Up to now, the read/write speeds only amounted to a few 100 MHz because the precession of the spins is very slow. Now, Hans Schumacher of the German Physical-Technical Institute in Braunschweig (PTB) has shown how this can be done faster: Very short impulses address the memory cell, adjusting the spin by 180 or 360° instead of precessing. Others have also had this idea, but the problem always was that the process only worked in one cell -- one bit -- and not in realistic memory chips in which multiple cells are addressed in one matrix. The short, but strong electric impulse affects neighboring cells.
Schumacher has now managed to develop a special series of pulses to solve this problem; he calls it "ballistic addressing." The step from the research lab to the development laboratory will reveal whether this new approach will make MRAM market-ready. Time is crucial, for the cells cannot keep getting smaller. They are now approaching what is called the superparamagnetic limit, above which a magnetic element is no longer constantly remagnetized by thermal energy. A smaller cell would not have any remanence and could therefore not be used as a storage module. Depending on the material used, this limit is around 10 nm -- in an area that lithography techniques should reach in about 15 years according to the semiconductor roadmap. In light of this deadline, MRAM proponents will have to produce a working product as quickly as possible.
MRAM record: memory cell runs at 2 GHz
Posted: Aug 01,2005 by Ron Mertens