On the spatial resolution of two-dimensional doping profiles as measured using secondary ion mass spectrometry tomography

Abstract

Knowledge of the lateral or spatial distribution of doping impurities is important for accurate process and device simulations of submicrometer silicon devices, since the edge effects of the electric fields can no longer be neglected. A newly developed technique, which utilizes the high sensitivity and high depth resolution of secondary ion mass spectrometry (SIMS), is capable of measuring two-dimensional (2D, i.e., in the depth and the lateral direction) doping distributions. The technique is based on a series of one-dimensional (1D) SIMS depth profiles obtained at different directions through a sample. The individual SIMS depth profiles are then recombined to generate a 2D doping profile using the expectation maximization algorithm, which was originally used in human body computer-aided tomography. The SIMS tomography technique gives a doping distribution as a function of position. The positional accuracy or spatial resolution of the technique needs to be fully understood in order to properly use the technique. During a reconstruction process, the area of interest is divided into small volume elements or voxels, and the doping concentration for each voxel is estimated. Therefore, the size of the voxels determines the spatial resolution of the reconstructed profiles. The smaller the voxels are, the better the spatial resolution is. However, the size of the reconstructed area cannot be made infinitely small, due to the limited number of SIMS measurements. Since each SIMS measurement provides only one linear independent equation, the total number of equations (or the total number of the SIMS measurements) should be equal to the number of voxels (or unknowns). Thus, the size of the voxels, or equivalently, the number of SIMS measurements available imposes a practical limit on the spatial resolution. In addition, since the 1D SIMS depth profiles are used as the input for the profile reconstruction, the depth resolution of the 1D SIMS measurements dictates the spatial resolution of the reconstructed profiles. Furthermore, the depth resolution of the 1D SIMS measurements is also affected by the sample skew, which is the measure of the lack of parallelism of the dopant line with respect to the beveled surface. And finally, the alignment of the 1D SIMS depth profiles relative to each other contributes to the positional uncertainty of the reconstructed distributions. A 20 keV boron implant through a 1.2 μm wide window into germanium preamorphized silicon has been reconstructed recently. The spatial resolution of the reconstructed dopant profile is evaluated to be 40 nm. It is found that the spatial resolution is predominantly determined by the depth resolution of 1D SIMS profiles, and the microtopography of the SIMS craters is the major cause of the poor depth resolution. It is expected that the spatial resolution of the 2D doping tomography can be greatly improved through the use of a simplified sample structure, which will reduce the SIMS profiling depth and provide better depth resolution for the 1D SIMS measurements.