Semicon West highlights laser silicon machining
Laser processes previously thought impractical for semiconductor-scale manufacturing applications have recently been developed for commercial use, and are likely to play key roles as the semiconductor industry proceeds toward sub-100-nm feature sizes, according to Semicon West exhibitors in San Francisco and San Jose, CA, last July.
Laser processes previously thought impractical for semiconductor-scale manufacturing applications have recently been developed for commercial use, and are likely to play key roles as the semiconductor industry proceeds toward sub-100-nm feature sizes, according to Semicon West exhibitors in San Francisco and San Jose, CA, last July. Attendance projections for the annual semiconductor industry convention at press time indicated more than 1500 exhibiting companies and almost 64,000 registered attendees, about a thousand attendees above the previous year. Exhibitors at the conference tended to describe the economic climate as "cautious" or "conservative." The general mood seemed a bit more upbeat than the previous year and hints of gradual economic recovery were definitely there (see figure).
For instance, Brian Klene, senior vice president for marketing and business development at Cymer (San Diego, CA) pointed to sales of the company's krypton fluoride and argon fluoride lithography light sources as an indication that Cymer's customers are anticipating increasing demand for semiconductor lithography equipment in six months or so.
Attendance at Semicon West 2002 was similar to 2001 but the mood on the floor seemed more upbeat.
Among laser light sources presented at the convention for semiconductor manufacturing applications, however, at least two were in applications in which laser optics seem to be finally defying conventional engineering wisdom and making it into commercial application for the first time. Engineers at Ultratech Stepper (San Jose, CA) described a new laser annealing capability targeted at sub-100-nm feature sizes, drawing upon a research group the company purchased from Lawrence Livermore National Labs (Livermore, CA) seven years ago (see "Laser annealing moves into semiconductor manufacturing," p. 32).
In addition, a multinational engineering team assembled by the Irish start-up company Xsil (Dublin, Ireland) has already begun volume shipments of laser micromachining tools for precision cutting and drilling silicon with order-of-magnitude throughput improvements over conventional methods.
One focus of the Xsil technology is on drilling vias in silicon with high aspect ratios needed by memory manufacturers, where 20- to 30-µm diameter vias must penetrate wafers as thick as 800 µm, according to chief technical officer Adrian Boyle. "With conventional plasma-reactive ion-etch or wet-etch techniques, the actual throughput from that process is extremely slow and costly," he said. "But with a laser and the tool that we've developed, we have extremely high throughput—orders of magnitude higher." Laser micromachining currently allows drilling of 16,000 vias on an 8-in. wafer in about one minute.
The other focus is on dicing very thin silicon wafers. Laser micromachining has proven competitive with dicing saws for wafer thickness of 150 µm or less and the process speed increases exponentially as thickness decreases, according to Boyle. "Below 100 µm, we are factors of two to ten faster than dicing saws," he said. Before now, the idea of using a relatively messy laser process for such precision work in silicon would have seemed counterintuitive, Boyle said.
"Previously, most people would try to do this with a high-power carbon dioxide laser or infrared YAG laser. They would scan at certain speeds; there would be a buildup of heat in the wafer, which would cause damage to local devices as well as to the wafer as a whole, and actually reduce the mechanical strength of the wafer, but also reduce the electrical functionality of the wafer.
"What we've done is develop a process in which we can do this without causing thermal damage and without causing mechanical stress on the wafer. So the mechanical integrity of the parts is better than even a saw-cut part but certainly better than is possible with a run-of-the-mill laser process. Also, we are certain that the electrical functionality of, let's say, microprocessors or smart-card devices is not impacted by the laser parameters we use."
While referring to technical details of the laser process as proprietary, Boyle did say the development cycle has taken about 18 months, and a large part of that has involved working with laser manufacturers to develop design parameters to meet the performance requirements.
"At the beginning, there was a kind of mentality that one laser was the same as the next, and if you put silicon in front of that, the results you would get would be the same as the results you would get from any other laser," he said. "But in fact the laser process we've developed has 17 possible parameters that are all interdependent. So there are effectively 17 factorial variables." Consequently, a major aspect of the interaction between Xsil and its customers has to do with fine-tuning the parameters for a specific production process, he said.
The laser is actually integrated within a fully automated production tool with robotic wafer handling that, because of the flexibility of laser scribing as opposed to mechanical sawing or chemical etching, can cut streets between devices as small as 30 µm and cut arbitrary shapes and depths on the same wafer, in 6-, 8-, and 12-in. wafer sizes. The rapidly growing company commenced volume shipments of its systems last March, according to chief executive officer Peter Conlon, and was awarded the European Semiconductor New Company of the Year Award in June.