Analytical Aberration-Corrected STEM of Ferroelectric Functional Interfaces

Abstract

In recent years progress in electron microscopy has pushed resolution to sub-Å values at 300 kV and below [1,2]. As an essential corollary to these instrumentation developments, significant increases in detection limits and signal-to-noise ratios were achieved such that the improved data collection ability and greater flexibility of the instruments have arguably proved the most beneficial advances for the materials sciences community, allowing users to utilise efficiently all available analytical signals. These benefits are particularly evident in the specific example of the chemical structure of dislocation cores in ultra-thin ferroelectric PbZr 0.2 Ti 0.8 O 3 oxide films [3]. The current trend of aggressive downsizing of the building blocks of silicon electronics has resulted in the functional material being confined to nanometric size volume and the presence of even a single defect and its associated longrange field could indeed adversely affect the performance and/or reliability of the device. Given the complex chemistry of the ferroelectric oxides involved, the vital and still largely unanswered question of how a dislocation core impacts the nanoscale local chemistry of the functional interface can only be addressed using a combination of analytical techniques and instrument settings. High acceleration voltage and ultra-fine probes are required for high-resolution imaging (Fig. 1a), lower, lessdamaging voltages and fine probes can provide atomically-resolved EELS information (Fig. 2a), while lower voltage and slightly larger probes can be used for high-resolution EDS work essential for the localisation of heavy metals such as Pb and Ru (Fig. 2b). This combination of techniques thus reveals O deficiency within the defect cores as well as cation non-stoichiometry in the vicinity of the dislocations. Additionally, the improvements in stability and signal-to-noise of the latest generation of aberration-corrected STEM instruments, along with much improved environmental control in microscope rooms, mean that strain mapping, a technique traditionally reserved to the processing of HREM images, can now yield quantitative data from Z-contrast images. Fig. 1b thus shows an example strain map in strong agreement with quantitative models that predict strain-assisted segregation of Pb around the dislocation cores. In conjunction with fully quantitative image simulations and image analysis, this wealth of high signal-to-noise analytical signals can thus provide a full sample characterisation. It is possible to observe directly and quantitatively phenomena such as stress-assisted diffusion and Cottrell atmospheres, which thus far were only conjectured to occur for dislocations in functional oxide interfaces.