Ball Semiconductor Presents &nIts Vision of the Future

John Baliga, Associate Editor -- Semiconductor International, 9/1/1998

T he weekend before Semicon West '98, Ball Semiconductor (Allen Texas) held a technology conference to present its spherical silicon processing technology. The technology for creating devices on silicon spheres spans the range from crystallization of polysilicon nuggets to placing solder balls onto finished spheres. From the time the nugget is sent to the crystallizing tool, it is processed in a non-contact fashion until it is ready for the solder balls.

Five enabling technologies were identified for making spherical processing possible: spherical single crystal formation, non-contact processing, spherical lithography, three-dimensional design and clustering. To date, the company has demonstrated a high degree of capability in four of the five areas.

The polysilicon spheres are dropped vertically through an inductively coupled plasma where they melt. Cooling of the balls is controlled to yield crystalline silicon. Disco Corporation (Tokyo, Japan) has provided Ball with technology to lap and polish the balls into spheres. A company representative claims that oxygen contamination of the silicon is nearly non-existent using this method, providing a potential performance advantage.

Non-contact processing is performed mainly in tubes, where the fluid mechanics have been engineered to keep the ball from touching the sides. Spiral flow generators are used to make the radial pressure gradient needed. No-contact catchers, feeders and receivers using engineered gas flow complete the set of tools needed.

One main advantage of using silicon balls is their robustness in high-temperature processing. Wafer processing has limiting thermal budgets for a number of reasons, one being warpage. If the temperature of a wafer is brought to 1300°C, it can warp to the point where further processing of 250 nm (0.25 µm) designs would be impossible. Ball has performed oxidation and diffusion at temperatures up to 1370°C, yielding good uniformity with no noticeable change in the shape of the ball.

One of the main disadvantages of using silicon balls is the inability to implant ions. Many modern design elements, such as shallow junctions and deep wells, can be realized only on wafers using ion implantation. On silicon balls, shallow junctions may be feasible using very high temperature diffusion and very short time intervals. Whether an added usable temperature range can enable diffusion to produce complicated doping profiles is a question that has not yet been answered.

The company has demonstrated promising results for CVD polysilicon deposition and good results for nitride and TEOS depositions. Aluminum deposition results using dimethyl aluminum hydride have been successful also. Wet etching results do not show any surprises, and an atmospheric plasma etching process is under development.

Lithography is performed using a single mask and a multi-faceted mirror "cup." The cup shape approximates a paraboloid, with the ball sitting at the focus. The resist is applied in sheet form, using surface tension. The most technically involved part of the lithography process is alignment. The ball is held in place by a stage that contacts the ball in a very small area. The stage is a small, six-axis stage that uses ultrasonic waves as drivers and was developed at the Toyama Laboratory at the Tokyo University of Agriculture and Technology.

The packaging used on individual balls is simply a layer of encapsulation material referred to as paint. The ball concept calls for components to be made on the balls, not finished devices. For example, a 16-bit multiplier would be made out of three balls. One would have the control circuitry and the other two the registers and latches. The balls would be connected in a three-dimensional arrangement, and they would be color-coded to help identify the components.

At present, Ball is working with 1 mm diameter balls to develop processes. The company plans to go to smaller sizes, and has no plans to use larger ones. The results demonstrated show potential for the ball concept to capture some of the semiconductor market. Also, without the need for a cleanroom or vacuum equipment, ball processing is much cheaper than wafer processing, making it more likely to be tried by a large number of manufacturers. Whether the ball concept merely holds on to a niche, or captures a large market may be answered in the next few years.