 51
5.2
Optical System Testing ................................................................................................... 58
5.2.1
Imaging Test ........................................................................................................... 58
5.2.2
Exposure Test ......................................................................................................... 63
Chapter 6:
Future Work ........................................................................................................... 68
References ..................................................................................................................................... 71 Appendix ........................................................................................................................................ 72
viii
List of Tables Table 1.1: Progression of lithography systems from 1970-1993 highlighting minimum feature size and pixel counts2................................................................................................................. 6 Table 1.2: Progression of step-and-scan lithography systems from 1990-2003 highlighting minimum feature size and pixel counts2 ............................................................................ 7 Table 2.1: Descriptions corresponding to component numbers from figure 2-3 .......................... 15 Table 2.2: Descriptions corresponding to component numbers from figure 2-4 .......................... 16 Table 2.3: Descriptions corresponding to component numbers from figure 2-6 .......................... 20 Table 2.4: Component descriptions corresponding to labels in figure 2-10.................................. 24 Table 2.5: Descriptions corresponding to component numbers from figure 2-11 ........................ 25 Table 2.6: Descriptions corresponding to labels in figure 2-12 ..................................................... 27 Table 4.1: Surface roughness of the chucks and wafers ................................................................ 43 Table 4.2: Results from a final mounting test for each wafer and chuck combination, the averages of five separate mountings for each combination, unspecified units in microns .......................................................................................................................................... 49 Table 5.1: Comparison of the results of axis deviation with and without stage motion, illustrating the role of vibration in the measurements....................................................................... 58
ix
List of Figures Figure 1.1: Single exposure vs. step and scan illustration3 .............................................................. 7 Figure 1.2: Illustration of DMD use in a projection system5 ............................................................ 9 Figure 2.1: Dover motion stage prior to modification ................................................................... 11 Figure 2.2: Top view of Dover motion stage prior to modication, demonstrating axis layout...... 12 Figure 2.3: Final iteration of superstructure for the T-slotted extrusions design family .............. 15 Figure 2.4: Final design chosen for the lithography superstructure .............................................. 16 Figure 2.5: Modular commercial optics mount from Edmund Optics ........................................... 18 Figure 2.6: Exploded view of the condenser lens subassembly .................................................... 19 Figure 2.7: Custom made spanner wrench for assembling condenser lens holders ..................... 20 Figure 2.8: Projection lens shown with adapter plate ................................................................... 21 Figure 2.9: UV lamp assembled and ready to mount on the light shelf ........................................ 22 Figure 2.10: Initial concept for DMD holder .................................................................................. 23 Figure 2.11: Concept for DMD holder prioritizing adjustability..................................................... 24 Figure 2.12: Final concept for DMD Holder ................................................................................... 26 Figure 3.1: Initial layout of the optical system, minus the projection lens ................................... 29 Figure 3.2: Projection lens mounted beneath the DMD on the lithography machine .................. 31 Figure 3.3: Initial Layout of the optical system, dimensions are in inches .................................... 31 Figure 4.1: Porous vacuum chuck top (left) and bottom(right): .................................................... 34 Figure 4.2: Grooved vacuum chuck (left), and spindle adapter (right).......................................... 35 x
Figure 4.3: Vacuum chuck created to supply vacuum to and hold the other two chucks for testing on the Profilometer .......................................................................................................... 36 Figure 4.4: Setup for wafer surface measurement ........................................................................ 37 Figure 4.5: Close-up of profilometer stylus ready to measure the glass wafer on the grooved chuck ................................................................................................................................. 37 Figure 4.6: Glass wafer surface profile for varying levels of vacuum on the grooved chuck ........ 38 Figure 4.7: Glass wafer surface profile for varying levels of vacuum on the porous chuck .......... 39 Figure 4.8: Silicon wafer surface profile for varying levels of vacuum on the grooved chuck ...... 40 Figure 4.9: Silicon wafer surface profile for varying levels of vacuum on the porous chuck ........ 41 Figure 4.10: Illustration explaining the different between surface roughness and waviness6 ..... 43 Figure 4.11: Three-dimensional image of glass wafer surface, taken with optical profilometer .. 44 Figure 4.12: Three-dimensional image of silicon wafer mounting surface, taken with optical profilometer ...................................................................................................................... 44 Figure 4.13: Three-dimensional image of a small portion grooved chuck surface, taken on one of the lands with optical profilometer .................................................................................. 45 Figure 4.14: Three-dimensional image of porous chuck surface, taken with optical profilometer .......................................................................................................................................... 45 Figure 4.15: Surface profile for glass wafer on the grooved chuck for five separate mountings, cleaned with Kimwipes between mountings .................................................................... 47 Figure 4.16: Surface profile for glass wafer on the grooved chuck for five separate mountings, cleaned with lens paper between mountings .................................................................. 47 Figure 4.17: Surface profile of silicon wafer on grooved chuck for five separate mountings, cleaned with lens paper in-between mountings .............................................................. 48
xi
Figure 5.1: Setup for measuring the straightness of the lithography machine axis, all dimensions are in inches ...................................................................................................................... 53 Figure 5.2: Mounting the laser for the calibration test ................................................................. 53 Figure 5.3: Calibration plot for horizontal laser displacement ...................................................... 54 Figure 5.4: Calibration plot for vertical laser displacement........................................................... 54 Figure 5.5: Plot of laser angular deviation vs. stage position in the vertical direction .................. 55 Figure 5.6: Plot of laser angular deviation vs. stage position in the horizontal direction ............. 56 Figure 5.7: Horizontal angular deviation of the axis measurement system due to vibration ....... 57 Figure 5.8: Vertical angular deviation of the axis measurement system due to vibration............ 57 Figure 5.9: Patterns input to the DMD, and the resulting image on the wafer surface when a laser is used as the system illumination source................................................................ 59 Figure 5.10: Field expedient alternative to the projection lens, a single lens mounted below the DMD on the projection shelf ............................................................................................ 60 Figure 5.11: DMD with the adapter plate and a black aperture .................................................... 61 Figure 5.12: Image showing the use of washers to change the DMD angle.................................. 62 Figure 5.13: Patterns input into DMD and their resultant images on the wafer plane, with the UV lamp as an illumination source ......................................................................................... 62 Figure 5.14: Surface image of the photo-resist structure produced in the first exposure test, with the triangle pattern on the DMD ...................................................................................... 64 Figure 5.15: Photo of wafer surface after second exposure test, increasing exposure time from left to right ........................................................................................................................ 65 Figure 5.16: Surface topography for second exposure test, 360 second exposure ...................... 65 Figure 5.17: Surface topography for second exposure test, 540 second exposure ...................... 66
xii
Figure 5.18: Surface topography for second exposure test, 720 second exposure ...................... 66
Figure A.1: Blueprint for Side Support Plate .................................................................................. 73 Figure A.2: Blueprint for Projection Shelf ...................................................................................... 74 Figure A.3: Blueprint for Light Shelf ............................................................................................... 75 Figure A.4: Blueprint for DMD adapter .......................................................................................... 76 Figure A.5: Blueprint for DMD mount stationary base .................................................................. 77 Figure A.6: Blueprint for DMD mount slider, page 1 ..................................................................... 78 Figure A.7: Blueprint for DMD mount slider, page 2 ..................................................................... 79 Figure A.8: Blueprint for projection shelf support, right side, page 1 ........................................... 80 Figure A.9: Blueprint for projection shelf support, right side, page 2 ........................................... 81 Figure A.10: Blueprint for projection shelf support, left side, page 1 ........................................... 82 Figure A.11: Blueprint for projection shelf support, left side, page 2 ........................................... 83 Figure A.12: Blueprint for y-axis adapter baseplate ...................................................................... 84 Figure A.13: Blueprint for main (front) vertical plate, page 1 ....................................................... 85 Figure A.14: Blueprint for main (front) vertical plate, page 2 ....................................................... 86 Figure A.15: Blueprint for projection lens adapter ........................................................................ 87
