Date of Award

Summer 2026

Language

English

Embargo Period

6-9-2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

College/School/Department

Department of Nanoscale Science and Engineering

Program

Nanoscale Engineering

First Advisor

Gregory Denbeaux

Committee Members

Gregory Denbeaux, Robert Brainard, Christophe Vallée, Vincent LaBella, Kandabara Tapily

Keywords

Photolithography, Chemically Amplified Resists, Chemical Stochastics, EUV Lithography, Chemical Segregation

Subject Categories

Nanoscience and Nanotechnology | Polymer and Organic Materials | Semiconductor and Optical Materials

Abstract

Photolithography is a critical manufacturing step in high-volume manufacturing (HVM) of semiconductor devices, where a photoresist layer is used to transfer nanoscale patterns onto the underlying stack materials. As the microelectronics industry continues to move toward smaller node sizes, driven by Moore's Law. As a result, the tolerances for photoresist performance have become increasingly demanding, which necessitates simultaneous improvements in resolution, defectivity, and roughness. At advanced nodes, these performance limitations are governed by the fundamental stochastic nature of the photochemical processes occurring within the resist film itself, rather than the optical or tool-level constraints. Photon shot noise, the statistical distribution of photoacid generation events, and the nonuniform spatial distribution of resist components all contribute to local variations in exposure, deprotection, and dissolution. This ultimately alters the roughness and defects of the patterned features. A deeper mechanistic understanding of these stochastic processes is therefore required for enabling the defect and roughness performance required by EUV and Beyond EUV (BEUV) lithography at future technology nodes. This thesis addresses the chemical stochastics of EUV chemically amplified resists (CARs) and investigates the origins, mechanisms, and mitigation strategies for chemical non-uniformity arising from molecular segregation during resist film formation.

This thesis investigates the chemical stochastics of EUV chemically amplified resists (CARs), focusing on the thermodynamic and kinetic origins of chemical segregation during resist film formation and on the process and formulation strategies available to suppress it. Chemical segregation which shows up as the spatially structured, non-random separation of chemically incompatible resist components during spin-coat drying, is identified as the dominant non-random contribution to chemical stochastics in multicomponent EUV resist platforms, producing a reproducible, domain-scale heterogeneity in chemical component distribution that directly modulates local acid generation efficiency, acid diffusion length, deprotection threshold, and dissolution rate throughout the resist film.

Five interconnected hypotheses are developed and tested experimentally using a model ESCAP-polystyrene-PGMEA ternary blend system, in which a low-molecular-weight polystyrene homopolymer serves as a chemically incompatible proxy for the chemical species in a commercial CAR formulation.

Hypothesis 1 establishes the thermodynamic framework: Flory-Huggins ternary phase diagrams are constructed from Hildebrand solubility parameters to predict the onset composition at which phase separation is initiated during spin-coat drying, and the predictions are validated by atomic force microscopy characterization of the segregated morphology as a function of ESCAP monomer composition. Systems with higher hydroxystyrene content, and hence larger Flory-Huggins interaction parameters χ₁₂ relative to the critical value χC are shown to produce larger, more fully developed segregated domains under equivalent processing conditions, confirming that the thermodynamic driving force can be reduced through deliberate monomer composition design.

Hypothesis 2 demonstrates that spin speed is the primary process lever for suppressing segregation. The Birnie-Meyerhofer-Strauss (BMS) evaporation model predicts tcoat ∝ ω-0.5, confirmed experimentally by in-situ reflectometry with a fitted exponent of −0.50 ± 0.03. Experimental results shows the possibility of reducing the segregated feature sizes with increasing the spin speed. Initial solid concentration is shown to control film thickness independently of drying time, enabling the two parameters to be adjusted together to achieve any target film thickness while minimizing segregation.

Hypothesis 3 reveals a counterintuitive temperature effect: contrary to the prediction that lower spin-coat temperature would extend the drying time and increase segregation, cold spin coating at −20 °C produced a factor of approximately 9.4 reduction in maximum domain diameter relative to room temperature processing, because the suppression of minority-phase molecular diffusivity Dmol dominates over the Clausius-Clapeyron extension of the drying time.

Hypothesis 4 establishes that increasing the ESCAP matrix molecular weight suppresses domain growth through Rouse-scaling reduction of Dmol, but identifies an optimal window of approximately 10–20 kDa that balances segregation suppression against the competing granularity contribution to line-edge roughness from the growing polymer radius of gyration.

Hypothesis 5 validates solvent vapor annealing at 130 °C as an accelerated screening methodology for ranking commercial resist formulations by their inherent segregation susceptibility without requiring EUV exposure or knowledge of proprietary formulation chemistry.

Together, these results establish a unified quantitative framework connecting molecular-level thermodynamics and polymer physics to macroscopic resist performance metrics relevant to EUV lithography, and identify four independent and combinable process and formulation levers: monomer composition, spin speed, spin-coat temperature, and matrix molecular weight. Each of these parameters can reduce the characteristic segregated domain size D₀. The combined application of all four levers is estimated to reduce D₀ by a factor of 3–5 relative to an unoptimized baseline, representing a meaningful reduction in non-random chemical stochastics and a corresponding improvement in stochastic yield at the N3 node and beyond.

License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download

Share

COinS