Design and prediction of refractive index of crystals used in the electro-optic sampling detection technique assisted by machine learning

Electro-optic sampling (EOS) detection technique has been widely used in measuring the electric field time waveform of terahertz pulses and few-cycle laser pulses, although it might produce some pulse distortions in the near- and midinfrared band. The performance of EOS technique is directly affected by the parameters of electro-optic crystal, including its thickness, refractive index, and absorption coefficient. In order to obtain a flat broadband frequency response function, machine learning is used to predict the refractive index of the crystal used in the EOS detection in the near-infrared band. The Sellmeier equation with five coefficients is used to calculate the refraction index of crystal, and these five coefficients are found assisted with machine learning. Then, three sample crystals with their coefficients and frequency response curves are given to prove rationality. After seeking the coefficients limited in the range of [0,10], two best curves and their refractive indexes are obtained. This work provides a reference for selecting electro-optic crystals used in the near-infrared band.