Last month, I had the opportunity to walk through the exhibits at the Semiconductor Equipment Manufacturers trade show (Semicon) in San Francisco, Calif. In some ways, it was a trip down memory lane. Many of the exhibits reminded me of just how far semiconductor processing has come—from the first planar transistors to the latest megagate ICs employing deep-submicron processes.
At the same time, I was overwhelmed by the dinner-plate-sized, 300-mm wafers beginning to make it into the latest fabrication facilities. These wafers promise to deliver close to 2 to 2.5 times the number of chips per wafer from today's 200-mm (8-in.) diameter wafers (for the same size chip). This allows companies to considerably increase their production capabilities. Of course, to use these larger wafers, the manufacturing facilities must revamp their production lines, installing new or upgrading old equipment in order to handle these bigger wafers.
In addition to the size of the wafers increasing, the cost of the equipment needed to process the wafers has skyrocketed. A new 300-mm fabrication facility that could handle over 20,000 to 40,000 wafers/month would cost almost $2 billion to build and equip. Plus, adding a second wrinkle, the use of copper metallization is growing in popularity, requiring the wafer fabs to install a second set of metallization equipment. That second set is necessary to handle the damascene copper technology and deposit the low-dielectric-constant insulators that are needed to optimize circuit performance.
It's therefore no wonder that many of today's semiconductor manufacturers have had to strike up alliances with foundries or other semiconductor manufacturers to help share the high cost of building a new fabrication facility. And it won't get any cheaper. As feature sizes drop below 0.13 µm, the key tools needed to fabricate the circuits, including the stepper lithography systems and masks, the pattern materials, and the various deposition and etch systems, promise to hit new highs in pricing. Along with all the equipment that will be necessary are all the chemicals to form resists, wash the wafers, etc. These chemicals must be ultra pure, with some impurity levels hitting lows of less than 100 parts per trillion.
The metrology required to make such measurements is astounding as well, because it must typically be able to resolve values one-half to one-tenth that of what the instruments are trying to measure. But how much further can we go? Will we be able to build equipment to create, resolve, and measure the various aspects for the next generation of wafers where dimensions shrink to less than 0.1 µm and impurity levels must be below 10 parts per trillion? Can it be done? Will the ability to measure the parameters cause us to slow down the move from one generation to the next? What do you think?