Banner
    The Memristor: Another Science Scandal
    By Sascha Vongehr | August 7th 2012 02:51 AM | 20 comments | Print | E-mail | Track Comments
    About Sascha

    Dr. Sascha Vongehr [风洒沙] studied phil/math/chem/phys in Germany, obtained a BSc in theoretical physics (electro-mag) & MSc (stringtheory)...

    View Sascha's Profile

    In 2008, The missing memristor found [1] was published in the respected science journal Nature, and this claimed discovery was announced on the front pages of most major newspapers. This “discovery” is simply a misinterpretation of devices that had been discovered many years before in India [2,3]. Those original inventors did not misinterpret their work in order to make it into the news. Given the serious doubts that have been presented in many places, one seriously wonders whether the fact that the cheated are ‘just a bunch of Indians in India’ has anything to do with the embarrassing situation of that science media do almost not care. The latter though cannot explain that criticism is effectively censored.


    Finally, a thorough and fatal criticism has found its way into an established scientific journal: The technical part of “The Missing Memristor: Novel Nanotechnology or rather new Case Study for the Philosophy and Sociology of Science?” [4] has appeared as “Missing the Memristor” in Advanced Science Letters [5]. Momentum is gathering, and the papers are already quoted in further critical scientific articles like the new Fundamental Issues and Problems in the Realization of Memristors*.  The memristor has still not been found and may actually be impossible, but let us first see what this “memristor” is all about anyway.


    The resistor, the capacitor, and the inductor are three well known, basic two-terminal circuit elements. In 1971, Leon Chua postulated a fourth, the “memristor” [7], on grounds of symmetry arguments, i.e. for the sake of completeness of circuit theory. The three well known ones correspond to extremely simple devices.


    Illustrations of the symmetry that led to proposing the memristor: (a) the tetrahedron spanned by the four fundamental circuit variables (current I, voltage U, electrical charge Q, and flux phi), (b) the relations and circuit symbols of the four basic two-terminal circuit elements that correspond to the four uncolored edges (thin instead of thick black or red) in (a). [Source: [4,5], do not reproduce]


    The discovered “memristic” devices [2,3] are indeed interesting news for basic physics, electrical engineering, and nanotechnology; their potential applications in integrated circuits immense: They may allow extended functionalities incorporating analog computation and self-programming neural networks and may thereby help approaching information processing the way human brains do it. However, these devices are actually not memristors.


    Naively, the memristor issue could be added to success stories like Le Verrier’s prediction of the existence and location of Neptune in 1846 and Dirac’s prediction of the positron in 1928. Mendeleev’s prediction of undiscovered chemical elements in 1870 rested on the still empty cells in the periodic table of the elements. The memristor is similarly a vacant spot in a two dimensional table, though a much smaller one. The case is somewhat similar to Murray Gell-Mann noticing missing pieces in the patterns of SU(3) symmetry representations, which led him to successfully predict some subatomic particles. These mentioned historically predicted entities were discovered relatively soon after having been foretold.


    The eventual discovery of something that was naively expected to be as simple as a capacitor or inductor, which can be constructed from just two metal plates or a metal coil, respectively, with a delay of almost 40 years, is intriguing by itself. The long delay is perhaps a symptom of there being no such thing as a real memristor. The predicted memristor involves magnetic flux, but the discovered devices involve none. As if this would not be enough, the following certainly submits the memristor as a case for the sociology of science: Leon Chua did not reject the purported discovery in 2008, but instead, the aging and otherwise almost forgotten man quickly changed his mind about what a memristor is! One wonders why, and particular suspicions are strengthened by the flood of papers that was generated under the catchy “memristor” label (as well as the number of patent applications that rest with certain, extremely competitive companies that can draw on large financial resources - the memristor is not called "HP Memristor" for no reason!).

    A memristic device, not a memristor.

    The memristor discovery reminds of the purported detection of gravitational waves by Joseph Weber in the late 1960s, which after an initial acceptance was discredited in the mid 1970s and subsequently led to a number of analyses. More sensitive detectors could not reproduce Weber’s claims. IBM physicist Richard Garwin built a similar detector but could only find one pulse in half a year, and that was noise. Some rebuttals reviewed Weber’s data analysis. The physicist David Douglass found an error in Weber's computer program which combined noise and artifices due to how the data was divided into batches which then resulted in daily coincidence signals.

