Chip making gets on the ball

Instead of manufacturing ICs on flat chips, why not use silicon spheres as a substrate? The result of this manufacturing paradigm shift would be manufacturing high-performance, low-cost devices in five days with a fab that costs only $100 million. Such agile, just-in-time manufacturing plants could respond to customer demand quickly and rapidly implement new technologies.

Keeping up with Moore's Law. Today, each new generation of integrated circuits provides more functionality, at less cost, than the previous generation. To keep up this cost/performance dynamic, semiconductor manufacturers have typically placed more functions on a single chip. Moore's Law, Gordon Moore's 1965 axiom on microchip performance, stated that chips double in capacity every 18 months.

As chips increase in complexity, the processes used to create them undergo a corresponding complexity increase. The industry is approaching technical limits on the number of circuits that can be placed in a given area without generating unacceptable electrical interference between the circuits.

Because chips are typically produced in wafer lots, one way to increase productivity and lower the cost per chip, while allowing chip sizes to increase, is to increase the diameter of the wafer. For this reason, the semiconductor industry, which only recently moved from 150- to 200-mm wafers, is now moving to 300-mm (12-inch) wafers. A 300-mm wafer yields more than double the number of chips of a 200-mm wafer.

Moving to larger wafer sizes, while imperative from a productivity standpoint, will antiquate existing manufacturing technologies and facilities. Semiconductor Equipment and Materials International (SEMI), an industry group comprising makers of chip-manufacturing equipment, estimates that the transition to 12-inch wafers will cost equipment makers and materials suppliers more than $21 billion.

The process for creating flat, rectangular integrated circuits today accounts for more than $1.5 billion in manufacturing assets. Because of the strict requirements for ultra-pure air and water in the fabrication process, and the precision equipment required to construct millions of circuits in a fingernail-sized space, wafer fabrication facilities cost about $1 billion to design, construct, and equip.

If silicon wafers were replaced with 1-mm balls, then manufacturing fabs could shrink accordingly. All fabrication steps could be carried out in ultra-clean tubing, and the processed balls would transfer from one process to the next under normal atmospheric pressure conditions.

Wafers produced by this process typically have many defects. These defects can be attributed to the difficulty inherent in having a single, highly pure crystal survive the cutting, grinding, and cleaning processes, as well as impurities associated with the containers used in forming the crystals.

The large capital outlays and short product life cycles characteristic of today's industry act as a huge barrier to innovation. Inventors with innovative algorithms, or new functions, but not endowed with several hundred million dollars to invest in capital, must rely on silicon foundries to bring their ideas to the marketplace.

Spheres in motion. BALL Semiconductor is leading a revolution in the semiconductor industry with a simple idea: a 1-mm spherical semiconductor in place of today's flat, rectangular chip. This technology will enable the production of spherical semiconductors in a single, enclosed process. Using a manufacturing line of small tubes and pipes, ball semiconductors will be manufactured in a single process using proven technologies of gases, chemical reactions, and solid-state physics of semiconductor manufacturing.

This process begins with the sorting of very small polycrystal granules that have been processed into single-crystal silicon balls. Initially, the company will attempt to produce 1-mm single-crystal balls; but we also intend to develop the capability to produce smaller balls. We anticipate that the balls will be in constant motion as they are continuously processed, treated, and transported at high speed through hermetically sealed pipes and tubes during various processes for the crystal-growing, grinding, and polishing steps and then for the repeated cleaning, drying, diffusion, film deposition, wet and dry etching, coating, and exposing steps of the integrated-circuit manufacturing process. The spheres are exposed to air only in photolithography; thus, there is no need for the traditional--and expensive--clean room. BALL Semiconductor's vision is to make semiconductors on 1-mm spheres, and then decrease 0.8 mm in diameter, 0.5, 0.1, and smaller.

System integration. The ball approach to system integration could have revolutionary impact on the semiconductor industry.

The current approach to system integration calls for placing more and more functions on one chip. This approach requires enormous capital infusions every few years for new processes and equipment to accommodate more complex systems on a chip and larger wafer sizes. The ball approach is to produce spheres with different functions in a uniform process, and then cluster the spheres needed to build a system. Instead of chips becoming larger, wafers becoming larger, processes becoming more complex, and equipment becoming more expensive, ball semiconductors offer the potential of equipment and process stability across several generations of technology.