PPT -- Time for a Reality Check?
While higher chemical purities are unavoidable, some short-term solutions are worth considering.
Alexander E. Braun, Associate Editor -- Semiconductor International, 6/1/1998
As the progress toward smaller linewidths and higher complexity continues, industry chemical and material purity and contamination requirements are becoming increasingly exacting. Chemical purity specs have risen from parts per million (ppm) to parts per billion (ppb), and now suppliers must provide parts per trillion (ppt) levels -- sometimes even parts per quadrillion (ppq). It is no better for materials. A 4N5 (99.995%) purity for elements such as those used in sputtering targets and other applications is considered adequate for most current applications, but it is theorized that 6N (99.9999%) may be needed for 0.18 µm (180 nm).
After interviewing several analytical chemists, fab process engineers and chemical suppliers -- all contamination experts -- under the promise of anonymity, an assessment of the real need for such exorbitant purity levels seems in order.
Undeniably, we are producing smaller, more complex architectures. Undeniably, minor impurities unnoticeable as little as five years ago can interfere, sometimes catastrophically, with wafer processing and device operation. Undeniably, higher purity seems the obvious solution.
However, current technology cannot meet these elevated purity requirements at industrial levels for anything but commodity chemicals and materials (some state nothing can be reliably measured at these levels with any degree of accuracy). The means to achieve and sustain these levels for specialty chemicals or exotic materials either do not exist or are still new, costly and eccentric in results.
Referring to materials, an engineer remarked, "'Cleaner is better' is not always right. Materials can be done at a 6N or 7N level, but are you getting value for twice the cost?" Techniques for volume lower-purity chemicals and materials testing are standardized. "Analytical equipment and techniques are not fine-tuned or standardized enough to detect consistently at 6N and ppt -- discrepancies are unavoidable. Test and detection technologies need to progress before working with 6N, much less 7N. Also, 5N materials have not run their course yet, as with titanium. Much remains to be done at 5N before looking at 6N."
|
|
Doing metrology at these levels is nightmarish and costly. The technology that guarantees 100% silicon atoms or extremely precise mixtures does not exist. As an analytical chemist put it, "One supplier makes a component, another a second and yet another a third. Each ships his product -- which he guarantees free from contaminants -- to the client fab, which then blends them to prepare a detergent, for example. The fab then sends a small sample to another company, which mixes it in their system and cleans some parts. Then they package the parts and send them to me to analyze. After three days of calibrating my equipment I can get ppt readings; but since I am not allowed to do enough sampling, when I tell someone, 'You have 40 ppt of chlorine,' it's a joke. It might be 300 or 400 ppt. If I told the customer we're unsure about the numbers' validity, he'd go to another lab but not be better off."
Few companies spend the money to analyze sufficient samples to enable a lab to establish the scatter in the numbers. Even then, if 40 ppt of chlorine are detected, where did it come from? Is it the suppliers' fault? Did the user mishandle the materials? Was the sample in a dirty bottle?
It is no secret that many a rejected die would work well if put in a lead frame package. As a process engineer said, "The question is, 'Will an extra chlorine atom affect device performance?' Maybe, to a certain degree, but we don't know what that degree is. We may have had one failure in the field, possibly traceable to chlorine, so we dump the wafer and up to ppb. Now, because we're big and you want our business, you must do better -- give us ppt. We're avoiding what we cannot prove."
Even if verifiable ppt purity is attained, it cannot be maintained indefinitely. At these levels, chemistries can become unstable, measurements are unreliable and transportation and handling are onerous. Very often, by the time an ultrapure chemical gets to point of use, it can be an order of magnitude more contaminated than when it came out. This can be true for elements, such as copper or gold. How do you get it truly pure, know that you have it and keep it pure, particularly given that the equipment in which the processes are run is nowhere near those contamination levels. A principal ppt driver is a "the purer the better" philosophy pertaining to longer bath life. However, after a few lots go through, metal levels rise to ppb or worse. If the only wafers exposed to ppt chemicals are the monitor wafers, what is the point?
Until specialty chemical raw materials are produced in volume, metrology advances, and handling and storing techniques are refined, it seems logical to investigate options. Some impurities are acceptable if they do not get into the dielectric and the oxides. If a solution's metal can be deactivated, that provides a short-term alternative to the ppt quandary. The move to ppt is unavoidable, but it can be moderated until the technology catches up. If the chemistry is done right, "dirty" chemicals can be used.
After all, the goal is not purity but working wafers.
