The Materials Gap

When consolidation thinned the ranks of semiconductor foundries and equipment makers, materials companies figured things were about to get better. They haven’t.

There are a couple of reasons for this. First, semiconductors are now so complex and difficult to develop that a slew of innovations are required on all sides. Everyone is familiar with transistor structures, interconnects and lithography issues, but materials make this all possible. This includes everything from strain in silicon and other changes to substrates to new films that provide insulation, improve conductivity, or which make it possible to bond two chips together. Those materials are becoming more complex in their own right, and much harder to develop with increasing levels of purity.

Second, despite what appears from the outside to be a fairly orderly progression to the next nodes—although at a slower pace than in the pre-finFET world—foundries, chipmakers, and IP developers are struggling to keep up with new process technology. It’s one thing to develop another new process. It’s quite another to make it stable and still offer reasons why chipmakers should spend more money per wafer to utilize those processes.

This is why there are so many process node numbers pouring out of the commercial foundry world these days. It’s difficult to reach the next full node (7nm, 5nm, 3nm), so there are quarter- and half-steps that are far less expensive and still provide some improvement in performance or lower power. At this point, various foundries have plans for the 12nm, 11nm, 8nm, 7nm, 6nm, 5nm, 4nm, 3nm 2nm and 1nm nodes. Some are announced, others are on the drawing board. There may be some decimals in there, as well, as the numbers shrink below 3nm.

To get to those nodes requires new materials, and the specs are not always well defined because the nodes themselves are not well defined. That has caused a fair amount of behind-the-scenes finger pointing about the quality of materials being developed, which could involve a mix of gases, complex chemistry in films, or different combinations of materials within a chip. There simply is not enough time to fully vet how these materials fit into a process, and there is a lack of communication up and down the supply chain to improve that.

“We find two issues that make collaboration difficult,” said Terry Brewer, founder and CEO of Brewer Science, during a panel at SEMI’s Strategic Materials Conference in San Jose, Calif. “One, the customer wants us to answer their questions and solve their problems with minimal or no information whatsoever. If you want collaboration from the customer, you have to do a better job of collaborating yourself.”

The second problem, according to Brewer, is that materials suppliers hear about a problem after the equipment company does, even though materials need to be included in the process to be effective. “So we’re supposed to solve a problem with a particular piece of equipment that may or may not be appropriate for a process,” he said. “You could select materials and equipment together, or you can select materials first. But you need to stop this hierarchy of selecting the equipment first and then the materials supplier has to meet whatever performance is required.”

Materials often fall under the radar of the electronics industry. Unlike a chip, which has a brand name emblazoned on it, or even IP, which appears in an online catalog, materials vendors live almost entirely behind the scenes. But they play an increasingly vital role in making electronics work properly, and a mismatch on specs can cause all sorts of problems for everyone.

This is an issue that needs to be solved for the semiconductor industry to move forward. Materials vendors need to be part of discussions much earlier in the process. A spec sheet or an e-mail after the fact is no longer sufficient.