Emerging U.S. Flat Panel Display Industry Embraces Automation
U.S. flat panel display manufacturers adopt robotics and minienvironments to control contamination, increase yields and gain a foothold on a rapidly growing worldwide market.
By Sheila Galatowitsch
After doing much of the seminal work to develop flat panel display (FPD) technology, U.S. players took a back seat to Asian manufacturers supported by well-monied corporate investors and industry expertise in mass production. Today, however, the U.S. industry is on the rebound, energized by several years of government and corporate investment, the promise of new technologies, and by its own manufacturing innovations.
New U.S. manufacturing facilities–both online and planned–are marrying minienvironments with robotics for a fully automated solution to the technology`s most troubling problem: increasing the yields of the large, cumbersome, hard-to-handle glass display substrates.
Currently at third generation sizes of 550 mm by 650 mm by 1.1 mm to 0.7 mm, and getting larger and thinner, the glass substrate dictates the size of the tools, room dimensions and the final cost of a facility. It necessitates mechanical transport, since assembly line workers cannot be expected to routinely lift and move the heavy substrates.
It is also the most vulnerable component in the manufacturing process. Glass substrates are extremely susceptible to particulate and ESD contamination and yields are directly related to controlling these contaminates. Unlike semiconductor manufacturing, where wafers are diced up and contaminated sections discarded, a single flaw on a glass substrate means the entire display is lost. “It comes down to addressing that zone in which the glass is held or contained rather than addressing the global room and environment,” says Abbie Gregg of Abbie Gregg Inc. (AGI; Tempe, AZ), an engineering consulting firm for FPD startups.
Japanese companies, which manufacture the majority of the world`s FPDs for consumer markets, offset low yields with high volumes. But the fledgling U.S. companies–targeting low-volume, niche markets for the most part–must maximize yields to increase cost efficiencies and stay alive. Many other variables factor into a manufacturer`s success, and displays are ultimately judged on visual criteria, such as brightness, contrast and color quality, but according to Gregg, “the current limiter on the success of the U.S. industry is defect density.”
The FPD market
The world is hungry for thin, low-power displays with superior image quality. FPDs helped create the market for notebook computers and lightweight projectors, and are now in demand for such applications as cameras, cellular phones, military and avionics instrumentation, and medical imaging.
Worldwide demand for FPDs was $10.8 billion in 1995 and will reach $23.7 billion in 2002, predicts Stanford Resources (San Jose, CA). Active matrix liquid crystal displays (AMLCDs), the technology in highest demand, will claim $6.3 billion of the market this year. Passive LCDs will take in $3.5 billion; plasma displays, $270 million; electroluminescent displays, $100 million; and the newest technology under development, field emission displays (FEDs), may garner a few hundred thousand dollars.
More than 60 companies worldwide manufacture LCDs, says David Mentley, Stanford Resources` vice president of research. The top 10 suppliers of AMLCDs are Asian companies. Although there`s still no U.S. high-volume AMLCD manufacturing plant comparable to an Asian plant, several low-volume facilities have just started to ramp up production–Optical Imaging Systems (OIS; Northville, MI); Dpix (Palo Alto, CA); and ImageQuest Technologies (Fremont, CA). All of these companies are aiming at the military, industrial and medical display markets.
The Japanese cornered the AMLCD market by making huge investments in manufacturing and infrastructure–as much as $11 billion by some estimates. Since 1990, the U.S. government has invested approximately $100 million a year to create a U.S. FPD industry and infrastructure. The funds have come primarily from the Department of Defense and Department of Commerce. The DOD has been especially interested in building a U.S. FPD industry after deciding several years ago to replace CRTs in most weapons systems with FPDs.
Insiders predict that U.S. manufacturers will get a share of the market by focusing on new technologies and new markets–but how much of a share is uncertain. “The bottom line is that the U.S. is going to have to get into the industry in niche markets, and contamination control has to be as good or better to be successful even in these niche markets,” says Doug Sinclair, head of materials reliability and component processing at Lucent Technologies (Murray Hill, NJ). Lucent`s Sinclair also chairs the USDC`s Material Handling Task Group, which is examining contamination control as part of its focus on material handling.
