Operation mechanism of ferroelectric HfO2-based transistor and memory has been elucidated
Advancement toward ultimate scaling and ultralow power consumption
Credit: Masaharu Kobayashi
As a part of JST PRESTO program, Associate professor Masaharu Kobayashi, Institute of industrial Science, The University of Tokyo, has experimentally clarified the operation mechanism of low voltage operation of a transistor with ferroelectric-HfO2 gate insulator. In addition, he has theoretically elucidated scalability of ferroelectric tunnel junction (FTJ) memory with ferroelectric-HfO2 down to 20nm diameter.
Negative capacitance FET (NCFET) with ferroelectric-HfO2 gate insulator is attracting interests as a steep subthreshold slope transistor for ultralow voltage operation, which can break the physical limit of 60mV/dec in the conventional MOSFET. However, its operation mechanism has not been fully clarified yet in terms of polarization switching dynamics. FTJ memory with ferroelectric-HfO2 is a promising high-capacity nonvolatile memory. However, its scalability considering resistance ratio between read current for access speed, resistance ratio between on-state and off-state for sensing margin, depolarization field for retention characteristics has not been fully elucidated yet.
In this study, he has experimentally demonstrated sub-60mV/dec subthreshold slope in a transistor with ferroelectric-HfO2 and clarified that the physical mechanism is attributed to charge injection assisted by polarization switching in ferroelectric-HfO2 gate insulator improves subthreshold characteristics as shown in Fig. 1. Furthermore, he has systematically investigated the relationship between subthreshold characteristics and polarization switching by monitoring gate current with high resolution and clarified that polarization switching dynamics causes transient negative capacitance by depolarization effect due to small depletion layer capacitance, as shown in Fig. 2.
As for FTJ, he has established non-equilibrium Green function method with self-consistent potential involving polarization charge and semiconductor surface potential to calculate current through FTJ memory. This simulation framework is calibrated by experimental results of ferroelectric-HfO2 FTJ memory in the form of metal-ferroelectric-semiconductor structure, which was previously developed and demonstrated in the same JST PRESTO program. By utilizing this simulation framework, he has theoretically predicted that ferroelectric-HfO2 FTJ can be scaled down to 20nm diameter size by systematically investigating the impacts of material properties and considering the trade-off among read current for access speed, resistance ratio between on-state and off-state for sensing margin, and depolarization field for data retention characteristics, as shown in Fig. 3.
The achievements in this study will largely contribute to guiding device design of ultralow power operating NCFET and high-capacity FTJ memory, which leads to enabling ultralow power IoT edge devices, deploying highly sophisticated network system, and thus providing more strategic social services utilizing big data.
This work was presented in IEEE International Conference on Electron Device Meeting on December 4th and 5th in 2018, which was held in San Francisco.