xiii
Chapter 1: Introduction
1
1.1 Project Overview The objective of this project is to design and build a projection lithography system that is based on the digital micromirror device (DMD). The average cost of a commercial lithography system was around $33 million in 2011.8 This machine does not possess the fine resolution obtainable with complex commercial systems, but the cost is much lower (around $12,000 for all components). In addition, the DMD offers a flexibility these commercial systems do not possess. Using the DMD as a dynamic mask, this system may quickly expose wafers using many different custom patterns. These patterns are simply loaded onto the DMD board in the form of a picture file, using a basic desktop PC. This is ideal for a research environment where one wants to experiment with creating many different types of small features. The low cost is attributable to the use of used commercial components, such as an obsolete Dover Instrument Corp hard drive tester (obtained on Ebay for ~$1200). This hard drive tester possesses the components of a high precision machine, and offers excellent positional accuracy and resolution. Although it may have outlived its intended function, it provides an excellent base upon which to build a lithography system. The raw materials used in the construction of new components were obtained from a local scrap dealer, who specializes in high quality materials. The most expensive component of the machine is the DMD, which costs around $10,000. Once completed, this machine will be evaluated for motion accuracy, ability to accurately hold a work piece (wafer), and ability to reproduce a pattern from the DMD onto a photoresist-covered wafer. 2
1.2 A Basic History of Lithography In the most general sense, lithography is the process of using a pattern or mask to reproduce itself (or its negative) upon another medium. The original process traces its origins to author Alois Senefelder, who sought affordable mass production of printed works.1 It is still a common practice in the printing industry today. Traditionally, the lithography pattern was created on a flat and smooth stone (usually limestone) or a metal plate. The pattern was etched into a wax or oil coating applied to the smooth surface. What remained were areas that are hydrophilic (will repel oil-based ink), and those that are hydrophobic. Ink was then applied to the surface and pressed against a sheet of paper. The result is the reproduction of the pattern (or its negative) onto the sheet. An alternative method is to press the lithography pattern onto an intermediate rubber surface, and then stamp the paper with the rubber. Modern lithography uses a similar process, except the wax coating has been replaced with photosensitive polymers (known as photoresists), and the substrate is typically a flexible sheet of aluminum, Mylar, or polyester. In addition, the pattern is no longer etched by hand but typically projected onto the pattern surface. The photosensitive polymer responds to the projected light pattern and is either chemically altered or prevented from chemically altering, dependent upon the type of polymer. The result is that part of the polymer may be washed away, leaving the desired pattern (or its negative). Beginning in the 1960’s, contact lithography was used to create integrated circuit (IC) and printed circuit board structures.2 Contact lithography uses masks in direct contact with the 3
silicon wafers onto which IC’s were produced. Light is shined through the mask and onto the wafer, which is coated with photoresist. Initially, the master patterns (or master masks) were cut by hand from a sheet of material known as Rubylith. 2 This master was then optically reduced onto a plate of glass which became what is referred to as a reticle. This reticle was then further reduced and replicated multiple times on the surface of a glass master mask. This master mask was then copied to create the masks used in the actual contact lithography production process. Although the cutting of the Rubylith was eventually automated, repairing the patterns and creating the final masks remained a time consuming and labor intensive endeavor. In the early 1970’s Bell Laboratories developed a more integrated approach. Primary patterns were generated using an Argon laser projected to a 7µm Gaussian spot, which was then scanned along the reticle using a stepper motor and ball screw. 2 The smallest line width this method could produce was approximately 35µm, which could later be reduced optically. The eventual minimum feature size after reduction and creation of the masks was 1µm. As feature size decreased, the optical systems became increasingly more complicated in order to account for distortion, variation in illumination, etc. The final product was still produced by contact lithography, however, which was an issue from both cost and quality standpoints. As the masks contacted the silicon wafers, they were inevitably damaged, which means the quality of features would diminish as the mask was used. Therefore, more masks had to be produced, which kept cost higher. During the mid to late 1970’s, projection lithography entered the scene. Initially, a 1:1 imaging lens was used between the mask and the wafer. At this time, wafer size was typically
4
50 mm. However, as the need for larger wafers and higher throughput arose, optical systems also had to increase in size. As optical size increased, overlay misalignment became a more serious problem. IC’s are made from multiple exposures in multiple layers on a wafer. As lens size increased, thermal distortion of lenses became an increasing problem. It was difficult to account for changing lens distortion from mask to mask (layer to layer). In addition, wafer size increases also increased the thermal distortion in the wafers. Film stresses on the wafers also had a greater effect on the larger wafers. The effective wafer size limit of 1:1 imaging turned out to be around 6”, in the early 1980’s. 2 It was about this same time that optical mask generation methods began to reach their limits, and electron beam exposure systems came into use. These systems eliminated the need for intermediate reticles, and a full size mask could be made in one step. The electron beam system worked much the same way as a cathode ray tube television. A beam of electrons was scanned across a surface (coated with e-beam sensitive resist) to create the desired pattern. Termed “electron beam lithography”, this system is still in widespread use for mask making. 2 However, it remains too slow and expensive to displace optical lithography in the exposure of the actual silicon wafers. The next major evolution was to forgo the creation of a 1:1 mask, reducing the mask and projecting it onto the wafer in the same step. This was known as reduction lithography. The projected image is smaller than the wafer itself, and therefore a mechanical motion and alignment system is required to expose the entire wafer. Such a system was known as a “stepper,” and the first wafer stepper was introduced in 1978.2 This system had automated
5
wafer alignment and handling systems, with a 10:1 image reduction ratio. The size of the projection field was 10mm x 10mm.2 Projection lithography has steadily evolved in order to keep up with the demands placed upon it by Moore’s law. Two examples of metrics of stepper lens technology are resolution and pixel count. Resolution is the smallest printable critical dimension of a feature (CDmin), and pixel count (N) is the number of pixels (of size CDminxCDmin) that will fit within the usable image field. Table 1.1 gives a progression of these values over about 20 years.2 Demand for even higher resolutions drove lithography from a step-and-expose method into a step-and-scan method. In step-and-scan, the actual projection area is quite small, but the image area may be larger by scanning the mask over the image. Figure 1.1 demonstrates the difference between step and step-and-scan projection lithography.3
Table 1.1: Progression of lithography systems from 1970-1993 highlighting minimum feature size and pixel 2 counts
Step-and-scan lithography has allowed CDmin and pixel count to steadily improve and keep up with Moore’s law. Table 1.2 demonstrates the progress that has been made using step-andscan lithography.2
6
Figure 1.1: Single exposure vs. step and scan illustration
3
Table 1.2: Progression of step-and-scan lithography systems from 1990-2003 highlighting minimum feature size 2 and pixel counts
With all of the mentioned advances in lithography, there is still widespread dependence upon the creation of an initial mask or reticle for each different exposure layer. Recently, efforts have gone into eliminating the need to create reticles altogether, by using what is called a dynamic mask. Two competing technologies for dynamic masks are the liquid crystal display (LCD) and the digital micromirror device (DMD). LCD and DMD lithography work much the same way as projection lithography. However, instead of the light interacting with a reticle prior to the wafer, the light either passes through the LCD or reflects off the DMD. Both may be used
7
with either step or step-and-scan methods. LCD’s, however, have issues with low contrast and transmittance.4 This is where the newer DMD technology has much to offer.
1.3 DMD The DMD was invented in 1987 by a team lead by Dr. Larry Hornbeck at Texas Instruments.5 The original goal was to develop a way to modulate light. Shortly after obtaining a patent in 1991, the DMD became the basis for DLP projectors, which are now commonplace in everyday life. The DMD is an array of tiny mirrors, each measuring less than 16 microns across. Every mirror may be controlled individually to tilt +12° or -12°. By reflecting light towards or away from the projector lens, each mirror may act as an on/off pixel. Grayscales are created by switching the pixel on/off very rapidly, much the same way brightness is controlled on an LED. Figure 1.2 demonstrates how a DMD mirror may act as a pixel in a projector.5 There are two main methods to display color. The first is to use a color wheel and produce images in 3 image sets, one in red, one in green, and one in blue. The second way is to use three DMD’s, one projecting in each of the basic colors (red, green, blue). Beginning in the early 2000’s, work began with the use of DMD’s as a dynamic mask for projection lithography. The basic principle is identical to that of the DLP projector mentioned above.