Recent private sector investments have led to a groundswell of optimism for the state of the U.S. industry. Candescent Technologies Corp. (San Jose, CA), formerly known as Silicon Video Corp., this year acquired $55 million in venture capital money, bringing its total funding to $150 million. The investment Xerox is making in its FPD spin-off, Dpix, is $140 million, and Motorola is spending $100 million on its new FPD division.
In 1993, the Advanced Research Projects Agency (ARPA) and private industry created the United States Display Consortium (USDC; San Jose, CA) to help bolster the fledgling U.S. industry. “Our mission was not to create a high-volume AMLCD manufacturing base in the U.S. to compete with the Japanese,” says Michael Ciesinski, USDC`s chief executive officer. “It was to create a supplier base. It was left indeterminate at the beginning as to what kind of FPDs U.S. companies would choose to pursue. The approach was to have an understanding of the different technology options.”
The task group began its work in 1993 by performing a benchmarking study of FPD manufacturing processes in Japan. “We believed we could learn from mistakes that might have been made elsewhere. We wanted to take a fresh look at everything,” Sinclair says. A thriving U.S. supplier infrastructure, serving worldwide manufacturers, would help the U.S. industry get off the ground. “You`ve got to have an established tool and material supplier base before any kind of manufacturing can be done–high or low volume, whatever technology,” says Bob Pinnel, USDC`s chief technical officer. The first tool resulting from USDC contracts, a FPD etch system developed by Lam Research Corp. (Fremont, CA), was delivered in May.
The Japanese model
FPDs are manufactured in a process similar to that of semiconductor manufacturing, involving a series of masking and metal deposition steps. In each of these processes, particles measuring between 1 to 5 microns can create defects, such as pinholes, line shorts and pixel shorts, that ruin the displays and decrease yields. ESD is an even greater problem for FPD manufacturers, because glass acts as an insulator, charging as it moves and attracting particles.
The Japanese spare no expense to build manufacturing facilities capable of producing the high volumes and throughput necessary to meet the world`s FPD demand. A new FPD facility in Japan today, with a capacity to turn out millions of displays a year, costs $1 billion. The manufacturing process itself is performed in one long line of machines–a “clean process tunnel”–as described by one expert. Each process is surrounded by a cleanroom environment appropriate to the work, ranging from Class 1,000 to Class 10. The rooms themselves have raised floors, high ceilings and enough width to accommodate automatic guided vehicles (AGVs).
The computer-controlled AGVs have been adopted over human-propelled pushcarts to transport cassettes from one group of process tools to another. The open cassettes carry between 10 to 25 glass substrates and can weigh as much as 50 pounds. The AGVs load the cassettes onto conveyors for intrabay transport, and robots transfer the glass substrates from the cassette into the process tool. Industry experts believe yields in Japanese plants are low, though improving thanks to the use of AGVs and more stringent contamination control.
Still, Japanese manufacturing processes “were not as clean as they could have been,” according to Gregg, who is also a member of the task group. Striving to meet the 1994-95 demand for laptop displays, the Japanese companies “just brute-forced it to a large extent. They were concerned with volume and doing it fast,” Gregg says.
With no deep pockets supporting them, U.S. manufacturers have been forced to modify the Japanese model in two key ways. First, they designed flexible production lines to accommodate a variety of display sizes for the niche markets they serve, which usually require low volumes. Although the Japanese can handle any display size in demand, they typically set up their facilities for a single display size in great demand. Second, the U.S. companies turned to isolation technology for local contamination control, which reduces the cost of building a facility. The U.S. companies use sealed pods rather than open cassettes. The process tools and robotic systems have been isolated within Class 10 or Class 1 minienvironments.
Pods are opened only inside the minienvironment, where the robot automatically loads the glass substrate onto the process tool and reloads the substrate into the pod after processing. Because U.S. players cannot afford the costly AGVs, pods are transported via pushcarts or conveyorized in-line systems. These manufacturing modifications are giving U.S. companies a chance to minimize costs, increase yields and compete with the rest of the world.
