Three-Dimensional Electron Tomography Of Integrated Circuit Devices
No Access Until
The three-dimensional structure of integrated circuit (IC) devices can be analyzed at the nanometer scale by electron tomography using projection images generated from a scanning transmission electron microscope (STEM). The instrument samples the projected mass thickness of an object, producing image intensities with contrast capable of diﬀerentiating materials. Resulting three-dimensional reconstructions provide direct analytical analysis of structures with nanometer scale resolution along all three spatial dimensions for accurate predictions of device performance and reliability. We apply electron tomography to reconstruct a 250nm-thick cross-section of Cu wires with Ta barrier layers and measure the conductor cross-sectional area at many points along the wire to calculate Cu resistivity. Measurements from traditional twodimensional projection images underestimate conductor cross-sectional area due to the overlap of line edge roughness defects along the wire's length in projection. The average cross-sectional area determined from tomography measurements is 5% higher than a measurement from a two-dimensional projection of the same wire, resulting in significantly different predicted resistivities. Reconstruction of thick material cross-sections allows many measurements of device variations for a statistical analysis of critical dimensions. Traditional STEM imaging techniques produce transmission functions with a nonmonotonic dependence of intensity on thickness for thick cross-sections of common microelectronic materials, which is unsuitable for electron tomography. We introduce a novel incoherent bright field (IBF) STEM imaging technique optimized to image cross-sections up to 1mu-m thick. Monte-Carlo simulations of beam scattering predict IBF-STEM is complementary to traditional incoherent STEM imaging techniques but provides superior signal-to-noise ratio and no image artifacts for ultra-thick specimens. We develop a general relationship from Monte-Carlo simulations and calculations of the Rutherford cross-section for elastic scattering that determines the suitable STEM imaging technique for any material thickness based on predicted signal-to-noise ratios. To test IBF-STEM's suitability to electron tomography, we reconstruct a stress-void in a 250nm thick Cu wire, where traditional imaging techniques fail to provide monotonic image intensities at all tilt-angles. Experimental STEM images of the ultra-thick cross-section verify the expected transmitted electron intensity from Monte-Carlo simulations of STEM detectors with different collection angles. An IBF-STEM reconstruction provides the location and size of a stress-void in three-dimensions that is not possible with two-dimensional projection images alone.