Researchers build room-temperature memory that doesn’t need a current to retain data
All modern computer memory works on basically the same principle — an electrical current is used to change the charge state of a cell. That charge state is then “read” by the memory controller. Whether we’re talking about NAND flash or RAM, this basic property is identical between them. Now, researchers at Cornell have announced and demonstrated a device based on bismuth ferrite that can store data and retrieve data without needing an electrical current to do so. The implications of such a room-temperature breakthrough could be profound — a long time from now.
Magnetoelectric memory is attractive precisely because it removes the electrical currents that currently (pun intended) account for a substantial fraction of power consumption in a system. If the DRAM or cache subsystems could draw less electrical power, total device battery life would increase. The impact of shifting to alternative ferroelectric or magnetoelectric memory systems could be substantial — more than equal to the traditional gains of adopting new process nodes.
Bismuth ferrite has long been known to exhibit both ferromagnetic and ferroelectric properties. This means it has its own magnetic field, like any common magnet, but also that this field’s polarity can be switched by applying an electrical field to the device. This combination of traits makes bismuth ferrite “multiferroic,” meaning it has more than one “ferroic order parameter.” Exploiting these capabilities in a room-temperature device is a major achievement; bismuth ferrite could potentially revolutionize modern electronics if someone can figure out how to build devices from it. This Cornell research demonstrates a single switching cell — meaning a structure that can hold one bit of data.
The agony and the ecstasy of Emerging Research Materials (ERD)
I’m going to borrow the term the International Technical Roadmap for Semiconductors uses to talk about these kinds of new structures and materials. ERMs are the cutting-edge materials that haven’t been proven commercially, but show enormous promise if we can figure out how to use and manufacture them in volume. This category is often linked to Emerging Research Devices (ERDs). So, for example, we’ve got ferroelectric RAM (FeRAM), an ERD, which might possibly be constructed with bismuth ferrite, an ERM.
In recent stories, some of you have complained that far too many of these technologies don’t pan out or seem to come to market, which is why I wanted to address this topic head-on. I write relatively few pie-in-the-sky articles on future manufacturing techniques or approaches precisely because it’s so difficult to forecast what will actually come to market. From OLED to Extreme Ultraviolet Lithography (EUV), the market is littered with technologies that have either ramped much more slowly than originally forecast (OLED), remain mired in development hell (EUV), or have effectively been canceled (the long-planned expansion to 450mm wafer manufacturing). At first glance, these issues may seem unrelated given that they touch on three completely different areas of expertise. What do they have in common? Complexity and scale. In each case, our materials engineering expertise — and bear in mind, we’re talking about the most advanced manufacturing technologies known to man — slam squarely into the fundamental difficulty of working with the materials or achieving the necessary levels of perfection.
The New Yorker had a recent piece on this as it relates to graphene. Graphene is brilliant, amazing stuff. It’s conductive at room temperature to a degree that copper wire only dreams of. It’s pliable, incredibly strong, and durable. It is, in many ways, absolutely as awesome as the hype-machine makes it out to be.
The problem? Graphene lacks a cheap means of production. As noted in the story, it’s common for elements and capabilities to languish for decades or even centuries before finding their absolutely must-have applications. Right now, much of the most promising work for graphene is based on finding ways to incorporate it into existing products rather than building new from-scratch applications. (Building an entire chip out of graphene or bismuth ferrite qualifies as a fundamentally new application, since it requires a complete redesign of the transistor structure and an entirely different assembly process.) For all its promise, the inability to open a band gap in graphene has kept it out of processors — so far.
Bismuth ferrite, for all its incredible promise, presents its own formidable stack of obstacles. It’s not yet clear if we can build stacks of such devices at scale or switch them at speed. The challenges between creating a single bit of bismuth ferrite at room temperature and an entire memory array (much less gigabytes worth of memory) can’t be understated.
We cover these announcements because they represent the innovations that could change the world of computing — even as we acknowledge that the short-term impacts of such innovations are likely to be small.
Now read: Graphene body armor: Twice the stopping power of Kevlar, at a fraction of the weight
- Thanks for reading Researchers build room-temperature memory that doesn’t need a current to retain data