Spherical semiconductors may become reality
08/01/1998
Spherical semiconductors may become reality
Ball Semiconductor Inc., Allen, TX, is on target to manufacture semiconductors on 1-mm spheres and commercially produce them by 2000, according to executives. The company claims this approach has the potential to reduce manufacturing costs and reduce production time compared to traditional chip manufacturing.
Spherical semiconductors are arguably the most radical idea in IC manufacturing today. However, two former Texas Instruments (TI) engineers have toyed with the idea for the past three decades. Silicon engineer Akira Ishikawa observed that granular polysilicon under a gas phase was falling into tiny shapes. He wondered why the silicon should be remelted into a rod and then sliced into wafers. Only today, with the huge capital investment and technological complications involved with ramping to 300- and 400-mm wafers, did Ishikawa, now president and CEO of Ball Semiconductor, believe the time was right to research the feasibility of ball semiconductors.
The concept behind ball semiconductors is to refine small polycrystal granules into single-crystal silicon balls. The balls are kept in constant motion through hermetically sealed pipes and tubes where the IC manufacturing steps of grinding, polishing, cleaning, drying, diffusion, film deposition, wet and dry etching, coating, and exposing are conducted. The spheres are exposed to air only during photolithography.
Ball`s approach to system integration will be to produce spheres with different functions - memory, logic, power, etc. - in a uniform process and then cluster the spheres needed to build a system. Ball executives say this approach will cost millions of dollars instead of the billions needed for chip investment, and take approximately five days cycle time compared to the conventional 100 days. Spheres can be produced at a rate of 2500/second, the equivalent of 20,000 wafers/month (see table).
Further cost benefits are achieved with the elimination of cleanrooms, as the balls are manufactured in clean tubes. "No matter how many meters of tubes you use, it isn`t going to cost a lot," said Hideshi Nakano, executive VP and chief operating officer. For lithography, the ball process uses a 1-cm mirror device that drives the cost of lithography systems to below $1 million, said Nakano. "We are confident we can accomplish full development and enter mass production for about $100 million."
Ball`s researchers have made progress in six technology areas critical for producing spherical semiconductors - spherical single crystals, spherical lithography, noncontact processing, 3D VLSI design, cubic VLSI by clustering, and automation/process integration.
Nakano discussed specific achievements, "Today we are already processing diodes and P/N junctions and are moving to the transistor level. We`re also on track with the application of much higher temperatures than conventional semiconductors. We don`t have to pay attention to warpage or deformation of wafers or wafer carriers." Nakano said the higher temperature yields high turnaround and sharper process results, and saves on oxidation resources. "This is very encouraging," he said.
 |
Ball says it has conducted characterization testing of a metal-oxide semiconductor diode and a P/N junction diode built on the surface of a sphere. The diode was built using dry oxidation and thin-film deposition. As the sphere was heated to 1300?C, dry oxidation was used to grow a layer of silicon dioxide on the surface of the sphere. A CVD process was then used to deposit a thin aluminum film on the surface. The P/N junction diode was built with a diffusion process of 1200?C. During these processes, the silicon spheres floated without touching the sides of a quartz tube, which was about 2 mm in diameter. The company`s next step is to build transistors and ICs on a 1-mm sphere, noted Ishikawa.
As with any new technology, obstacles must be overcome. A significant challenge is to keep the silicon ball moving without coming in contact with any part of the 1.6-2.0 mm tube. "Fortunately, we developed our own technology to keep the ball in the middle of the tube, using fluid mechanics. In addition, we can put an electrical coolant outside the tube, yet keep the environment of the ball at a high temperature," explained Nakano.
Packaging a spherical semiconductor calls for a nontraditional solution. "Protective overcoat technology is improving year by year," noted Nakano. "We are also trying to put a coating on the top of the protective overcoating, primarily for color coding purposes for individual devices and for a mechanical buffer from the outside." Nakano questions whether the industry will accept unconventional packaging, and admits that most big Japanese consumer manufacturers are more aggressive in accepting new forms of packaging and new forms of semiconductors and are accustomed to handling different shaped semiconductors.
Backed by Japanese and Asian investors, Ball Semiconductor is on target to introduce a true engineering sample to potential customers by the end of 1999, according to Nakano. A $52 million prototype plant is planned, and the company has applied for numerous patents on processes used in ball semiconductor manufacturing.
If traditional tool manufacturers fear they may be put out of business, Nakano countered, "We are not trying to be equipment manufacturers. They have a good history of how to make a system more reliable and easy to operate. They have their own know how. We need to demonstrate that a prototype is workable."
Nakano acknowledged the reluctance of traditional manufacturers to accept new technologies. "If I wasn`t directly involved, I would be one of the 9999 people out of 10,000 who are skeptical about this idea," he admitted. "But we live in a technological world - we should demonstrate our theories and concepts and ideas. [At Ball Semiconductor], we`re not talking about just theory here, we`ve moving from mere concept to reality." - L.S.