    The earliest and most damaging rebuttals however were of a theoretical, even philosophical nature, similar to those that apply to the memristor. Garwin pointed out that if Weber's detections were real, the universe would convert all of its energy into gravitational radiation in only 50 million years (the universe is much older than that!).


    In both cases, the entity in question was predicted on theoretical grounds but may conceivably not be verifiable. At the time of the first claimed detection of gravitational waves in 1969, their measurability was doubted even by those comfortable with general relativity: Would not the measure tape contract exactly along with the space contraction due to gravity waves?

    The originally proposed memristor may be impossible as a really existing device. Both times there are devices and real observations involved. As far as we know, Weber did not consciously make up any data; he just honestly reported what his visual system’s acute pattern-recognition told him (when confronted with noise in what is a naïve way of data analysis). Later, he reported what flawed computer software calculated. However, the cases are also quite different. Gravitational waves have been inferred from convincing astrophysical observations, but they still have not been detected. But if you ask for a memristor, you may be given an actual device and the claim “here it is, you are holding one in your own hands”. That object sitting in your palm is certainly not a misinterpretation of noise by software. If that object is not a memristor, something other than engineering has gone astray.


    Physics itself refused Weber’s claims relatively fast. Particularly those differences that make the memristor issue interesting to sociology seem to ensure that science proper will not deliver speedily again without some outside prodding. Not just one Joseph Weber fears for his reputation, but a well funded scientific sub-field wants to believe into the memristor story.


    The memristor debacle needs to be brought to the attention of a wide audience. There must be a focused and solid argument on the grounds of which the case can be widely accepted as controversial. “The Missing Memristor” [4,5] does precisely this. A self contained introduction on what the purported discovery is and why it is questionable reduces technicalities* to the bare minimum. Memristance and the memristor are carefully introduced and the discovered thin film devices are described in a selfcontained fashion. Explaining the memristor as a model of circuit theory and discussing its stability cast doubts on the necessity and viability of a memristor as an actual device.


    The real death sentence to the claimed “memristors” is spoken by a thought experiment that assumes a world without magnetic fields. In this hypothetical world, the discovered devices would have been discovered, too, but inductors would obviously not exist. Identifying the devices as memristors would predict a missing “inductor” on the same grounds as the memristor was historically predicted. And an “inductor” would be found although we know that inductors in that world are impossible.


    The early attempts at an analysis in terms of the sociology of science points toward a deepening crisis that has repercussions for the public trust in science and important related issues, say global warming denial: The publish-or-perish crisis that has corrupted science via the evolved peer review mechanisms.


    * There are many technicalities, for example [6] remarks already in the abstract that


    “The device will operate for a certain time like a "chemical capacitor"”.


    [4,5] anticipated this via:


    “That the independence requirement is only satisfied for limited (and practically short) times results in an intermediateness that we do not accept elsewhere. For example, the capacitor supplies energy when discharging and would not even count as passive if we were allowed to observe it for properly selected, short durations in order to categorize the device. Batteries are not capacitors and do explicitly not belong to the B2TCE [Basic 2-Terminal Circuit Elements], although they show similarities to de-charging capacitors. The latter is not a far fetched analogy given the necessary chemical (e.g. redox) reactions in the memristic thin film devices, which render the discovered devices indeed similar to batteries.”


    Our criticism should not focus only on such technicalities too much. The technicalities are a support, but the general argument is to my understanding much more powerful and understandable to a wider audience.


    --------------------------------------------

    [1] Dmitri B. Strukov, G. S. Snider, Duncan. R. Stewart, R. Stanley Williams: “The missing memristor found.” Nature 453, 80-83 (2008)

    [2] H. M. Upadhyaya, Suresh Chandra: “Polarity dependent memory switching behaviorin Ti/Cd Pb S/Ag system.” Semiconductor Science and Technology 10, 332-338(1995)

    [3] Suresh Chandra: “On the Discovery of a Polarity-Dependent Memory Switch and/or Memristor(Memory Resistor).” IETE Technical Review 27, 179-180 (2010)

    [4] Sascha Vongehr: “The Missing Memristor: Novel Nanotechnology or rather new Case Study for the Philosophy and Sociology ofScience? http://arxiv.org/abs/1205.6129

    [5] Sascha Vongehr “Missing the Memristor.AdvancedScience Letters 17, 285-290 (2012)

    [6] Paul Meuffels, Rohit Soni: “Fundamental Issues and Problems in theRealization of Memristors.http://arxiv.org/abs/1207.7319 (2012)

    [7] Leon O. Chua: “Memristor – the missing circuit element.” IEEE Trans. On Circuit Theory CT-18, 507-519 (1971)

    --------------------------------------------

    More from Sascha Vongehr sorted Topic for Topic

    Comments

    Nice article. Thanks for reporting facts and not hype.

