Invented and pioneered in the United States (US), the transistor has been an enabling technology in our way of life, a source of pride, and a symbol of American ingenuity. As a result, it is not surprising that the transistor is the heart of information processing equipment, ranging from computers to networks and data center equipment. However, as fantastic as the transistor has been, charging and discharging a growing number of transistors consumes a significant amount of energy. It requires an applied voltage of the order of 1 Volt to operate a single bit of information. This contrasts with the fact that electronic circuits operate with acceptable signal-to-noise at millivolt levels, resulting in much lower power consumption. As a result, a challenge facing America is the growing fraction of total U.S. energy consumed by the information processing sector, which scales as the square of the operating voltage of transistors. Clearly, the scientific search for low-energy bit-manipulation presents an exciting opportunity to make a significant difference in U.S. competitiveness in information processing and energy consumption.
Addressing this challenge, CODEC is a collaboration between the University of Arkansas (UA) and The National Research Energy Laboratory (NREL) teams to discover and develop novel materials that dramatically meet the need for faster, more energy-efficient U.S. information systems. The approach is based on the growth, characterization, modelling, and fabrication of nitride semiconductor alloys and nanostructures. Some of these semiconductors, which are called ferroelectrics, have the desirable properties of possessing a substantial electrical polarization (which can serve as a "bit" of information) which can be switched up (ON state) and down (OFF state) very quickly by applying electric fields. Among the entire class of ferroelectric materials, only a few show good compatibility with current silicon-based electronic platforms. Nitride semiconductors, such as (Al,Sc)N, the focus of this project, belong to this limited class of silicon-compatible ferroelectrics. However, switching the polarization/bit still requires large electric fields.
Understanding, at the atomic scale, what governs the speed of reversal in the electrical control of the polarization, is key to designing materials with smaller operating voltages and integrating them into faster and/or less energy-consuming computing architectures. To do so, the University of Arkansas and NREL will combine their modelling and growth expertise to reveal, in ferroelectric (Al,Sc)N, the dynamics of the electrical control of polarization, and thus optimize composition and structuration to enable reduced voltage operations. Furthermore, the University of Arkansas and NREL will investigate how the atomic structuration, at the nanoscale, of these ferroelectric nitride semiconductors may lead to exotic effects allowing local voltage amplification (enabling lower energy consumption) or mimicking brain-like computing operations (to develop novel efficient hardware for bio-inspired computing and its machine learning applications on large datasets).
We anticipate that CODEC will, beyond its scientific achievements, train the next generation of scientists and engineers who will design the computing hardware of the future and promote US leadership in this domain.