Full text loading...
(a) Monte Carlo simulations of beam spreading in Cu show that the FWHM of a buried Au feature is smaller than the resolution predicted by both the simulated 90% beam radius and the formula of Goldstern et al. (shown as fit line). (b) Transmission function simulations of Cu for dark field detectors with different inner detection angles and (c) a detector collecting for three elements. These determine the thickness at which contrast reversals occur. The contrast reversal in Ta occurs at a thickness below the limits of many sample preparation methods, such as focused ion beam.
(a) The simulated transmission function for BF STEM detectors with different outer collection angles (in milliradians) does not significantly deviate from Beer’s law for large half angles. The signal does not undergo a contrast reversal and allows thick sections of highly scattering material to be reconstructed in three dimensions. (b) HAADF and IBF line profiles along a Cu line tilted to 70°. The approximate material thickness is shown by the dashed line . The HAADF transmission function, unlike the IBF, is nonmonotonic and shows a contrast reversal, as predicted in Fig 1.
Imaging a stress void in a copper interconnect. (a) The original HAADF two-dimensional (2D) image at zero tilt shows both the real stress void and second voidlike artifact at the bottom of the Cu interconnect (labeled with an arrow) created by contrast reversal of the Ta liner. (b) The original 2D IBF image of the sample showing the stress void, but the contrast reversal artifact is eliminated. (c) Slices from the HAADF three-dimensional (3D) reconstruction show that the artificial void persists in the reconstructed data, but only the true void is present in (d), a slice from the IBF-3D reconstruction. (e) The thresholded HAADF reconstruction, which should map the copper surface, still shows the artificial void. (f) The thresholded IBF reconstruction correctly maps the Cu surface and also shows that the void is faceted near the interconnect. Both projection images, (a) and (b), suggest that the interconnect possibly undercuts the via, but (d), a slice from the IBF 3D reconstruction, reveals that the interconnect is not undercut, which could only be determined using electron tomography.
Article metrics loading...