    This article is just "piling on" on the anti-memristor bandwagon.

    HP did not claim to be the first to create resistive oxide RAM or polarity-dependent memory switching, etc.

    What they claimed was recognition of a principle or law: How memristor theory simplifies understanding of, FOR EXAMPLE, the switching phenomena observed in TiOx memory cells at nanoscale dimensions.

    They noted that there were numerous observations of pinched hysteresis loop phenomena in physics and engineering literature, but everyone overlooked the connection with memristors and, consequently, missed out on the simplicity that connection provided for modeling.

    They also provide their own, convincing, explanation of why the wait time between theory and discovery: physics at nanoscale dimensions had to become common. For this is a prerequisite for a large and important class of memristive phenomena.

    I'm waiting to see proof or refutation of whether memristive modeling sped up the development of resistive memories.

    That is my necessary and sufficient criterion for the veracity of HP's claims.

    vongehr
    Thank you for your interest and your on topic comment. Allow me to reply straight to the point:
    This article is just "piling on" on the anti-memristor bandwagon.
    "anti-memristor bandwagon" seems to be a political term. The memristor is a well defined scientific concept.
    HP did not claim to be the first to create resistive oxide RAM or polarity-dependent memory switching, etc.
    I never claimed such.
    but everyone overlooked the connection with memristors
    No - they did not overlook (perhaps some did not focus on the modeling aspect via memristor theory as much as you think is necessary in your personal opinion). The simply understood that these memristic devices are not the fourth basic circuit element called "memristor".
    physics at nanoscale dimensions had to become common.
    The memristic nano devices need to be nanoscale because their memristic behavior depends on the square of the size D as explained clearly in [4,5]. This does not imply that the memristor has anything to do with that.
    whether memristive modeling sped up the development of resistive memories ... sufficient criterion for the veracity of HP's claims.
    Use of the memristor as a physical modeling tool (known for a long time), as well as the success of nanodevices in the marketplace, both have nothing to do with that a predicted scientific entity, here the memristor, has still not been discovered and may be impossible. It is simply unethical corruption of science to claim such a discovery in order to generate hype. This is no different from the hyping of arsenic life forms. It harms the recognition of good science by other conscientious scientists and it harms the public perception and trust in science as an important input into society's decision processes.
    Mathguy
    Great article.  I was just having a discussion the other day about how it's hard to tell whether what you read is "good" science.  Also great, direct responses to the above comment.
    Mark
    It seems like my last comment didn't make it past moderation so let me try again. In order to recognize a memristor I will need to know what it is and how it works. The problem with this is that Chua only defined it in terms of what it does and not by how it works. It is true that he was motivated by a symmetry between electricity and magnetism but he gave no mechanism involving either electricity or magnetism.

    For example batteries and capacitors are defined differently by their different working mechanism. If we didn't know how either worked then our motivation for defining them differently would be suspect.

    The only important question for me is can HP's circuits do everything a memristor can do? Think of it as a memristor version of the Turing test. Is there any test in terms of input and output that would allow you to tell the difference between HP's device and a real memristor? I ask because I really don't know.

    Also what if I connected an electric motor shaft to a potentiometer and connected a voltage across them in parallel. Now the more current that flows in one direction the higher the resistance will be. Reverse the flow and the resistance is reduced. Is this a memristor? It even uses a magnetic field.

    Again, I need a precise definition of a memristor in terms of mechanism before these questions even have an answer.

    vongehr
    The definition as the fourth basic two-terminal circuit element that completes the symmetry is quite rigid, especially since we already know all the other three. If there were a detailed description of what exactly a memristor is in terms of mechanisms, one would just build one according to that description; one would not have to discover anything.
    The 1971 description asks for a certain interaction between magnetic and electric fields as being crucial (there are clear quotes on that), which means that whatever a memristor is, those thin film memristic devices are not it.
    I have reason to suspect that a memristor is impossible as a real device, so I am neither surprised about there being no precise prescription of its mechanisms, nor is it strange that it has not been discovered for such a long time.
    I am not sure what can satisfy you if you "need a precise definition of a memristor in terms of mechanism" and the necessity of magnetism is not enough to exclude anything that works without magnetism.
    But is the potentiometer motor combination a memristor? It involves magnetic fields. Your "rigid" definition would seem to say that it is.

