The Curvilinear Bottleneck in Advanced Node Manufacturing

The shift toward advanced semiconductor nodes is breaking the traditional Manhattan geometry paradigm. As manufacturing pushes toward tighter feature sizes, the industry is adopting curvilinear masks to achieve larger process windows that vertical and horizontal edges cannot deliver Siemens EDA reports that this transition introduces a computational challenge that threatens to slow the entire optical proximity correction (OPC) workflow.

Traditional Manhattan masks constrain shapes to vertical and horizontal edges, forcing algorithms to approximate curves using many small straight segments. Curvilinear masks use cubic Bezier splines—smooth mathematical curves—to represent shapes naturally. This enables precise control around corners and curved features where lithography is most challenging.

The bottleneck lies in mask rule check (MRC), the validation step ensuring designs can be manufactured without defects. For curvilinear masks, MRC can represent a large portion of overall OPC runtime. When teams cannot validate manufacturability quickly, they face extended iteration cycles and delayed convergence. This creates schedule risk at a time when advanced nodes demand faster time-to-market.

The complexity stems from the math. With Manhattan geometries, calculating the minimum distance between two parallel mask segments is efficient and exact; the answer is simply the absolute value of the difference between two coordinates. Curvilinear masks lack this simplicity. Computing the minimum distance between arbitrary cubic Bezier curves requires solving a multivariate optimization problem with no efficient closed-form solution. Current methods rely on iterative approximation methods that are both computationally expensive and inherently approximate.

To address this, OPC methodologies are transitioning to GPU acceleration to reduce computation time and cost. This shifts the paradigm from breaking designs into chunks across CPU cores to assigning a grouping of cores to share a GPU machine. While GPUs offer massive parallelism for specific tasks, the primary limitation is GPU memory (VRAM), which must hold all relevant data for all tiles being processed. Moving data on and off the GPU adds significant overhead.

The industry's ability to scale depends on solving the MRC computational tax. If the transition to curvilinear masks remains tethered to inefficient approximation methods, the gains in lithographic precision will be offset by the cost of processing time.

Determine if your current OPC workflow is prepared for the shift from Manhattan geometries to spline-based mask representations.

Source