Monatomic glassy antimony might be used as a new type of single-element phase change memory. This is the new finding from researchers at IBM Research-Zurich and RWTH Aachen University who say that their approach avoids the problem of local compositional variations in conventional multi-element PCMs. This problem becomes ever more important as devices get smaller.
New-generation non-volatile memory
The worldwide volume of digital information is doubling every two years and could reach 160 zettabytes (109 terabytes) by 2025 according to the latest whitepaper from the International Data Corporation (IDC). Phase change memories are one of the new types of non-volatile memory being studied to meet this demand. These memories are based on a material’s ability to switch between two “0” and “1” states: a crystalline state with high electrical conductivity and a meta-stable amorphous state with low electrical conductivity. They are switched using electrical pulses that heat up the material and drive the transitions. The energy of the electrical pulses is lower when there is less material to heat up.
Conventional PCMs are usually made from a complex mix of alloys doped with additional chemical elements to tune their physical properties. While such materials can be used to make chips with good data storage densities, these could be increased further by scaling down the cell size of memory units. There is a problem, however, in that the smaller the device, the more sensitive it becomes to local compositional variations in the alloy, which deteriorates the cell’s properties.
“Our work shows that we can solve this problem by making the PCM from just one simple element instead of these complex doped alloys,” explains Martin Salinga, lead author of this study. “Antimony (Sb) is semi-metallic in its crystalline phase and semiconducting as an amorphous thin film and shows a large contrast in resistivity between these two states. It can also crystallize very easily and quickly. This makes it a good choice for a PCM in a highly-confined structure, which usually slows down the crystalline kinetics.”
Rapid melt-quenching in a nanoconfined volume
The researchers, reporting their work in Nature Materials 10.1038/s41563-018-0110-9, made pure Sb films that are between 3 and 10 nm thick and confined inside thermally and electrically insulating SiO2 layers that are 40-200-nm thick. They were able to electrically switch between the amorphous and crystalline states in these films in just 50 nanoseconds.
Until now, it had been difficult to make amorphous Sb because the element rapidly crystallizes at room temperature. Salinga and colleagues have now managed to do this by rapidly cooling (or quenching) the material from the melt at a rate as high as 1010 kelvin per second in a nanoconfined volume. The result: amorphous Sb that is stable for nearly 51 hours at 20°C.
Immediate applications
“The first applications that could benefit from a ‘monatomic PCM’ might be in the area of ‘in-memory’ computing, ‘memory-type storage class memory’ or ‘brain-inspired computing’,” IBM scientist and study co-author, Abu Sebastian tells Physics World. “These devices could be operated with 10-ns-long electrical pulses. We will likely be able to scale these devices down to ultra-small dimensions that will consume very little energy. Their monoatomic nature might also make them more robust to repeated switching cycles.”
It is not all plain sailing though: the amorphous state of Sb only lasts for around 100 seconds at 60-70°C, which is the typical operating temperature inside electronic devices, so the researchers say that this will have to be improved. “This may be achieved, for instance, by further reducing the Sb film thickness, confining Sb in all three dimensions, and designing better confinement materials,” suggest Wei Zhang and Evan Ma at Xian Jiaotong University in China and Johns Hopkins University in the US in a related Nature news & views article. “The voltage pulse (currently 50 ns) required for amorphization (also) needs to be shortened to become competitive with DRAMs and SRAMs.
“What has been achieved by Salinga and colleagues is nevertheless unprecedented and eye-opening, in terms of the perspective that monatomic PCMs are indeed feasible, and that an elemental glass, usually considered impractical due to its poor glass-forming ability, may be rendered useful in memory devices,” they add.
