White light microscopic interferometry is a traditional method for non-destructively measuring the three-dimensional (3D) topography of step microstructures. However, its tendency for low-pass filtering smoothens the 3D topography along sharp step edges, making precise edge detection challenging and affecting the calculation and extraction of linewidth parameters. Moreover, when the depth of the microstructure is smaller than the coherence length of the light source, the batwing effect may generate sharp pulses in the recovered 3D topography at these step edges. Despite these pulses being spurious signals, their high-frequency traits are beneficial for identifying step edges. The microstructure linewidth is determined by the distance between the peak positions of the batwing pulse heights corresponding to the two step edges. The spatial sampling frequency of the 3D topography is constrained by the size of the Airy spot and the detector pixel. Typically, only a few pixels in the linewidth measurement direction are influenced by the batwing effect. It is crucial to note that the accuracy of linewidth measurement is confined to the pixel level, and is determined by the peak location of the batwing pulse heights. In this study, we propose a super-resolution measurement method for microstructure linewidth. It is based on precisely locating the peak positions of batwing heights in white light interferometry, surpassing the system’s lateral resolution. We anticipate that our findings will enhance 3D topography measurement of step microstructures and advance our understanding of the batwing effect.

We introduce a numerical simulation model for white light interferometry signals, with a specific focus on the batwing effect and its discrete sampling characteristics. The model unveils the slow-varying attributes of the spatial domain orthogonal to the linewidth measurement direction, offering theoretical backing for improving the sampling frequency of batwing pulses in that direction. By harnessing the spurious high-frequency information provided by batwing pulses, we can attain super-resolution measurements for microstructure linewidth. This breakthrough surpasses the optical resolution limitations of the system. First, we apply the center of gravity method to process the interference signals captured by white light interferometry from the step sample. Next, for each step edge, we utilize wavelet transform to identify the peak key pixels in the direction parallel to the edge, which signifies the edge position. Subsequently, precise step edge positions are determined through linear fitting of multiple key pixels corresponding to each edge. Furthermore, we extract linewidth parameters based on the positions of the left and right step edges. In essence, we leverage the high-frequency information of batwing pulses in the orthogonal direction of the linewidth, presenting an effective approach for achieving super-resolution in linewidth measurement.

We conduct measurements on an RSN standard plate (provided by Physikalisch-Technische Bundesanstalt) featuring a standard grating step structure using a self-developed white light interferometry system. The test samples have linewidths of 6, 4, 3, and 2 μm. The step height falls within the coherence length of the light source, while the Airy spot radius of the measurement system is 0.590 μm. Our method yields linewidth measurement results with deviations from the calibration value of 0.011, 0.016, 0.021, and 0.015 μm, respectively (Table 1). These experimental findings demonstrate that our proposed method enables super-resolution linewidth measurement, surpassing the system’s lateral resolution. Furthermore, we explore the effect of sample orientation on measurement accuracy and provided recommended values for both the pitch angle of the sample and the tilt angle within the field of view based on measurement uncertainty. As for the pitch angle, optimal measurement conditions are achieved when interference fringes are aligned nearly perpendicular to the step edge during experimentation. Simultaneously, to ensure precision in linear fitting and peak point positioning, the sample tilt angle should be constrained within a specified range [Eq. (6)]. By meeting these criteria, linewidth measurement accuracy can attain a resolution within a few tenths of the pixel scale of the object surface.

We make full use of the spurious high-frequency information provided by batwing pulses at step edges, proposing a super-resolution measurement method for microstructure linewidth using white light microscopic interferometry. This method surpasses the optical resolution limit of conventional systems. We develop a numerical simulation model to elucidate the relationship between batwing heights and step edge positions. Unlike traditional approaches, we focus on microstructure topography in the vertical step-edge direction. Additionally, we transform abrupt step topography into a gradual change process across multiple pixels, enabling precise positioning of batwing heights at step edges. This facilitates the calculation of high-precision linewidth measurements. In our experiments, we measure four grating regions with varying linewidths in the PTB standard plate, confirming the effectiveness of the method. Overall, our approach offers rapid calculation speeds and broad applicability in post-processing white light interferometry signals.