Supply Chain Needs Paradigm Shift in How to Look at Defect Control

The future of contamination control in the next-generation supply chain for beyond 14nm-node semiconductor processes faces stringent challenges. While Moore’s Law is driving scale reduction, the industry is also facing ever-increasing process sensitivity, integration challenges of new materials and the need for unprecedented purity at process maturity.
“The supply chain needs a paradigm shift in thinking about defect control. What was just process variation for previous technology nodes can now be an excursion!” says Dr. Archita Sengupta, Intel senior GSM Technologist, leading the filtration and related supply chain contamination control program, who will discuss these challenges and possible solutions at SEMICON West 2017 on July 11 in San Francisco at Moscone Center.

There are new materials being used for the first time, and even familiar materials need to be treated with new and different specifications. Even if the needed parameters are correctly specified, there may not be an accurate way to measure those parameters under HVM conditions, at least that most material suppliers can afford. Chemicals, advanced filtration and purification, chemical delivery systems and equipment manufacturing can all be sources of wafer contamination. “The interaction between the tool and the chemicals is also increasingly important,” she notes. “All this is going to add more cost for the industry supply chain for quality control, but it will cost more in the end if we don’t proactively work together throughout the supply chain to figure out what matters to control and how!”

Stability is key

The most important thing material suppliers can do to meet customer quality demands is to maintain absolute stability of everything about their material and manufacturing process, suggests Jim Mulready, VP Global Quality Assurance, JSR Micro, who will also present at SEMICON West. “Traditional quality control, where the QC data at the end of my line only has to meet the customer’s specifications, doesn’t work,” he says, noting that the material supplier doesn’t have the same process tool, the same substrate, or the same process conditions as the customer, so the testing can’t duplicate the customer’s result. Moreover, the process sensitivity is getting tighter at every generation, with the tolerance of defects often being beyond the supplier’s ability to detect them. So, no specification can ever be precise enough to capture everything the customer really needs. “Often tightening the specs doesn’t solve the problem,” he notes. “There are plenty of examples of material that was well within spec but didn’t function properly. The problem is not inadequate specs, it’s inadequate attention to other quality tools. The spec is necessary, but not sufficient.”

“The systematic (as opposed to technical) root cause of the material problems I faced as fab materials quality manager at Intel almost always came down to a problem in stability,” says Mulready, where there was a change to the material the supplier didn’t think was important, a change in the processing that they didn’t catch, or a change in the incoming raw material that they didn’t detect. “Material suppliers have to accept that the customers’ definition of quality becomes their definition of quality, and the main rule is to make sure that a material that’s working does not change at all. Consistency is the key for the end user, so it must be for us as well. A spec alone will not measure or ensure that. It takes robust change control, process control, and incoming raw material control.”

Semiconductor makers meanwhile, need to start paying attention not just to their immediate suppliers, but also to their suppliers’ supply chain; for example, not just the resist but also the resin and even the monomers used to make it. While the material suppliers need to qualify the incoming material, and serve as a kind of safety valve between the chemical industry and the IC makers, it can be difficult for them to control the supply quality when they are a very minor customer for the commodity chemical suppliers. Those suppliers in turn may have no interest in investing in the tools needed to measure the particular properties of concern, and there may be a need for the IC customer to help inflict some pressure.