Exploring the Role of Artificial Intelligence in Precision Photonics: A Case Study on Deep Neural Network-Based fs Laser Pulsed Parameter Estimation for MoOx Formation

Ultrafast pulsed laser technology presents unique challenges and opportunities in material processing and characterization for precision photonics. Herein, an experiment is conducted involving the use of an ultrafast pulsed laser to irradiate a molybdenum film, inducing oxide formation. A total of 5...

Descripción completa

Detalles Bibliográficos
Autores: Paredes-Miguel, Jose R., Cano-Lara, Miroslava, Garcia Granada, Andres Amador, Espinal, Andres, Villaseñor-Aguilar, Marcos-Jesús, Martínez-Jiménez, Leonardo, Rostro Gonzalez, Horacio
Tipo de recurso: artículo
Fecha de publicación:2025
País:España
Institución:Universitat Ramon Llull (URL)
Repositorio:DAU Arxiu Digital de la Universitat Ramon Llull
OAI Identifier:oai:dau.url.edu:20.500.14342/5389
Acceso en línea:http://hdl.handle.net/20.500.14342/5389
https://doi.org/10.1002/adpr.202400113
Access Level:acceso abierto
Palabra clave:Deep neural networks
Material characterization
Molybdenum thin films
Oxide formation
Ultrafast pulsed lasers
Molibdè
Làsers d'impulsos ultracurts
Òxids
621
Descripción
Sumario:Ultrafast pulsed laser technology presents unique challenges and opportunities in material processing and characterization for precision photonics. Herein, an experiment is conducted involving the use of an ultrafast pulsed laser to irradiate a molybdenum film, inducing oxide formation. A total of 54 experiments are performed, varying the laser irradiation time and per-pulse laser fluence, resulting in a database with diverse oxide formations on the material. This dataset is further expanded numerically through interpolation to 187 samples. Subsequently, eight different deep neural network models, each with varying hidden layers and numbers of neurons, are employed to characterize the laser behavior with different parameters. These models are then validated numerically using three different learning rates, and the results are statistically evaluated using three metrics: mean squared error, mean absolute error, and R2 score.