Innovative U.S. AMLCD plant
Three years ago, a typical high-volume facility in Japan cost $300-$400 million to build, nearly four times what ARPA and OIS spent three years ago to build and equip a new state-of- the-art facility in the U.S. OIS is perhaps the oldest AMLCD manufacturer in the U.S. Targeting primarily the military market, it has had few competitors since its launch in 1984. Until recently, the company manufactured its displays out of a small engineering facility converted into a production plant, which had 4,000 ft2 of Class 100 and Class 10 cleanrooms with an annual capacity of 2,000 to 3,000 displays. In 1993, the company won a $48 million contract from ARPA for construction of a pilot production manufacturing testbed facility. OIS supplied the balance of the funds for the $105 million plant.
The plant was designed to accommodate the company`s 15 display geometries for commercial and military avionics, transportation and medical applications. “This facility is a teething ring for us, a beginning,” says Vincent Cannella, vice president of advanced technology planning. “We are not competing in that large consumer market–yet.”
The new 108,000 ft2 plant, which delivered its first products to customers earlier this year, has an annual capacity of 50,000 to 160,000 displays, depending on the sizes produced. The 33,000-ft2 cleanroom complex has a bay/chase layout. The nested cleanrooms include Class 10,000 service areas and chases, Class 1,000 bays for operators and material transport, and Class 1 minienvironments installed around the loading stations of the individual process machines. Additional Class 10 areas are reserved for final production steps and R&D.
The plant is currently working with glass substrate sizes of 350 mm ¥ 400 mm ¥ 1.1 mm. The substrates are contained in sealed cassette pods, which are opened only in the Class 1 minienvironments. Each minienvironment has its own particle, humidity and temperature detection tracking system. Automated docking units provide the mechanical interface of the pods to the minienvironments. The units open the pods and route tracking information to the central network. Robots transfer substrates from the pods into the minienvironments and process machines.
“The idea behind our entire process is yield preservation,” Cannella says. “Since we need extra-high yields, we chose the `ultimate in isolation` that we could find, similar to the closed SMIF process in semiconductors.” Gregg, who helped OIS with the engineering and conceptual design of the new plant, says that much of the technology had to be developed from scratch. “One thing we realized after issuing the RFP was that there wasn`t a lot out there on material handling at that time.”
Progressive System Technologies (PST; Austin, TX) designed the automated, isolation systems for each of OIS` 88 process tools. “The challenge was to build around the 350- ¥ 400-mm substrate size,” says PST president Tony DiNapoli. “The technology required an extremely high level of design and integration between the robotics, the minienvironments, the software and the tools.” The success of its OIS work has kept the company busy. PST`s FPD market revenues went from zero in 1993 to roughly half of its 1995 $10 million revenue. It has several patents pending on designs resulting from the OIS project, and is aggressively targeting the Asian market with its new products. The company has also won two USDC contracts to design standardized cassettes and manually guided vehicles.
Since the new plant is a testbed manufacturing facility with partial government funding, OIS is keeping an open dialogue with industry about its manufacturing innovations. Meanwhile, the company`s business is growing, says Cannella, and it has already purchased a major land parcel adjacent to the plant for future expansion.
Xerox AMLCD spin-off
The second major U.S. AMLCD manufacturer is Dpix, a wholly owned company that the Xerox Corp. launched earlier this year. Xerox is hoping Dpix will capitalize on a decade of FPD research by its Palo Alto Research Center. The company ultimately wants to compete with the Japanese in large commercial markets for AMLCDs–not in the notebook market, but for the nascent desktop FPD market and other markets demanding high-resolution AMLCDs. But initially, Dpix is targeting the military, avionics and medical imaging markets and has already shipped its first prototype, a 5- ¥ 5-in. military avionics display.
Dpix is using a converted development lab for its manufacturing facility. It has a capacity of 30,000 plate starts a year for 360 mm ¥ 465 mm ¥ 1.1 mm glass substrates. The 60,000 ft2 facility includes a 10,000 ft2 fab with 6,500 ft2 of Class 100 or better cleanroom space. Dpix`s fabrication process uses modules that can be adapted to accommodate low-volume orders and custom products.
