The piece by Ashley Smart on page 16 of the January 2018 issue of Physics Today highlights the ongoing effort to develop rapidly switchable phase-change materials for computer memory and storage. Our work at IBM Research in the early 1980s helped jump-start that effort: It demonstrated for the first time fast crystallization and thermally stable data storage in a reversible phase-change material and clarified the underlying physics.1 

The key to short crystallization times was picking materials that do not require phase separation and associated diffusion to crystallize. The material must also have a glass transition temperature high enough to guarantee stability of the amorphous phase, into which the data are recorded. And it must be put into a properly designed thermal structure so that heat pulses of different duration can alternatively crystallize and amorphize it.

We demonstrated high-speed crystallization with non-phase-separating compounds such as germanium telluride and antimony telluride. We were aware of materials—pure aluminum, for instance—that crystallize so fast that they could not be thermally quenched to an amorphous phase. In conventional thin-film structures, GeTe suffered the same problem. Building it into a higher-cooling-rate structure, however, increased the quench rate and made amorphization possible. We also understood that amorphous GeTe, with a structure similar to a close-packed liquid, would often be kinetically disposed to form a metastable face-centered cubic crystalline phase. That extended the range of workable compositions beyond those suggested by an (equilibrium) phase diagram.

We shared our understanding with researchers at Matsushita Electric Industrial (now Panasonic), who then expanded the set of fast-crystallizing materials to include the GeSbTe-type materials. The smaller bit geometries associated with modern nonvolatile memories allow for higher quench rates and thus enable faster-crystallizing materials to be considered.

The latest research from China, as highlighted by Smart, is notable in introducing “rational design,” such as the doping of SbTe with scandium to seed heterogeneous nucleation. It raises an interesting question: How would the picture change if Sc were deposited in a discontinuous layer rather than dispersed throughout the SbTe film?

K. A.
R. W.
Appl. Phys. Lett.
Ashley G.
Physics Today