by Hank Hogan
For cleanrooms and contamination control, the push into liquid crystal on silicon (LCOS) microdisplays presents challenges that are similar to those found in IC or liquid crystal display (LCD) manufacturing. Differences arise, however, because such microdisplays have one plate of silicon and one of glass with liquid crystal material in between. It's a combination not seen in either the IC or LCD industries, and, additionally, every defect is magnified.
 |
RIGHT. A cleanroom environment and automation are key elements for high yields.
Microdisplays, as the name implies, are small displays that magnify an image 10 to 1,000 times, making them prime candidates for use in cell phones, HDTVs, desktop monitors, wearable computers and high-performance projectors.
Kopin Corp. (Taunton, MA) has spent more than a decade mastering the combination of technologies behind LCOS. The company has found that contamination control challenges, like microdisplay images, can also be enlarged.
 |
RIGHT. Microdisplay cleanrooms require modifications from standard flat-panel display equipment.
“Any contamination, whether it comes from the cleanroom or the process or for any non-uniformity in the cell gap, really gets magnified. So you have to be very, very careful in the cleanroom practice. The practice has to be equal or better than the IC [integrated circuit] practice,” says John C. C. Fan, president and CEO of Kopin. “Any little particle gets blown up into a boulder. Your eyes are very sensitive.”
In addition to Kopin, variations of LCOS are being pursued by a number of companies, including Colorado MicroDisplay Inc. (Boulder, CO), Three-Five Systems Inc. (Tempe, AZ), and the MicroDisplay Corp. (San Pablo, CA).
Keeping things clean
Fan says that the cleanroom used by Kopin is a converted, standard semiconductor cleanroom, with a top floor full of the air handling plant, a bottom floor full of infrastructure equipment, and a cleanroom in between. The layout is a bay-and-chase arrangement, with a Class 10 cleanroom. Fan notes that originally Kopin ran a Class 1000 cleanroom, but the company eventually decided that only a Class 10 would work for volume production.
Farid Durrani, vice president of operations at MicroDisplay, reports ceiling-to-floor laminar flow using standard HEPA and ULPA grade filters and a raised floor. This is done in a cleanroom that ranges from Class 10 to Class 100. He notes that MicroDisplay tries to keep the high particle area where individual devices are separated from the wafer at Class 1,000. There are two chases that are used for return air.
Rainer Kuhn, director of product marketing at Colorado MicroDisplay, says his company uses a ballroom arrangement with most of the area at Class 100. A few specific areas are Class 10.
Some of the equipment used in microdisplay manufacturing is standard IC production equipment. Measuring about a half inch diagonally, a microdisplay is the same size as an integrated circuit, and the wafers used in the LCOS approach are standard wafer sizes. Therefore, standard cassette-to-cassette equipment works, as does most of the other semiconductor manufacturing gear. Just as importantly, many of the same cleaning techniques such as chemical etches and mechanical scrubs can also be used. This ensures that the silicon surface is defect free prior to the application of the spacer and cover glass.
However, not everything in microdisplay manufacturing is standard semiconductor gear. For instance, the liquid crystal material handling equipment is not found in run-of-the-mill integrated circuit cleanrooms. Nor can microdisplay cleanrooms use standard flat-panel display equipment, which handles large panels that measure ten or more inches across in a relatively defect tolerant environment. For this reason, some microdisplay suppliers have had to make proprietary modifications to tools for both yield and contamination control reasons.
“We had equipment built to our specification so we could do this,” says Al Davis, senior director of microdisplay sales and marketing for Three-Five Systems.
However, that's not been the case for every vendor.
“We have not modified the hardware that is used to assemble the liquid crystal displays. Everything is off the shelf,” notes Kuhn of Colorado MicroDisplay.
One key element of microdisplay production is the liquid crystal material itself. All of the microdisplay vendors work with liquid crystal material suppliers to ensure that the incoming source is particle and inclusion free. Care is taken with the dispensing system so that particles are not introduced at that point.
