Memristive Devices for Computing

Jianhua (Joshua) Yang
Hewlett Packard Laboratories / USA

Existing technologies for the current computing system are approaching their physical limits, and novel device concepts are required as device sizes continuously decrease. Under these new concepts, the devices need to be not only increasingly infinitesimal and simple but also increasingly capable¹. Memristive devices (also called RRAM when used for memory) seem to fulfill these goals well for the next generation computing system. These devices are electrical resistance switches that can retain a state of internal resistance based on the history of applied voltage or current. Memristive devices can store and process information, and offer several key performance characteristics that exceed conventional integrated circuit technology².

An important class of these devices is two-terminal resistance switches based on ionic motion, which are built from a simple conductor/insulator/conductor thin-film stack³. The switching can be bi-stable or analog, making these devices useful for both digital and analog/neuromorphic computing. The bi-stable switches can be used to achieve implication logic and NAND logic for digital computing⁴. In these new Boolean logic devices, logic states are stored as resistance states of the devices (thus nonvolatile) rather than voltage levels as in conventional logic devices. The new logic has a much smaller footprint. The crossbar array of bi-stable switches can serve as the routing network to store configuration information in a Memristor/CMOS hybrid FPGA circuit, leading to significantly improved density and performance⁵. The crossbar array of analog switches can be used for analog computing. An example is a crossbar array based dot-product engine, which can greatly accelerate some computation intensive tasks, such as vector matrix multiplications. In addition, these analog devices can be used to emulate synapses in bio-systems. Memristors with negative differential resistance (NDR) behavior have been be used to build neuristors that exhibit the important neural functions of all-or-nothing spiking⁶. Therefore, memristive devices are also very attractive in neuromorphic computing.

1. Yang, J. J., Borghetti, J., Murphy, D., Stewart, D. R. & Williams, R. S. A Family of Electronically Reconfigurable Nanodevices. Adv. Mater. 21, 3754-3758 (2009).

2. Yang, J. J., Strukov, D. B. & Stewart, D. R. Memristive devices for computing. Nature Nanotech. 8, 13-24 (2013).

3. Yang, J. J. et al. Memristive switching mechanism for metal/oxide/metal nanodevices. Nature Nano. 3, 429-433 (2008).

4. Borghetti, J. et al. 'Memristive' switches enable 'stateful' logic operations via material implication. Nature 464, 873-876 (2010).

5. Xia, Q. F. et al. Memristor-CMOS Hybrid Integrated Circuits for Reconfigurable Logic. Nano Lett. 9, 3640-3645 (2009).

6. Pickett, M. D., Medeiros-Ribeiro, G. & Williams, R. S. A scalable neuristor built with Mott memristors. Nat Mater 12, 114-117 (2013).