Fundamental modeling of ferroelectric thin film switching for high-density, low-power non-volatile memories

Ferroelectric (FE) materials are dielectrics with spontaneous bound charges that can be electrically switched by an applied field, resulting in a switchable polarization. The voltage-switchable characteristic of ferroelectrics provides an ideal material platform for low-power, high-density, Back-End-Of-Line (BEOL)-compatible memories, in particular with the recent discovery of CMOS-compatible, Hf-based ferroelectric thin films. Example technologies include FeRAM (Ferroelectric Random Access Memory), FeFET (Ferroelectric Field-Effect Transistor) and FTJ (Ferro-electric Tunnel Junction), with the FeRAM being technologically the maturest. The latest development of non-destructive readout (NDRO) operating paradigm for FeRAM further brings the read endurance to quasi-infinite, promising to unlock its full potential as non-volatile Dynamic RAM(NV-DRAM) and energy-efficient compute-in-memory (CiM) technology.

Notwithstanding the multitude of explorations of the device- and application space so far, the modeling of the precise field-induced response of FE thin films remains far from evident. Theoretically, the most common approaches to date resort to the mean-field, Landau-Devonshire, free energy minimization. A major drawback of these approaches, however, is that they generally overlook the thermal fluctuation in the system at finite temperatures and hence fail to account for the dominant nucleation-growth based FE domain reversal and the inherent thermal switching stochasticity. As a result, the community has often had to resort to using phenomenological, Kolmogorov-type, descriptive models to match experiments, with the thermal stochasticity left out of the picture. This inevitably undermines the predictability of the service operation of FE-based memories, e.g., the retention time of FeRAM at elevated temperatures extending to the tail of the statistical distribution.

Another conundrum in understanding the dynamics of realistic Fe thin-films rests with its multi-domain, polycrystalline nature. It is known that Hf-based ferroelectrics make use of a metastable, orthorhombic phase out of a multiphase matrix, formed during deposition and heat treatment, with various crystal defects (vacancies/dislocations/grain boundaries/etc.) abound. Given that there exists no “exchange interaction” in FE materials (unlike in ferromagnetism), the multi-domain effects essentially entail a proper treatment of the electromechanical (e.g., strain) coupling arising from the discontinuity of lattice periodicity. Today’s state-of-the-art studies only consider “multi-domains” under the Landau-Devonshire framework with an effective “domain wall energy”, thus lumping entirely the various, experimentally observed, phenomena like domain wall pinning/depinning, which nevertheless play critical roles in accounting for the polarization dynamics in FeRAM write/read operation. Stopgap models such as nucleation-limited switching (NLS) provide practical post-silicon data analysis capabilities but fall far short of pre-Si guidance for material selection or processing optimization.

Against the existing challenges above (among others), this thesis seeks to establish a mechanical modeling framework to elucidate the fundamental field-stimulated response of FE thin films, and as such to provide predicative guidelines for the processing and operation of FE-based memories. This could include (but not limit itself to): thickness/stress engineering for remnant polarization/coercive field; retention at various temperatures; temperature-dependent variability and stochasticity, etc. The modeling effort in this thesis will be supported by experimental measurements on FeCAPs fabricated at imec. The resulting fundamental models will be used to benchmark and optimize the design of individual FeCAPs and full arrays. These learnings will be further transferred to other FE-based devices (FeFET, FTJ) for guiding/gauging their application-specific power/area/reliability metrics.

What will you do?

• Developing inhouse simulation framework (or customized commercial programs) for the FE thin film system.
• Numerical (finite-element/finite-difference) modeling of FE-thin films, with focus on thermally activated domain dynamics (e.g., Brownian motions) and on polycrystalline defects (grain boundaries, electrode interfaces).
• Linking modeling results with both inhouse and external (collaborators, literature) experimental data: electrical / microscopic /…, and ab initio atomistic simulation results.
• Abstracting the numerical modeling findings into compact, circuit-compatible models of FE memory devices and explore basic figures of merits (write/read latency, error rate, etc.).

Who are you?

• A materials scientist, or applied physicist, or nano-scientist/-engineer, or electrical/electronic engineer with an emphasis on physics
• Strong analytical mindset, enthusiasm for critical thinking
• Excellent knowledge of solid-state physics, preferably with reasonable understanding of materials science (equilibria/phase transition/defects)
• Being comfortable with programming / coding