Efficient optimization of SHG hotspot switching in plasmonic nanoantennas using phase-shaped laser pulses controlled by neural networks.

We present a novel procedure for manipulating the near-field of plasmonic nanoantennas using neural network-controlled laser pulse-shaping. For our model systems we numerically studied the spatial distribution of the second harmonic response of L-shaped nanoantennas illuminated by broadband laser pulses. We first show that a trained neural network can be used to predict the relative intensity of the second-harmonic hotspots of the nanoantenna for a given spectral phase and that it can be employed to deterministically switch individual hotspots on and off on sub-diffraction length scale by shaping the spectral phase of the laser pulse. We then demonstrate that a neural network trained on a 90 nm × 150 nm nano-L can, in addition, efficiently predict the hotspot intensities in an antenna with different aspect ratio, after minimal further training, for varying spectral phases. These results could lead to novel applications of machine-learning and optical control to nanoantennas and nanophotonics components.