The case for rejecting the memristor as a fundamental circuit element (2018) | Hacker News Reader

The case for rejecting the memristor as a fundamental circuit element (2018)

nature.com

191 pointsg0xA52A2A46 comments

I'm not, I apologize, an electronic engineer, so I mainly only get to view the whole memristor discussion as one of competing narratives - and certainly the 'whatever HP are pitching, it's not what was originally envisaged, and there's no certainty that's even a thing' narrative is certainly compelling. But there is a slight feel in this paper of an attempt to reconcile an unclear new concept with an orthodoxy, and that rubs me the wrong way. It reads a little like an alchemist dismissing the concept of phlogiston. Maybe you're both wrong?

"The basic question of the “missing element” is whether we can we have a new passive element that cannot be made from the combination of existing passive elements. Water (W), Fire (F), Earth (E) and Air (A) are the four fundamental passive elements that a contemporary alchemical engineer is familiar with.... "

I think that the exposition of the manipulation and deployment of the fab three differentiates this discussion from the descriptive and mystical narrative of alchemy.

Oh what a thoroughly impressive academic hatchet job, and I mean that in the kindest possible sense. It's intended to kill the idea of memristors as fundamental passive circuit elements, and it's very good at doing just that.

TLDR: by re-deriving the periodic table of passive elements from first principles, it shows that memristors don't fit. The "missing fourth element" from Strukov's table is not really a missing type of a passive element. Rather, it shows that Strukov's table itself is wrong, that it's just a view of the standard C/R/I relationships that conflates two different ways of looking at resistance specifically (see especially https://www.nature.com/articles/s41598-018-29394-7/figures/3 )

FWIW this does not diminish memristors as potentially industry-changing inventions, like transistors. But it denies them the status of fundamental, irreducible passive circuit elements, specifically, partly by showing that while fundamental elements are all connected by charge-voltage relationships (specifically, changes over time, like position/velocity/acceleration are in classical physics), memristors don't have a unique spot there; and also because fundamental elements can't be composed from each other, while the memristors demonstrated so far are "nonlinear composition of resistors with active hysteresis".

Is it not possible to create the other "fundamental" components using combinations of other components? I know with logic gates one might think of AND/OR/NOT as fundamental but yet one can make anything out of NANDs or NORs, so they could be considered fundamental instead. It's a thing in mathematics as well that you can often define essentially the same system using different sets of prepositions.

I saw a comment later on in here that explained that memristors, resistors, capacitors, and inductors round out all four combinations of relating charge or the derivative of charge over time with flux or the derivative of flux over time.

I'm not sure the logic gate metaphor works here, because the domain is so different, the questions relate to the charge-voltage relationship and its first and second derivatives.

Maybe a different metaphor would be better: a bank statement involves stored money (like a capacitor), and a profit/loss sheet involves effects of money moving over time (like a resistor), and you can see how they're related through integration/differentiation, but they're fundamental because you can't create a bank account by nailing a bunch of P&Ls together.

Ok, so maybe that's not a great metaphor either? ;)

The paper does discuss equivalence categories (along the lines of nth derivatives of charge and voltage) so it's probably best to refer directly to that!

> Is it not possible to create the other "fundamental" components using combinations of other components?

Unlike with logic gates it's just not possible.

Resistors have real impedance and capacitors and inductors have purely imaginary impedance so you can't bridge those two domains (it's like saying can you make a complex number by adding/subtracting real numbers).

For capacitors and inductors one have an impedance that increases with frequency and the other has an impedance that decreases with frequency. So for a given frequency they are interchangable, but as soon as the frequency changes that breaks down.

The article says that the two diagonal elements in the table of those four combinations, U = RI (or dPhi/dt = R dQ/dt, the classical resistor) and Phi = RQ (the purported memristor) describe the same fundamental element, the resistor. Which is sort of obvious, as one is just the time derivative of the other.

[Edit: got the direction of which is the derivative of the other wrong on the first try]

I enjoyed this look at the memristor. Coming at the question from first principles is one that works for me. I've saved it in my notebook for forward to folks who go on and on about how revolutionary they are.

However this work doesn't say anything about how revolutionary for our ordinary use cases some device based on something commonly called memristors could be.

Like iPhone: it is not "fundamental" in any sense, but one can really see the difference "before and after."

The mobile phones existed before, but the way people use them hugely changed.

Another problem is that we read for years how memristors will be revolutionary, but then just some time passes and we read it again... I haven't researched the background to that, maybe somebody here has some insider info?

The article is about whether, using your example, to add iPhone to the Periodic Table of Elements. They may be useful but they aren’t a fundamental element.

