Abstract
The period of time in which this dissertation is undertaken falls within an era of overlap where Si technology is reaching the physical limits of device scaling and new approaches, such as 2D graphene nanoelectronics are being explored. Low frequency noise (LFN) is one of the reliability challenges for sub-10nm Si transistors, and presumably also for future graphene nanoelectronics. Although the broad picture of defect based noise, and the related problem of defect generation in gate dielectric breakdown, is known, traditional measurements carried out on micron sized test structures often do not provide enough spatial resolution to probe individual defects of gate dielectric material, which are crucial to understand the effects of LFN and dielectric breakdown in ultra-scaled transistors. In this thesis, we address this shortcoming by the use of conductive atomic force microscopy (CAFM) and apply the technique to both HfO2, a standard gate material for Si transistors, and hexagonal Boron Nitride (h-BN), a new material for novel 2D electronics. In the first half of the thesis, we addressed experimentally and theoretically the thermal drift in CAFM, which is crucial for its application in noise spectroscopy, which typically involve a long measurement time. A method is demonstrated to prolong the dwell time of a CAFM tip at a given location for reliable acquisition of electrical data over extended time scales. We have then demonstrated the successful application of CAFM for nanoscale defect spectroscopy using random telegraph noise (RTN) measurements in high-?? HfO2 gate dielectrics. Specifically, we have employed the technique to correlate the hardness of breakdown and RTN in HfO2 thin films. With single defect resolution, CAFM-based RTN spectroscopy has been used to directly probe the electron capture/emission time constants of a defect. In the second half of the thesis, we have investigated the time dependent dielectric breakdown (TDDB) in h-BN. This is an important aspect of h-BN which must be carefully addressed for integration as a gate dielectric in emerging graphene nanoelectronics. Based on the electrical, statistical, and transmission electron microscopy (TEM) analysis carried out in this work, we find that h-BN shares many similarities with the defect generation and breakdown observed in conventional SiO2 and HfO2 gate dielectrics. This includes the existence of charge trapping, stress induced leakage current (SILC) and soft and hard breakdown phases in the TDDB. The measured breakdown field strength on crystalline h-BN is ~15MVcm-1 which is comparable to SiO2 (~13MVcm-1). However, we also observed some differences such as discrete steps in a gradual reduction of the tunnel resistance in multi-layer h-BN, which may point to a layer-by-layer progressive defect generation. Signatures of LFN have been observed for the first time in h-BN and it becomes more prominent after soft breakdown because of defect generation. Also, for the first time, we have measured temperature dependent electrical data and by suitable modelling and analysis, have obtained relevant defect energy values for h-BN. Using CAFM for localized breakdown measurements on h-BN, we find that progressive defect generation is the dominant breakdown mechanism for low current compliances, whereas material removal is commonly observed for hard breakdown conditions. The layer-by-layer defect generation and removal of material under hard breakdown conditions are the major differences compared to conventional gate dielectrics. TEM analysis confirmed that metal migration between electrodes occurs during hard breakdown in h-BN dielectrics. Indirect evidence for significant metal migration was also obtained from the CAFM data where it was observed, using a novel CAFM adhesion approach, that after hard breakdown, most breakdown locations showed metal-metal contact, with ohmic conduction and very high adhesion forces. Statistical analysis of the breakdown electrical data of h-BN showed a positive correlation of the ramp rate with breakdown voltage and the suitability of a defect clustering model rather than a Weibull distribution to describe the breakdown statistics in CVD grown h-BN was demonstrated. Finally, statistical analysis showed that the percolation model and Poisson area scaling are valid to describe the breakdown in h-BN, as in conventional dielectrics.