Spatially Resolved Electrical Analysis of Epitaxial Defect-Stabilized Hafnium Oxide Thin Films

Hafnium oxide (HfO₂) is a transition metal oxide which receives great research interest due to its compatibility with complementary metal-oxide-semiconductor (CMOS) technology, as well as its ferroelectric and memristive properties. One key to improve the performance of hafnium oxide based non-volatile resistive random access memory (ReRAM) devices is specifically tuning the electronic properties. Here, the precise engineering of defects, such as doping with other transition metals or the variation of the oxygen content, plays a major role. The objective is to improve the formation and disruption of a conductive filament by diffusion of oxygen vacancies in valence change mechanism (VCM) type of ReRAM devices. Scanning probe techniques, such as conductive atomic force microscopy (c-AFM), are powerful tools to investigate the switching behavior and electronic properties at the surface of transition metal oxides with highest local precision.[1]<br/>The purpose of this study is to identify filament locations on the nanoscale and to examine their minimal size with respect to the sample stoichiometry of hafnium oxide thin films. The hafnium oxide samples with varying oxygen content were grown by molecular beam epitaxy (MBE) on c-cut sapphire substrates with an intermediate TiN layer acting as bottom electrode. The oxygen flow during evaporation was varied to grow a series of samples, starting from stoichiometric, monoclinic HfO₂ and going to substoichiometric, highly oxygen deficient HfO_2-x, where a low-temperature cubic phase could be identified.[2] Subsequently, the samples were analyzed without breaking the vacuum by c-AFM and scanning tunneling microscopy/spectroscopy (STM/S). The RMS surface roughness of the hafnium oxide films, which have a thickness of 10nm, is about 1nm or less, measured by c-AFM. By performing a series of conductivity maps, a gradual increase in conductivity of the hafnium oxide layer with decreasing oxygen content and increasing amount of the substoichiometric cubic phase can be determined. The onset behavior of the conductivity for each case of the different oxygen deficiencies has a similar appearance. In particular, while the first conductive pathways are identified along grain boundaries and expand with increasing bias voltage, distinct grains show a gradual onset behavior in a voltage range of about 1V. Analogous to the onset of the general conductivity, the forming behavior of grains is likewise dependent on the mean oxygen vacancy content defined by the growth process. Moreover, the switching behavior for the different stoichiometries is analyzed with one focus on the smallest switchable structure. For a more profound perception of the electronic properties around the Fermi level, the most conductive samples are investigated by STS. Thereby, the target is the identification of possible in-gap states caused by introduced defects.[3]<br/> <br/>References:<br/>[1] R. Dittmann <i>et al.</i>, Proc. IEEE 2012, <b>100</b>(6), 1979-1990<br/>[2] N. Kaiser <i>et al.</i>, ACS Appl. Mater. Interfaces 2022, <b>14</b>, 1290-1303<br/>[3] N. Schmidt <i>et al.</i>, ACS Appl. Nano Mater., <i>submitted</i>