Maintaining Moore’s Law without Silicon
At the International Electron Device meeting in San Francisco last week, MIT scientists presented research that revealed a possible silicon-free future for electronics. The team showed that a transistor made of a compound semiconductor called indium gallium arsenide could operate more than two times faster than a silicon transistor of the same size. The findings could keep Moore’s Law alive after silicon has reached its limit, says Jesús del Alamo, professor of electrical engineering and lead researcher on the project.
First postulated by Gordon Moore, of Intel, Moore’s Law holds that the number of transistors on a chip will double every two years. “There are some people who believe they can get more mileage out of silicon,” says del Alamo. “But there are other people who believe that [pushing silicon] looks difficult, and the better approach is to look for different material with much better properties.”
One problem with silicon, he says, is that it is not the best-performing semiconductor for making transistors: electrons in silicon move slowly compared with those in compound semiconductors. And another issue involves a transistor component called a “gate,” a sort of switch that controls the flow of electrons. On silicon transistors, gates are made of an insulating material called silicon dioxide that gets thinner with each generation of microprocessor. As the silicon-dioxide layer thins, it loses its insulating properties, and electrons tunnel through it, resulting in power dissipation and excess heat.
“Moore’s Law is predicated on the notion that, as transistors get smaller, they also get better,” says del Alamo. “But what we see coming ahead with conventional silicon design is that, as you make [transistors] smaller, they’re not getting better; they’re getting worse.”
Compound semiconductors are, for del Alamo, an attractive alternative because “all things being equal, they allow you to get more performance for the same physical dimensions and voltage.” Essentially, he says, as the device shrinks, the performance loss of transistors made of compound semiconductors is less severe than those of silicon because there was so much more performance to begin with.
The MIT team’s most recent result is a transistor that replaces silicon with indium gallium arsenide, and the gate material, silicon dioxide, with indium aluminum arsenide. Crucially, the length of the gate is 60 nanometers, which is comparable to today’s mass-produced silicon chips (with gate dimensions of 65 nanometers). The new transistors can carry 2.5 times more electrical current than the comparable silicon transistors, translating into faster operation.
Using these materials is not entirely unique, del Alamo says. A number of other universities–including University of Illinois at Urbana-Champaign; University of California, Santa Barbara; Purdue University; and University of Texas at Austin–are looking at similar approaches. However, he adds, his group “did a lot of things right” in order to get such high performance at such small dimensions. The major advance, he says, is that his group developed a technology to thin down the insulating material, indium aluminum arsenide, to the point that it is extremely thin but still avoids leaking much electrical current.
These initial results are impressive, says Milton Feng, professor of electrical and computer engineering at University of Illinois at Urbana-Champaign. “Professor del Alamo has provided the first experimental evidence that a 60-nanometer indium-gallium-arsenide [transistor] can outperform a 65-nanometer silicon [device].” He adds that the work has “significant implications,” and he indicates that compound semiconductors could be a good solution to some of the problems the chip industry faces as device dimensions shrink.
But this speedy transistor is still just a demonstration of the potential of compound semiconductors, and it’s far from overtaking silicon on the manufacturing line. While faster than today’s transistors, compound semiconductor transistors will only be economically feasible when silicon gate lengths are roughly 20 nanometers, near its physical limit. Therefore, to really extend Moore’s Law, the compound semiconductors need to be fabricated with gate lengths of at least 20 nanometers. One of the hurdles is to find an appropriate gate material that works well at the smaller dimensions, as indium aluminum arsenide may not be the best choice.
Additionally, the microprocessor industry has spent decades and billions of dollars perfecting silicon manufacturing. A complete materials transition would be horrifically costly. Some researchers are investigating ways to smoothly integrate compound semiconductors into this process by using standard silicon wafers as a substrate and building the new chips on top of them. But silicon and compound semiconductors don’t easily stack on top of each other, so the challenge is to find a buffer material to go in between them.
More and more researchers are beginning to seriously look at these problems, says del Alamo, whose research is funded in part by Intel (see “Beyond Silicon”). He expects that as the ranks grow, many issues will be resolved. Del Alamo estimates that he could have a working prototype for a 20-nanometer device in two to three years. When this happens, he says, it will show that materials other than silicon could have merit in the microprocessor fabrication plants. “If the promise is there, and there’s a chance to continue Moore’s Law, then the industry will be able to work on these [manufacturing] problems and solve them.”
Geoffrey Hinton tells us why he’s now scared of the tech he helped build
“I have suddenly switched my views on whether these things are going to be more intelligent than us.”
ChatGPT is going to change education, not destroy it
The narrative around cheating students doesn’t tell the whole story. Meet the teachers who think generative AI could actually make learning better.
Meet the people who use Notion to plan their whole lives
The workplace tool’s appeal extends far beyond organizing work projects. Many users find it’s just as useful for managing their free time.
Learning to code isn’t enough
Historically, learn-to-code efforts have provided opportunities for the few, but new efforts are aiming to be inclusive.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.