Inverse modeling applied to Scanning Capacitance Microscopy for improved spatial resolution and accuracy

Abstract

Scanning Capacitance Microscopy (SCM) is capable of providing two-dimensional information about dopant and carrier concentrations in semiconducting devices. This information can be used to calibrate models used in the simulation of these devices prior to manufacturing and to develop and optimize the manufacturing processes. To provide information for future generations of devices, ultra-high spatial accuracy (<10nm) will be required. One method, which potentially provides a means to obtain these goals, is inverse modeling of SCM data. Current semiconducting devices have large dopant gradients. As a consequence, the capacitance probe signal represents an average over the local dopant gradient. Conversion of the SCM signal to dopant density has previously been accomplished with a physical model which assumes that no dopant gradient exists in the sampling area of the tip. The conversion of data using this model produces results for abrupt profiles which do not have adequate resolution and accuracy. A new inverse model and iterative method has been developed to obtain higher resolution and accuracy from the same SCM data. This model has been used to simulate the capacitance signal obtained from one and two-dimensional ideal abrupt profiles. This simulated data has been input to a new iterative conversion algorithm, which has recovered the original profiles in both one and two dimensions.In addition, it is found that the shape of the tip can significantly impact resolution. Currently SCM tips are found to degrade very rapidly. Initially the apex of the tip is approximately hemispherical, but quickly becomes flat. This flat region often has a radius of about the original hemispherical radius. This change in geometry causes the silicon directly under the disk to be sampled with approximately equal weight. In contrast, a hemispherical geometry samples most strongly the silicon centered under the SCM tip and falls off quickly with distance from the tip's apex. Simulation of the expected signal for each tip geometry shows significant differences in the expected resolution. This has also been explored experimentally with the SEMATECH # 1 sample. This sample has a staircase dopant profile with 50 nm steps. Simulation of the expected signal for the SEMATECH # I using a fiat tip model shows good agreement with measured data. However, the flattened tip reduces the steps in the profile to inflections.

Influence of Interface Traps and Surface Mobility Degradation on Scanning Capacitance Microscopy Measurement

“…The use of the inverse modeling technique based on MOS capacitor physics to extract the dopant profile [9], [10] has not been successful to date due to the fact that many physical effects influencing the experimental SCM data are not accounted for in the associated forward modeling. The presence of a significant amount of interface traps at the oxide-silicon interface and degradation of surface carrier mobility are two such effects, their influence on dopant profile extraction have not been previously investigated.…”