Construction of a Wide-Field MOKE Microscope for Domain Analysis with CNNs

This thesis presents the development of a wide-field magneto-optical Kerr effect (MOKE) microscope and a machine learning approach for magnetic domain analysis. The microscope was constructed from fundamental optical components, incorporating both 662 nm laser and LED illumination sources. A significant challenge of laser speckle patterns was overcome through the implementation of a Laser Speckle Remover, enabling clear image formation. The system achieved a theoretical resolution of 0.6 μm using a 20× objective lens. In experimental observations of 25 nm cobalt thin films, polarization-dependent contrast was detected, though its magnetic origin requires further verification. Complementing the experimental work, a convolutional neural network was developed to predict magnetic parameters from domain images. The network was trained on simulated magnetic domain patterns generated using micromagnetic simulations. The initial architecture achieved an R2 score of 59.11 % for two-parameter prediction. Subsequent improvements through the introduction of specialized regression heads and batch normalization significantly enhanced performance, reaching R2 scores of 83.5 % for two-parameter prediction ( , ) and 74.4 % for three-parameter prediction ( , , temperature). This integrated approach establishes a foundation for comprehensive magnetic domain analysis and characterization.

• Autores: Rueda Torres, Juan David
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2025
• Institución: Universidad de los Andes
• Repositorio: Séneca: repositorio Uniandes
• Idioma: eng
• OAI Identifier: oai:repositorio.uniandes.edu.co:1992/75920
• Acceso en línea: https://hdl.handle.net/1992/75920
• Palabra clave: Microscopy; MOKE; Kerr; CNNs; Magnetic domains; Microscopía; Dominios Magnéticos; Física
• Rights: openAccess
• License: Attribution-NonCommercial-NoDerivatives 4.0 International