Thursday, July 26, 2012

Memristor or not - discussion in Wired.

Earlier in the month, Wired reported that HP is planning to bring their TiO2-based resistive memory to market in 2014.  Resistive memory is composed of a bunch of two-terminal devices that function as bits.  Each device has a resistance that is determined by the past history of the voltage (equivalently current) applied to the device, and can be toggled between a high resistance state and a low resistance state.   In HP's case, their devices are based on the oxidation and reduction of TiO2 and the diffusion of oxygen vacancies.

This announcement and reporting apparently raised some hackles.  Wired has finally picked up on the fact that HP's use of the term "memristor" to describe their devices is more of a marketing move than a rigorous scientific claim.  As I pointed out almost two years ago, memristors are (in my view) not really fundamental circuit elements in the same way as resistors, capacitors, and inductors; and just because some widget has a history-dependent resistance, that does not make it a memristor in the sense of the original definition.


David Cobden said...

It's moderate hype by modern standards. At least there's something substantial there - I think!

Nevertheless, it seems to me that the basic distinction is simply that the 'memristor' is nonlinear: voltage goes as current times integral of current. Impedance is by definition a linear property. Once you allow nonlinearity the game changes (to a vastly more complicated one). To claim that the memristor is on a par with the three linear elements is just ridiculous.

Anonymous said...

Anonymous said...

Is there any response of HP's research to the recent publication "Fundamental Issues and Problems in the Realization of Memristors" (on There are strong arguments that the physics behind the "memristive systems" approach might be in conflict with Landauer’s principle of the minimum possible amount of energy required to change stored information in a system.

Anonymous said...

It's really astonishing that up to now nobody seems to have refuted the arguments given in the publication "Fundamental Issues and Problems in the Realization of Memristors" (arXiv:1207.7319v1): The physics behind the generalized concept of memristive systems such as that described by the HP memristor model seems to be in conflict with fundamentals of irreversible thermodynamics. Following the dynamic state equations, one would be able to violate Landauer's principle of the minimum possible amount of energy required to erase "information" states in a system.

Those who are familiar with Landauer's principle and irreversible thermodynamics should have understood what can be read between the lines in the above publication: There are serious concerns that the generalized concept of memristive systems might be related to a modern, electronic version of a "perpetual motion machine of the second kind".

Anonymous said...

Where is "science" moving to?
Call a resistance-varying device "memristor" and you get all the public. Sadly for those scientists who do serious and solid research and avoid any catchy/advertising labeling.

Anonymous said...

Some in the scientific community seem now to catch on to the problems of the theoretical “memristor” concept:
"On the physical properties of memristive, memcapacitive, and meminductive systems" by M. Di Ventra and Y. V. Pershin (Nanotechnology 24, 2013, ).