Ferroelectric HfO2 Based Materials and Devices: Current Status and Future Prospects

Ferroelectric Hafnium Oxide Based Materials and Devices: Assessment of Current Status and Future Prospects [OPEN ACCESS]

J. Müller, P. Polakowski, S. Mueller and T. Mikolajick

ECS J. Solid State Sci. Technol. 2015 volume 4, issue 5, N30-N35

Abstract

Bound to complex perovskite systems, ferroelectric random access memory (FRAM) suffers from limited CMOS-compatibility and faces severe scaling issues in today's and future technology nodes. Nevertheless, compared to its current-driven non-volatile memory contenders, the field-driven FRAM excels in terms of low voltage operation and power consumption and therewith has managed to claim embedded as well as stand-alone niche markets. However, in order to overcome this restricted field of application, a material innovation is needed. With the ability to engineer ferroelectricity in HfO₂, a high-k dielectric well established in memory and logic devices, a new material choice for improved manufacturability and scalability of future 1T and 1T-1C ferroelectric memories has emerged. This paper reviews the recent progress in this emerging field and critically assesses its current and future potential. Suitable memory concepts as well as new applications will be proposed accordingly. Moreover, an empirical description of the ferroelectric stabilization in HfO₂ will be given, from which additional dopants as well as alternative stabilization mechanism for this phenomenon can be derived. 

Comparison of the two major flavors of FRAM. 1T-1C: (a) Working principle illustrating the sensing margin / switched polarization P_sw derived from switched charge Q_sw and non-switched polarization P_nsw in the P-E-hysteresis. (b) DRAM-like architecture of FRAM adding a plateline to word- and bitline for bipolar ferroelectric switching. (c) TEM-micrograph and related P-E-hysteresis of a FE-HfO₂ based deep trench capacitor array proving the concept of 3D-integration capability. To illustrate the advantage of this area enhancement, the polarization density is calculated with respect to the lateral footprint of a comparable planar capacitor. 1T: (d) Illustration of the working principle by a graphical representation of the charge neutrality condition in a MFIS stack. Position 1 and 2 of the insulator-semiconductor loadline represents the transition from the ON-state to the OFF-state of the FeFET or vice versa. Accordingly, the gate voltage difference to turn on/off the FeFET can be approximated by 2 · V_C = 2 · E_c · d_FE, i.e. the memory window MW. (e) Disturb resilient AND architecture of the FeFET. (f) TEM-micrograph and related I_D-V_G-hysteresis of a FE-HfO₂ based 28 nm high-k metal gate transistors proving the concept of advanced 1T FRAM scalability

The recent success of smartphones and tablet computers has accelerated the R&D of fast and energy efficient non-volatile semiconductor memories, capable of replacing the conventional SRAM-DRAM-Flash memory hierarchy. These so called emerging memories usually leverage on the fact that certain materials possess the capacity for remembering their electric, magnetic or caloric history. For the extensively investigated ferroelectrics this ability to memorize manifests in atomic dipoles switchable in an external electric field. This unique property renders them the perfect electric switch for semiconductor memories. Consequently, only a few years after the realization of a working transistor the first ferroelectric memory concepts were proposed.

However, more than 60 years and several iterations later it is now clear that the success or failure of FRAM is mainly determined by the proper choice and engineering of the ferroelectric material. Perovskite ferroelectrics and related electrode systems underwent an extensive optimization process to meet the requirements of CMOS integration and are now considered the front up solution in FRAM manufacturing. Nevertheless, those perovskite systems require complex integration schemes and pose scaling limitations on 1T and 1T-1C memory cells that until now remain unsolved. This creates an unbalance between memory performance on the one side and manufacturing and R&D costs on the other side. This dilemma has ever since restricted FRAM to niche markets. 

With the recent demonstration of ferroelectricity in HfO₂-based systems (FE-HfO₂) a CMOS-compatible, highly scalable and manufacturable contender has emerged, that significantly expands the material choice for 1T and 1T-1C ferroelectric memory solutions as well as nanoscale ferroelectric devices. 

In this paper we will review and expand the current understanding of ferroelectricity in HfO₂, as well as discuss future prospects of ferroelectric HfO₂-based devices with respect to scaling, reliability and manufacturability. Opportunities and drawbacks of this disruptive development in ferroelectric material science will be critically examined. 

Continue reading in the full paper with Open Access here.