8
Figure 1.2: Illustration of DMD use in a projection system
9
5
Chapter 2: Mechanical Systems
10
2.1 The Dover Stage The mechanical system is based upon a Dover Instruments Corp. hard drive tester. This device was cheaply obtained on the used market, and already contains all the components of a basic, high precision 2-axis motion stage. This Dover stage consists of a granite block attached to a stout weldment through four self-leveling pneumatic vibration isolators. The weldment rests on four wheels for easy transportation around the lab, but also has four jacks to enable a stationary foundation when desired. A picture of the system prior to any modification is shown below in figures 2.1 and 2.2.
Figure 2.1: Dover motion stage prior to modification
11
Figure 2.2: Top view of Dover motion stage prior to modication, demonstrating axis layout
Mounted to the granite slab are two linear air-bearing ways with approximately six inches total travel. This is ideal for traversing the entire surface of a small silicon wafer. Air bearings are a good choice for this application because they offer very high accuracy and smooth motion. The ways are driven by linear motors, which contribute to high positioning accuracy. Linear motors tend to have a high potential for accuracy because there is no dependence upon a mechanical lead screw or other method of converting rotary to linear motion. The other contributor to this system’s high accuracy potential is the RSF Elektronik 6905-4 linear encoder. This particular encoder has a scale pitch of 10 micron, which allows a position resolution of 0.1 micron per encoder count.
12
One of the ways also contains a rotary axis, which may be used as either a spindle or indexing table.
2.2 The Superstructure The structure that would mount on the existing Dover stage went through several iterations. The major concerns in design were cost and flexibility. The immediate goal is for this system to be a projection lithography system, but the desire was expressed for it to be adaptable to laser lithography, laser cutting, or micro-milling. One of the biggest challenges was to work around the non-conventional layout of the axes on the Dover stage. Many two-axis stages will have one axis stacked upon the other. This allows the projector to be stationary, while the workpiece may move in two dimensions below it. The Dover system, however, has two axes that are completely independent of each other. Therefore, in order to obtain two-axis motion, the projection lens would have to be mounted on one of the axes and the workpiece on the other. This is less than ideal, because the projection system may be quite cumbersome to fit onto a relatively small slide. In addition, there are concerns about too much weight on the axis which must bear the projection system. In order to address these concerns about weight, it was first determined that as much of the projection system as possible should be stationary, off of the axis stage. This basically amounts to mounting the illumination lamp and perhaps collimating the light off of the stage. In order to achieve this without a complicated system of mirrors, the light would have to be arranged to shine in parallel to the motion axis. The granite table of the stage is not large enough to accommodate such a setup, and therefore this would require adding a mounting surface either cantilevered off of the granite 13
surface, or attached independently from it. The problem with attaching independently from the granite is that one loses the vibration isolation benefits of the table. On the other hand, if the light is cantilevered off of the granite, the large off center weight taxes the capability of the auto-leveling isolators. In addition, UV light is the planned light source for this project. Therefore, some type of shielding is desired to protect bystanders and operators. In order to both mount the light independently of a motion axis, and shield it, a shielding bellows would be required. When coupled with the additional mirrors required to mount the light off of the motion axis, this quickly takes a large chunk from the minimal budget of this project. Furthermore, this will be a step-and-expose system. In other words, the system will move into position and then expose the wafer before stepping to another position. Therefore, the dynamic response of the system is not critical. It may accelerate slowly as long as the final position is correct. Therefore, despite the weight concerns, it was decided that the entire projection system would be mounted on the stage’s Y-axis (horizontal in figure 2.2). From this point, a few different designs were composed. The first set was designed with t-slotted aluminum extrusions as an external frame. The entire frame would rest upon the Dover’s y-axis, and all components would attach via the t-slots. The final concept using extrusions is shown below in figure 2.3. This design was ultimately rejected for two primary reasons. The first was concerns over vibration, as all of the mirrors and lenses would have to be connected to the frame via cantilever mounts. The second reason was mostly aesthetic. The final design did not look particularly refined from a machine standpoint, which might affect obtaining funding down the road.
14
Figure 2.3: Final iteration of superstructure for the T-slotted extrusions design family
Item Number Description 1 DMD 2 Dover Base 3 Illumination Lamp 4 Projection Lens 5 Mirror Table 2.1: Descriptions corresponding to component numbers from figure 2-3
15
Figure 2.4: Final design chosen for the lithography superstructure
Item Number 1 2 3 4 5 6 7
Description Adapter Base Plate Main Plate Light Shelf Projection Shelf Projection Side Supports DMD Assembly Light Side Support Plates
Table 2.2: Descriptions corresponding to component numbers from figure 2-4
16
The design that was ultimately chosen is shown above in figure 2.4, minus the optical system. Aluminum was chosen as the structural material due to its affordability, lower weight (as opposed to steel), and dimensional stability (compared to plastics or wood). The thickness of the aluminum plate was mostly arbitrary, and was chosen based on what was available the most cheaply. The only constraints on thickness, from a design standpoint, are that the plate must be thick enough to drill and tap the sides, but not so thick as to be excessively heavy. A large quantity of ~0.75” 6061-T6 aluminum plate was available at low cost from a local supplier, and therefore this was chosen. The core of the design is the vertical front (main) plate, which divides the system into the projection side and the light source side. This plate has two sets of hole patterns, each with holes spaced 0.75” apart. One set is for the projection shelf, and one set is for the light shelf. The two shelves may be adjusted independently in 0.75” increments, depending on which holes they are bolted through. The large slot through the center of the plate is so that the path of light may be adjusted vertically. When the final position of the shelves and path of the light is determined, a sheet metal plate with aperture may be bolted into the pocket on the projection side. This will help prevent scattered light from reaching the wafer or leaving the system. In order to provide support for the shelves and make the system more rigid, side support plates were added to both the projection and light source sides. The supports on the projection side were cut from the same 0.75” aluminum as the other plates in the system, while those on the light source side were cut from 0.1875” aluminum plate.
17
In order to mount the superstructure to the machine axis, an adapter base plate was made. This plate was made larger than the original top area of the axis slide, to increase the available mounting space. The light shelf will hold the UV lamp, the light tunnel and light tunnel aperture, and the condenser lens assembly. The projection shelf contains the appropriate threaded holes to accept one of the reduction lens holders, the DMD holder, and the projection lens. These items will be described in further detail in later sections.
2.3 Mechanical Aspects of the Optical System 2.3.1
General Lens Mounting The projection lens, light tunnel, and condenser lens assembly are all held by
commercially available lens holders, which are intended for use with a threaded light table. Holes were drilled and tapped in the plates in order to attach these commercial holders. One of these holders is shown below in figure 2.5.
Figure 2.5: Modular commercial optics mount from Edmund Optics
18
These holders are adjustable for the size of the lens, and also its height, rotation, and lateral position (along the optical axis). 2.3.2
Condenser Subassembly The condenser lens assembly contains commercially manufactured lenses, but the
tubular housing and lens holders were designed and manufactured in house. The general design of the condenser lens assembly is shown below in figure 2-6.
Figure 2.6: Exploded view of the condenser lens subassembly
19
Item Number 1 2 3 4 5
Description Tubular Housing Set Screw Locking Ring Lens Lens Holder Main Body
QTY 1 2 2 2 2
Table 2.3: Descriptions corresponding to component numbers from figure 2-6
The condenser subassembly contains two lenses and two lens holders. Each holder consists of a main body into which the lens sits. The main body has a shoulder for the lens to set against, and female threads to accept the locking rings. Once the lens is placed in the main body, the threaded lock ring is then placed over the lens and tightened with a custom made spanner wrench, shown in figure 2.7. The radial clearance between the tubular housing and the main body is ~0.001”, which allows for the lenses to be slid through the tubular housing without tilting.
