Homemade Memristor (2011)

The strange looking curve is really kind of two parts forward biasing/reverse biasing. Assuming you're starting from a virgin device you move along the curve where you're sweeping the voltage upwards but not really getting much increase in current (since there's no real channel yet). In these materials the mechanism is the establishment of little "fingers" of metal migrating and completing the circuit. As the "fingers" move forward you suddenly start seeing more and more current without increasing voltage (your effective resistance is dropping) and then jumps up as the connection is made. Eventually you sweep the voltage backwards. Something similar happens in the part of the AC curve in reverse.

There is a really good diagram in this paper: "Mechanism for resistive switching in chalcogenide-based electrochemical metallization memory cells" from Zhuge et al http://scitation.aip.org/content/aip/journal/adva/5/5/10.106...

A resistor would give you a straight line where the slope is its resistance. Voltage goes up or down it traces out the same line of current at any instant voltage.

This dude gives a different current based on previous voltage, it remembers its past state, basically.

It has to pass thru zero otherwise we call it a "rechargeable battery". A nicad would have an offset of 1.2 volts at zero current, not passing thru zero (unless it was totally dead and leaking) Also battery frequency response is pretty dismal, being chemical electroplating devices.

If you look at a magnetic hysteresis loop there is some family resemblance. Sticking a cap or inductor on a curve tracer looks quite a bit different and often will not pass thru origin point either (unless you run the device at resonant freq which is whole nother kettle of worms)

Its also kinda frequency dependent and the limiting stage of a memresistor at super high frequencies would be a boring resistor and the way the gap shrinks with freq is important to prove its a "real" memresistor. What's interesting about "recent" news is that its possible to make some pretty fast ones now a days in a research lab and some prototypes are supposedly shipping etc. I would have to think for a second, what is the limiting behavior of a 2 pin semiconductor device... if a memresistor is a resistor I think a diode would at high freq be basically a capacitor, at least when reverse biased (the famous varactor diode). There is a weak varactor effect of forward biased diodes too BTW, although its not terribly useful.

Funny that browsers now need to have a special "reader mode" specifically to make pages look like this.

I think that's a little unfair. They were first shown just 7 years ago, and have already been used to build (extremely) simple neural networks.

Transistors took ~6 years before they were sold in commercial products, and you can already buy a memristor on a chip.

When transistors were invented, ENIAC was around and had ~18,000 vacuum tubes, and was the kind of thing governments built. In 2008 you could wander into a shop and pick up something like this: http://ark.intel.com/products/37147 which has 730 million transistors.

I think it's understandable that it'll take a bit longer for memristors to overtake transistor based processors.