Contamination concerns dull copper`s benefits
By Hank Hogan
East Fishkill, NY — The advent of copper interconnects is blazing some trails in semiconductor performance at the same time it`s raising contamination concerns. IBM`s recent shipment of microprocessors made with copper interconnects — a PowerPC 740/750 device operating at 400 MHz — was preceded by researchers grappling with a potential “killer” type of contamination defect.
Thanks partly to copper-based processing, developed at the company`s Advanced Silicon Technology Center (East Fishkill, NY), the new device reportedly operates 25 percent faster than its aluminum-interconnect counterparts.
However, this performance increase comes with a price, part of which lies in the need for improved contamination control.
“One of the big issues with introducing copper was as the wafer moves through the line, will the copper metallization affect the front-end devices?” says Ghavam Shahidi, senior manager in R&D and also silicon-on-insulator (SOI) program manager at the center. The back end of fabrication is where metallization occurs while the front end is where transistors are created. Shahidi says IBM was so concerned about this, that initially fabrication was divided into two separate areas. Once wafers crossed the copper demarcation line, they never went back. Processing tools were, if necessary, duplicated on both sides of the line.
The contamination control concerns involve what copper does to transistor characteristics. As a metallic contaminant, copper introduces levels deep within the forbidden energy band of silicon. One result is increased electrical leakage by affected transistors, which at some level is fatal to circuit operation. The problem arises if copper should somehow find its way from the back end of fabrication to the front end.
“That is a primary concern because that will actually be a killer type of defect,” notes David Jensen, program manager for defect reduction technology at the research consortium Sematech (Austin, TX).
Making the problem all the worse is such cross-contamination may not be detected until the affected wafers have completed processing. By then an entire line`s worth of material may be impacted.
Copper processing differs from today`s aluminum standard in a number of ways. Aluminum metallization is sputtered onto wafers and then etched off. Following that, dielectrics are deposited that both insulate the metal and planarize the surface.
Copper is put down by a dual damascene process. Interconnect traces, including through holes, are etched in an oxide dielectric. After a barrier metal is put down, a copper seed layer is sputtered over the wafer. That is followed by an electrofill, where copper in solution flows over the surface, filling in the etched traces and via holes. Chemical mechanical polishing (CMP) then removes the copper everywhere except for the traces and vias. After a cleanup of extra copper, the interconnect is encapsulated in a dielectric, and the planar surface is ready for the next metallization layer.
There are a number of possible avenues for copper to end up where it shouldn`t. These include maintenance tools, operator gowning, air streams, and other mechanisms. However, investigations by Sematech indicate that airborne contamination is not a measurable problem. Separate gowns, maintenance tools, and wafer boxes take care of many of the other possibilities. This leaves one primary concern.
“The main transport mechanism that we see, or the greatest likelihood of copper transport, is in effect limited in our mind to backside transfer of contamination,” notes Jensen of Sematech.
Such contamination arises because once a chuck for something like a measurement tool is sullied by copper, the next wafer to touch the chuck can pick up the metal. Indeed, Jensen says that work at Sematech shows such contamination will take place. If metrology tools are shared by front and back end processes, cross-contamination is possible. The same is true for shared photolithographic tools.
Hence the decision by some fabrication lines to segregate copper from non-copper processing. The only question is how much separation there should be. That is something no one has a definitive answer to, and so each company is coming up with its own guidelines. Equipment vendors are looking to help the situation by eliminating the problem at the source.
As Jeff Venzang, vice president of engineering and product development of Novellus Systems Inc. (San Jose, CA), says “A properly designed copper deposition system takes measures to control backside contamination.” CR
Hank Hogan is a freelance writer based in Austin, TX. He has written for New Scientist, High Technology Careers, Electronic Components, and Multichannel News International. He was previously a semiconductor process engineer and holds a process technology patent.