Weeding Out Immersion-Type Defects

Engineers from Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC, Hsinchu, Taiwan) have found ways to trace defect sources from optical microscope images to their scanning electron microscope (SEM) counterparts and reduce the average defect count on immersion-exposed 300 mm wafers from 19.7 to 4.8 particles/wafer. The researchers, including Lin-Hung Shiu, Fu-Jye Liang and several colleagues at TSMC, reported on their immersion defect reduction method at the recent SPIE Advanced Lithography conference in February.¹ It involved construction of a defect library and a novel exposure routing approach to reduce overall defectivity and improve yield.

Typical immersion-related defects include bubbles, watermarks, particles and printed particle defects. The classification of immersion-related defects requires a methodology that is more efficient than the traditional approach that uses optical inspection followed by SEM review. This approach is time-consuming, and can also make it difficult to determine whether unresolved resist patterns are caused by watermarks or obscured by fall-on particles during exposure, according to Shiu. The group's method clearly identified the primary particle source as particle obscuration during exposure. In addition, if the printing of particle defects can be reduced significantly, defect count can also be significantly reduced. A new exposure routine was developed to limit the number of printed defects in relation to the immersion hood (active diameter of immersion fluid).

From previous studies,² the TSMC group determined that the immersion hood residue caused by water leakage from previous exposure positions has a tendency to leave particles on the current exposure field, which then become printed defects. Through simulations and tests on resist-coated wafers and bare silicon wafers, they were able to model the water leakage traces as a function of position on the wafer during exposure. They also identified the primary defect type associated with the immersion hood, the source of the residuals, and the critical immersion settings.

For this study, typical resist pattern defects were captured with inspection tools following exposure, then classified as particles, watermarks or polymer residues. The wafers were covered with a bottom antireflective coating (BARC), resist and topcoat, and then exposed on the immersion scanner. Attenuated phase-shift masks were used, and wafers were processed in a clean track after exposures, scanned by a conventional defect inspection tool and reviewed with a SEM. Most defects were <2 µm in size. Time-delay images (TDIs) of the defects were correlated with the SEM images, and a library of the TDIs was built. After post-exposure bake and develop steps, most of the defects (except bubbles) could be matched using the TDI library.

Many of the defects found were caused by immersion residues, which became printed upon exposure. A distorted pattern-type defect can form when a water droplet leaks from the immersion hood and diffuses into the interface between the immersion topcoat and resist, causing swelling of the topcoat. Exposure light hits this water “bump,” causing exposure light scattering and pattern distortion. In many cases, the pattern at the center is not resolved because of dilution of the photo acid by the water. This leads to watermark, water stain, particle, polymer residue and bubble defects.

Because the immersion hood is larger than the exposed field, it will cover parts of neighboring fields during exposure. The overall water stain (~5 µm) becomes dried on an unexposed area, impacting the next exposure. In the printed image, only the center of the water stain is seen because water in the immersion hood quickly moves from the exposure area. Even with a topcoat, the engineers admit that the particle printing defects are very difficult to eliminate. One successful strategy, however, involves changing the direction of the exposure sequence. The initial step-and-scan routine would step in the X direction and scan in the perpendicular, Y, direction. The leaked water droplets carry particulates and leave them on unexposed fields. They are likely dried immediately by the air knife used to confine the immersion water. The new routine orients the stepping direction parallel to the exposure scan direction (both in Y direction), so the residues typically left are not printed by the subsequent exposures. This approach significantly reduced the defect count from 19.7 to 4.8 particles/wafer.