Homemade Memristor (2011)(sparkbangbuzz.com)
sparkbangbuzz.com
Homemade Memristor (2011)
http://sparkbangbuzz.com/memristor/memristor.htm
23 comments
The family of these materials are often referred to as chalcogenides (although strictly speaking they don't always include only metals from that periodic group). They've been investigated quite a bit for use in phase change memory if you have access to academic journals (I could probably root around later and see what's not behind a paywall).
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...
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...
https://en.wikipedia.org/wiki/Memristor#Pinched_hysteresis
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.
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.
This circuit simulator has some nice memristor examples: http://www.falstad.com/circuit/
It's just a hysteresis curve, which I believe is effectively the "mem" part of "memristor".
I love it! "I found some smelly corroded metal objects, so I must investigate their electrical properties."
Serendipitous science at its best! Or, how to increase your luck surface area.
Serendipitous science at its best! Or, how to increase your luck surface area.
Cat's whisker diodes and the initial discovery of the LED are other good examples of this. Connect a small wire to the right kind of rock (or rusty old razor blade) just right and you get a diode. If you get it just perfect (and use the right kind of rock) you'll even get a faint LED.
For those who want to experiment with their own LEDs, they should check out:
https://web.archive.org/web/20120913060200/http://neazoi.com...
https://web.archive.org/web/20120913060200/http://neazoi.com...
This is also a neat place to start with making your own OLEDs. https://www.youtube.com/watch?v=CAgRF8TibJ0
Indeed! That brought back a vivid childhood memory of moving the cat's whisker around on a crystal to find just the right spot to pick up the local radio station.
It was mind-blowing to realize that the acoustic waves vibrating my eardrums were directly powered by the radio waves transmitted from the broadcast antenna.
It was mind-blowing to realize that the acoustic waves vibrating my eardrums were directly powered by the radio waves transmitted from the broadcast antenna.
Beautiful. We need more serendipitous moments like this. Like the 19th century experimentalists you read about as a kid.
Be sure to check out the rest of this guys stuff, like the homemade CRT, flame triode, magnetic amplifiers, etc.
http://www.sparkbangbuzz.com/
http://www.sparkbangbuzz.com/
Ah, 90s web design, replete with a link urging the reader to "Join the Blue Ribbon Online Free Speech Campaign!"
I miss the old web.
I miss the old web.
The page is under 5KB and loads quickly, it's content-full and isn't over-designed. A quick look at the source shows that it's hand-written. Very nice indeed.
Funny that browsers now need to have a special "reader mode" specifically to make pages look like this.
Funny that browsers now need to have a special "reader mode" specifically to make pages look like this.
Completing the circle isn't it. The web 2.0 experiment concludes that old mail and html3 were actually fine for actually useful browsing.
One step away from joining the "Maker Web-ring"
The IBM 1401 restoration team found some corroded germanium transistors that seemed to show memristor-like behavior: http://ibm-1401.info/GermaniumAlloy.html#CommentsGarner (scroll to the p.s.) If anyone can explain this behavior, they would be grateful.
To change topic, progress on memristors seems painfully slow. In comparison, it took just 10 years from the invention of the transistor until IBM's president forced the company to give up tubes and switch to transistors. And it was just 4 years from the invention of the first IC until NASA decided to use them in the Apollo Guidance Computer and landed them on the moon a few years later.
To change topic, progress on memristors seems painfully slow. In comparison, it took just 10 years from the invention of the transistor until IBM's president forced the company to give up tubes and switch to transistors. And it was just 4 years from the invention of the first IC until NASA decided to use them in the Apollo Guidance Computer and landed them on the moon a few years later.
> To change topic, progress on memristors seems painfully slow.
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.
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.
Associated youtube: https://www.youtube.com/watch?v=MlswP_qXbdA
not sure why it's not linked on the website
not sure why it's not linked on the website
Neat, he's even got the pinched hysteresis loop going on.
The first talk I ever gave was on Memristors at a Rochester Bar Camp in '09 as a college freshman: https://docs.google.com/presentation/d/1GVqEeFUJyI7VmzqYPabQ...
The first talk I ever gave was on Memristors at a Rochester Bar Camp in '09 as a college freshman: https://docs.google.com/presentation/d/1GVqEeFUJyI7VmzqYPabQ...
Do memristor have MTBF ?
Yes. Since the mechanism for memristors is either phase-change or filament production (by ion conduction), they eventually fail, basically just becoming a normal resistor. The number of cycles of some experimental devices is very high, though I don't know if anyone has done a true MTBF study on one.
Is anyone aware of a good simulator for one of these curve tracers? I understand the idea of a memristor, but it's not immediately intuitive to me why it should make this particular shape (and I'm unsure which "arm" of the eight is taken when the voltage is going up and when it is going down -- I assume the bottom is traveled upwards and the top is traveled downwards). It would be nice to see one animated or with a third time dimension. (The figure eight is more apparent in the link I mentioned above, although you can see it in this submission once you know what you're looking for.)