Professor Qiangfei Xia of the Electrical and Computer Engineering (ECE) Department at the University of Massachusetts Amherst leads a multi-institutional group of researchers who have invented a new kind of memristor entirely based on silicon materials, which could act as a promising building block for the next generation of memory and neuromorphic computing systems. “The current work opens up opportunities for low-cost mass production of 3D memristor arrays on large silicon and flexible substrates without increasing circuit complexity,” as the research team summarizes its research.
A memristor, short for memory resistor, is an electrical component that limits or regulates the flow of electrical current in a circuit. Most memristors use materials that are incompatible with the silicon-dominant CMOS technology and require external selectors in order for large memristor arrays to function properly. However, the team’s newly developed device is a fully CMOS-compatible, all-silicon-based, self-rectifying memristor that negates the need for external selectors in large arrays.
The multi-institutional team – containing researchers from UMass Amherst, the Center for Functional Nanomaterials at Brookhaven National Laboratory in Upton, New York, and the Air Force Research Laboratory Information Directorate in Rome, New York – recently had its manuscript on an all-silicon-based memristor and arrays published in the prestigious academic journal Nature Communications. The other UMass Amherst co-authors are Can Li, Lili Han, Hao Jiang, Moon-Hyung Jang, Peng Lin, and J. Joshua Yang, all from the ECE department.
As Xia and his colleagues observe in their paper, memristors are two-terminal devices that exhibit voltage- and/or current-actuated resistance switching. With proven advantages in scalability, switching speed, power consumption, endurance, and three-dimensional stackability, these devices have been proposed and demonstrated for applications such as non-volatile memory, reconfigurable logic, neuromorphic computing, and radiofrequency switches.
A memristor is usually composed of two metallic electrodes that sandwich a layer of switchable material, which to date has usually been manufactured from transition metal oxides and perovskites.
Unfortunately, say the researchers, these materials and associated fabrication processes are costly and not fully compatible with the CMOS platform. On the other hand, resistance-switching phenomena in silicon oxide have been observed since 1960s and have attracted revived interest recently because of their CMOS compatibility. However, their device performance needs further optimization, the switching mechanism is still under intensive debate, and devices and arrays made of all-silicon-based materials have yet to be demonstrated, let alone 3D stacking of the crossbar arrays.
“Building high density, massively parallel memristor crossbar arrays is critical for most of the aforementioned applications,” as the researchers observe. “One of the prominent challenges to access an individual device in such an array is to solve the so-called ‘sneak path’ problem that leads to operational failure or high power consumption. To address this issue, ‘selector’ devices such as rectifying diodes, non-linear bidirectional selectors, and complementary resistive switches have been proposed and demonstrated to suppress the ‘sneak path’ current through the unselected devices.”
Silicon-based selector devices, such as diodes and transistors, provide promising solutions because of materials compatibility and process maturity.
“In this paper, we report for the first time an all-silicon-based memristor with a built-in rectifying selector that is fully compatible with CMOS platform,” Xia’s group writes. “The device employs p- and n-type doped single crystalline silicon transferred from fluid supported membranes as top and bottom electrodes and a thin layer of chemically produced silicon oxide as the switching layer. The device exhibits repeatable self-rectifying resistive switching with high rectifying ratio, high ON/OFF conductance ratio, and long data retention at elevated temperature.”
The self-rectifying effect is due to the different doping types in the silicon electrodes that form a self-assembled diode within each junction.
As the researchers explain in their paper, “We further construct 3D crossbar arrays of Si/SiO2/Si memristors and demonstrate that the built-in selectors effectively alleviate both intra- and inter-layer sneak path problem. The all-silicon-based materials and room-temperature fabrication process are fully compatible with CMOS technology, opening up opportunities for low-cost volume production of large arrays of memristor devices and hybrid memristor/CMOS circuits on a variety of substrates.”
Professor Xia received his Ph.D. degree from Princeton University in 2007. He joined the ECE faculty at UMass in 2010, after spending three years as a Postdoctoral Research Associate at Hewlett Packard Labs. Upon his arrival at UMass, he quickly established his research group and initiated research on the fabrication of nanoscale memristive devices and their integration into electronic circuits.
As the ECE Department Head Christopher Hollot has written, “I’ve heard it said from folks, more qualified than I, that Dr. Xia is a young superstar who has the potential and best credentials to become a future scientific leader.” (June 2017)