The first new passive circuit element since the 1830s might transform computer hardware
The device that has sparked all the recent excitement over memristors was created in 2008 by R. Stanley Williams and several colleagues at Hewlett-Packard Laboratories. The Williams memristor consists of two metal electrodes separated by a thin film of titanium dioxide, or TiO2. This substance, also known as titania, is familiar to artists as a white pigment and to beachgoers as an ingredient of sunscreen.
In its natural form titania is an electrical insulator, presenting very high resistance to the flow of electric current. In the memristor, part of the titania layer retains this natural insulating character, but the rest is altered during deposition by restricting the amount of oxygen available. The resulting oxygen vacancies in the crystal lattice reduce the resistance of the material by supplying mobile electrons that can carry a current. The oxygen-starved layer is said to be “doped.” (This term usually refers to added impurity atoms, but the effect of the oxygen deficiency is the same.)
An electric current passing through the memristor has to cross both the doped and the undoped regions, and so the total resistance is the sum of contributions from the two layers. The total depends on the relative thickness of the layers, or in other words on the position of the boundary between them. What gives the memristor its special traits is that this boundary can move.
Consider what happens inside the titania film when a voltage is applied to the terminals of the memristor, so that a current flows through it. The current is carried by conduction electrons—mostly electrons liberated by the oxygen vacancies. Electrons have a negative charge, and so they are repelled by the negative terminal and attracted to the positive one. In the background, meanwhile, another process is going on. The oxygen vacancies also have an electric charge; they act as positive ions, which drift toward the negative electrode. Movement of the vacancies requires physical rearrangement of the crystal lattice, and so it is much slower than the flow of electrons.
The relatively slow drift of the oxygen vacancies makes no significant contribution to the electric current flowing through the memristor, but by shifting the boundary line between doped and undoped layers, it alters the overall resistance of the device. Depending on the polarity of the applied voltage, the resistance can either increase (if the doped region is squeezed into a narrower layer) or decrease (if the doped region expands to include more of the total thickness). When the external voltage is removed, the boundary line stays put in its new position.
It is the migrating boundary between doped and undoped regions that gives the memristor its memory. And it’s not hard to see how this property can be put to work for information storage. One simple scheme defines a low-resistance state as a binary 0 and high-resistance state as a binary 1. To write a bit into the memory cell, apply a strong voltage pulse of the appropriate polarity, thereby setting the resistance either high or low. To read the stored state of the cell, use a lower voltage or a briefer pulse, which can measure the resistance without appreciably altering it.
A notable advantage of the memristor is that it can be made very small. As a matter of fact, it must be small, at least along one dimension—the thickness of the TiO2 film. The ratio of maximum to minimum resistance varies inversely as the square of this thickness. In practical devices the film might be as thin as 10 nanometers, which is just 25 or 30 atomic diameters.
It’s also notable that the memristor offers nonvolatile storage: The device retains its memory even when the power is turned off.