Surface potential and morphology issues of annealed (HfO2)x(SiO2)1−x gate oxides

Abstract

The surface morphology and surface potential variations of annealed (HfO2)x(SiO2)1−x films were investigated by noncontact atomic force microscopy (AFM) in ultrahigh vacuum. Additional modes of data acquisition included contact potential difference (CPD) and differential capacitance. Two types of samples were investigated. The first, a set consisting of 4 nm thick samples with (HfO2)x(SiO2)1−x compositions of x=0.4, 0.6, and 0.8, were annealed at 1000 °C for 10 s in N2 gas. The second, a 2.2 nm thick sample of composition (HfO2)0.78(SiO2)0.22 was annealed in vacuum at 50 °C intervals from 850 to 1000 °C. The anneals resulted in a microstructure consisting of phase-separated HfO2 crystallites and amorphous silica, as observed in high resolution transmission electron microscope (HRTEM) images. The crystallites appear to be responsible for most of the morphology observed with the AFM, with surface features for the hafnium rich x=0.6 and 0.8 compositions generally agreeing with the crystal sizes observed by HRTEM. The AFM images for the x=0.4 sample showed substantially broader features than the 5 nm crystallites seen by HRTEM, with evidence for inclusions of low dielectric constant (κ) material, presumably silica, on portions of the surface. The vacuum annealed sample showed an initial trend to lower roughness and CPD fluctuation range, with a minimum in both (rms roughness=0.077 nm and ΔCPD=0.2 V) for a 10 s anneal at 900 °C. Thereafter both measures increased substantially. The 1000 °C vacuum annealed sample compared favorably in structure, roughness, and to a lesser extent in the CPD fluctuation range with the 1000 °C N2 annealed sample. The N2 annealed samples for x=0.4 and x=0.8 exhibited CPD fluctuations as large as 0.4 V, with a smaller value of 0.22 V observed for the x=0.6 sample. CPD fluctuations consist of a small amplitude substructure that correlated with the microstructural features of the surface, superimposed on long range CPD fluctuations (20–>50 nm) unrelated to any surface features. Their origin is speculative, but could be associated with bulk and/or interface fluctuations in the density of trapped charge. Their potential adverse impact on device performance is discussed.