Memristors are fundamental in that they relate charge w/ flux.

Capacitors relate charge to voltage. q = Cv

Resistors relate voltage to current. v = Ri

Inductors relate current to flux. theta = Li

Before memristors, nothing related charge to flux. theta = Mq

(also, voltage is the derivative of flux over time, and current is the derivative of charge over time, which makes a nice symmetry that memristors complete)

(skimmed the article briefly, seems like they're arguing that memristors are not fundamental because hysteresis is a non-linear relationship? Can I get a TL;DR HN?)

More that it requires active and time dependent history to know the state of the device, making it not fundamental in the sense that the resistance doesn't depend on history.

It's a fair point I'd say, and they're clear that this isn't something that diminishes utility of any practical device. Just that it isn't fundamental. This seems clear if you define fundamental things as not requiring history or energy to be defined. If you don't define it that way I'm sure you can come to a different conclusion.

I only understand your comment enough to reformat it and discover the symmetry you're describing.

capacitors: charge to (flux/time)'

resistors: (charge/time)' to (flux/time)'

inductors: (charge/time)' to flux

memristors: charge to flux

It's every combination of charge or the derivative of charge over time with flux or the derivative of flux over time.

For people like me who are slow:

    cap - res      f/t
     |     |    chg   c/t
    mem - ind      flx

Squares are neat.

(Not my area of knowledge; have no real clue. Hoping someone comes along and explains further.)

The article's argument is that those are not the right dimensions to look at. The way to look at fundamental electronic components is as charge and the derivatives with regard to time:

1/C * q = v (Capacitors relate charge to voltage)

R * q' = v (Resistors relate the rate of charge to voltage)

L * q" = v (Inductors relate the rate of rate of charge to voltage)

There's no room in this view for the memristor

"fundamental" refers to mathematical symmetry. Your comparison between a passive circuit element and a consumer product strikes me as bizarre.

It's a context that matters. I actually haven't compared "a passive circuit element" to an iPhone.

I answered to the following (one level above): "I've saved it in my notebook for forward to folks who go on and on about how revolutionary they are."

My claim is that iPhone was "revolutionary" even if "it is not 'fundamental' in any sense."

Moreover, just "a passive circuit element" that you used is surely way too simple to be used to describe a memristor:

https://en.wikipedia.org/wiki/Memristor

Also:

https://arxiv.org/abs/1207.7319

The way I read it, a resistor is still the only "real" passive circuit element.

My view of a "memristor" was always more as a "marketing name." As here:

https://www.newscientist.com/article/mg21328535-200-online-s...

Which doesn't mean something marketed as them won't be useful. But, as I've also said, up to now, year after year the announcements happened, the time passed... Still, I know nobody would care "if it's fundamental or not" if it would allow us to do the things we couldn't before.

You kind of lost me with your emphasis on the differential forms of the v-i relations. As I’ve advanced, the integral relations are more fundamental. The accumulation of charge on a pair of conductors increases the voltage. The volt-time integral across an inductor gives rise to the flux therein. Etc.

Aren't they equivalent? what makes one "more fundamental" than the other?

It’s kind of a cause and effect thing that designers use. Of course, from calculus, you are right. But for instance in practice, you charge a capacitor with voltage as the outcome. You don’t look at the voltage, and then differentiate it to infer the current. Indeed, you get into trouble if you think you can put an independent voltage source across a capacitor. With an inductor, you build up the flux by applying a voltage. In both cases, the fundamental idea is the electric or magnetic field stored energy in the component, which the component acquires by accumulation, i.e. integration.

The credit for the first functional memristor goes to the Hewlett-Packard Company—in particular, researchers R. Stanley Williams, Dmitri B. Strukov, Gregory S. Snider, and Duncan R. Stewart—for building a bi-level titanium dioxide thin film containing dopants (impurities) on one side that migrate to the other side when a current is applied and back when the opposite current is applied, changing the resistance in each case. Hewlett-Packard is working on incorporating memristors into traditional integrated circuits.

https://www.britannica.com/technology/memristor

In my opinion whatever thing you want to call fundamental circuit element should have a relationship to the physical reality strong enough that theoretical model consisting of some idealized conductive solids of trivial-ish geometry works for most of the practical problems. L, C and R satisfies this test with various levels of "-ish" (with R being the hard one) and handwaving, while nobody even knows how to model an memristor in that view.

> while nobody even knows how to model an memristor in that view.

Isn't hysteresis the model? Pretty sure that model was in the original paper introducing the concept of memristors by Chua, IIRC.

