PROCESS PROXIMITY CORRECTION (PPC) METHOD BASED ON DEEP LEARNING, AND SEMICONDUCTOR MANUFACTURING METHOD COMPRISING THE PPC METHOD

Smart overview of the Invention

A deep learning-based Process Proximity Correction (PPC) method is designed to enhance the reliability and process margin in semiconductor manufacturing. This method involves several key steps, starting with receiving a first layout from After Clean Inspection (ACI), which contains multiple patterns. The process includes extracting and sampling features from these patterns, optimizing the sampling through an algorithm, and developing a deep learning model based on the derived optimization and dedose conditions. The model is then used to perform corrections on the layout.

Technical Process

The PPC method begins by receiving a first layout of ACI, which is crucial for the subsequent steps. It involves extracting features from unique patterns within the layout and sampling these features. An optimization algorithm is applied to refine the number of unique patterns sampled. The optimization and dedose conditions identified by this algorithm are crucial for creating a robust deep learning model. This model is pivotal for executing corrections on the semiconductor device layout, ensuring enhanced accuracy and efficiency in the manufacturing process.

Manufacturing Method

Incorporating the PPC method into semiconductor manufacturing involves generating a second layout using the deep learning-based PPC on the initial layout. This second layout, representing After Development Inspection (ADI), is further refined through Optical Proximity Correction (OPC) to produce a third layout. This sequence ensures that the semiconductor device is manufactured with high precision, aligning with the intended design specifications and reducing errors.

Mask Manufacturing

The method also applies to mask manufacturing, where the initial layout undergoes PPC to form a second layout, followed by OPC to create a third layout. This third layout is used to generate Mask Tape-Out (MTO) design data, which is crucial for preparing mask data. The mask substrate is then exposed to light based on this data, ensuring that the mask accurately reflects the desired patterns, which is critical for the photolithography process in semiconductor production.

Detailed Analysis

The detailed process involves various analytical steps, such as feature extraction using Principal Component Analysis (PCA) and Locally Linear Embedding (LLE). These analyses help in understanding feature effectiveness and optimizing the sampling process. The method also includes binning and sorting features to enhance the accuracy of the PPC model. These steps ensure that the final semiconductor device meets high standards of precision and functionality, minimizing potential deviations during the manufacturing process.