Scatterometric defect detection of nanowire surfaces: evaluating the effect of sensor position density

Fast and robust defect detection is needed in nano-surface manufacturing processes for high production quality. While scatterometry offers a suitable in-situ-capable measurement approach in principle, the design of a measurement setup with maximal sensitivity has to be clarified. This holds in particular for the detection of stochastically occurring structural defects, requiring a large number of measurement data sets to statistically design/train and test/optimize the measurement chain including the signal processing. To this end, the data is preferably generated by virtual experiments, using rigorous optical simulations solving the Maxwell equations. Currently, the complete scattering response of the nano-surface is sampled by evaluating the entire scattering sphere with a high spatial resolution. However, such a highly dense photodetection limits the method’s applicability for fast and in-line measurements, due to a long scanning time or the large space for many sensors. To overcome this limitation, a sensor position density reduction is investigated regarding its effect on the defect detection capability for a surface with GaN-InGaN nanowires. A statistically sufficient data set was generated by repeating a virtual experiment 240 times, simulating a randomly redistributed nanowire surface in each repetition. For the defect detection task, highly sensitive regions of the full scattering sphere are observed with an angular dimension of about 90° and 30° in azimuthal and elevation directions, respectively. As a central result, a reduction of the sensor position density up to 85 % leads to almost no changes in the defect detection sensitivity and the distinguishability between surfaces with 40 % and 70 % defects. Hence, a better applicability of the scatterometric nanosurface defect measurement method is feasible without a loss of sensitivity. Future investigations will also consider nonequidistant sensor positions, to reduce the number of required sensors or sensor positions even further.