Asyst Technologies (Fremont, CA), designed and built four new products for Dpix: a cassette that holds 12 substrates; a pod enclosing the cassette; a standard mechanical interface that opens the pod and exposes the substrates; and a manual transport cart. The battery-powered, motor-driven cart pulls two pods and mechanically lifts the 30-pound payload onto itself. The cart transports pods from tool to tool, and robots load and unload the substrates from the cassettes to the tools. This was Asyst`s first FPD project, according to Ray Martin, director of advanced technology development.
Minienvironments enclose certain pieces of equipment that do not integrate with the pods. “In the end, something like a minienvironment is critical to this type of conversion,” says Bob Bortfeld, vice president of operations for Dpix. Operations managers anticipate building a new facility when the company moves to the next-generation glass substrate sizes.
Four-year-old ImageQuest Technologies, a U.S.-based subsidiary of Hyundai Electronics America, plans to migrate to minienvironments as it ramps up production of AMLCDs for the military, avionics and medical imaging and display markets. The company`s 35,000 ft2 manufacturing facility, with 10,000 ft2 of Class 100/Class 10 cleanrooms, has a capacity of about 10,000 displays per year.
Other manufacturers
In addition to AMLCDs, there are other FPD technologies that serve different applications, as well as new technologies in development that may compete with AMLCDs. There are no winner or loser technologies, says USDC`s Pinnel. “The question is the size of the pie.”
Standish Industries (Lake Mills, WI) has been manufacturing high volumes of low-cost passive LCDs since the 1970s. These FPDs are used in industrial applications, such as the gas station pump market, and in agriculture equipment and instrumentation applications. “Our forte is high-end passive matrix displays used in environmentally tough situations,” says vice president of operations Jerry Van Auken. Standish manufactures approximately 250,000 custom displays per month that range from sub-square-inch sizes up to a 13- ¥ 13-in. single display. The glass substrate sizes are 360 mm ¥ 360 mm, and 360 mm ¥ 365 mm, with thicknesses of 1.1 mm to 0.7 mm.
Standish`s 60,000 ft2 facility has approximately 15,000 ft2 of cleanroom space. The rooms were originally designed for Class 10,000, but the company has modified the design to meet Class 5,000. Glass is exposed only under Class 100 conditions, and minienvironments are used for the entire process. Glass is moved via an in-line conveyorized system. The company continually measures particle counts within the minienvironments and the room itself with equipment from Particle Measuring Systems (Boulder, CO).
Planar Systems Inc. (Beaverton, OR) is the leading manufacturer of electroluminescent (EL) and active matrix EL (AMEL) FPD technology, producing approximately 150,000 displays yearly. “Our process is much simpler and has fewer steps than AMLCD manufacture,” says John Laney, director of display process technology. Planar`s process is performed in several different cleanrooms totaling 15,000 ft2 ranging from Class 100 to Class 100,000.
Planar is currently working with 195 mm ¥ 265 mm ¥ 1.1 mm substrates, transporting them via process totes and racks throughout the facility. The company does not use minienvironments and only a small number of conveyors, but is looking closely at minienvironments and robotic handling to update its manual processes, Laney says. Plans are on the drawing board for a new facility and process line, but no date has been set for construction. The company is also planning to migrate to larger substrate sizes.
FED startups
Much of the excitement within the U.S. industry surrounds the potential of FEDs to compete with or leapfrog the dominant AMLCDs. The five-year-old startup Candescent Technologies Corp. plans to introduce a full-size color, video-rate notebook computer display based on its FED technology, called Thin CRTs, in 1998. According to David Bergeron, vice president of manufacturing technology, the company is in a scale-up engineering mode, simultaneously developing its first products and manufacturing processes. Pilot production of the color notebook displays is scheduled to begin late 1997, with production volumes ramping up to one million units yearly in 1998.
Twenty percent of Candescent`s manufacturing processes will be new, Bergeron says, with the rest coming from existing tooling and equipment technology. The Thin CRT displays use the same photolithography, thin film and etch systems as AMLCD, but use fewer of them, and are less sensitive to defects in some processes.
