Slashdot Mirror
Nanoscale Switches in Memory
Frans Faase writes "At the university of Boston, researchers are using nano-scale mechanical switches as a novel technology for building memory. These switches are extremely small, require only femtowatts of power to switch, but still can switch at speeds of 23.57 megahertz. And they are expected to become even smaller and faster and are expected to overcome the theoretical limit of 100 gigabits per square inch capacity for magnetic media."
femtowatt is one quadrillionth (10 ^ -15) of Watt
sorry to nitpick but its "Boston University" not "the University of Boston"
BU Alum '84
I don't know about you, but I've never heard of a "University of Boston". Are they neighbors with Caltech University or Georgia Tech University perhaps?
Do editors even know what I'm talking about?
At the university of Boston
Its Boston University, my alma mater.
Indeed, electronic devices fail due to having moving parts. It's called Electromigration.
The Tao of math: The numbers you can count are not the real numbers.
While the article mentions these switches being extremely robust, what have they done to address some of those older issues?
Nanoscale is ATOM scaled, and the mechanical issues are difficult but ENTIRELY different from the mechanical issues of machines 10000 times larger.
At the atom level, chemistry is the right frame of mind, not thermodynamics/friction/dust/surface quality.
Since a Femto-watt = 10^-15, 10 billion or more would be required to reach even 20 milli-watts. I believe that even then a moderate amount of shielding would keep this from radiating too far.
You have enemies? Good. That means you've stood up for something, sometime in your life.
The effect they are using is a non linearity in the restoring force of a doubly clamped beam. It is well known that if you have a nonlinear restoring force F = kx + k_3 x^3, for sufficiently large driving power, the amplitude close to resonance becomes bistable. (This system is called a Duffing oscillator).
switching from one stable state to another is accomplished by driving the system at sufficiently high or low power, such that the bistability vanishes and the system is force into the high or low state. A simple hysteresis problem ...
For an example of a Duffing oscillator in a related system, look at figure 3 in this publication http://www.its.caltech.edu/~postma/pdf/APL83_1240. pdfAppl. Phys. Lett 83, 1240 (2003) (pdf file) ... yes, this is a shameless plug ;-)
The thing is, with any form of solid state data storage, a transistor is the basic unit of an IC, and a 32 megabit FLASH (or other NVRAM) chip is about as complex (in terms of number of transistors) to produce as a 32 megabit DRAM chip.
What that means is that the cost of NVRAM drives will always remain on the same order of magnitude of cost / capacity as RAM, whereas you need better than 1/100th of that cost/capacity to compete with hard drive storage. That $70 for one GB of FLASH storage simply doesn't stack up well when compared to spending the same $70 or so to get 80GB or more of storage.
Also, the speed needs to be improved markedly. Right now, I've seen CF cards rate themselves at 4x, I think to compare itself with CDs, that means a paltry 600kBps. Solid state storage in this case needs to improve by 100x as well.
In short, don't look for a solid state hard drive replacent for a good long while.
Each of these switches is probably smaller than any particle shedded by normal wear-and-tear, and also smaller than the surface features that the whole concept of friction is based on.
Actually, friction increases with decreasing size. For nano-sized particles friction is one of the dominant effects and often cause microelectromechanical devices to fail due to "stiction"; one piece of the machinery semi-permanently sticking to another due to van-der-Waals forces.