Microstructured polymer fibers have been widely studied for terahertz waveguides, sensing, environmental monitoring, and medicine. Time-consuming methods including finite element method and finite-different time-domain have been used to design and simulate microstructure polymer fibers. Deep learning method enabled by artificial neural networks provides an effective alternative to accurately predicting the core mode of optical properties within a short time, which avoids the waste of data resources. Designing optimal artificial neural networks with various hyperparameters for each optical property considerably improve prediction accuracy. To assess whether high-accuracy artificial neural networks were trained by checking absolute percentage errors between predicted values and actual values on the validation set. We utilized the successfully trained artificial neural networks to accurately evaluate the optical properties of the unknown geometry including effective refractive index neff , and effective mode area (Aeff ) and took the logarithm to form a more uniform distribution also had been successfully predicted the confinement loss (αc). We also demonstrated artificial neural networks that predict the output for unknown geometric parameters, with several orders of magnitude speed-up compared to the finite element method simulation. Moreover, this approach can be easily applied to other similar types of optical fibers which accelerate the optical function device design process.

No abstract

No abstract

We proposed and simulated a surface plasmon resonance (SPR) temperature sensor with two loss peaks in a hollow core negative curvature fiber (HC-NCF). Inner walls of the anti-resonant tubes in HC-NCF were plated with gold films to stimulate SPR, while the thermo-optic mixture of toluene and chloroform was filled in the air holes in HC-NCF to modulate the coupling between core modes and surface plasmon polaron modes (SPPMs). Simulation results showed that two SPPMs with opposite thermo-optic respond effects were excited at two separate wavelength bands due to their different dispersion characteristics. Temperature measurement sensitivities of −3.976 nm/℃ and 1.071 nm/℃ were obtained for the two SPPMs, while the sensitivity reached −5.047 nm/℃ when detected the wavelength interval between the two SPPMs loss peaks. The two separate loss peaks could also be utilized in self-verification. The designed temperature sensor based on HC-NCF and SPR depicts high sensitivity and self-verification, which could be utilized for high precision and stable temperature monitoring.

No abstract