Time for
01/01/2002
 |
For years the industry has talked about putting metrology instruments into process tools, or at least hooking them up in-line, so that wafers could be monitored either during processing or right afterwards. This could prevent running a new batch of wafers on a faulty process. It might also eliminate taking one monitor wafer from a batch of 25 to the lab and putting it through some rigorous measurements to ensure that processing is within acceptable limits. The trouble with that traditional approach is that it wastes too many good test wafers and takes too long (another wafer lot could already be in the tool).
Metrology companies have run many experiments showing how their tools could be built into a process line and thus provide early detection of any processing flaws. The tool companies might have been impressed, but when they asked the chipmakers how much more they were willing to pay for tools with built-in or in-situ metrology, they found that it was not close to enough. There are a few exceptions, such as residual gas analyzers, some laser probes peering through chamber windows, and thin film analyzers to check the results of CMP, but in general, built-in metrology never really caught on.
As the industry moves to 300mm wafers, each holding potentially hundreds of thousands of dollars in chips, the traditional approach will no longer do. No fab will want to waste one test wafer out of every 25 processed, and a lot more of them during process development and yield ramp. But tool and process performance will need to be continuously monitored, somehow. Good metrology tools are expensive, and hooking them to every tool in the fab would be cost prohibitive. Because of this, there has been little demand for built-in tools, so vendors didn't waste time developing them. There's the dilemma: The industry badly needs built-in metrology to avoid catastrophic misprocessing of millions of dollars worth of wafers, but it would just cost too much to build metrology equipment into every process tool.
Now, however, the potential solution is emerging: It might be termed "quick-and-dirty metrology." It certainly would have to be much less costly than conventional metrology tools providing extensive detail. Instead, these small devices would be designed to take a quick look only at some critical aspect of the processing. They would be "dirty" because they would get right up close to the action, perhaps right inside a plasma chamber or maybe a discharge line.
Traditional metrology is often, by its nature, slow and painstaking. Process engineers can wait for the intricate data needed not just to find out that something is wrong, but to track down potential causes. That's not the case in the factory. All the operator needs to know is that some critical operation is not right. Instantly, processing can be stopped for troubleshooting. A wafer can be taken to the lab for detailed examination of film thicknesses, densities, interfaces, or particle types and distributions. Thus, the need for conventional metrology will not go away. In fact, with a new materials set and aggressive shrinks, it will be more important than ever. But a new class of quick-look selective tools is already beginning to emerge for duty right on the production line.
Let's give some examples, mostly from relatively new companies just getting into the metrology game. One tool operates somewhat like the golden model concept used with vision systems that scan printed circuit boards to ensure there are no errors. This laser-based tool can be put right into a loadlock to scan the film stacks on a small section of a wafer to ensure that their thicknesses are what they should be. Like a stepper, it can scan around to check uniformity across the wafer. It doesn't analyze densities or interfaces, or check for contaminants: that can be done in the lab if trouble is spotted. Another new idea is a charged-coupled device (CCD) array on a chip at the top of a plasma chamber. It takes a quick look at the wafer surface after processing to determine whether there are excessive particles. If not, fine. If there are too many, then troubleshooting can begin. Other tools look at plasma parameters, within a chamber or in the immediate effluent, to spot any deviations from pre-determined limits.
In the lithography arena, Dr. M. David Levenson, editor of Microlithography World, recently described an emerging class of tools to measure subtle CD changes that indicate process variations (MLW, Nov. 2001, p. 28), referring to them by the more delicate term "secondary metrology." Airborne molecular contamination will become a greater problem with 193nm and 157nm lithography, and silicon sensors with microbridges can provide ultrasensitive chemical detectors. This suggests that microelectromechanical systems (MEMS) devices might prove valuable as future monitors for specific tasks such as trace chemical detection.
Our own semiconductor technologies provide many candidates for these quick-look, fast-response devices for spotting trouble on the fly. Minimal but careful design will ensure they they won't cost a lot, but they'll do a good job of spotting trouble. With passivation and well-chosen coatings, they will last a long time, even in the most hostile environments. Opportunity beckons for quick-and-dirty metrology. Let's go for it.
Robert Haavind
Editor in Chief