Surface Potential Measurements of Reconfigurable p-n Junctions in Graphene

Manipulation and control of electron current in a graphene p-n junction (e.g. electron waveguiding, reflection, focusing) is directly determined by the spatial gradient of the Fermi level across the junction. Sharp Fermi level gradients are associated with negative index 'lensing' of electrons in graphene while broader gradients are predicted to form reflective boundaries. Quantitative metrology of the Fermi level gradient at p-n junctions is thus essential to determine device performance, validate models for device design and switch architectures, and quantitatively determine the impact of defects on device function and leakage.;In this dissertation work, direct surface potential measurements of reconfigurable p-n junctions in graphene were carried out via Kelvin Probe Force Microscopy (KPFM) with an embedded split-gate test structure to quantitatively characterize the electrostatic doping profile (spatial variation of the Fermi level) across the junction. The KPFM technique was adapted, optimized and calibrated for graphene p-n junction devices. Ambipolar electrostatic doping in both exfoliated and CVD grown graphene was observed to change according to the gate voltage polarity (i.e. switching from positive to negative values). The p-n junction electrostatic doping profile width for exfoliate and CVD graphene devices was extracted from a Gaussian fitting of potential profiles taken from section analysis of surface potential maps. The Fermi velocity renormalization model and charge neutrality point shift model were applied to quantitatively analyze the gate modulated Fermi level shifts in exfoliated and CVD grown graphene devices, respectively. The application of these models demonstrated very good agreement with experimentally acquired graphene p-n junction electrostatic doping profiles. A Finite Element simulation was also performed to investigate junction width broadening due to KPFM tip-sample coupling. It was shown that the finite element simulation combined with a simplified deconvolution method can restore the true surface potential distribution of a given KPFM experiment geometry without the necessity of knowing the exact form of the point spread function of the tip. For the graphene p-n junction devices studied here the true junction width is approximately 25% smaller than the measured width.