Figure 2.7: Custom made spanner wrench for assembling condenser lens holders
20
When the position of the lens holder inside the tube is satisfactory, the set screw in the main body is tightened. These set screws are nylon tipped to prevent damage to the inside of the tubular housing. 2.3.3
Projection Lens The projection lens was obtained from a commercial DLP projector. In order to simplify
mounting, the holes originally intended for the DMD holder were drilled and tapped completely through the projection shelf. An adapter plate allows the projection lens to mount to the underside of the shelf using these holes. In other words, the DMD holder and the projection lens use the same threaded holes (therefore short screws must be used to avoid interference). Short flat-head screws were chosen because their tapered underside helps ensure that the projection lens is pulled into alignment with the DMD. The projection lens and holder is shown below in figure 2.8.
Figure 2.8: Projection lens shown with adapter plate
21
2.3.4
UV Lamp The UV lamp must be mounted in such a way that the center of the lamp is in line with
the optical axis of the system. The UV lamp in use originally came with a different system and has no provisions for mounting in the manner required here. The solution is to use the holes originally intended for the lamp’s sheet metal housing. Commercial male/female threaded standoffs were inserted into the holes (to clear external features of the lamp). In order to raise the lamp to the proper height, mounting posts were made from 0.1875” thick, 1.5” leg aluminum 90° angle stock. Angle stock was chosen for a relatively high stiffness and low weight. These posts are mounted to the light via the standoffs mentioned above. Another small piece of 0.1875” thick aluminum was welded to the bottom of the posts and drilled such that they may be bolted to the light shelf. The resulting light and mounting post assembly is shown below in figure 2.9.
Figure 2.9: UV lamp assembled and ready to mount on the light shelf
22
2.3.5
DMD Holder The main design concerns for the DMD holder were adjustability and stability. Since
there was some uncertainty in the final optical design, the DMD had to be adjustable for both tilt and vertical position. Regarding stability, the mount must not vibrate lest the definition and resolution of the projected images suffer. At first, it was decided that the DMD would be cantilevered off of the aluminum extrusions, when extrusions were still being considered. The concept for this style of DMD holder is shown below in figure 2.10.
Figure 2.10: Initial concept for DMD holder
23
Item Number Description 1 DMD 2 DMD to Rod Swivel Mount 3 Hanger Rod 4 T-Slot Frame 5 T-slot to Rod Adapter Table 2.4: Component descriptions corresponding to labels in figure 2-10
However, after extrusions were ruled out for the superstructure design, two alternative DMD holder designs were put forth. The first is shown below in figure 2.11.
Figure 2.11: Concept for DMD holder prioritizing adjustability
24
Item Number Description 1 DMD Rod Base 2 DMD Slider Rods 3 Projection Lens Plate 4 DMD Adapter 5 DMD Model 6 DMD swivel 7 Vertical to Horizontal Pin 8 Bushing Half 9 Slider Side Pin
QTY 1 2 1 1 1 1 2 2 2
Table 2.5: Descriptions corresponding to component numbers from figure 2-11
The advantage to this design is that all of the close tolerance parts, namely the bushings and guide rods, are relatively cheap and commercially available. Hard coat anodized aluminum guide rods and bushings are available to a fitting clearance of 0.0005” to 0.001”. This would allow the DMD to be adjusted vertically while maintaining its angular alignment and lateral position. Although this design possesses many small components in addition to the rods and bushings, they are all relatively simple and easy to manufacture. The second design (shown below in figure 2.12) possesses fewer components and is simpler mechanically, but each component must be manufactured in-house. In addition, the components themselves are more complex and difficult to manufacture than in the previous design, and also require much larger pieces of raw material. This design, however, should be very stable and perform well. It was decided that the stability and mechanical simplicity of the second design were desirable, and therefore it was chosen for manufacture.
25
The fit clearance between the sliding mount and the stationary base is specified as 0.0004” to 0.0016” (g5 H6 ANSI). This fit was the tightest free-sliding fit obtainable with available manufacturing methods. A tight fit is desired to constrain the mount to vertical translation, without affecting the DMD’s angle or lateral position.
Figure 2.12: Final concept for DMD Holder
26
Item Number Description 1 Stationary Base 2 DMD Sliding Mount 3 Shoulder Bolt 4 DMD Adapter 5 DMD 6 Projection Lens Plate Table 2.6: Descriptions corresponding to labels in figure 2-12
This vertical adjustment is obtained by sliding the sliding mount over the stationary base. The mount is constrained rotationally by means of the shoulder bolt (5/16”x5/8” shoulder and ¼-20 UNC threads) and the vertical slot. The clearance between the width of the slot and the diameter of the shoulder bolt is ~0.002”. The vertical position may be locked by tightening the shoulder bolt. Angular adjustment may be achieved by different methods. Shims may be used under the DMD adapter, or an angled DMD adapter may be produced. This is somewhat cumbersome, but will provide the most stable mounting solution. The initial angle is set at 45 degrees, which is what is believed to be appropriate.
27
Chapter 3: Optical System
28
The optical system (except for the projection lens) as originally designed is shown below in Figure 3.1.
Figure 3.1: Initial layout of the optical system, minus the projection lens
Beginning from right to left, item number 1 is the UV illumination lamp. This lamp was taken from a Porta-Ray 400R UV curing system. It is a 400W metal-halide lamp that outputs light in the wavelength range of 320-390nm. Item 2 is a “light tunnel” combined with a shield to block extra UV light from escaping the system. The light tunnel consists of four inward-facing mirrors arranged in a rectangular tube. The purpose of this element is to make the intensity of the light field coming from the UV lamp more uniform. This element also serves to make the light field rectangular to match the shape of the DMD.
29
Item three is the condenser lens assembly. This consists of two 75mm diameter B270 aspheric condenser lenses from Edmund Optics. Item four is an f/3.5 projection lens that was removed from a commercial projector. The purpose of this lens is to further reduce the area of the light field before it hits the DMD. By the time the light passes through this element, it should be as close as possible to a collimated beam with a light field roughly the size and shape of the DMD. Item five is the DMD assembly, which consists of the DMD board mounted to the DMD holder. The DMD board in this system is a Texas Instruments Discovery 1100. It consists of the DMD chip mounted to a controller board, which is provided with software to drive the DMD. This particular DMD chip consists of an array of 1024x768 mirrors with a pitch of 13.68 micron. Item six is the projection lens, which mounts to the underside of the projection shelf as shown below in figure 3.2. The purpose of the projection lens is to image the light from the DMD, such that the system is not limited to 1:1 size conversion from the DMD to the imaging plane. In other words, the projection lens should enable reducing the DMD pattern to obtain smaller features on the photoresist. The initial layout of the optical system is shown below in figure 3.3. The dimensions between the DMD and projection lens are controlled by the DMD holder position, which is set such that the DMD is in line with the optical axis for a given projection shelf setting. The distance between the projection lens and the wafer is controlled by the position of the projection shelf, and will have to be adjusted depending upon the current work-holding fixture and projection lens.
30
Figure 3.2: Projection lens mounted beneath the DMD on the lithography machine
Figure 3.3: Initial Layout of the optical system, dimensions are in inches
31
As discussed in section 5.2.1, the projection lens was eventually replaced with a generic convex lens found in the lab. In addition, a ground glass optical diffuser was placed immediately in front of the light tunnel. Otherwise, the final optical system is as shown in figure 3.3 and described above.
32
Chapter 4: Work-Holding
33
4.1 Work-Holding Devices Two different work-holding fixtures (chucks) were decided upon for the lithography system, both of which depend upon vacuum to hold the part securely. The first vacuum chuck (shown in figure 4.1) is commercially made, although the manufacturer information is not available.
.
Figure 4.1: Porous vacuum chuck top (left) and bottom(right):
The top surface is porous, and the bottom surface is lapped flat and parallel to the top. The bottom surface contains a hole, which allows vacuum to travel from the underside to the porous surface on top. This chuck must be placed upon another vacuum chuck, which both holds the porous chuck securely and supplies vacuum (through the aforementioned hole). This chuck should excel at holding thin or otherwise flimsy workpieces, given the small distance between its pores. The second workholding method is also a vacuum chuck, but was designed and produced at Ohio State (Figure 4.2). This design borrows heavily from the chucks that may be
34
purchased with Moore Nanotechnology ultra-precision lathes.
