Abstract
As smaller structures are being increasingly adopted in the semiconductor industry, the performance of memory and logic devices is being continuously improved with innovative 3D integration schemes as well as shrinking and stacking strategies. Owing to the increasing complexity of the design architectures, optical metrology techniques including spectroscopic ellipsometry (SE) and reflectometry have been widely used for efficient process development and yield ramp-up due to the capability of 3D structure measurements. However, there has been an increasing demand for a significant reduction in the physical spot diameter used in the SE technique; the spot diameter should be at least 10 times smaller than the cell dimension (~30 × 40 μm(2)) of typical dynamic random-access memory to be able to measure in-cell critical dimension (CD) variations. To this end, this study demonstrates a novel spectrum measurement system that utilizes the microsphere-assisted super-resolution effect, achieving extremely small spot spectral metrology by reducing the spot diameter to ~210 nm, while maintaining a sufficiently high signal-to-noise ratio. In addition, a geometric model is introduced for the microsphere-based spectral metrology system that can calculate the virtual image plane magnification and depth of focus, providing the optimal distance between the objective lens, microsphere, and sample to achieve the best possible imaging quality. The proof of concept was fully verified through both simulations and experiments for various samples. Thus, owing to its ultra-small spot metrology capability, this technique has great potential for solving the current metrology challenge of monitoring in-cell CD variations in advanced logic and memory devices.