One-third of the company`s manufacturing facility will be equipped, demonstrated, and the processes proven before replicating the set-up in the remainder of the plant. “Part of what we are doing in the initial manufacturing stage is to determine what percentage of the fab has to be Class 10 or better, and where we can move to a minienvironment with pods or conveyors or a monorail system,” Bergeron says. Costs will play an important part in the decision, says Curt Ward, director for advanced manufacturing. “If I can run the cleanroom at Class 1000 with minienvironments, the costs will be less than running the cleanroom at Class 1. We have to reach the lowest operating cost and not impact our yields,” Ward says.
The current 70,000 ft2 prototyping and development facility has a 21,000 ft2 Class 100 cleanroom operating at Class 10 specifications, Ward says. This month the company will begin processing 320 mm ¥ 340 mm substrates to produce 5-in. prototypes. With an eye toward the eventual one-million display ramp-up, the company is looking to buy or lease a 200,000 ft2 manufacturing plant in the San Jose area. Half of that space will include cleanrooms.
Contamination control suppliers will factor into the success of the industry, especially as substrates get larger and more transistors are packed per square inch for greater resolutions. Smaller transistors mean smaller line widths and smaller particle sizes of concern, necessitating more stringent environments.
Lepco Inc. (Houston, TX) has already had sales in the FPD market. The facilities engineering and construction firm built the process utility infrastructure and cleanrooms for OIS, Candescent Technologies Corp. and Planar. Daw Technologies (Salt Lake City, UT) has recently completed a turnkey 43,000-ft2 FPD facility for Prime View Intl., located in the Science Based Industrial Park in Hsinchu, Taiwan. The company is now bidding on construction of two other U.S. FPD facilities, says Marketing Director Ryan Young. “FPD is a very small percentage of our business now, but we see it growing substantially over the next few years,” Young says.
Other companies targeting the market include Dexon Manufacturing (Rush City, MN), a fume hood, workstation and portable cleanroom supplier. Customized portable cleanrooms have been the company`s biggest seller to the FPD market, says Darik Nelson, sales engineer. Ion Systems (Berkeley, CA), a manufacturer of air/nitrogen ionizers and instrumentation for controlling static charge and particle contamination, has been selling into the global market for 10 years and is now seeing more sales in the U.S., according to Clark Taylor, director of marketing. The company provided cleanroom and minienvironment ionization for the OIS facility.
There`s still a lot to learn about manufacturing FPDs. Gregg says that the industry needs methodologies for accurately measuring glass cleanliness and identifying processes contributing to contamination, and that will require a thorough understanding of the material handling issues related to glass substrates. “The companies that can provide both of those solutions–contamination control within material handling constraints–will do the best,” says Gregg.
“We need a commitment from the whole contamination control industry to supply into the flat panel business,” says Dpix`s Bortfeld. “We are a smaller business than the IC business, but we have our own unique needs. We can`t just take something the IC industry uses and transfer it over. Equipment suppliers need to set up their own capability for understanding and characterizing the cleanliness–just as they have done for IC.”
Contamination control suppliers who target the FPD industry, however, must set their sights on the global market to be successful. “If some of the technologies being worked on now pan out and niche markets pan out, there would be a large U.S. market, but vendors have got to focus on the international market as well,” says Sinclair. “To concentrate just on the U.S. market would be a big mistake.” n
 |
In an effort to preserve yields of its AMLCD flat panel displays, OIS installed Class 1 minienvironments around the loading stations of individual process machines at its new facility. Each minienvironment has its own particle, humidity and temperature detection tracking system.
 |
Electroluminescent (EL) displays are used primarily for applications where image quality and viewing characteristics are crucial. Planar supplies to the medical, instrumentation, industrial, defense and transportation markets.
 |
Dpix, a flat panel display manufacturer launched by Xerox earlier this year, hopes to compete in the market for desktop displays. It is currently demonstrating an active matrix liquid crystal display with an ultra-high resolution of 7 million pixels.
 |
Candescent Technologies Corp., currently in a scale-up engineering mode, plans to ramp up production of its FED displays to one million units yearly in 1998.
Semiconductor, Flat-Panel Display Industries Comparable
The FPD industry draws heavily from the semiconductor industry–both in equipment and processes, as well as people. That`s because the two industries compare in several ways, according to Doug Sinclair, who chairs the United States Display Consortium`s (USDC) Material Handling Task Group. Sinclair lists the similarities, and differences, between the two industries.
