Groove microstructures are widely applied in micro-electro-mechanical systems, semiconductors, and other fields, and are rapidly evolving in extreme directions, including miniaturization and high integration. The linewidth is one of the most important geometric parameters determining the physical properties of groove microstructures. Currently, the scanning electron microscopes used for micro-nano-scale linewidth measurement fall short of meeting the demands of online linewidth measurement as a result of inherent limitations such as their complex and destructive measurement process. Through-focus scanning optical microscopy (TSOM), which is a super-resolution linewidth measurement approach, has drawn extensive attention thanks to its non-destructive, fast, and non-contact measurement characteristics. In current studies, TSOM images are stacked with the focal plane as the center and the same scanning range at the top and bottom. However, given the depth of focus of actual microscopes, the images obtained within the depth of focus range are clear, and it is challenging to ascertain the axial position of the optimal focal plane through subjective evaluations using the human eye. This directly impacts the imaging range and measurement accuracy of TSOM. Hence, there is an urgent need for a method that can precisely locate the axial position of the focal plane during TSOM scanning to guarantee the stacking of the most accurate TSOM images.

This paper proposes a TSOM measurement method that locates the focal plane by utilizing the axial high-sensitivity characteristic of low-coherence interference fringes. First, a low-coherence microinterferometer is used for the through-focus scanning of a microstructure linewidth sample. This process helps to obtain an image sequence that contains surface information about the sample to be measured. Second, the high-definition vertical-scanning interferometry (HDVSI) algorithm is employed to analyze and solve the envelope and peak index of the low-coherence microscopic interference signal collected in the experiment. Then, with this position as the focal plane, the TSOM image of the sample to be measured is established. Meanwhile, the finite difference time domain (FDTD) algorithm is used to calculate the surface return light field of the incident light after it is modulated by the sample to be measured. An ideal TSOM system is established through the introduction of the through-focus scanning process. The near-field data calculated by the FDTD algorithm is transmitted to the imaging plane of the optical system based on the angular spectrum theory. The low-coherence microscopic interference signal of the sample to be measured is obtained by superimposing the reference plane light field data and combining using the coherence equation, and a standard-size linewidth model database is established. Finally, the library matching algorithm of TSOM is applied to measure the linewidth of the sample.

The experiment constructs a three-dimensional point spread function (3D-PSF) model of the low-coherence interference TSOM system. This is done with the goal of analyzing the effects of the numerical aperture (NA) and spectral bandwidth of the light source on the low-coherence interference TSOM system. Additionally, it proves the feasibility of the through-focus scanning of the microstructure sample using a low-coherence microinterferometer (Fig. 9). A comparison of the morphology restoration results of microstructure samples with different linewidths obtained using low-coherence microinterferometry shows that for a microstructure with a true linewidth of 10.97 μm, the measured linewidth is approximately 11.00 μm (Fig. 13). For a microstructure with a true linewidth of 2.01 μm, the measured linewidth is approximately 2.50 μm (Fig. 14). The experimental results show that for microstructures that are close to or even exceed the measurement limit of the low-coherence microinterferometer, their lateral resolution ability is restricted by the diffraction limit of the optical system and thus cannot meet the measurement requirements. Moreover, when measuring a microstructure sample with a 2.01 μm linewidth using the traditional TSOM method, the error of a stack of TSOM images centered on front and back depth-of-focus-plane images is as high as 0.15 μm. In contrast, the TSOM measurement result when using a stack with the focal plane image as the center is more accurate (Fig. 18). After improving the traditional TSOM technology using the proposed method, a standard groove sample with a linewidth of 2.01 μm is used as the sample to be measured. For 10 measurements, the average linewidth is 2.005 μm, the absolute deviation is 0.005 μm, and the standard deviation is 0.009 μm (Fig. 24). This indicates that the method can break through the diffraction limit of the low-coherence interferometry system and achieve super-resolution linewidth measurement.

Because of advancements in micro-nano fabrication technology, the traditional microstructure linewidth measurement method can no longer meet the current requirements for online measurement in industrial production. The proposed method combines the advantages of low-coherence microinterferometry and through-focus scanning optical microscopy. It employs a low-coherence microinterferometer to perform through-focus scanning on a sample and utilizes the defocus information of the sample in TSOM technology to achieve the super-resolution measurement of the linewidth. Compared with the traditional TSOM technique, this method effectively eliminates the error caused by the measurement of the system depth of focus and breaks through the linewidth measurement limit of the low-coherence microinterference system. Experimental results indicate that the method exhibits good stability and repeatability and holds broad application prospects for the precision measurement of precision machinery, integrated circuits, and micro-optoelectronic systems.