Detection of defective chips from nanostructures with a high-aspect ratio using hyperspectral imaging and deep learning

We have developed an imaging spectroscopic reflectometry (ISR) method based on hyperspectral imaging and deep learning to detect defects in the bottom region of high-aspect-ratio nanostructures. ISR enables fast and non-destructive imaging of the bottom critical dimension (BCD) of channel holes (CHH) on a chip die of vertical NAND (V-NAND). A supervised learning model is built to predict the BCD by associating a pre-measured hyperspectral cube with scanning electron microscopy images after decapsulation of the top of the sample. The BCD predicted by ISR shows a high correlation of R2=0.72 with the actual BCD, and the distribution of CHH not open (NOP) defects on the chip die identified by bright field inspection after decapsulation is consistent with the BCD image obtained by ISR. In addition, ISR can detect defects that occur at arbitrary positions relative to the optical critical dimension (OCD) of the die. On fully integrated V-NAND chips, the ISR result showed a high correlation (R2=0.82) with the failure rate caused by CHH NOP, while the conventional spot OCD showed only R2=0.41. Thus, ISR can be used to optimize the etch process for weak wafer edge regions and to detect defective etch equipment.