White Paper: The Memristor

Posted 04/01/2009 at 7:00pm | by Michael Brown

5

Comments

Print

Integrated-circuit design is currently based on three fundamental elements: the resistor, the capacitor, and the inductor. A fourth element was described and named in 1971 by Leon Chua, a professor at the University of California, Berkeley’s Electrical Engineering and Computer Sciences Department, but researchers at HP Labs didn’t prove its existence until April 2008. This fourth element—the memristor (short for memory resistor)—has properties that cannot be reproduced through any combination of the other three elements.

Chua first theorized the memristor’s existence based on symmetry. There are four fundamental circuit variables—current, voltage, charge, and flux (changes in voltage), but until now, relationships had been defined for only three of those variables: A resistor opposes the flow of an electric current, so it relates voltage to current; a capacitor stores energy in an electric field between two conductors, so it relates charge to voltage; and an inductor stores energy in a magnetic field created by the electrical current running through it, so it relates flux to current. Chua believed that there must be an element that relates charge to flux, and he dubbed this undiscovered element the memristor because it would “remember” changes in the current passing through it by changing its resistance.

Resistance is Futile

A memristor is an element in which the magnetic flux between its two terminals is a function of the amount of electrical current that passes through it. If a charge flows through a memristor in one direction, the memristor’s resistance increases; if the charge flows through in the opposite direction, its resistance decreases. And if the charge is removed altogether, the memristor retains whatever level of resistance it exhibited when the current was present.

A memristor operates on the principle of hysteresis: Its rate of change increases as it moves from one state to another, i.e., from “on” to “off” and back again. This phenomenon has been observed and even used for commercial purposes, although engineers didn’t thoroughly understand why it happened. The propensity of titanium dioxide to change its resistance in the presence of oxygen, for example, led to its use in the manufacture of oxygen sensors.

An HP research team led by R. Stanley Williams used what was known about titanium dioxide as a jumping-off point to develop the first memristor. The researchers passed an electrical current through a thin film of titanium dioxide that they’d doped to be missing some oxygen atoms. The current pushed the holes created by the missing atoms from one side of the film to the other. Passing a charge in the opposite direction pushed the holes back through to the other side. Repeating this process essentially turns the memristor on and off, changing its state from one to zero, but the key is that when electricity stops flowing through the memristor, the memristor remains in its current state.

No Flash in the Pan

Since a memristor doesn’t depend on the presence of electrical current to maintain its state, it is widely expected that one of its first commercial uses will be in the manufacture of non-volatile memory—perhaps as a replacement for flash memory. HP’s team has already created a very simple memristor storage device capable of storing 100 gigabits of data in a one square centimeter die. A single flash memory chip, in comparison, is capable of storing just 16 gigabits. Memristors could also eventually replace DRAM, leading to the development of instant-on computers. A computer using memristors instead of DRAM would need to load its operating system only once—the first time it’s powered up. You’d be able to power off the computer for the night and all your application software and work in progress would be waiting for you when you powered it up the next morning.

Chua noted that the properties of a memristor are very similar to those of synapses, the junctions between nerve cells in the brain. When there is a smaller change in charge, there is a smaller change in resistance. Chua and Williams predict that memristors will eventually be used to manufacture new types of devices that no one has yet thought of. If you used memristors to build an analog computer, for instance, you’d have a computational device that instead of relying on just ones and zeroes could utilize all the values in between.

I See, Therefore I Am

One application for such a device, according to Williams, would be to build computers capable of making decisions based on size comparisons, pattern recognition, and similar forms of analog input. Digital computers are capable of doing this today, but the task consumes prodigious amounts of processing power. And unlike today’s computers, a memristor-based computer would be able to learn from experience because it’s capable of retaining the information it acquires.

Integrated-circuit designers are excited about one other characteristic of memristors: Memristance exhibits a tendency to become stronger as circuits become smaller. That’s one reason so much time passed between Chua’s theory and Williams’s discovery.

No one was building devices small enough for the phenomenon of memristance to manifest itself sufficiently enough to not be dismissed as an anomaly. Since conventional circuits experience more problems with power leakage and heat as they’re shrunk, memristor technology should enable the development of ever-smaller microprocessors for a long time to come; it could also prove to be a key milestone in the development of commercial nanotechnology.