Novel material and device for ferroelectric memory: thin Hf1-xZrxO2 film and tri-states memory

Abstract

Ferroelectric random access memory (FeRAM) has been considered as one of the best candidates for universal memory. On the other hand, difficult scaling of the memory cell size has hindered the realization of high density FeRAM. Given that size scaling is inherently limited by the complicated crystal structure and difficult processing of ferroelectric materials, exploring new material or new device might be the best solution for that. In this dissertation, Hf1-xZrxO2 thin films and tri-state memory is studied as new material and device for next-generation high-density ferroelectric memory. As the first step, the origin of the ferroelectricity in Hf1-xZrxO2 thin films is examined based on their crystallographic structure, micro-structures and resulting in-plane stress. Although it seems evident that the formation of the non-centrosymmetric Pbc21 orthorhombic phase causes ferroelectricity in the doped and alloyed HfO2-based films, the origin of such evolution has not yet been elucidated. From the electrical and physical characterization on Hf1-xZrxO films with various film thickness (tf) and composition, ferroelectric orthorhombic phase is formed with the composition of ~0.4 - 0.6 and tf of < 25 nm. Even though these conditions are appropriate for the formation of tetragonal phase due to the size effect, the unexpected ferroelectric orthorhombic phase is formed. To elucidate the origin of the phenomenon, the film stress calculated from the change of substrate curvature and that of interplanar distance calculated from the change of diffraction peaks of X-ray diffraction spectra were analyzed, and it could be noticed that the unexpected orthorhombic phase is formed with the tensile strain of >1.5% and the grain size of < 25 nm. The large tensile strain is formed due to the huge tensile stress during the stage of island coalescence of Volmer-Weber type growth. In addition to that, the Hf0.5Zr0.5O2 (HZO) films on Pt bottom electrode (BE) were deposited with tetragonal phase with (111)-prefered orientation, and they hardly showed ferroelectric behavior. On the other hand, the in-plane strain of HZO films on Pt BE was almost equivalent with that on TiN BE, which is large enough for the formation of orthorhombic phase. However, the stress along a-, b-, and c-axis are almost equivalent for the case of (111)-oriented films, so it does not meet the condition of asymmetric stress for orthorhombic phase formation. In addition to that, the effects of annealing temperature (Tanneal) and film tf on the crystal structure and ferroelectric properties of HZO films were examined. The HZO films consist of tetragonal, orthorhombic, and monoclinic phases. The orthorhombic phase content, which is responsible for the ferroelectricity in this material, is almost independent of Tanneal, but decreases with increasing tf. In contrast, increasing Tanneal and tf monotonically increases (decreases) the amount of monoclinic (tetragonal) phase, which coincides with the variations in the dielectric constant. The remanant polarization was determined by the content of orthorhombic phase as well as the spatial distribution of other phases. As the next step, the effects of forming gas annealing (FGA) on the ferroelectric properties of HZO films were examined. Although the H-incorporation during FGA degrades the ferroelectric properties of HZO films, the degree of degradation was much lower compared with other ferroelectrics, such as Pb(Zr,Ti)O3. Pt worked as a catalyst for H-incorporation, and maximum 2Pr loss of ~40% occurred. However, the insertion of a ~20-nm-thick TiN layer between Pt and HZO decreased the degradation to ~12%. HZO is more resistant to degradation by FGA compared with the conventional ferroelectrics, which is a highly promising result for next-generation ferroelectric memory. In addition to that, the effect of the top electrode (TE) on the ferroelectric properties and switching endurance of thin HZO films was examined. The TiN/HZO/TiN capacitor can endure up to 109 times the electric cycling, which is promising for the next-generation memory. RuO2 TE was reduced during annealing due to the reactive TiN BE, resulting in the degradation of the ferroelectric properties and endurance. In addition, the endurance of the TiN/HZO/TiN capacitors was optimized by changing the film thickness and the post-annealing temperature. Finally, this thesis presents a feasible structure and actual operation of a tri-state memory function for high density FeRAM using stacked ferroelectric Pb(Zr,Ti)O3/insulating Al2O3/semiconducting ZnO layers with Pt top and BEs. The complicated electrical responses of the stacked structure to external stimuli were well understood based on the separated trapping of the compensating charges at the Pb(Zr,Ti)O3/Al2O3 and Al2O3/ZnO interfaces and the discrete dissipation of the trapped charges during polarization switching in one direction. This unique function of the structure induced three discrete charge states that can be used to increase the memory density by 50% compared to conventional FeRAM at a given cell size. It is believed that this thesis presents new pathways for the next-generation ferroelectric memory by exploring the new material system of Hf1-xZrxO2 films and by suggesting the new structure and operation of ferroelectric capacitor with novel ferroelectric-insulator-semiconductor heterojunction. Even though the ferroelectric memory is still far from the adoption as a universal memory, the results in this thesis could shed light on this field by suggesting new pathways different from the conventional approach.