Figure 4.2: Grooved vacuum chuck (left), and spindle adapter (right)
This chuck is a two piece assembly; the first piece (figure 4.2, right) is simply an adapter, to go from the spindle face on the Dover system to the actual vacuum chuck. The chuck itself (figure 4.2, left) contains grooves cut into the surface with the vacuum supply coming through holes in each groove. The number of grooves connected to the vacuum may be changed by an internal set screw, which blocks off all of the holes beyond its position. The front and back faces of this chuck were diamond-turned on a Moore Nanotech 350FG in order to ensure flatness and parallelism. For more design details, please see the blueprint which was electronically submitted with this paper.
4.2 Evaluation and Performance of Work Holding Devices In order to quantify the performance of these chucks and their suitability with different wafers, several tests were performed. These tests had to be conducted off of the Dover system (in order to use a profilometer), and therefore a temporary vacuum supply system had to be created. This involved a standard diaphragm vacuum pump, a vacuum regulator, and a third
35
vacuum chuck. The third vacuum chuck is simply a method of supplying vacuum to the ports in the chucks being tested. This third chuck is shown in figure 4.3.
Figure 4.3: Vacuum chuck created to supply vacuum to and hold the other two chucks for testing on the Profilometer
In order to perform the tests, the third vacuum chuck was placed on a 3-axis tilting table on the Mitutoyo Surftest SV-3100 Profilometer, and leveled as much as possible (within about 10 µm). The vacuum was then hooked up to the port (right side in figure 4.3), and the chuck being tested placed on top, followed by the workpiece. The desired vacuum was set, and the profilometer was run from the wafer center along a radial path for 35mm. Tests were conducted both ascending and descending from 10 KPavac to 60 KPavac in 10 KPa increments (tests with the porous chuck went up to 70 KPavac). The workpiece was given about 30 seconds to settle after each vacuum change. The overall setup for the experiment is shown in figures 4.4 and 4.5. 36
The first test was conducted with a piece of Schott thin glass type D 263 T, 400µm thick and 4” square, mounted on the grooved vacuum chuck. The results are shown in figure 4.6.
Figure 4.4: Setup for wafer surface measurement
Figure 4.5: Close-up of profilometer stylus ready to measure the glass wafer on the grooved chuck
37
Glass Wafer Distortion vs. Holding Vacuum, Grooved Chuck 1.5 10 KPa 20 KPa 30 KPa 40 KPa 50 KPa 60 KPa
1
Y Deviation (micron)
0.5
0
-0.5
-1
-1.5
0
5
10
15 20 X Distance from Center (mm)
25
30
35
Figure 4.6: Glass wafer surface profile for varying levels of vacuum on the grooved chuck
As expected, waviness increased with the vacuum level, and the wafer surface began to conform to that of the vacuum chuck. No hysteresis was observed. The profile is the same regardless of the direction from which the test pressure is approached. It was not expected that the wafers would deform so much even at low vacuum pressures. In addition, the average flatness of the wafer was much less than expected. A surface diamond-turned on the Nanotech 350FG typically deviates less than 200 nanometers from perfect flatness. The glass surface showed deviations on the order of 2 µm which may not be entirely attributed to deformation into the lands of the chuck. The second test was with the same glass but using the porous vacuum chuck. The results are shown below in figure 4.7.
38
Glass Wafer Waviness Test, Porous Chuck 2.5 10 KPa 20 KPa 30 KPa 40 KPa 50 KPa 60 KPa 70 KPa
Y Deviation (micron)
2
1.5
1
0.5
0
-0.5
0
5
10
15 20 25 X Distance from Center (mm)
30
35
Figure 4.7: Glass wafer surface profile for varying levels of vacuum on the porous chuck
The effect of increasing vacuum with the porous chuck produces very different results than with the grooved chuck. As the vacuum is increased, the glass becomes flatter. The cause of the large bump at ~17 mm is uncertain, but will be investigated later. On subsequent runs of this test, the glass was never without some similar anomaly. Even with the bump, however, the surface peak-to-valley variation is on the order of 0.75 micron at 70 KPa vacuum. This is much better than with the grooved chuck. The third test was conducted using a 1mm x 100mm diameter silicon wafer and the grooved chuck. A vacuum level of 50 KPa was the maximum attainable with this setup and, therefore, was the maximum vacuum tested. Otherwise, this test was performed in an identical fashion to test one. The results are shown below in figure 4.8.
39
Silicon Wafer Waviness Test, Grooved Chuck 3 10 KPa 20 KPa 30 KPa 40 KPa 50 KPa
Y Deviation (micron)
2
1
0
-1
-2
-3
0
5
10
15 20 25 X Distance from Center (mm)
30
35
Figure 4.8: Silicon wafer surface profile for varying levels of vacuum on the grooved chuck
The silicon wafer does not conform to the shape of the grooved chuck like the glass wafer. At 10 KPa, the vacuum is not large enough to even pull the wafer against the surface of the chuck. After 20 KPa, the increase in vacuum has only a minor effect on the surface waviness. This is most likely due to the relatively large thickness of this wafer. It is believed that a thinner wafer would more closely mirror the results with the glass wafer; however, no thin silicon wafers were available for this test. With this setup, the least surface variation attainable seems to be about 0.6 micron peak-to-valley, which is very good. The fourth test was conducted using the same silicon wafer and the porous chuck. Once again, the porous chuck allowed higher vacuum levels to be obtained, in this case 70 KPa. The results are shown in figure 4.9.
40
Silicon Wafer Waviness Test, Porous Chuck 0.5
Y Deviation (micron)
0
-0.5
-1 10 KPa 20 KPa 30 KPa 40 KPa 50 KPa 60 KPa 70 KPa
-1.5
-2
0
5
10
15 20 X Distance from Center (mm)
25
30
35
Figure 4.9: Silicon wafer surface profile for varying levels of vacuum on the porous chuck
Similarly to test three, 10 KPa does not appear sufficient to even pull the wafer against the chuck surface. Unlike test three, increasing vacuum improves the flatness of the wafer up to 60 KPa, where the trend levels off. The peak-to-valley surface variation at 60 KPa is about 0.3 micron, which is the best of any test. The glass was the poorest performer in these first set of tests. The thinness of the glass is one of the likely culprits in its poor performance. The glass seems to deform around any defects or variations on the chuck surface. The porous chuck has obvious pits and knicks from abuse by previous users, and these may be the cause for some of the anomalies seen in the glass surface profiles. However, these are random in nature and difficult to quantify.
41
When it comes to the glass and the grooved chuck, however, an analysis is more straightforward. The spacing between groove centerlines on this chuck is ~6.6 mm (0.260”) and, therefore, one would expect to see corresponding dips in the glass surface. Looking at figure 4.6, this is confirmed. There are larger waves, however, which do not correspond to the structure of the chuck. It is suspected that the high smoothness and flatness of both the chuck and the glass may be to blame. Based on previous precision machining experience, it is normal for extremely flat surfaces to trap oil and debris between them, such that they never sit true to each other within less than a micron (without wringing). If two surfaces must mate very closely and wringing is not possible, at least one of the surfaces must have a roughed texture. Typically, this is achieved through machining very small grooves into the surface, such that ~75 micron lands remain. The grooves provide a place for trapped fluids (oil, air) to go, and the end result is that the two surfaces may mate together within less than a micron. It may seem counterintuitive that a rougher surface could offer better flatness. However, the surface peaks and valleys are extremely small, and the surface peaks are very close to the same level. A mating surface will rest on these peaks, while debris and fluids may move into the valleys. If the peaks were not on the same level, then the flatness of the wafer would suffer. Furthermore, if the peaks were too far apart, then the wafer would begin to deform into the valleys as vacuum is applied. Therefore, there is an ideal size range for these surface features. Figure 4.10 offers a visual explanation of these ideas.6 It is worth noting that much of the difference between waviness and roughness depends upon the scope of one’s view. At low magnification, a surface may appear very flat. However,
42
when zoomed in very close, the surface is full of mountains and valleys. This is part of the ambiguity to surface measurement, and why variables such as cutoff frequency are important in distinguishing between waviness and roughness.
Figure 4.10: Illustration explaining the different between surface roughness and 6 waviness
In to order evaluate the effect of surface roughness on the results, sections of the mounting surfaces of the glass and silicon wafers and grooved and porous chucks were measured with a Veeco NT9100 optical profilometer. The surface profiles are shown below in figures 4.11 through 4.14, respectively. Surface roughness values are shown in table 4.1.
