Optical near-field study of ferroelectric tunnel junctions

点击翻译按钮获取中文摘要

This program merges two active research areas of my group as we are deploying tip-enhanced Raman spectroscopy (TERS) to ferroelectric tunnel junctions (FTJs) of HfxZr1-xO2, which are only  three to four unit cells thick (approx. 2 nm). In 2017, we were the first to demonstrate the operation of these FTJs for non-volatile memory applications, thus paving the way for a resistance-based readout memory combining the cost efficiency and scaling potential of DRAM with a write speed that considerably exceeds SRAM while at the same time being non-volatile like Flash, all this at an energy consumption about four orders of magnitude inferior to Flash. The discovery of ferroelectricity in HfxZr1-xO2 in 2011 was unexpected, in particular as both HfO2 and ZrO2 had been individually used as dielectrics over several decades. Despite the tremendous interest in this first fully cmos-compatible ferroelectric, all attempts to explain the physical origin of this ferroelectricity are still in their infancy and lack nanoscale information through a direct, non-invasive technique. Tip-enhanced Raman spectroscopy is currently the only non-invasive imaging technique to yield chemical, structural and functional information at a nanometer scale. The technique is based on the enhancement and confinement of an optical near field mediated through a localized surface plasmon resonance in direct proximity of a noble metal tip that scans as part of an atomic force microscope in feedback at a controlled distance above the sample surface. We have chosen a shear-force tuning fork configuration to be compatible with electrochemically-etched gold and silver tips on virtually any Raman active surface, regardless of whether it is conducting or not and we have achieved a spatial resolution of 3 nm for hyperspectral imaging (i.e. a complete Raman and gold luminescence spectrum at each pixel). In HfxZr1-xO2, neither the bulk monoclinic phase nor the thin film tetragonal phase are ferroelectric and neither is susceptible to epitaxial biaxial strain to induce ferroelectricity as e.g. observed in conventional perovskite ferroelectrics. With a  typical grain diameter between 10 and 20 nm, the current working model is that an interfacial energy contribution from grain boundaries induces ferroelectricity; the model is however yet to be experimentally verified. The nature of ferroelectricity in HfxZr1-xO2 has tremendous implications for all optimization procedures regarding the material as well as the deposition process. We therefore intend to deploy tip-enhanced Raman spectroscopy and to back up our experiments with DFT perturbation calculations in order to predict the Raman spectra for all phases in a TERS geometry where the local k-vector is complex and where conventional Raman selection rules are violated due to strong local field gradients at the scale of single bonds. This research program relates to two ongoing strategic partnership grants and several other collaborative research projects.