Associate Professor Jon Ihlefeld’s Unique Expertise will be Key to Stacking Memory and Processing Chips
mkw3a@virginia.eduIf you were to open up and peer inside your computer, you would see that the bank of memory cards and the processor are centimeters apart. The memory bank and processor are connected by wires that transmit logic rules, data and instructions back and forth in the form of electric charges or current.
Like a light switch on the wall, a transistor in the computer processor releases current onto the wire on one end and seeks to learn the state of the memory on the other end, in the form or a 1 or a 0.
This arrangement has two distinct disadvantages. First, the processing chip must wait after sending the signal to receive a reply from the memory chip, which constrains processing speed. Second, every time the transistor gate is opened and releases a signal, this consumes energy that is proportional to the length of the wires in the circuit. The longer the wire, the more energy is consumed.
Computer design has hit this bottleneck because the behaviors and preferences of the materials used for memory and processing are sometimes at odds.
Jon Ihlefeld, University of Virginia associate professor of materials science and engineering and electrical and computer engineering, has joined a newly formed Energy Frontier Research Center for 3D Ferroelectric Microelectronics aimed at overcoming these challenges. Funded by the U.S. Department of Energy's Basic Energy Sciences Program, this center seeks to boost computer processing performance by stacking memory and computational chips in a single component, while also adding new functionality enabled by ferroelectrics.
Susan Trolier-McKinstry, who is the Evan Pugh University Professor, the Steward S. Flaschen Professor of ceramic science and engineering and a professor of electrical engineering at Penn State, serves as principal investigator for the $10 million, four-year center, which includes researchers at Penn State, University of Pennsylvania, Rochester Institute of Technology, Purdue University, Oak Ridge National Laboratory, and Sandia National Laboratories.
Ihlefeld will help prepare and test memory materials that can be stacked above the processor chip—the part of the fabrication process called back-end-of-line integration. A successful design will also enable further miniaturization of computer components, reduce the energy needed to operate and provide for new functionality by enabling the development of new devices.
Ihlefeld’s multifunctional thin film group at UVA fabricates and tests electronic oxide materials; their thin films increase the efficiency of computer chips and can serve as low-power computer memory. He specializes in a sub-class of materials known as ferroelectrics, which have applications in memory. Ferroelectrics remember which way an electric field is applied. They can either serve as the 1 or 0 of a persistent memory element or essentially keep the transistors in an open state without repeated application of electric charges or voltage.
The challenge is that most ferroelectrics require preparation at high temperatures. This is not a problem when the memory and computing chips are fabricated as independent units. However, stacking the memory on top of the computing chip needs to be done at low temperatures. If the temperature exceeds 400 degrees Celsius, stacking could damage the pre-existing layers.
Ihlefeld’s research group looks at ferroelectricity in hafnium oxide, one of the ferroelectric materials that the Energy Frontier Research Center team chose for theoretical and experimental research. Hafnium oxide is attractive because it is compatible with silicon. It is already present in your computer in a different form and can be prepared and applied alongside and on top of the processing chip. Additionally, the methods of its application may assure a uniform coating.
Ihlefeld’s group members have been investigating the properties and means to synthesize hafnium oxide for six years in efforts funded by the Semiconductor Research Corporation, an industry consortium, and Sandia National Labs.
Ihlefeld’s collaborations through the Energy Frontier Research Center takes this work a step further, to overcome operational barriers facing the material.
Hafnium oxide holds great promise but faces some challenges to becoming the answer to the high-speed, low-energy computing problem. The research team that Penn State has assembled will shed light on these fundamental questions: why hafnium oxide functions as a ferroelectric, what makes it work and how its preparation can change the oxide’s properties and performance.
The voltage required to make the hafnium oxide switch between a 1 and a 0 is very high. Repeatably subjecting the material to high voltages will cause it to eventually break down. Also, hafnium oxide’s performance tends to vary with the number of voltage pulses it receives, which makes repeated and consistent computing functions more difficult.
“If we learn what we can do to hafnium oxide to overcome these challenges, we will gain greater reliability and lifetime of the memory chip and integrated circuit as a whole,” Ihlefeld said.
Hafnium oxide is the most mature ferroelectric that the center will investigate—and it’s only 10 years old in a research field that has been around 100 years.
“The team is really sitting on the cutting edge of ferroelectric thin film research. If we can successfully improve hafnium oxide’s performance, while keeping it compatible with contemporary semiconductors and processes, then there is the real possibility of successfully integrating ferroelectric-based memory and computing elements in the third dimension of a computer chip in the near-term,” Ihlefeld said.
Lower-power consuming, higher-performance devices using these 3D integrated ferroelectric microelectronics may be just around the corner.