Item Glass Wafer Silicon Wafer Grooved Chuck Porous Chuck
Surface Roughness, Ra 9.83 nm 638 nm 14.3 nm 295 nm
Table 4.1: Surface roughness of the chucks and wafers
43
Figure 4.11: Three-dimensional image of glass wafer surface, taken with optical profilometer
Figure 4.12: Three-dimensional image of silicon wafer mounting surface, taken with optical profilometer
44
Figure 4.13: Three-dimensional image of a small portion grooved chuck surface, taken on one of the lands with optical profilometer
Figure 4.14: Three-dimensional image of porous chuck surface, taken with optical profilometer
45
It may be seen that the glass wafer and grooved chuck are extremely flat and smooth. The grooved chuck shows minor surface damage, which may contribute to some of the waviness seen in test one. However, in general, the surface is very smooth as indicated by the Ra values. Therefore, it is plausible that the smoothness of the two surfaces is contributing to the flatness issues. The relatively good performance of the silicon wafer with the grooved chuck may support this notion further, as figure 4.12 shows that the silicon wafer’s mounting surface is relatively rough. In an attempt to investigate this further, experiments were conducted to test alternative cleaning methods and their effect on the wafer surface waviness. It is suspected that if oil or fluids are responsible for waviness, then the profile will change significantly with successive surface wiping. The first of these tests was conducted using Kimwipes laboratory tissues and 99% Isopropyl alcohol. The wafer was wetted with alcohol and then wiped in a circular motion, starting from the center and working outward. Once completely dry, the wafer was mounted on the chuck and the Mitutoyo profilometer used to measure its profile (as with tests one through four). The piece was then removed from the chuck, and the process repeated. This was done until the wafer was mounted and removed five times. The vacuum level was chosen to be 10 KPa, because that yielded the best results in previous testing. The results are shown in figure 4.15. The test was then repeated, except lens paper was used instead of Kimwipes. The results for this run are shown in figure 4.16. In figures 4.15 and 4.16, each line represents the the profile of the wafer surface for a given mounting. With Kimwipes, each mounting results in
46
Mounting Test Using Kimwipes and Alcohol 3 test test test test test
Surface Height (micron)
2
1 2 3 4 5
1
0
-1
-2
-3
0
5
10
15 20 25 Distance From Chuck Center (mm)
30
35
Figure 4.15: Surface profile for glass wafer on the grooved chuck for five separate mountings, cleaned with Kimwipes between mountings Mounting Test Using Lens Paper and Alcohol 3 test test test test test
Surface Height (micron)
2
1 2 3 4 5
1
0
-1
-2
-3
0
5
10 15 20 25 Distance From Chuck Center (mm)
30
35
Figure 4.16: Surface profile for glass wafer on the grooved chuck for five separate mountings, cleaned with lens paper between mountings
47
a significantly different surface shape. The peaks and valleys are in different locations and vary in magnitude. This is confirmation that much of the surface deviation is due to random variables such as fluid or debris between the wafer and the chuck. If the deviation were due to the chuck itself, the locations of the peaks and valleys would be more consistent as they would represent fixed deformities in the chuck surface. When lens paper was used to clean the wafer, the profile still changes between mountings. However, the magnitude of the surface variation is significantly reduced. One of the prominent characteristics to distinguish lens paper from Kimwipes is the presence of lint. Therefore, much of the variation seen with Kimwipes may be due to lint sticking to the glass. The test was again repeated using lens paper and the silicon wafer. For this run 50 KPa was chosen as the vacuum pressure, as it yielded the best results from the previous testing with this wafer and the porous chuck. The results for this test are shown below in figure 4.17. Mounting Test Using Lens Paper and Alcohol, Silicon Wafer 3 test test test test test
Surface Height (micron)
2
1
1 2 3 4 5
0
-1
-2
-3
0
5
10 15 20 25 Distance From Chuck Center (mm)
30
35
Figure 4.17: Surface profile of silicon wafer on grooved chuck for five separate mountings, cleaned with lens paper in-between mountings
48
The profile of the silicon wafer is extremely consistent between mountings. This is confirmation that a little surface roughness aids in mounting consistency. If glass must be used as a wafer material, it is evident that it must be cleaned with methods better than those presented here. Based on what has been learned thus far, a final mounting consistency test was performed for all four combinations, using the ideal conditions for each. Each combination was remounted five times, and the surface profile measured each time. The results of this test are shown below in table 4.2. All combinations, except for the glass with the grooved chuck, performed very close to the same in this test.
Wafer Material Chuck Used glass glass silicon silicon
Grooved Porous Grooved Porous
Vacuum Pressure 10 KPa 70 KPa 50 KPa 70 KPa
Average Maximum Peak-toValley Deviation 1.0695 0.6764 0.6145 0.6677
Average Standard Deviation 0.2511 0.1452 0.1606 0.2029
Table 4.2: Results from a final mounting test for each wafer and chuck combination, the averages of five separate mountings for each combination, unspecified units in microns
49
Chapter 5: Machine Performance and Evaluation
50
5.1 Measuring Axis Straightness In order to gain more understanding of the lithography system’s accuracy potential, the deviation of a linear axis from true linear motion was measured. Testing was only performed on the axis upon which the superstructure is mounted. The drive system for the other linear axis is still in need of repairs, and therefore could not be tested. In order to perform this measurement, a laser was mounted on the superstructure, and, as much as possible, lined up parallel with axis. This was done by shining the laser on a flat surface, and adjusting it until the spot remained stationary (visually) as the axis was moved. A CCD camera was then placed such that the laser would shine on its sensor. The camera used for this measurement was a Pixelink PL-B741F. This camera records in 8-bit grayscale and has a 1280x1024 pixel sensor, with a pixel size of 6.7 µm. The basic idea is that the CCD camera will record the position of the laser spot. As the motion stage is moved, the laser spot on the CCD will also move due to three primary causes: misalignment between the laser axis and the motion axis, angular deviation (yaw, pitch, and roll) of the axis, and lateral deviation of the axis. The motion of the spot due to laser misalignment will be linear, and can be filtered out using a linear fit to the data. The angular deviation of the stage will be magnified by the distance between the laser and the CCD detector, and is therefore the easiest to measure. The lateral motion is the most difficult to measure and will be neglected. It is assumed that the laser spot motion due to lateral deviation will be much less than that due to angular motion. In addition, the components of angular motion will not be measured independently. The equipment required to measure them independently is more complex than what is available. For example, yaw results in a horizontal movement of the laser
51
spot, while roll results in both a horizontal and vertical movement (because the laser is not mounted collinear with the roll axis). The horizontal effect of yaw may not be distinguished from the horizontal effect of roll with the equipment available. In order to track the motion of the laser, an image of the laser spot is first recorded from the CCD. The picture is then loaded into MATLAB and converted into a binary image using a threshold of 50%. The threshold value may be changed, but 50% gave good results for this experiment. The function “regionprops” is then used to find the centroid of the illuminated region in the picture (the laser spot). Matlab outputs the position of the centroid in pixel numbers. In order to convert this to microns, one must multiply the pixel coordinates by the pixel size of 6.7 microns. It is the position of this centroid that is used to track the position of the laser. In basic concept, the measurement system only requires the laser and the CCD camera. In order to achieve practical results, however, additional elements had to be included. Polarizers were added to control the strength of the laser beam so that it would not oversaturate the CCD sensor. A lens also had to be placed in front of the camera, due to the poor quality of the laser and the limitations of the image processing software. The purpose of the lens is to focus the laser and make the spot more uniform. The setup for the experiment is shown below in figure 5.1. Adding the lens not only focuses the laser beam, it also scales the motion of the spot by an unknown amount. Because of this, a calibration test had to be devised to convert the scaled spot motion to actual laser motion. This calibration test consisted of mounting the laser to a 3axis manual motion stage (Figure 5.2), which was in turn mounted on the lithography system
52
superstructure. The laser was mounted as close as possible to its position for the straightness test. The optical system and the relationships between its elements were identical to those in the straightness test.
