Cerfe Labs: Spin-On Memory

Arm has spun off one of its more intriguing semiconductor research projects, a new non-volatile memory type called correlated electron materials RAM (CeRAM) that holds the potential to substantially reduce the cost of memory in everything from edge devices to high-performance computing.

Headed by two former Arm Research insiders — Eric Hennenhoefer, who will serve as CEO and Greg Yeric, who will serve as CTO — the goal of the new company, Cerfe Labs, is to explore a new type of memory and its potential applications, and also to build on work that the group has been doing into ferroelectric FETs.

“We’re focusing as a startup company on continuing the work we’ve done with Symetrix, which has a deep history in ferroelectric memory,” said Yeric, a former Arm Research fellow. “They happened upon a new type of memory that is based on a correlated electron switching mechanism. The field of strongly correlated electronics is not new. It’s been around since the 1940s, and people have been playing around with it since then. But they never cracked the code. We don’t have a direct proof yet that we have a bulk-switching correlated electron switch, but we have a lot of data that’s basically every part of the elephant minus the actual DNA test.”

Yeric points out this is still early research. However, the results are intriguing enough to continue the research that Arm has been doing with Symetrix. “If you look at the things you want out of memory — speed, power, dimensional and cost scaling — we feel strongly that it’s going to move the needle on all of those. We already have data on speed and dimensional scaling, and we’re excited about the potential. We don’t have a time for a 5nm SRAM replacement, but at the same time this is a fundamental material switch. So you can put it on a wafer with ALD, with PVD, or you can spin it on. A spin-on memory device would be very low cost at older nodes, and it’s just a one-layer depth between electrodes. But you get this thing that is nanosecond fast, low power, and it has potential across the board from edge AI to HPC.”

Fig. 1: CeRAM photos and structure. Source: Cerfe Labs

Time frames for actually producing a working version of CeRAM aren’t yet defined. But the attraction of being able to spray on memory that is low voltage is intriguing. The film itself switches in a bulk manner, said Yeric, who describes it as an electron orbital, quantum mechanical switch. “So it’s not a filament, it’s not an MRAM. It needs a very precise layer. If a film varies from 70nm thickness, you’re going to get some current variation across the wafer with variation and thickness of the film, but it will still operate, voltage-wise, and it will operate quite well. Symetrix invented this, and it’s a lot to get your head around. I’m a semiconductor person, and on page one of my textbooks in grad school we did approximations that ruled this stuff out. The correlated electron is opposite of the independent electron approximation.”

Part of the initial funding for this memory came through a DARPA program for CeRAM, which has multi-level cell capability. That also opens it up to potential neuromorphic computing applications, as well as possible use in quantum computing in extreme cold, and in high-temperature applications.

How this ultimately rolls out into the semiconductor industry remains to be seen. Hennenhoefer, who until recently was vice president of research at Arm, stressed this is a research effort. “We have the team, we have all the IP and contracts, and we have enough funding to get out there and get this thing started. Arm’s on board, and coming from the research side at Arm, we always liked this memory from the standpoint of what we coud do with it one day. The world needs a better memory, so it would be hugely impactful if something like this works. It’s a risk, and if you look at the history of emerging memory launches, they tend to disappoint and take a long time. So while we’ve made great progress over the last five years, this still has a long way to go.”

Nevertheless, it’s interesting enough that work will continue in this space. Arm has taken a minority ownership position in Cerfe Labs, and Arm’s chief strategy officer, Jason Zajac, will join the company’s board of directors. As part of the deal, Arm will transfer about 150 patent families to Cerfe Labs, which will become the foundation for a roadmap.

Related
Startup Funding: September 2020
Big investment in EV, batteries, and data center chips as 26 companies raise $2.6B.
More Startup Corner Articles
All our latest startup funding reports and new startup articles