Dopant and Defect Physics for Device Optimization for Hafnium Oxide based Devices

点击翻译按钮获取中文摘要

Devices realized with ferroelectric hafnium oxide are silicon compatible, power-efficient, and can be cost-effectively integrated into advanced technology nodes for sensor, nonvolatile memory, logic, and neuromorphic applications. Currently, hafnium-zirconium mixed oxide (HfxZr1-xO2) offers the widest stoichiometry window for fabricating ultrathin ferroelectric films with large remanent polarization. Still, the film requires oxygen vacancies (VO) to stabilize the ferroelectric phase and has reliability issues. VO concentration is difficult to control and can affect electrical properties such as polarization stability, jeopardizing the prospect of hafnia-based memory and low-power logic. An alternative could be to start from stoichiometric, quasi-vacancy-free hafnia and use suitable dopants to optimize the ferroelectric properties. This would have the significant advantage of being more reproducible than the rather uncontrolled generation of VO in HfxZr1-xO2. We will explore this possibility by studying the atomic and electronic structure of selectively and controllably doped hafnia using ab initio calculations and phase-field simulations to describe the influence of the dopant modulated atomic and electronic structure on the ferroelectric properties. A range of dopants, concentrations, and process conditions will be considered to provide an initial assessment of the correlations between dopant chemistry and the ferroelectric properties. The chosen materials will be characterized on large area capacitors and optimized by successive simulation, processing, and characterization iterations. Then they will be integrated into scaled capacitor arrays to provide statistically significant results on ferroelectric capacitor performance. At a fundamental level, D3PO will give a better understanding of the influence of dopants on local chemistry, electronic structure, phase composition, and their effects on material and ferroelectric parameters, including recrystallization temperature and remanent polarization. We will then elaborate physical models based on real devices, using parameters obtained from ab initio simulations and structural and electrical characterization to predict, through statistical analysis, key metrics such as wake-up, endurance, retention, leakage, and breakdown using vacancy-free doped hafnia.