    vongehr
    It needs to involve the magnetic field in a certain way crucial to the mechanism. You can use a motor that does not use magnetic fields. 1971-Chua asked for EM-interaction that suggest a memristor, if possible at all, is an AC device.
    I am sorry but maybe you have to read the older Chua articles a bit more carefully. In the definition of the four basic variables, charge, flux, current, and voltage, the flux variable does not necessarily need have the physical meaning of magnetic flux. It is just defined as the integral of voltage. A similar argument holds for the charge. In that sense memristors are devices that integrate upon the voltage across it and the currents running through it - hence the memory effect.

    Nevertheless, there appear to be some recent results that proof that in the "HP memristor" magnetic flux is actually present; see e.g. arxiv.org.

    vongehr
    the flux variable does not necessarily need have the physical meaning of magnetic flux.
    Wrong, and 1971-Chua insists on magnetism. I wrote this further article to explain this in detail: The important aspect is not some re-interpretation of some flux definition, but as usual with such predictions of scientific entities, a fundamental symmetry! You can define as many fluxes as you want and then get as many devices, but you will not get precisely 2 times 2 equals 4 devices; you cannot sell your "flux" thingy as the fourth basic two terminal circuit element that completes circuit theory.
    The nonsensical arxiv article you mention is also discussed there.
    The problem here is that you mix up two different frameworks by persisting in considering the flux and charge as defined in circuit theory as real physical quantities. Flux linkage is a generalization rather than an equivalent of magnetic flux. In circuit theory it is simply defined as the integral of voltage. Nothing more, nothing less.

    Prior to Section IV of Chua's 1971 paper (and all his papers that follow**), Chua defines the time-integral of voltage (current) as flux linkage (charge) without the specific reference to "magnetic" flux ("electric" charge). What is perhaps a bit confusing is the electromagnetic field interpretation of Section IV in that same paper. As circuit theory is a (very) limiting case of electromagnetic field theory, things derived or predicted from circuit theory do not necessarily pass on (back) to the electromagnetic domain.

    Furthermore, in circuit theory we just consider the terminal behavior of an element by observing (measuring) its voltage across it and the current running through and do not care about the physics. In that sense the flux and charge variables are just dummy variables to enable us to equate functional relationships and classify the elements. And this is precisely what the people from HP did. They observed a pinched hysteresis loop in the voltage-current plane, which has an underlying well-defined functional relationship in terms of the integrals of this voltage and current. Hence providing an example of a memristor as defined by Chua.

    I do, however, agree that a memristor is not the fourth and final missing element that completes circuit theory. This is indeed misleading. There are other elements with similar properties, like the meminductor and the memcapacitor. I would prefer to refer the resistor, inductor, and capacitor as the trinity of basic circuit elements, and the memristor, meminductor, and memcapacitor as the trinity of memory circuit elements (or mem-elements).
    ________________
    **A comprehensive overview of Chua's work on the family of circuit elements, including the circuit-theoretic definition of a memristor prior to HP's Nature paper, can be found in "L. O. Chua, “Nonlinear circuit foundations for nanodevices,. Part I: the four-element torus” Proc. IEEE 91, 1830–1859. (2003)".

    vongehr
    I do, however, agree that a memristor is not the fourth and final missing element that completes circuit theory.
    The memristor is !defined! that way. Being the fourth and last element is what the symmetry is all about, the memristor is all about, and what the big hype in 2008 was all about. Do you think that the New York Times would have cared if it is not about the last predicted basic element that completes an important theory but instead some stinking little engineering device in a long row of however many flux-definitions?
    Why even all this blah blah if you 100% agree that the 2008 hype about finding the last missing element is wrong? Because HP has money? Why writting about whatever other flux definitions that whoever came up with and memristic devices or memistor-like objects or what not? The memristor was defined a certain way, and it was hyped that way and that is an outright deception, period. The memristor has not been found!
    Chua's later papers about memristic whatevers are beside the point! That he did not in every single line where he used the flux-voltage formulas explicitly wrote "this is magnetic flux" is beside the point. It is clear from the whole paper that magnetism is crucial as far as the memristor, i.e. the fourth basic element, is concerned.
    we just consider the terminal behavior of an element by observing (measuring) its voltage across it and the current running through and do not care about the physics.
    Perhaps you actually do that, sure, science is in decline, but if you want to tell me that there is an inductor in a world without magnetism because you guys define flux integrals however you feel like or just measure voltage and current at a black box (perhaps with a boy inside who plays with a potentiometer), let me tell you something: There is no inductor without magnetism and if you derive such nonsense, it is high time to stop.
    What is the importance of the distinction between linear and non-linear mem-element?