A final area where microdisplays differ from semiconductors is the spacer and sealant steps. The spacers are attached to the glass, and then the entire assembly is glued to the silicon using an adhesive. This step is critical for particle control, because any contamination will be trapped between the two layers. While a defect may cause only a visual problem, that is enough to render a microdisplay useless. Consumers expect defect-free display technologies, and for mass-market success microdisplays have to meet that criteria.
Two hybrid processes
The basic LCOS manufacturing flow starts with silicon wafers that have CMOS circuits fabricated on them. All of the companies mentioned use foundries for this, with most refusing to specify which foundries. Only Kopin names its source, which is UMC in Taiwan. In all cases, the CMOS silicon wafers are shipped to the microdisplay companies, which complete the manufacturing process in their own cleanrooms.
 |
There are two flavors of LCOS manufacturing flows. Kopin uses a transmissive approach. That is, light shines through the silicon and the liquid crystal material. To do this, Kopin perfected a proprietary technique that enables the top 0.2 to 0.3 micron of the surface of a 150-millimeter wafer to be peeled off. That thin layer of silicon is then transferred to a glass plate. Fan of Kopin notes that by using a 0.5-micron linewidth process, some 50 percent of the surface does not have metalization or other opaque elements. That, combined with the thinness of the silicon, allows light to shine through without difficulty. According to Fan, a transmissive approach is easier optically and that's why Kopin has selected it. He notes that others have come to the same conclusion, because all laptop displays are transmissive.
 |
The contamination downside to a transmissive approach, however, is that thin layer of silicon. It must be peeled off and transferred to a glass plate without creating particles or ruining already completed electrical circuits. Fan says it took Kopin years to perfect this.
“We can do that at very high yield and very high throughput. But we're the only company in the world that can do that right now,” remarks Fan.
All the other companies use a reflective approach. Light shines through the liquid crystal material, is reflected off the silicon surface, and then comes back out through the liquid crystal material. In the reflective approach, the wafer surface is planarized and mirrored at the foundry by coating it with a metal.
No matter what method is used, however, a glass plate, complete with a conductive layer, is mated with the circuit-bearing wafers. The plate is held off of the surface with spacers that leave the final cell open. An adhesive ensures the plate and wafer stick together. Up until this point, handling is done in wafer form, but now as the manufacturing process continues individual die are scribed and broken free. From there the individual units are filled with liquid crystal material and then sealed.
The end seal is the last step where contaminants can enter the microdisplay device. The scribe and break step just before that can be a source of particles because it involves separating a continuous sheet into individual devices. This singulation step is trickier than the same operation for silicon integrated circuits, which are embedded into a single, solid wafer.
“There are some differences because it is sort of a flexible sandwich rather than a silicon piece, which can be diced easily,” observes Durrani of MicroDisplay.
Durrani notes that MicroDisplay has applied for a patent on its proprietary scribe and break solution. He also points out that vacuum and other air handling techniques are used to reduce the number of particles. The cell gap also serves as a barrier to particle entry by setting an upper size limit to the particles that can enter the cell.
That's a good thing, because as Carl Derrington, chief manufacturing officer of Three-Five Systems, notes it's impossible to clean out a particle that's trapped inside the microdisplay.
The final manufacturing step is the connection of the microdisplay to the outside world so that individual pixels can be addressed. This step uses the same techniques, and faces the same contamination control issues, that integrated circuits do during final assembly.
A composite device
Complicating this manufacturing is the composite nature of the LCOS device. These microdisplays consist of a silicon plate which has CMOS circuitry embedded in it for one electrode, a glass plate for the other electrode, and liquid crystal material between the two. Spacers hold the glass off of the silicon. Changing the voltage across the cell gap switches the liquid crystal material from one polarization state to another, or from transparent to opaque.
