Inspection technology and equipment are crucial in the manufacturing of integrated circuits (ICs), and photomask inspection is key in lithography for ensuring IC manufacturing with high reliability and yield. Advanced semiconductor manufacturing has entered the 5 nm mass-production era, as represented by smartphones. However, the demand for mature processes (28 nm and above) remains high. This process technology is cost effective and can be widely used in appliances, consumer electronics, automotive electronics, and 5G communication. Therefore, methods that satisfy the requirements of photomask inspection in mature processes must be identified.

Various photomask inspection methods have been employed. E-beam inspection (e.g., scanning electron microscopy) has high-resolution (subnanometer) and high-sensitivity but low-throughput. Optical inspection is widely adopted because of its acceptable sensitivity at high throughput and its nondestructive nature. Owing to the optical diffraction limit, the spatial resolution of conventional wide-field microscopy is limited by the wavelength and numerical aperture (NA) of the objective. To achieve high resolutions, high-NA objectives are used; however, their design is complex and costly.

Structured illumination microscopy (SIM) can overcome the diffraction limit bytransferring the high-frequency information of samples into the detectable frequency range of the imaging system via frequency mixing. The principle of transmission SIM is illustrated in Fig. 1. Illuminating samples with interference patterns enables previously inaccessible high-frequency components to be encoded into the observed image, thereby improving the spatial resolution. In this study, a digital micromirror device (DMD) was used to create illumination patterns and switch beams with a spatially controlled intensity and phase. Four-orientation illumination (0°, 45°, 90°, and 135°, with each orientation used in three phases) was used to obtain a near-isotropic resolution and resolution enhancement. The DMD is based on microelectromechanical system technology. Each mirror of the DMD was controlled using a computer to modulate the projected beam in real time.

Figure 2 shows a schematic illustration of the DMD-based SIM experimental setup. The illumination source is a He-Ne laser with a wavelength of 632.8 nm, and the laser power was adjusted using a neutral density filter. The laser beam illuminates the active area of the computer-controlled DMD. As shown in Fig. 3(a), DMD line patterns were designed. To obtain light and dark stripes with high contrast, one must satisfy the blaze criterion of the blaze grating [Fig. 3(b)]. The DMD-modulated beam was reflected from the micromirrors and a mask was used to select the desired diffraction orders. The two selected diffracted beams passed through the lens and were focused onto the USAF 1951 resolution test target, which was mounted on a three-dimensional translation stage. Subsequently, the image was formed on the camera using an objective, a focusing lens, and a mirror. Twelve raw images (three phases × four orientations) were reconstructed using the MAP-SIM reconstruction algorithm to generate high-resolution images [Fig. 4(b)].

Imaging resolution was investigated comprehensively while considering various diffraction orders. By selecting the diffraction orders, various diffraction angles were obtained, which altered the spatial frequency of the fringes (i.e., fringes were generated by the laser beam interference of the selected diffraction order). The spatial resolution of the imaging setup was evaluated using a USAF 1951 resolution test target. First, we compared the imaging results using an unmodulated 0th order light with those of diffracted light from the first to the fourth order incident on the sample (Fig. 5). The results indicate that selecting higher diffraction orders as the illumination for SIM effectively enhances the resolution (Fig. 6). In addition, high resolution was achieved in this imaging system with a low NA. Finally, a comparison between the theoretical and experimental resolutions is shown in Fig. 7(a). The enhancement of SIM resolution with increasing γ (interference angle at the sample) is consistent with the theoretical result. The errors may originate from aberrations, deformations in the optical components, sample drift and vibration, image noise, reconstruction algorithms, or discrete values of the resolution test target. Therefore, by selecting a light-source wavelength and an objective NA that offer reasonable cost performance, as well as the corresponding interference angle at the sample, the photomask-inspection requirements of mature processes can be fulfilled.

We propose a simple and flexible SIM imaging system based on a DMD to perform photomask inspection in mature processes. By selecting different diffraction orders, the interference angle at the sample can be controlled (from 2° to 8°), thus altering the spatial frequency of the two-beam fringes. Additionally, high-order diffraction light is used as the illumination light of the SIM system for resolution enhancement. The relationships among the wavelength, NA, interference angle at the sample, and resolution are discussed as a guideline for system improvement. This technology offers the advantages of rapid imaging, large field of view, noncontact compatibility, and low cost. It is a promising approach for inspection applications in IC manufacturing.