[sdiy] Memristor, posited in 1971, built by HP in secret, revealed today

[Joan Touzet]

Sorry in advance for the crosspost...follow-ups probably should go to sdiy.

A new basic type of electronic device, first postulated in 1971 by
Leon Chua (1), has been fabricated by HP and published today in Nature
Magazine and a few other notable publications:

  http://www.physorg.com/news128786808.html (with photo)
  http://www.hpl.hp.com/news/2008/apr-jun/memristor.html
  http://www.eetimes.com/news/latest/showArticle.jhtml?articleID=207403521&pgno=1
  http://technology.newscientist.com/article/dn13812-engineers-find-missing-link-of-electronics.html
  Original 1971 article reference:
http://ieeexplore.ieee.org/Xplore/login.jsp?url=/iel5/8147/23413/01083337.pdf?tp=&isnumber=&arnumber=1083337
(requires membership)

In short, a memristor is a current-controlled resistor. The device has
the unique property of being capable of remembering the last voltage
applied to it, and for how long it had been applied. The device has
been fabricated out of a bilevel titanium dioxide film, where one
layer has been doped through oxygen vacancies. The oxygen vacancies
lead to the hysteresis that is predicted by Chua, an indication of
flux-linkage (aka flux, indicated by lower-case phi). The
characterization of the device is (forgive my horrid notation):

q(t) = integral (-inf, t) i(t) dt
phi(t) = integral (-inf, t) v(t) dt

The point being that resistance arises from a relationship between v
and i, inductance arises from a relationship between phi and I,
capacitance arises from a relationship between q and v, and
*memristance* as a relationship between q and phi.

Chua's 1971 article, which I can't share here legally (check with your
local library to see if you have full IEEE Xplore access) goes into
great detail on the presumed device, even providing a complete
schematic diagram of a memristor tracer for tracing the device's phi-q
(flux-linkage vs. charge) curves. It's an accessible article if you
have some background in device physics, even just one course at the
university level.

The application everyone in the media is talking about is based on the
hysteresis-induced memory effect - zero-power memory storage, that
sort of thing - but I can't be the only one thinking "Hm,
current-controlled resistance...sounds like something I can make music
with! Sounds like a new device that implements what has been taking
many many components to date!"

There's also natural applications in delay lines, sequence playback
(set "samples" on each memristor, then play back in sequence of your
choice), neural net modeling (anyone remember the Intel integrating
chip a while back that did something similar? I can't recall its #),
crossbar matrices (one of HP's intended applications, actually)...
Even HP is saying that the interesting applications of this device are
in its analogue characteristics, not in the binary "on" and "off"
states. This is where they think it will act most like a neural
network.

Anyone up for a discussion of device physics and potential synth applications?

-Joan
-- 
Joan Touzet | joant at ieee.org | wohali at efnet