The future in a silicon-crystal ball

For Ball Semiconductor Inc., the name says it all. The Allen, TX-based start-up plans to revolutionize semiconductors by going round instead of flat.

"Basically, we're trying to get an IC on a sphere," says Ram Ramamurthi, vice president for R&D operations at Ball.

The company, founded in 1996 by former Texas Instruments executives Akira Ishikawa and Hideshi Nakano, aims to have a process in place by 2000 that will put integrated circuits on silicon spheres of one millimeter in diameter or less. To prove that circuitry can be built on a silicon-crystal sphere, Ball fabricated a five-micron transistor on such a surface. Eventually, the company hopes to create specialized spheres covered by circuitry. Systems would be built by cobbling spheres to each other like so many snap-together building blocks. The company is still developing an interconnect and packaging system to enable this.

Ball won't reveal who is backing the company, although Japan's Mitsui Hi-tec, an IC assembly company, invested $26 million. The rest of the initial $52-million funding came from unnamed Asian sources, according to information on Ball's Web site.

The company's backers say that going spherical will lower capital spending for semiconductor plants, speed up manufacturing cycle time, and enable more flexible and smaller electronics products. Capital spending would be less because the spheres are smaller than today's 200mm-diameter wafers. That means smaller, and presumably less expensive, manufacturing tools.

Clean spheres

And because the spheres are continuously transported from tool to tool in clean tubes, they won't need a large and expensive clean room. Faster cycle time results from continuous processing and easier system assembly. Today, circuits must be fabricated on a wafer, cut free and then assembled into chips. The spheres will connect to each other with fewer steps. The snap-together nature of the spheres also leads to more product flexibility and miniaturization. Whole circuit boards could be reduced to a spherical cluster.

Although the concept is interesting, it's far from commercial viability, say analysts. Big questions remain about most aspects, including how to mass produce the sphere and change existing process equipment to handle it.

"I don't see this as being a technology that's going to be in mainstream manufacturing anytime soon," says Dan Rose, president of Rose Associates, an electronic materials consulting firm in Los Altos, CA. "I think it's going to take quite a while to evolve this technology into the level of sophistication that current flat-wafer technology is at, assuming it ever does. But certainly, it's a very interesting approach and one that's well worth exploring a little bit to see if it will go anywhere."

Once developed, Ball hopes to license the technology to traditional semiconductor manufacturers and other companies. The spherical nature of the devices lends itself to certain applications. For instance, Nakano envisions a spherical device that, after being swallowed by a patient, would transmit information about the patient's body to his doctor. But observers don't see these uses becoming commonplace.

"I don't see it replacing existing devices. I think that it will supplement existing devices," says George A. Lee, a partner in Lee.Horwath and Associates, a Portola Valley, CA, consultancy.