On-product Overlay Improvement for a Back-End-of-Line Immersion Layer

For On-Product Overlay (OPO) control, the process induced lot-to-lot (L2L) and wafer-to-wafer (W2W) variation is a big challenge. In this paper, we investigate the variation source and search for methods to cover at a back-end-of-line (BEOL) immersion layer. Study shows that the variation mainly comes from wafer high order, 3
^rd
and 5
^th
order terms, which inherited from early layers. Alignment strategy and overlay control method optimization would be effective ways to minimize it. Compared with inline linear alignment (6-par) model, high order wafer alignment (HOWA3) model simulation shows overlay x/y 45%/48% improvement. We also demonstrate a new feedforward (FF) method – prelayer de-correction fingerprint (FP) FF is effective in reducing L2L variation. Compared to the feedback (FB) only strategy, the FF-and-FB method is predicted to bring overlay x/y 36%/22% and 29%/13% improvement under 6-par and HOWA3 model, respectively. With HOWA3 and FF-and-FB implemented, we predict on product wafer will show overlay x/y 61%/54% improvement and ~81% rework rate reduction.