    What is the importance of the distinction between linear and non-linear mem-element?

    Linear mem-elements coincide with the linear basic elements, like a linear memristor is just an ordinary resistor. Mem-elements are essentially nonlinear.

    >>Linear mem-elements coincide with the linear basic elements, like a linear memristor is just an ordinary resistor. Mem-elements are essentially nonlinear<<

    Really?!! I did not expect that answer. I will have to study on that.

    I like the symmetry of your mem-elements conception. Was it an error that led Chua to his memristor theory? Why did he not postulate the other two at the same time?

    vongehr
    A linear memristor is in some sense just a resistor, but that does not mean that a non-linear resistor is a memristor. The possibility of a certain non-linearity (relative to a certain parameter space - everything is non-linear in some space) in the memristor case is one way to see that it is independent from a resistor (the four basic elements need to be intependent just like vectors of a basis of a four dimensional space, because otherwise it is not four dimensional at all). Apart from that, the linearity issue is not important.
    Chua did not do any mistake then. He defined the memristor and he afterwards published about memristic devices, and at times he was not too careful with keeping them apart. What he did wrong is to help HP to hype an already much earlier discovered memristic thin film device as the finding of the memristor. That was a stunt much like the arsenic life forms, and we see more and more of this kind of gaming in science, so no wonder the public does not trust it anymore. Why should anybody trust cheaters telling us the climate is warming?
    Chua came up with the 4-element symmetry in 1971 and refined his ideas later on, which finally led to the universe of higher-order elements or so-called (alpha,beta)-elements (see e.g. the 2003 paper cited in my earlier post). Recently it was realized that the (-2,-1)- and (-1,-2)-elements correspond to, respectively, inductors and capacitors that, like the memristor, retain memory of their past state. Call it progressive insight ;-)

    Being the fourth and last element is what the symmetry is all about, the memristor is all about, and what the big hype in 2008 was all about.

    Well, don't believe the hype, haha. Blame the blunt media people, throw out your TV and don't buy any newspapers. Its all b*llshit, I agree. Just read a little further than the 1971 paper and be amazed by the universe of circuit elements!

    ...but if you want to tell me that there is an inductor in a world without magnetism because you guys define flux integrals however you feel like or just measure voltage and current at a black box (perhaps with a boy inside who plays with a potentiometer), let me tell you something: There is no inductor without magnetism and if you derive such nonsense, it is high time to stop.

    Of course magnetism is involved in a real inductor. You can all nicely describe and predict its behavior by electromagnetic field theory or visualize the magnetic field lines using iron powder. But again, if you like it or not, from a circuit-theoretic perspective an inductor is essentially a device that causes current to lack on the applied voltage, whereas a capacitor does the opposite. Flux in circuit theory is unambiguously defined as the time-integral of voltage. Why don't you accept this fact and accept that there is a difference between physics and circuit theory?

    From a circuit-theoretic perspective, an example of a memristor has been found. Not only by HP, but it characteristics (pinched hysteresis in the i-v plane) can be found in many devices.

    vongehr
    First of all: Science/Nature are not TV and blunt media. They are supposedly the most reliable, finest scientific journals out there, so thank you for agreeing with me that too much complete nonsense is found in those journals.
    Now to your argument. If you do not need magnetism for inductors, then why only one "inductor" in your "non-physics circuit theory"? After all, there could be a different kind of x-magnetism with an entirely different x-inductor. You would have to come up with a new kind of flux definition in order to show that also that physical x-magnetism is not necessary for circuit theory to have all the inductors. Well do it then! Where are your 2*2*2=8 basic circuit elements (or at least 6 if the symmetry between different x-magnetisms should be broken). The fourth element is defined and was still in 2008 sold as a completion! Your argument, although utterly unscientific (Electrical engineers do not think their stuff is nothing to do with physics), supports my point, namely that what they found is not that completion.
    You are absolutely correct. When I first read of the memristor, I wondered what new physics had been discovered. It turned out to be some non-linear device built out of wodge. Just one step away from simulating the whole thing in Mathematica. Talk about dishonest hype! <spit>....