HP's Memristor: Revolutionizing Technology?

[sanman]

HP's memristor continues to catch attention as a fundamentally new and disruptive technological component:

http://www.technologyreview.com/computing/25018/?a=f

The memristor can function as low-power non-volatile memory, and is also capable of performing both logic and memory operations. It also seems capable of reproducing some aspects of a neuron's behavior.

What kind of impact will this fundamentally new electronic device have on our technology?
Will future system designs look more FPGA-like?

I'm thinking that the trend towards mobile devices will benefit, because of the memristor's low-power features.

I think that deep thinkers like Google should push to get involved, because of the deep benefits that memristors could provide them. Google has always said the thermal management issues are a top challenge for their data centers, and perhaps memristor-based systems could help address that. Google is also trying to get in on the trend towards mobile devices. I'm wondering if memristors could make search algorithms more efficient. They'd probably be a better bet than D-Wave's poorly understood quantum weirdness.

If the Google generation aspire to turn the planet into one big wired brain, then maybe memristors could give a significant boost to that.

 

[Antiphon]

I don't like the memristor as a fudamental circuit element. Why you say?

Because it's non-linear. It is the only "fundamental" circuit element which is *not* linear.

I'm not saying it's not interesting; it is. But linearity is something I've come to expect from my fundamental circuit elements.

 

[sophiecentaur]

Like Diodes, you mean?

 

[sanman]

And yet its non-linearity enables its amazing abilities. It's like a neuron. Just imagine if it had been discovered way back along with the others. Then AI technology would be ubiquitous today.

Electromigration is the enemy of classical components like transistors, and yet the memristor seems based on electromigration. Does that mean that memristor components would be less likely to exhibit sudden breakdown, and instead show gradual erosion of performance? (ie. their resistance range would gradually drift or skew due to erosion)

I am wondering if we could one day interface our brains to memristor-based systems, so that they could augment/extend our brains. Gradually, we would migrate fully onto them.
What is the difference if our neurons are made of waterbags or TiO2, as long as they function harmoniously with the whole?

In the meantime, memristors can give us smarter gadgets that use less power.

 

[Antiphon]

Memristors are cool. Very cool. But a fundamental circuit element? (I love diodes btw. Use them every day. But they're not even close to being a fundamental circuit element.)

I think memristors are great, just not linear and therefore not fundamental.

 

[sanman]

Why is linearity a prerequisite for being fundamental?

Flux and charge are fundamental properties, and the memristor embodies the relationship between the two. It's not the memristor's fault that relationship isn't linear.

 

[Antiphon]

Well then what about the Memcapacitor and Meminductor?

How many new fundamental circuit elements do you think will be "discovered"?

Linearity feels right. Not a satisfactory answer, I know.

 

[Deoxyribose]

And yet its non-linearity enables its amazing abilities. It's like a neuron.

How is a memristor like a neuron? Is it a subunit of a neuron that has memritor-like properties, like a synapse or an ion channel, or is it the neuron as a whole? I'm interested in biophysics and want to get a better idea of how memristors can allow nervous systems to adapt and learn from environmental stimuli. Can a voltage-gated sodium channel act as a memristor? Can a ligand-gated sodium channel? Since voltage-gated sodium channels open in response to current from other channels, is memristance a property of the surface of a neuron rather than a single protein?

Since the article mentions several times that memristors have variable resistances & can remember resistance after the power is turned off, I'm guessing that it is synapses that can act like memristors due to their ability to control conductivity of specific ions by controlling the amount of ion channels on the synapse.

Maybe we could learn something by studying the role of memristance in the nervous system. The nervous system is capable of self-modification, which allows it to remember and learn from experiences. Perhaps a better understanding of this mechanism in the CNS could lead to better AI.

 

[elect_eng]

Well then what about the Memcapacitor and Meminductor?

How many new fundamental circuit elements do you think will be "discovered"?

Linearity feels right. Not a satisfactory answer, I know.

I think that there is a general misconception going on here. This seems common and I don't mean to single out this one post, but it brings up the points that many people seem to think about.

There is no special nonlinear/linear difference between resistors, capacitors, inductors and memristors. These 4 quantities are the fundamental parameters that relate voltage, current, charge, flux. There are actually 6 ways to relate 4 quantities, but two of them are definitions for current and voltage.

There are no more fundamental elements to be considered (at least in this context) and leaving out the memristor creates a gap. The reason why we tend to think of the memristor as nonlinear is because a linear memristor looks identical to a resistor and is redundant. However, we need to remember that resistors, inductors and capacitors can be (and often are) nonlinear too. Hence, there is no reason to exclude the memristor. Nor is there a reason to expect new fundamental components to crop up.

Probably the biggest reason why memristors have been mostly ignored until recently is that there were no practical devices that achieved significant nonlinear memristor operation (note that linear memristors are widely available as resistors). Perhaps that is changing now.

 

[freydawg56]

Anyone who thinks resistors/caps/inducts are always linear are not looking at them at higher frequency's e.g. Giga Hertz and above. V=IR breaks down, take an EM course.

 

[elect_eng]

Anyone who thinks resistors/caps/inducts are always linear are not looking at them at higher frequency's e.g. Giga Hertz and above. V=IR breaks down, take an EM course.

Considering the high frequency effects and EM theory is probably an unfair comment in this context. It is also misleading to say that high frequency effects necessarily imply nonlinearity. The assumption here is lumped circuit elements which auotmatically restricts the considerations to the low frequency domain where circuit equations provide a reasonable approximation. At high frequency the idea of a simple pure inductor, capacitor or resistor makes no sense at all. Even at low frequency the idealization is questionable, but this classification is commonly done in circuit analysis and we understand the implications of it.

However, you are correct to point out that all components are inherently nonlinear. Capacitors will break down at higher voltage. Inductors with magnetic materials for a core will saturate and resistors will experience nonlinearity due to thermal drift at higher currents and all three devices burn out if current is too high. True linearity only exists for EM fields in vacuum.

 

[fourier jr]

I just watched these two clips:





I didn't understand much of it but it looks like the wave of the future. Now that a working memristor has been built will Chua & Williams have one of those Nobel thingies coming their way or is it still too soon? It wouldn't be the first time the physics prize has been given out for a tech or engineering-related discovery.