A recent paper from Feynman Prize winner James Tour’s group at Rice
relates an interesting new form of memory based on a bistable 2-terminal
graphitic switch. Once developed, the switch could form the basis
of a high-density non-volatile storage which might replace flash devices
(which are already beginning to replace magnetic disks).
The way the device works is straightforward: put one volt across the
switch, and if it’s on, it will conduct several microamps; if it’s
off, it conducts a few picoamps, an easily-measurable factor of a
million (or more) difference. Put 6 volts across it, and it turns
off, as if you were blowing a fuse. But put 4 volts across it and
it turns back on again! The paper proposes that there is a physical
configuration change, and the device acts as an electrostatic relay.
The group did extensive testing on the devices and showed that they
are remarkably robust, operating for thousands of cycles at a wide
range of temperatures and even after having been zapped with x-rays.
The devices themselves are essentially just nanocables made by depositing
a layer of graphene on a SiO2 nanowire by CVD. The interesting point
is the switching behavior appears to occur at defects in the graphene;
if the wire is too perfect, it doesn’t work. Since the defect size
is comparable to the cable width, the entire active part of the switch
fits in a 100 nanometer cube or so.
How soon are we going to have these in our computers? It’s important
to understand the amount of development that has to be done between
any laboratory advance and commercialization. So note that the following
remarks apply to virtually every such promising development you hear
about:
- The devices are essentially made, or at least processed, by hand one
at a time – the lab work involved picking 35 working devices out
of 48. Commercialization would require some way of reliably making
billions of devices with properties similar enough that the same driving
voltages, sense amps, and other supporting circuitry could be used
for all of them. - The physics are not completely understood. (For example,
a paper out of Bockrath’s group at Caltech
proposes a somewhat different mechanism for a similar switching behavior
in a non-nanocable graphene device.) This isn’t actually as important
as you might think, given that the useful behavior is robust and well-characterized.
But it would help with the ultimate reduction of the devices to the
lower limits of scale. - Given a cheaply-manufacturable array of a billion perfectly-working
devices, you’d still have to develop the read-write circuits, interfaces,
packaging, standards, and so forth.
All of this takes time, not to mention the investment of substantial
development resources. A general rule of thumb is that from a lab
demonstration, even one as extensively tested and well-characterized
as this, expect a decade or so before you have it on your desk.
The long-term promise of this kind of discovery is that there are
structures accessible to current-day fabrication techniques such as
CVD which exhibit this switching/memory behavior at what is apparently
quite close to atomic scale. This can only improve (particularly in
density — probably approaching the 10-atom dimensions of the graphene layer)
as fabrication technology approaches and ultimately attains
atomic precision.