My point is that you can take length of insulated wire, do something relatively trivial to it and end up with single port device that is decidedly mostly L, C or R. You can not construct practical memristor that way.

Just to make sure I'm following you:

R would be a long piece of wire, L would be a coiled piece of wire, and C would be two wires in parallel, for some distance?

In my original idea, C would be twisted pair, but from the theoretical standpoint, that does not matter, so yes.

The idea is that the whole circuit theory is motivated (as is the rest of physics) by ability to describe phenomena that can be practically observed, and should not invent things that only should exists to preserve mathematical purity when such concepts have no real world equivalent nor are useful as simplified model for something.

I’m not sure I agree with you on this. I once read that pure maths tends to preceed applied maths by about 50 years. I don’t know how true it is, but if the practitioners aren’t constantly chasing some ideal version of reality, I’m not sure we would get the same major advances in industry.

I think you may also have a dated idea as to ‘real world equivalent’. We’re in a very interesting time for meta materials. Things like negative index of refraction was pure science fiction 20 years ago, but today we can make multilayered substances that behave as though they have a negative index. Which is to say, that if we can make it, it now has a real world equivalent. Even your example of a wire ignores the enormous amount of materials processing necessary to produce wire. Wire was unimaginable maybe as recently as 10k years ago.

I'm on the fence between you two. On one hand, it's true that missing pieces in a mathematical symmetry often motivate us to go find the real-world thing we haven't observed.

On the other hand, your example of a multi-layered substance that exhibits negative refraction does seem like "cheating". We can make transistors by layering different doped silicone variants together, but it's not considered a fundamental electrical component. If memristors require the same amount of engineering, are they "natural"?

What would be fascinating is, what if the missing puzzle piece implies an exotic shaping of "just" wire that would act as a memristor? Some crazy contortion like the magnetic field designs I've seen for containing fusion reactions.

Indeed you can't. You also can't construct a diode or a transistor.

There are two parallel arguments here. One is yours. The other is whether or not it's possible to construct a resistor with memory at all.

It clearly isn't if you all you have is copper wire. But copper wire is only fundamental in the 19th century pre-semiconductor sense.

None of the practical memristor models assume that. They all rely on junction effects, spin effects, and other exotica.

You are arguing that the memristor is an interesting active device like a transistor (which it may be, and which isn't addressed by the article). What the article argues against is the claim that is a the long missing last fundamental passive device.

The paper has an interesting perspective but much of it felt wierd. I looked in my old Basic Circuits textbook and the phrase fundamental circuit element isn’t used at all.

From an EE perspective, the usual basic groupings are passive and active, and linear and non-linear, because that directly relates to how to mathematically solve circuits including the components. They ignore diodes in the paper because they do not cleanly fit any of their categories. Also the frequency domain behavior/description of each component is perhaps more important.

Someone asked me once if the resistor was really needed as a circuit element. I postulated that it was not desired, nor needed for any circuit that didn’t intend to waste energy.

Also, who needs brakes? They only slow you down!

Jake Brake (compression braking), regenerative braking. Those slow you down as well but do not use friction as the main working component.

Motion in, heat out, either way.

Regenerative braking is motion in, electrical power out. It is really just generation applied for reducing motion, but the power could as well be turned directly into heat if all you wanted was to avoid using up your brake shoes.

There is a third method, where you shift down and rely on friction in your engine block to heat up your oil, and indirectly the water jacket, and dump the excess heat to the radiator.

Engine braking actually supplies the engine with less fuel than is required to maintain the current speed, the engine then compresses the inlet air and releases the compressed air to the outside, this is the main element, the ones you listed are side effects of a much smaller magnitude.

You do understand that the air comes out hotter than it went in, right? And that the added heat came from the kinetic energy of the truck?

Yes I do. In the end it is all converted to heat and lost to the environment. The only question is when, not if. Even regenerative braking is postponing the inevitable.

I will bite and counter that resistors are needed to prevent wasting energy. Current limiting resistors are used in many different places like low power electronics (these are always very high resistance) and drivers for electric motors (a powered electric motor blocked at rest is the same as a short circuit).

Though you can do the same current limiting with a pulse width modulated switch (similar to a switching power supply).

My point being you can take a resistive element and replace it with a switched transistor. All ideally of course, since it’s not practical many times.

How would you make a constant-current power supply without resistors?

Any practical circuit has lots of implicit resistors of very high value (between any two unconnected points on the schematic).

If Memristors aren't a thing yet, they're never gonna be. This is HP CEO #3 to be pushing the cursed things when they're still not even market viable.

Could someone provide a context to this article? Why is it important?

But how does those results compare to capacitor?

Reply on news.ycombinator.com