#407  from Innovative Leader Volume 8, Number 6          June 1999

A “Dirty” Risk
by Klaus D. Beyer, Ph.D.

Dr. Beyer is a senior engineer at the IBM Microelectronics Division, Hopewell Junction, New York.  His inventions have led to thirty patents.

In computer chip manufacturing, a flat chip surface is essential because subsequent imaging steps would be distorted if the chip would have even tiny bumps.  I discovered a process to achieve that flat surface.  All the major silicon chip manufacturers now use that process, whose invention couldn’t have been planned.

A silicon wafer is produced by growing a cylinder of crystalline silicon from a melted silicon solution.  This cylinder is then cut into thin slices, or wafers.  A mechanical grinding process removes the saw damage of the silicon slice. The final wafer is then polished.  After processing, in order to form the many semiconductors that become the “working parts,” the wafer is cut into small chips, the heart of your computer.

The polishing step involves abrasion through mechanical action.  That’s a technology that’s been around since ancient humans made jewelry.  But this kind of activity releases particles. As you may know, processing of silicon chips requires ultra-clean environments, since a speck of dust can interfere with the chip’s operating ability.  You wouldn’t want such a dirty chip in your computer.  Therefore, after the wafer is polished, a final brush cleaning results in a perfect shiny surface.

In 1981, I was asked to solve a problem in the silicon wafer manufacturing facility.  The brush that was used to clean silicon wafers frequently became clogged with particles, and ended up scratching the wafer.  It looked like an impossible assignment, and I had no idea of what to do.

While browsing in the library, I stumbled upon a five-line note in a patent review that described a megasonic cleaning machine that uses ultra-high-frequency vibrations to remove particles when the vibrations are directed parallel to a surface.  I wondered if this machine might be useful to remove particles in the wafer production process.  Perhaps I could use the vibration technique, after polishing, to shear off particles from the wafer surface into a cleaning solution that would then carry the particles away.

Not an Innovation-Fostering Environment

Before I had the chance to purchase and test the megasonic cleaner, I was asked to work half-time on an entirely different project.  I wasn’t too happy about this situation, having to report to two different managers.  Besides working on trying to keep particles out of polished silicon wafers, I now had the responsibility for figuring out how to fill the small scratches--we call them trenches--that are necessary to separate sections within each chip in a wafer.  I was asked to try coating the silicon surface with a thin layer of glass, which hopefully would fill the trenches and produce a smooth, flat and stress-free chip.

The reason I was assigned to the second project was that many others, using the latest techniques, tried in vain to obtain a stress-free chip trench process.  So I was asked to try an old technique to deposit glass over the etched silicon.  This method involved centrifuging a suspension of tiny glass particles onto the wafer, and then melting the glass particles together.  Unfortunately, the trenches didn’t fill in a predictable way and the wafer ended up with an irregular bulging glass surface. 

With two half-time jobs, I knew that, together, they would add up to much more than “full-time.”  I was assigned to a small lab attached to the development laboratory because working with small glass particles--a potential dust-producing step--kept me away from the facilities dedicated to clean chip processing. 

I was also a half-mile away from the wafer manufacturing area, where some of the equipment (including the polisher and the megasonic cleaner) was located. My working situation certainly didn’t seem conducive to innovation. What I didn’t know then, was that one assignment would help to solve the other.

The first project went very well, as the megasonic cleaner got rid of all contamination introduced during the silicon polishing step. 

But I didn’t know of any method that would create an even glass surface over the uneven patterns that are integral to the wafer.  I feared that I would fail on this project.  Out of desperation, I turned to my fresh experience with the polishing process. 

I speculated that simple polishing might smooth out the glass bulges.  Polishing, of course, produces particles, and particles are “forbidden.”  However, I was now an expert on using the megasonic cleaner and it seemed obvious to use that machine on polished glass-coated wafers.  So I took a gamble and walked the half mile to manufacturing, polished the bumpy glass surface and made some “forbidden” dust.  But I used the megasonic cleaner to remove the particles, with a procedure that I had tried many times on bare silicon wafers covered with polishing residue.

When I examined the surface with an electron microscope, I was astonished as to how flat that polished surface was.  But, I was especially surprised to see that the entire glass surface covering the silicon wafer was clean! That discovery, in early 1983, became even more important when we found that a number of materials, with properties better than glass, could fill the trenches appropriately.  In these cases, as well, the megasonic cleaner would effectively remove particles, after polishing. A few months later, a “whole army” of engineers now joined my efforts and, what was originally a side project for IBM, became mainstream. This was an important breakthrough as it led to silicon device wafers having far greater reproducibility and quality. 

Reflections

Thank goodness I was assigned two very different projects which, by fate, were complementary to each other in respect to my invention.  Could this particular smooth surface-generating invention have been planned?  Probably not.  If it would have been planned, it would have required the cooperation of managers and engineers, from different areas, none of whom would have felt comfortable inserting a particle-producing process into the sequence of clean chip-processing steps.  I’d bet that I would have been talked out of this “dirty” approach.

My story goes to show that sometimes an activity that “shouldn’t be done” (a dirty process near a clean site) can end up as, “Thank goodness it was done!”  I guess this is just another example of the unpredictability of innovation.  And that’s what makes work so exciting!