The memristor is a remarkable device that has the potential to change the world of electronics and computers. It was first theorized back in 1971 by Leon Chua, a professor at the University of California, Berkley. Until then, there were three fundamental passive electronic components: the resistor, the capacitor, and the inductor. Based on the symmetry between these three devices, Dr. Chua predicted that there should be a fourth fundamental device, which he named the memristor. The name is a contraction of “memory resistor” because the memristor is a two terminal device that behaves like a resistor. However, unlike a normal resistor, the memristor has the ability to “remember” the previous applied voltage. In other words, its resistance is a function of the magnitude and length of time that a voltage is applied across it.
It was not until 2008 that scientists at Hewlett Packard labs were able to fabricate the first functioning memristor. Their device consisted of two layers of titanium dioxide (TiO2) sandwiched between two conducting layers of platinum (Pt). An applied voltage changes the boundary between the two layers of titanium dioxide by moving the number of oxygen deficiencies scattered throughout the material. This causes the conductivity of the memristor to change and to “remember” this state even when the applied voltage is removed.
(a) Structure of the memristor from HP Labs and (b) the theoretical current-voltage characteristics of a memristor.
The memristor has the potential bring a revolution to the field of electronics for two reasons. First, unlike most electronic devices, the memristor remembers its previous state when the power is switched off. Combined with the memristor’s ability to be integrated with existing transistor technology will lead to the new computer designs that combine logic circuits with non-volatile memory. Our personal computers will be able to instantly turn on and off in addition to performing computations more efficiently. Second, the memristor can not only remember the two binary states (1 and 0) found in digital circuits but a whole continuum of states that are used in analog circuits. This will allow engineers to design amplifier circuits that can be fine-tuned for performance requirements – an important consideration as scaled technologies exhibit more process variation and require some post-process calibration. Even more exciting, the memristor can function like an artificial synapse for mimicking how biological neurons process information. This moves us closer to creating artificial intelligence that can rival the capabilities of the human brain.
My research involving memristors focus on applications involving analog circuits:
· Analog Circuit Security
The memristor’s unique ability to remember a range of continuous resistive values has important application in tuning the performance of analog circuits. We leverage this capability for security applications where the user must provide an authentication code in order for a device to function properly (see figure below). Current work has investigated the sense amplifier of a memory array as the target circuit for securing the contents of a digital memory.
· Neuromorphic architectures
The ability of a memristor to function as an electronic synapse is the foundation for building an artificial brain. Our work here focuses on the analog support circuits for building neuromorphic circuits. This research will also leverage our work in FPGAs and GPUs for providing massive parallel computation to allow efficient simulation of these architectures.
Scheme for Memristor-based Analog Circuit Security
For this research in memristors, we are collaborating with professors at research-intensive institutions with top-rated graduate programs.
Dr. Ramesh Karri, Professor in Computer Science at the New York University.
Dr. Sungyong Jung, Associate Professor in Electrical Engineering at the University of Texas at Arlington.
Representative Memristor Publication
D. H. K. Hoe, J. Rajendran, R. Karri, “Towards Secure Analog Designs: A Secure Sense Amplifier Using Memristors,” IEEE Computer Society Annual Symposium on VLSI, July 2014 (PDF)