As semiconductor technology nodes continue to shrink, tighter error budgets make uniformity (including CDU) increasingly critical. D2S takes a “uniformity first” approach, reducing variation while also delivering the most robust linearity correction.
For manufacturing of wafers and masks, uniformity and linearity are both important. Uniformity measures how consistently a feature of a given size is reproduced across the wafer or the mask. Linearity describes how small a feature on the mask or the wafer can get and still maintain the target size. Even if average linearity is very good, manufacturing variations can cause nominally identical features to print at different sizes, consuming a part of the available error budget. Controlling this variation therefore becomes increasingly important as tolerances tighten on both the wafer and the mask in increasingly advanced technology nodes.
Traditionally, while the industry waited for EUV lithography, wafer manufacturing increasingly relied on software-based solutions to expand process windows and improve resolution. Optical proximity correction (OPC), followed by inverse lithography technology (ILT), helped improve both uniformity and linearity. With eBeam-based writing of leading-edge masks at 4x scale and mask write times in hours or even tens of hours per mask, depth of focus presented no concerns. Until recently, mask making was more concerned about linearity correction than uniformity enhancement. For software correction, mask process correction (MPC) also has focused primarily on linearity: bringing the average manufactured feature size closer to its target. At the leading-edge nodes now, mask uniformity is equally a concern. While depth of focus remains well under control for masks, smaller features have worse dose margin (exposure latitude in wafer terminology) making uniformity a larger concern. This variation consumes the available error budget and cannot be recovered simply by improving the mean with linearity correction. D2S therefore takes a “uniformity first” approach: reducing feature-to-feature variation while applying linearity correction to bring the resulting distribution onto target.
The D2S solution for the best uniformity leverages pixel-based computing on graphical processing units (GPUs), and includes inverse lithography technology (ILT) that creates Entirely Manufacturable™, and therefore entirely curvilinear mask designs and pixel-level dose correction (PLDC) that provides an MPC solution that improves both uniformity and linearity on the physical mask. Together, this flow produces mask shapes with the best uniformity that, in turn, help produce wafers with the best uniformity.
Mask processing or wafer processing can’t create 90-degree angles – all shapes on mask and wafer are curvilinear, even if the design is Manhattan. Mathematically, trying to manufacture shapes that aren’t achievable will always make for worse uniformity. Shapes that are achievable will always be more reliably manufacturable. Entirely Manufacturable designs (whether for mask or wafer) are entirely curvilinear. Enabling curvilinear patterns enables Entirely Manufacturable designs and thereby enables more uniform manufacturing of masks and wafers.
Even if the process window is large and the MEEF is reasonable, every variation on mask is a systematic variation on every chip on every wafer. So, reducing mask variation should be one of the first steps to reduce wafer variation.
Mask uniformity is not just about mask processing. When the mask shop is asked to produce unmanufacturable shapes, the best the mask shop can do is to come as close to the target shape as possible, with as much uniformity as possible. A far better flow is for ILT to produce Entirely Manufacturable shapes that the mask shop can manufacture faithfully. In turn, such mask shapes will also be more uniform in both local and global uniformity. And therefore, the wafer target will also be closer and wafer uniformity will be better. It is a virtuous cycle of collaboration between ILT and mask shops for mask-wafer co-optimization for uniformity.
