The Smallest Bit: Ultimate Limits of Phase Change Memory William P. King and Eric Pop University of Illinois Urbana-Champaign
The objective of this research is to investigate the ultimate scaling of phase change memory de-vices, below 10 nm bit size. Phase-change materials (PCM) undergo a reversible phase change accompanied by a drastic change in resistivity, induced by electric and temperature fields. PCM are prime candidates for fast, high-density memory with ultra-low power consumption. Such a technology would enable scaling of memory devices much beyond the present state of the art, represented by Flash memory or other charge storage devices like DRAM or SRAM. The ap-proach is to investigate the fundamentals of nanometer-scale electric and temperature fields that induce phase change, resulting in an understanding of the smallest data bits that can be formed in PCM. Specifically, the proposed work will perform experiments and simulations that determine the smallest addressable PCM bit using scanning probe techniques and carbon nanotubes as the electrodes.
The intellectual merit of the proposed research lies in its thorough approach for achieving inde-pendent control of nanometer-scale electric fields and temperature distributions in PCM. In turn, these will allow a significant advance in understanding the behavior of materials used in phase change memory.
The research will achieve broad impact by providing information about the ultimate speed, size, and longevity limits of future data storage systems. This new understanding could bring about radical changes in consumer electronics devices. The research will achieve additional broad im-pact through web-enabled communication, and personal interactions with high school teachers, undergraduate students, graduate researchers, and U.S. industry.
A central issue of nanoelectronics concerns their fundamental scaling limits, i.e. the smallest and most energy-efficient devices that can function reliably. Computing and data storage based on spin, resistive switching or phase change has received considerable attention as charge-based electronics approach their fundamental scaling limits, particularly due to charge leakage and power dissipation issues. Phase-change materials (PCMs) such as the chalcogenide Ge2Sb2Te5 (GST) are particularly interesting for applications in electrical or optical data storage and reconfigurable electronics. With traits such as potential for high density, fast access time, high endurance, robust data storage and non-volatility, PCM has the potential to become a ‘universal’ memory, which could replace all data storage from the non-volatile but slow flash memory to the high-speed yet volatile DRAM. Unlike conventional charge-based electronics, PCM-based devices store data as the state of the material which can be reversibly switched between a high-resistance amorphous and a low-resistance crystalline phase. This behavior renders them immune to leakage as non-volatile memory for low-power electronics, and relatively impervious to radiation damage for remote terrestrial or space applications. PCM switching is induced by Joule heating through voltage pulses for electrically-programmable PCM devices. However, one of the main drawbacks associated with PCM technology is its relatively high programming current (~0.1–0.5 mA) and power. This limitation is inherent in the layout of current PCM devices where heat must be transferred to a significant volume of the PCM bit. In this project, we used carbon nanotubes (CNTs) as nanoscale electrodes to induce reversible phase change in an extremely small volume of the GST bit. With diameters less than 2 nm and excellent electrical conductivity, CNTs could serve as one dimensional nanoscale electrodes for PCM devices. This approach significantly lowers the programming current and power of PCM devices and provides an excellent platform to study their scalability. Reversible switching between the crystalline and amorphous phase in GST is typically driven by Joule heating; however, Peltier, Seebeck, and Thomson effects have been observed to contribute to phase change. Previous studies have shown the thermopower for bulk and thin-film depends on the phase of the GST with a large thermopower (200-400 µV K-1) for face-centered cubic (fcc) GST and a smaller thermopower (15-50 µV K-1) for hexagonal close-packed GST.. Electrical contacts and thermal interfaces to GST are also important for heat generation and thermal confinement of GST devices. Recent work has indirectly and separately measured the role of interfaces and thermoelectric effects in GST devices. These are essential, because electrical and thermal interfaces could reduce PCM programming power by 20-30%, and thermoelectric effects may reduce power consumption an additional 20-40% depending on the thermopower of thin GST films. However, direct observations of Joule, Peltier, and current crowding (interface or contact) effects at the nanometer-scale in a PCM device are still lacking. We measured the nanometer-scale temperature distributions in lateral PCM test devices. We used an atomic force microscopy (AFM) based thermometry technique known as scanning Joule expansion microscopy (SJEM). The measurement’s spatial and temperature resolutions were ~50 nm and ~0.2 K. We also improved the sensitivity of the SJEM technique for observing thermoelectric effects; allowing for observation of Peltier effects in hcp GST devices with relatively low thermopowers. Joule heating dominated the temperature rise of the GST; while the temperature rise at the GST-TiW contacts consisted of Joule, Peltier, and current crowding effects. Transfer length method (TLM) measurements on devices with varying lengths yielded GST electrical resistivity and GST-TiW electrical interface resistivity for our mixed amorphous, fcc, and hcp GST films of 11, 22, and 25 nm thickness. We observed uniform channel heating in thick (25 nm) and predominately fcc GST devices, and we observed heterogeneous channel heating in thin (11 nm) and mixed fcc-hcp GST devices. Comparing SJEM measurements to a finite element analysis (FEA) model, uncovered the thermopower of the thin GST films. The large measured thermopower of GST could reduce the energy consumption by >50% in highly scaled PCM devices due to Peltier heating, compared to scenarios which only utilize Joule heating.