With a few exceptions, the same contamination-sensing control strategies work for both manufacturing methodologies, and that includes particle and gas sensing. Methods to minimize particle formation and deposition in the process tool are also similar. A few processes, such as diffusion and oxidation, used extensively in semiconductor manufacturing are less important or not used at all in FPDs. Other processes, such as photolithography, are less critical in FPD processing.
Models developed for particle formation, transport and deposition in various process tools will apply with minor geometry-related modifications for FPDs. The models have been developed by researchers at Sandia, the University of Illinois, and the University of California, Berkeley.
Many wet cleaning techniques for wafers will also work for cleaning glass at various stages of processing, with the difference being in simply handling the larger substrates.
Both industries appear to be headed toward the use of closed cassette, pods or minienvironment manufacturing. If open cassettes are used, both industries will have to have a Class 1 cleanroom. If closed, the room can be Class 100 or Class 1000. FPD manufacturing will also likely use more single substrate processing, moving the substrates on conveyors, which would have to be Class 1.
The differences between the two industries are more interesting, Sinclair says. They begin with the sizes of the substrates, which create a significant difference in material handling and contamination control. Silicon is now 200 mm moving to 300 mm. The leading-edge FPD substrates are 550 mm x 650 mm x 1.1 mm. Future generations of cassettes for FPD transport will weigh 100 pounds–an order of magnitude greater than a silicon cassette. The size, on the order of a meter cubed, dictates material handling and fab design. While semiconductor fab designers are concerned about minimizing the footprint of all process tools, FPD fab designers are starting to think in terms of “volume reduction.
The particulate sizes of concern are also vastly different. The semiconductor industry is worried about 0.10 micron and even smaller, while in the FPD world, 1 micron is the minimum level of concern for most process steps. And because glass is a good insulator, ESD poses a much more critical threat to FPD manufacturing than in semiconductor manufacturing. Glass will charge naturally in a laminar flow environment, leading to particle deposition, microprocessor lockup and ESD.
Another fundamental difference is that silicon wafers are round and glass substrates are square and rectangular. The corners create risks such as cracking, chipping, jamming and improper alignment, and handling edges is not as well understand as handling cylinders.
Dozens of chips are typically produced from a single silicon wafer, and if a particle falls on a wafer, only the contaminated chip is lost. FPD substrates produce from one to 12 panels, but one particle on a panel, wrecking a single pixel, may ruin the entire panel. Yield loss has a severe financial penalty in the FPD world.
The challenges of transporting a heavy piece of glass are also much different than moving a 12-in. wafer, not only in moving the glass but keeping it clean. “`Clean and large` is harder to design than `clean and small,`” Sinclair says. “This goes well beyond what silicon has had to face.”
The entire process to make a chip is typically contained in one fab. In FPD manufacture, components may come from different fabs for final assembly. For example, glass, front and back plates, and color filters may be transported to fabs thousands of miles away. The task group is looking into cassette technology that allows ultraclean transport without mechanical damage. “These are areas we are struggling with, and we don`t have an answer,” Sinclair says.
Another significant challenge to the industry is finding an alternative to light scattering methods for particle detection, the methodology used in silicon manufacturing. “Light scattering methods won`t work for glass–at least the same approach won`t work,” Sinclair says. “We need alternative technology to detect particulates on glass.” Working under a USDC contract, Display Inspection Systems (Wixom, MI) has developed a system that uses a proprietary black beam technology to inspect glass substrates for contamination and defects. The system is being shipped to OIS for beta testing and is nearing commercial availability.
“There are issues with respect to tool design where FPD manufacturing takes its cue from large glass manufacturers. Contamination control vendors selling to this industry need to start thinking about that,” says AGI`s Abbie Gregg, who also sits on the task group. “Those industries use chain drives and conveyors to move glass, and cleanliness is not that critical.”
And, Gregg says the semiconductor industry might look to the FPD industry for ways to handle larger substrates. “The size of the FPDs has given that industry challenges. The 300-mm wafer manufacturers are going to have a lot more similarities to flat-panel manufacturers. It puts them over the edge to larger substrates that cannot be ergonomically handled. The two technologies can learn from each other.”–SG