Inverse design of Photonic crystal fiber for Four-Wave-Mixing application

We demonstrate a method of photonic crystal fiber (PCF) inverse design for nonlinear wavelength conversion based on four-wave mixing (FWM), where deep learning neural networks (DNN) is applied to predict PCF structure parameters. By applying empirical formula of PCF dispersion instead of numerical simulation, a large dataset of phase-matching curves is generated of various PCF designs. The average running time of DNN prediction is 0.2s. With the help of DNN, we design and fabricate a PCF for wavelength conversion via FWM from 1064 nm to 770 nm. Pumped by a microchip laser at 1064 nm, signal wavelength is detected by optical spectrum analyzer at 770.2 nm.