Finding Hidden Joules

Ferroelectric material could help the arc of Moore’s Law reach new heights and radically reduce energy consumed by digital devices

by Gordy Slack

For the past several decades digital devices have grown more plentiful and more efficient at roughly the same pace. As transistors and the devices they serve have shrunk, packing a greater processing punch per square centimeter every year, our reliance on more and more of them has grown. The average American now has four digital devices, up from two just five years ago. And while the coming “internet of things” may introduce innovative new energy-conserving efficiencies into our daily lives, every networked “thing” will require a battery or some other supply of power.

In an effort to reduce energy consumed by the growing number of digital devices, we envision a paradigm we call “Attojoule-Per-Bit,” says Ramamoorthy Ramesh, professor and a pioneer in the field of ferroelectric memory and multiferroics. This effort seeks to radically reduce the amount of energy required to manipulate a single bit of information; that is, to store a one or a zero in a digital medium.

A joule is a standard unit of energy; very roughly speaking, the amount it would take to raise an apple one meter off the ground. An attojoule, the project’s target expenditure per bit, is 10-18 of one joule. Since many technologies currently use a picojoule (10-12) per bit, this would represent a reduction of six orders of magnitude. “To get there,” says Ramesh, Berkeley professor of materials science, “we need to take advantage of all of the different possible pathways.”

Professor Sayeef Salahuddin  has

Professor Sayeef Salahuddin is working with new materials that could drastically decrease the amount of power consumed by our electronics.

One promising approach that may cover a hefty portion of that distance is described this month in a paper titled “Negative Capacitance in a Ferroelectric Capacitor” in Nature Materials. The work, led by CITRIS researcher and Berkeley professor of electrical engineering Sayeef Salahuddin, and Ramesh, directly observed for the very first time this elusive phenomenon called “negative capacitance.”

Capacitance is the ability of a material to store an electrical charge. Ordinary capacitors—found in virtually all electronic devices—store charge when a voltage is applied to them. The new phenomenon, negative capacitance, derives its name from its paradoxical response: when applied voltage is increased, the charge in the material actually goes down.

“This property, if successfully integrated into transistors, could reduce the amount of power they consume by at least an order of magnitude, and perhaps much more,” says Asif Khan, the paper’s lead author and a Berkeley PhD candidate in electrical engineering.

Without a major breakthrough of this sort, the decades-long trend toward miniaturization and increased function is threatened by the physical demands of transistors operating at a nano scale. Even though the tiny switches can be made ever smaller, the power they need in order to be turned on and off can be reduced only so much. That limit has been defined by what is known as the Boltzmann distribution of electrons—often more vividly called the Boltzmann tyranny. Because they must be fed an irreducible amount of electricity, ultra-small transistors that are packed too tightly cannot dissipate the heat they generate quickly enough to avoid self-immolation.

In another decade or so, engineers will exhaust options for packing more computing power into ever tinier spaces. This consequence is viewed with concern by device manufacturers, sensor developers, energy conservation advocates, and a public addicted to ever smaller and more powerful digital tools and toys.

The new research provides a possible way to overcome the Boltzmann tyranny and give new legs to the famous Moore’s Law, which has long predicted a doubling of the number of transistors in a dense integrated circuit every two years. “Moore’s Law is done,” says Costas Spanos, CITRIS director and professor of electrical engineering at Berkeley. “But that gives us an opportunity; we don’t just have to spend all our resources keeping up with the law. We can step back and really innovate.”

The CITRIS researchers’ innovative work relies on the ability of newly synthesized materials to store energy intrinsically and then exploit it to amplify the input voltage. This could, in effect, potentially “trick” a transistor into thinking that it has received the minimum voltage necessary to operate. The result: less electricity is needed to turn a transistor on or off, the universal operation at the core of all computer processing.

The material used to achieve negative capacitance falls in a class of crystalline materials called ferroelectrics, which was first described in the 1940s. Ramesh’s decades of seminal work on ferroelectric materials and device structures for manipulating them underlies the group’s findings as well as currently employed memory applications and storage technologies. Ferroelectrics are also popular materials for frequency control circuits and many MEMS applications. However, the possibility of using ferroelectrics for energy efficient transistors was first proposed by Salahuddin in 2008, right before he joined Berkeley as an assistant professor.

Over the past six years, Khan—one of Salahuddin’s first graduate students at Berkeley—has used pulse lasers to grow many kinds of ferroelectric materials and has devised and revised ingenious ways to test for their negative capacitance. Now, ready to receive his PhD in just a few months, Khan is also poised to be part of what may prove to be another revolution in the miniaturization of processors.

“The next step,” says Salahuddin, “is to try to make actual transistors such that they can exploit the new phenomenon. We need to make sure they are compatible with silicon processing, that they are manufacturable, and that the measurement techniques we’ve now proven in principle are practical and scalable.”

If Khan’s and Salahuddin’s predictions bear out, this work could catalyze a technology that enables further development in mobile tech while allowing batteries to run longer per charge. It may also spawn whole new industries based on ferroelectrics, with the added benefit of conserving energy and thereby reducing greenhouse gas emissions.

Read the related press release about the researchers’ work (December 15, 2014).