Will Materials Derail Moore’s Law?


Is Moore’s Law slowing down? Clearly, chipmakers are struggling to keep up with Moore’s Law these days. But one sometimes forgotten and critical technology could easily derail Moore’s Law–materials.

In fact, the cost and complexity for electronic materials are increasing at each node. “Chemical and gas commodity procurement spends are growing rapidly due to process complexity and unique chemicals,” warned Tim Hendry, vice president of Intel’s Technology and Manufacturing Group and director of the company’s Fab Materials unit, in a recent presentation. “Process and material complexity (are) driving total cost higher, so what can be done to bend the curve? There is no silver bullet. But there are some areas (the industry) can work jointly together.”

IBM, GlobalFoundries, Samsung and TSMC face the same challenges. So what are fab materials managers up against at the 16nm/14nm node and beyond?

In the front-end-of-the-line (FEOL) process flow, for example, chipmakers are moving towards new and more complex patterning films, resists and high-k/metal-gate stacks. Deposition, etch, ion implantation and wafer cleaning also involve complex materials. Meanwhile, in the back-end-of-the-line (BEOL), IC makers must contend with new interconnect materials and low-k schemes. The chemical mechanical polishing (CMP) process is becoming more complex and expensive in both the FEOL and BEOL. And let’s not forget a new wave of specialty gases hitting the market.

IC packaging has its own set of complex materials. If that isn’t enough, materials managers must worry about a growing list of other technologies–particle filtration; purification methods; materials blending; metrology/process control methods; and the supply chain.

Chipmakers and the materials industry must look at new and different ways to deal with the challenges. According to Intel’s Hendry, here are a few solutions to the problem:

*Affordable cost models. The days of developing new materials and gases at any cost are over. And chipmakers would prefer not to remove the cost of a material during the mass production phase. That’s time consuming and expensive. So going forward, materials vendors must focus on one buzzword —“design for cost.” In other words, the industry must focus on developing affordable and production-worthy materials much earlier in the process, possibly even in the R&D phase.

*Making faster decisions. Chipmakers and suppliers can no longer afford to haggle or negotiate for a year or longer on a given material or transaction. Time to market is critical. Negotiations and decision making processes should be compressed, possibly down to weeks or a couple of months.

*Getting out products faster. At one time, materials took forever to develop. Product delays were tolerated, at least to a degree. Going forward, time to market is the rule, not the exception. Here’s one way to deal with the issue– Materials suppliers may need to consider developing small-scale/prototype manufacturing capabilities as a means to speed up product development.

*More collaboration. Chipmakers and suppliers can’t afford to live in silos. Collaboration is an overused term. But there needs to be an earlier engagement between vendors. It should take place in the early stages of R&D.

*Better supply chain management. Logistics is no longer an afterthought. To support a global manufacturing network, chipmakers are looking for on-site materials manufacturing and regional sourcing. Capacity planning is also key.

*Process control. Chipmakers are reluctant to open up their purse strings when it comes to process control and metrology. That type of thinking is short-sighted. Process control/metrology are required in the supply chain to ensure quality.

Those are just some of the challenges for materials managers. So what does this mean for Moore’s Law? Based on one set of logic, Moore’s Law is indeed slowing down. Chipmakers, including Intel, are not keeping up with the traditional two-year process technology cadence. What this implies is that the number of transistors in a chip cannot possibly double approximately every two years, as defined by Moore’s Law.

Intel, on the other hand, looks at it from another perspective these days. The company insists it is keeping up with Moore’s Law, simply because it can maintain the traditional cost-per-transistor curve. Since the 90nm node, the cost-per-transistor has fallen about 29% at each node. This, in turn, drives down chip prices and makes electronic products more affordable.

At 14nm, though, Intel saw its overall wafer costs increase. But thanks to the finFET, Intel is able to offset higher wafer costs with more transistor density. In materials, though, the trends are going in the wrong direction. And if the trends continue, materials could be the problem, and not the solution. And this oft-forgotten technology could ultimately undo Moore’s Law.