Figure 5.1: Setup for measuring the straightness of the lithography machine axis, all dimensions are in inches
Figure 5.2: Mounting the laser for the calibration test
53
Using the manual stage, the laser was translated in steps of 500 micron and the spot position recorded at each step. This was done for both horizontal and vertical translation. The results are shown below in figures 5.3 and 5.4. Horizontal Beam Shift vs. Horizontal Displacement Of Stage 50 data fit line
y = - 0.0698*x + 0.202
Beam Position on CCD (micron)
0
-50
-100
-150
-200
-250
0
500
1000
1500 2000 Stage Position (micron)
2500
3000
3500
Figure 5.3: Calibration plot for horizontal laser displacement
Vertical Beam Shift vs. Vertical Displacement Of Stage 250 y = 0.0692*x + 0.446
Beam Position on CCD (micron)
200
150
100
50 data fit line 0
0
500
1000
1500 2000 Stage Position (micron)
Figure 5.4: Calibration plot for vertical laser displacement
54
2500
3000
3500
The slope of the linear fit line is a conversion factor between actual laser motion and motion of the laser spot on the CCD. The sign of the slope is not important but is dependent upon arbitrary factors. The absolute value of the slope should be the same for both plots, since the pixels are square. In fact, they are very close, within less than 1% difference. The conversion factor used for the straightness test is the average of the two slopes from these calibration tests, or 0.0695. After calibration, the actual straightness measurement was performed. The machine was jogged in 1mm increments, and the laser spot position recorded at each position. After filtering out the motion due to misalignment, the results for angular displacement of the axis are shown below in figures 5.5 and 5.6.
Angular Deviation of Stage in Vertical Direction 1.5
Stage Angle (arcsec)
1
0.5
0
-0.5
-1
-1.5
0
5
10
15
20 25 30 35 Stage Position (mm)
40
Figure 5.5: Plot of laser angular deviation vs. stage position in the vertical direction
55
45
50
Angular Deviation of Stage in Horizontal Direction 5 4 3
Stage Angle (arcsec)
2 1 0 -1 -2 -3 -4 -5
0
5
10
15
20 25 30 35 Stage Position (mm)
40
45
50
Figure 5.6: Plot of laser angular deviation vs. stage position in the horizontal direction
Due to both equipment and space limitations, the CCD camera and lens had to be mounted off of the Dover platform and were not vibration-isolated. Therefore, given the extremely small nature of the displacements being measured, there was some concern about the effect of ambient vibrations on the results. In order to quantify the effects of vibration, the position of the laser was recorded over a period of time while the axis was held stationary. The results are shown below in figures 5.7 and 5.8. It is clear from these results that vibration has a significant impact on the straightness measurements. The standard deviation of the results is summarized in table 5.1.
56
Static Angular Deviation of Stage in Horizontal Direction 5 4 3
Stage Angle (arcsec)
2 1 0 -1 -2 -3 -4 -5
0
5
10
15
20
25 30 Time (s)
35
40
45
50
Figure 5.7: Horizontal angular deviation of the axis measurement system due to vibration
Static Angular Deviation of Stage in Vertical Direction 1.5
Stage Angle (arcsec)
1
0.5
0
-0.5
-1
-1.5
0
5
10
15
20
25 30 Time (s)
35
40
45
50
Figure 5.8: Vertical angular deviation of the axis measurement system due to vibration
57
Direction Horizontal Vertical
Standard Deviation Of Axis Angular Orientation (arcsec) Straightness Test Vibration Test Difference % Attributable to Vibration 1.850 1.214 0.635 65.7% 1.214 0.556 0.658 45.8%
Table 5.1: Comparison of the results of axis deviation with and without stage motion, illustrating the role of vibration in the measurements
The difference category represents the deviation attributable to causes other than vibration, such as axis form error. These errors are in the single digit arc-seconds, which is extremely small. It is worth noting that the measurement system is more prone to vibration in the horizontal direction. This is to be expected, given the mounting method of the camera and lens. The tall mounts will be more prone to swaying than bouncing vertically.
5.2 Optical System Testing 5.2.1
Imaging Test The main purpose of the optical system is to produce an image of the DMD pattern on
the top surface of a photo-resist covered wafer. Therefore, before any exposures were done, a piece of white paper was placed on the chuck surface to see if a clear image was visible. This initial test proved that the light system, as designed, was not going to be sufficient. The intensity of the light field was not uniform, and still resembled the shape of the UV bulb. In order to remedy this, a piece of translucent glass (an optical diffuser) was added after the light tunnel. In addition, it was discovered that the light field coming out of the optical system was not well collimated. When combined with the fact that the projection lens was found 58
unsuitable for this setup, no image was obtained on the working plane of the wafer. Therefore, the projection lens was removed. In order to verify that the cause of failure was poorly collimated light, the UV lamp was replaced with a red laser. Two different DMD patterns were tried, and both showed up on the paper (Figure 5.9). Although these images also exhibited diffraction patterns and other issues, they proved that collimated light would produce an image. This also verified that the DMD was working properly.
Figure 5.9: Patterns input to the DMD, and the resulting image on the wafer surface when a laser is used as the system illumination source
59
If the UV light were well collimated when hitting the DMD, an imaging lens would not be necessary to produce an image on the work surface. However, since the light was poorly collimated, a lens was required. A single lens was found in the lab which proved suitable to focus the light coming from the DMD into a usable image. This lens replaced the projection lens on the underside of the projection shelf, after the DMD (Figure 5.10).
Figure 5.10: Field expedient alternative to the projection lens, a single lens mounted below the DMD on the projection shelf
After these modifications, the system was re-tested with the UV lamp. This time, reasonably clear images of the DMD pattern were obtained. The images, however, suffered from poor contrast, which is a major problem for exposing photoresist. It was theorized that poor contrast may be due to light scattering around inside the DMD holder and then shining out onto the work surface. This scattered light would effectively “wash out” the DMD image. In response to this theory, all of the reflective surfaces inside the holder were painted a non-
60
reflective matte black. In addition, an aperture was placed over the DMD to restrict the directions from which light could hit it or reflect off (to avoid scattering) (figure 5.11)
.
Figure 5.11: DMD with the adapter plate and a black aperture
These modifications did little to improve the contrast of the image. In fact, the aperture only interfered with the image and was removed. The next solution was to modify the angle of the DMD. It was discovered that tilting the DMD, a few degrees away from the nominal 45°, significantly improved the image contrast. Therefore, four ~0.050in washers were used as shims under the DMD to tilt it, as shown in figure 5.12. The light system was re-tested with this modification. The resulting images are shown in figure 5.13, along with the corresponding pattern loaded into the DMD. There is some distortion in the DMD images because of the ~45° tilt of the DMD. The distance of the DMD to the lens changes depending on where you are on the DMD surface;
61
Figure 5.12: Image showing the use of washers to change the DMD angle
Figure 5.13: Patterns input into DMD and their resultant images on the wafer plane, with the UV lamp as an illumination source
62
therefore, the lens focus is slightly different for each point. However, the contrast and clarity of these images were deemed sufficient to try an exposure test. For the results in figure 5.13, the DMD holder was set at 0.75” above its bottom-most position, and the top bolts of the projection side supports were in the fourth hole from the topmost position. The vacuum chuck adapter was used alone, with a 3/16” shim under the wafer. The light shelf bolts were in the 16th hole from the top-most position. This allowed for the image to be in the best focus when the lens from figure 5.10 was mounted as shown. 5.2.2
Exposure Test The exposure test was conducted using 500µm thick, 100mm diameter silicon wafers.
They were spin-coated with SPR 220-7 positive tone photoresists. The spinning speed was 3000 rpm, which should result in a 7.5µm coating thickness. Coated wafers were baked at 115°C for 60 seconds in order to drive out the solvents. The wafers were placed on the vacuum chuck adapter plate without vacuum, as the chuck was too high for the focal length of the improvised lens. Then, using one wafer for each of the patterns in figure 5.13, separate exposures were conducted for 15, 30, 60, 120, and 180 seconds. In order to protect the rest of the wafer during a given exposure, an aluminum mask with a 7/16” hole (for the pattern to shine through) was laid on the wafer. The results of this first test were poor. Only the 180 second exposure of the triangle resulted in any structure in the resist. The shape merely resembles a triangle; the definition is poor. Figure 5.14 shows a surface picture taken with the Veeco Profilometer.
