Effect of deprotection activation energy on lithographic performance of EUVL resist

Abstract

As the feature size becomes smaller, it is difficult for the lithography progress to keep pace with the acceleration of design rule shrinkage and high integration of memory device.Extreme Ultra Violet Lithography (EUVL) is a preferred solution for the 32nm node. In this paper, we have synthesized two types of polymers. One is based on hydroxy phenol, the other is based on hydrocarbon acrylate type polymer. We have diversified each polymer type according to different activation energies for deprotection reaction. In this experiment, we have observed on the resist lithographic performance such as resolution, LER (Line Edge Roughness), photo-sensitivity, and out-gassing during exposure. Different properties according to activation energy were well explained by acid diffusion and polymer free-volume.

Point cloud synthesis with stochastic differential equations

“…18 In this section, we focus on 3D point cloud generation methods, specifically in the area of synthesizing diverse and high-fidelity 3D shapes, which are synthesized point cloud data that are real-world examples Roughly speaking, previous works can be divided into four categories based on their learning manner: auto-encoder based generation, 5,6 autoregressive-based generation, 10 GANs-based generation (generative adversarial networks) [7][8][9][19][20][21] and flow-based generation. 13,16,18,22,23 The success of the diffusion probabilistic distribution approach has inspired many follow-up works that extend the diffusion probabilistic approach to 3D point cloud generation. They have viewed 3D point cloud generation as a probabilistic distribution transform.…”

“…Generally speaking, this task relates to many fields in computer graphics and computer vision, such as 3D shape reconstruction based on 2D image, 16 depth map generation 17 and 3D shape transform 18 . In this section, we focus on 3D point cloud generation methods, specifically in the area of synthesizing diverse and high‐fidelity 3D shapes, which are synthesized point cloud data that are real‐world examples Roughly speaking, previous works can be divided into four categories based on their learning manner: auto‐encoder based generation, 5,6 autoregressive‐based generation, 10 GANs‐based generation (generative adversarial networks) 7‐9,19‐21 and flow‐based generation 13,16,18,22,23 …”