Silicon’s Long Game

The era of all-silicon substrates and copper wires may be coming to an end. Progress in the future increasingly depends on more exotic combinations of materials that are developed for specific applications. But after years of predicting the death of silicon, it appears those predictions may be premature.

That’s not always obvious, given the growing number of chemical combinations being created for various reasons. III-V materials such as gallium arsenide offer higher electron mobility, which helps alleviate RC delay, a problem that has been on the rise since 16/14nm. And there is indium phosphide for photonics and various germanium-based compounds for applications that need to withstand high heat, such as in power electronics and industrial applications. There also are a number of new compounds coming into use based on transition metal dichalcogenides, which are extremely thin and highly conductive.

Put in perspective, however, these new materials are incremental steps to extend the use of silicon. There are two key drivers for this. First, so much money and effort has been invested—and still is being invested—in the development and manufacturing of chips that the best plan is to build on what already has been proven to work. The problems at 5, 3 and 2nm are numerous and difficult to solve, but there is no reason to believe they won’t be solved. And the advantage of using existing approaches and supplementing those with new materials, rather than completely replacing them, is a half-century of refined design tools and flows, along with proven manufacturing processes and methodologies.

Bottlenecks will need to be solved everywhere to move increasing amounts of data through hardware at faster speeds than ever before. That effort will require everything from additional materials such as cobalt for interconnects and ruthenium liners to new chip architectures, including an army of processors and accelerators located throughout a system. It also will include an array of packaging solutions to shorten distances that signals need to travel, as well as new memory types and different approaches to where data is processed. But what has been developed so far still appears to be the best starting point for all of these changes, even if it looks daunting and expensive.

Second, while chips have been built with a combination of electrical engineering, chemistry and applied physics, the push into new materials is just a first step in an evolutionary progression with silicon as a starting point. Increasingly, chipmaking will include a wider range of disciplines such as biology and quantum physics. There already is work underway to store large quantities of data for long periods of time using biological structures such as DNA, which will become particularly useful in areas such as medical electronics. And as chips reach down into 5nm and beyond, chipmakers will begin dealing with quantum effects—until recently this was considered theoretical physics—which affect the speed and predictability of how electrical signals behave.

Ultimately, this will demand entirely new cross-disciplinary training, both within companies and at universities. What the most sought-after graduate degrees will look like isn’t obvious at this point, but they almost certainly will be some combination of electrical engineering, material science, chemistry, biology and physics. It’s not clear how these disciplines will need to be fused together so they are useful in the semiconductor industry, or what else is needed to round out these curriculums as machine learning and artificial intelligence begin rolling out.

What is certain, however, is that engineering and scientific teams at 3nm and beyond will require completely different mixes of skills than today. And if you thought it was hard to get analog and digital engineers to talk, wait until biologists and perhaps even philosophers (think AI and machine learning algorithms) begin entering into discussions about the best path forward. But at the base of all of these efforts, silicon will remain a fundamental building block. And that is unlikely to change anytime soon.

Related Stories
Chipmakers Look To New Materials
Silicon will be supplemented by 2D materials to extend Moore’s Law.
Ruthenium Liners Give Way To Ruthenium Lines
Materials to replace copper at 5nm and beyond are starting to come into focus.
Materials For Future Electronics
Flexible electronics, new memory types, and neuromorphic computing dominate research.