63
Figure 5.14: Surface image of the photo-resist structure produced in the first exposure test, with the triangle pattern on the DMD
The maximum height of the feature above the wafer surface was ~1µm, which is considerably less than the initial resist thickness of 7.5 µm. It is suspected that the intensity of the UV light is too low to properly expose the resist. This was initially suspected based on the amount of light escaping the system and the dimness of the images themselves. Based on these results, a second test was conducted using a different resist and longer exposure times. For this test S1813-D positive tone resist was spin-coated at 3000 rpm, which should result in a coating thickness of ~1.3 µm. Coated wafers were again baked at 115 °C for 60 seconds in order to drive out the solvents. The exposure times were: 180, 360, 540, and 720 seconds. Only the triangle pattern was used in this test. The wafer after development is shown below in figure 5.15. The patterns are in order of increasing exposure time, with 180 seconds on the right and 720 seconds on the left.
64
Figure 5.15: Photo of wafer surface after second exposure test, increasing exposure time from left to right
Surface images taken with the optical profilometer are shown in figures 5.16 through 5.18 below, for the 360, 540, and 720 second exposures, respectively. A good image could not be obtained from the 180 second exposure. The triangle becomes progressively more complete at the exposure time increases. Even at 720 seconds, however, the corners lack definition.
Figure 5.16: Surface topography for second exposure test, 360 second exposure
65
Figure 5.17: Surface topography for second exposure test, 540 second exposure
Figure 5.18: Surface topography for second exposure test, 720 second exposure
66
One conclusion from this test is that the intensity of the UV light is severely reduced by the time it reaches the wafer. An exposure time of 720 seconds is much longer than normal for a lithography system. A more typical exposure time would be less than 15 seconds. This is an indication that the optical system needs to be redesigned to increase its efficiency. Based upon the specifications of the photoresist, a rough estimate of the efficiency of the system may be obtained. According to the resist manufacturer, for the 1.3µm expected thickness, the required energy density to cure the resist should be
. Taking into account
the 720 second exposure time for the best result, the light intensity is somewhere around . The light field as it hits the wafer is about 4cm in diameter, which translates into approximately
=2.44 W. This means that,
of the 400 Watts required to power the bulb, only 0.6 % of that power is making it onto the wafer as UV light. Granted, this is an extremely rough calculation, but it still gives an idea of how much light is being lost between the lamp and the wafer. A second conclusion is that the system does in fact work. A pattern may be loaded into the DMD, and approximately replicated by exposing a photoresist coated wafer.
67
Chapter 6: Future Work
68
This machine is a work in progress. As it sits, the machine will produce basic patterns in a photoresist. However, in order to reach the ultimate goal of micro-manufacturing capability, some improvements are necessary. The biggest area for improvement on this machine is the optical system. First of all, the efficiency of the lens system should be improved. This will require finding a more suitable illumination lamp, as well as modifying the lens system. It is clear, from visual observation, that much light is being lost and escaping the system entirely (if not enclosed). This could be improved by modifying the light tunnel and, perhaps, obtaining larger lenses to capture more of the bulb’s energy. This will allow for shorter exposure times and improve the definition of the resulting structures. Secondly, a proper projection lens needs to be obtained, in order to allow the patterns to be scaled. As the system sits, the reproduced patterns are roughly 1:1 with their size on the DMD. If the desire is to create large circuit patterns, or smaller micro-features, the DMD image must be scalable. Lastly, the optical system could use some attention to improve the uniformity of the DMD image on the wafer surface. As previously mentioned, the tilt of the DMD means that the distance to the projection lens is different for each pixel row. It may be possible to reduce the effect of this issue by rearranging or improving upon the lens system. One possible solution is to simply increase the distance between the projection lens and the DMD, although other solutions may exist. Once these improvements are implemented, the system should be capable of manufacturing features on the micro scale. It is recommended that the machine be re69
evaluated to quantify CDmin after the improvements. Additionally, given the very high motion accuracy of the machine, it would be prudent to develop the DMD control such that scanning lithography may be implemented. This would greatly expand the machine’s capabilities.
70
References 1. "lithography." Encyclopædia Britannica. Encyclopædia Britannica Online. Encyclopædia Britannica Inc., 2012. Web. 13 Jul. 2012. . 2. Bruning, John H. "Optical Lithography … 40 Years and Holding." SPIE 6520 (2007): 652004-1-52004-13. Print. 3. N.d. Photograph. Step-n-scan Lithography. Web. July 2012. . 4. Ha, Young Myoung, Jae Won Choi, and Seok Hee Lee. "Mass Production of 3-D Microstructures Using Projection Microstereolithography." Journal of Mechanical Science and Technology 22 (2008): 514-21. Print. 5. DMD 101:Introduction to Digital Micromirror Device (DMD). Rep. Texas Instruments, July 2008. Web. July 2012. . 6. N.d. Photograph. Flatness and Roughness. Graham Optical Systems, 27 Nov. 2011. Web. July 2012. . 7. Shipley. Microposit S1800 Series Photo Resists. N.p.: Shipley, n.d. Web. Aug. 2012. .
8. ASML Announces 2011 Third Quarter Results; ASML Confirms Expectation for Record Sales Year. ASML, 12 Oct. 2011. Web. 5 Aug. 2012. .
71
Appendix
Note: Blueprint scales are not correct due to modifications required to meet The Ohio State University Graduate School’s formatting requirements. All prints have been reduced in order to fit within the mandatory margins. Any prints too large to be included in the Appendix will be filed electronically with this document.
72
73
Figure A.1: Blueprint for Side Support Plate
74
Figure A.2: Blueprint for Projection Shelf
75
Figure A.3: Blueprint for Light Shelf
76
0 .200
Figure A.4: Blueprint for DMD adapter
5
IS PROHIBITED.
4
2.675
USED ON
2.165
APPLICATION
1.336
.825
DRAWING IS THE SOLE PROPERTY OF Edition. SolidWorks Student . ANY IN PART OR AS A WHOLE ForREPRODUCTION Academic Use OFOnly. NEXT ASSY WITHOUT THE WRITTEN PERMISSION
THE INFORMATION CONTAINED IN THIS
PROPRIETARY AND CONFIDENTIAL
2.675 2.850 3.300 3.500
0 .200 .650 .825
.140 THRU x 8
R.250 TYP
3
DO NOT SCALE DRAWING
Glass Bead Blasted
FINISH
Al 6061-T6
MATERIAL
INTERPRET GEOMETRIC TOLERANCING PER:
DIMENSIONS ARE IN INCHES TOLERANCES: FRACTIONAL 0.05 ANGULAR: MACH BEND TWO PLACE DECIMAL 0.05 THREE PLACE DECIMAL 0.01
COMMENTS:
Q.A.
MFG APPR.
ENG APPR.
CHECKED
DRAWN
UNLESS OTHERWISE SPECIFIED:
3.300 3.500
DM
2
DATE
5/21/12
NAME
.188
SCALE:2:3
A WEIGHT: 1
SHEET 1 OF 1
REV
DMD Adapter SIZE DWG. NO.
TITLE:
ISOMETRIC SCALE 2:3
77
Figure A.5: Blueprint for DMD mount stationary base
78
Figure A.6: Blueprint for DMD mount slider, page 1
79
Figure A.7: Blueprint for DMD mount slider, page 2
80
Figure A.8: Blueprint for projection shelf support, right side, page 1
81
Figure A.9: Blueprint for projection shelf support, right side, page 2
82
Figure A.10: Blueprint for projection shelf support, left side, page 1
83
Figure A.11: Blueprint for projection shelf support, left side, page 2
84
Figure A.12: Blueprint for y-axis adapter baseplate
85
Figure A.13: Blueprint for main (front) vertical plate, page 1
86
Figure A.14: Blueprint for main (front) vertical plate, page 2
87
Figure A.15: Blueprint for projection lens adapter