Spatial distribution of reaction products in positive tone chemically amplified resists

Abstract

The perpetual advancement of materials and equipment for microlithography has resulted in the ability to print critical dimensions that approach the size of the molecules that make up photoresists. As a result, molecular scale effects such as line edge roughness have become a concern for both resist manufacturers and process engineers. In this work we have investigated the increasing importance of molecular level effects, especially in terms of the contributions of the exposure and postexposure bake (PEB) steps to spatial variations in film composition. A mesoscale simulation of the PEB was used to model the discrete mass transport and reaction events that create the changes in film composition responsible for resist function. Local irregularities in resist composition are generated during the PEB, the magnitude of which can be related to the local concentration of acid. This study is focused on the establishment of an understanding of the effects of process and composition variables on the reaction product distribution. The reaction product distribution was calculated for an APEX®-like resist under a variety of exposure and bake conditions. These process variables have a profound influence on spatial irregularities in the composition gradient. Ultimately, it is the interaction of this reaction product distribution with the development process that will determine line edge roughness.

Effect of copolymer composition on acid-catalyzed deprotection reaction kinetics in model photoresists

“…The current approach on blanket exposed films is a special case of the conventional model [21]. It is difficult to model a variable diffusivity because the system must be treated heterogeneously as in the mesoscale simulation methods of Schmid et al [5] or stochastic simulation methods of Houle et al [13]. In some cases, the photoacid diffusion coefficient can vary by several orders in magnitude in protected versus deprotected polymers [12].…”

“…However, the performance of actual photoresist systems requires recognition of the spatial heterogeneity of the reaction-diffusion process that arises from the discrete concentration of the acid and the local composition of the material. Schmid et al [5] used mesoscale simulations to understand the role of photoacid distribution on this heterogeneity. They concluded that the contrast between regions of high versus low photoacid concentrations was needed to minimize the deprotection gradient to enable sharper image profiles.…”

A review of line edge roughness and surface nanotexture resulting from patterning processes

“…It takes into account the molecular structure of the photoresist, the photoacid generator presence, its initiation, diffusion and reaction to create deprotected sites as well as the dissolution of the exposed areas using a quasistatic fast dissolution algorithm [32][33][34]. Excellent process simulation and experimental studies have also appeared in the literature for the process effects on LER, such as aerial image contrast [35][36][37], shot noise [38], development process [39][40][41][42][43], lithographic process conditions [44][45][46][47][48], lithographic materials [32,33,[49][50][51][52][55][56][57][58][59] and others. Our simulation results are consistent with these studies.…”

Relevance of technology computer aided design (TCAD) to process-aware design

“…This evolving status of resist modeling, coupled with the urgent demands of the semiconductor development schedules, has led to a hierarchy of resist models in the complexity-run time space, as shown in At one end of this map are some of the most physically based resist models that track individual molecules and rely on molecular dynamics or Monte Carlo techniques to simulate the reactions and transport phenomena that occur during resist processing. 1 These models are designed to accept intrinsic resist properties ͑such as polydispersity͒ as input parameters, and provide a link between such properties and lithographic performance. However, such models are both computationally expensive and relatively nascent.…”