The cell gap is critical to final display performance, because the space strongly influences the display uniformity and the display speed. A smaller cell gap means less voltage is needed, and that leads to some substantial visual benefits.
“The response time is much quicker so the display efficiency is much better,” remarks Durrani of MicroDisplay.
MicroDisplay uses a 1-micron cell gap. Other microdisplay companies use up to a 5-micron cell gap. However, the smaller cell gap means that smaller particles in either the liquid crystal material, on the silicon or glass surface, or generated during manufacturing can cause problems. Also, the absolute tolerance on the cell gap scales with the cell gap itself. So if one device has a fifth the cell gap of another, it has five times as tight an absolute tolerance. Therefore, cost effectively manufacturing a smaller cell gap device in volume is more challenging because it requires tighter cell gap control and is susceptible to smaller particles.
Another composite complication arises from the nature of dissimilar materials used in microdisplays. The two plates will expand and contract at different rates in response to temperature changes. If this difference is large enough, it can lead to stress and generate contaminants. It can also lead to electrical failures of the silicon circuitry.
“You have to watch the glass and silicon thermal expansion coefficients, so we choose glasses that are identical or very close to the silicon,” explains Henning Stauss, director of LCD manufacturing for Three-Five Systems.
In a real sense then, LCOS microdisplay manufacturing is a blending of both semiconductor and flat panel techniques, and that requires some special expertise.
As Fan of Kopin says, “You need to know how to make LCDs and you need to know how to make ICs. And use good practice on both to turn out good microdisplays for volume manufacturing. Without that, one would have great difficulty getting consistent products.”
The magnified market
Today, Kopin ships tens of thousands of microdisplays a month, most notably to the Victor Co. of Japan for its JVC video cameras. In published statements, officials at JVC said the smaller size and lower power of the LCD microdisplay were the reason it was chosen over a more traditional cathode ray tube (CRT) approach.
Kopin is one of only a reported 30 or more companies working on microdisplays. Several technologies compete in this area. All make use of magnification to enlarge an image, and so all can be considered a form of microdisplays. LCOS is by far the smallest in terms of market penetration. The leading technology is a high-temperature polysilicon-on-quartz LCD approach favored by such large Japanese companies as Epson, Sony and Sanyo. However, the micromachined Digital Micromirror Device (DMD) from Texas Instruments (Dallas, TX) is making microdisplay inroads, according to McLaughlin Consulting Group (Menlo Park, CA) President Chuck McLaughlin.
“They've got a thriving business, and it's growing,” says McLaughlin. “That's where today's market is.”
McLaughlin sees the total microdisplay market expanding four-fold over the next five years, with most of the growth in revenue coming from projection systems and most of the growth in volume coming from camera viewfinders.
Fortunately for LCOS advocates, the other approaches labor under several disadvantages. McLaughlin notes that high-temperature polysilicon is an expensive technology, while TI's DMD method is currently limited by the 17-micron square micromachined mirror cell. For the highest HDTV resolution, 1920 x 1200 pixels, McLaughlin points out that a 17-micron cell size leads to a chip that measures 4.6 centimeters (1.8 inches) diagonally. Making a device of that size yield is not an easy task.
 |
RIGHT. Chip-sized devices can magnify images from 10 to 1000 times to make them readily visible.
LCOS, on the other hand, is potentially less costly than high-temperature polysilicon, and pixel sizes that are 12 microns or less have already been demonstrated. Therefore, LCOS may lead to less expensive displays that are of higher resolution. That is, at the moment, only a possibility. Barry Young, a principal with the market researcher DisplaySearch (Austin, TX), puts the total LCOS market today at less than $50 million. He notes that there are no large volume applications yet, but demand could explode in 2001 and beyond. That's when Internet-enabled cell phones with 2-Mbps data rates are expected to hit the market. It's thought that the display demands of such devices may push manufacturers to LCOS microdisplays.
LCOS revenue
1999 — $48 million
2000 — $80 million
2001 — $140 million
Source: DisplaySearch