HP’s Memristor: Now Slower and Less like Unobtainium

Back in May, I wrote about HP Labs’ development of the memristor, the fourth basic passive circuit element after the resistor, capacitor, and inductor. (See Move over Resistor, Capacitor, and Inductor: HP Labs Builds the Fourth Passive Circuit Element—the Memristor—in Nanoscale.) The memristor was predicted by UC Berkeley Professor Leon Chua based on mathematical symmetry arguments back in 1971. After existing only in theory for 37 years, HP Labs announced that they’d built one early this year. It seemed good to be true: very small devices built with nanoscale imprint technology and—best of all—“too fast to measure.” Great stuff for a memory. Literally Unobtainium. However, “too fast to measure” is now 50 nsec according to this article in EETimes.

The memristor’s memory mechanism seems to be based on the drift of oxygen impurity vacancies from one layer of titanium oxide to another. Without vacancies, titanium oxide is an insulator. With vacancies, it’s a semiconductor. So adding oxygen vacancies to pure titanium oxide causes it to start conducting current. A voltage bias across the layers induces the vacancy drift between layers.

Now the prototype memristor’s 50 nsec switching time is still pretty fast for a nonvolatile memory. It’s nowhere near fast enough to serve as RAM for today’s fastest on-chip microprocessor cores, which need access and write times measured in a very few nanoseconds, but 50 nsec is really, really fast compared to Flash EEPROM, which has read times measured in the hundreds of nanoseconds and write times measured in milliseconds. So it looks like memristors, should they prove manufacturable, could give on-chip Flash memory some real competition. And who knows? There might be ways to go faster.

So far, memristors are just lab experiments. But at least they do exist. HP Labs claims it will have prototype chips next year. Stay tuned.