Development of angle-resolved light scattering system and computation techniques for ultrafast surface spectroscopy

Abstract

This work focused on the development of a time- and angle-resolved hemispherical elastic light scattering (tr-ARHELS) optical apparatus, which was applied to monitor the phase transition and complex pathways of photoinduced nonlinear optical dynamics in V3O5 thin film. This setup enabled us to perform spectral imaging of optical diffraction and ultrafast diffraction conoscopy with femtosecond resolution. The investigation of ultrafast photoinduced processes in phase-change materials is of special interest since methods of ultrafast spectroscopy can potentially track electron and phonon lattice dynamics separately and detect small structural variations on the surface of thin films. To achieve higher flexibility in optical measurements, a non-collinear optical parametric amplifier was built to generate broadband pulses with variable wavelengths by using an optical parametric amplification process. Since our main goal was to study short events, a pulse compressor was built in order compress the pulse from 130 to 20 femtoseconds. All the components of the tr-ARHELS setup such as charge-couple device (CCD) camera, translational stages, stepper motors, and photo detector, were integrated and programmed. Using a vacuum system and a refrigerator with a cold finger, this scatterometer operated at cryogenic temperatures, reaching vacuum level of 10−6 Torr and a temperature of about 7K. The implementation of the Gerchberg-Saxton error reduction and hybrid input-output phase-retrieval algorithms allowed for the reconstruction of the power spectral density function and the correct calculation of the surface autocorrelation function, providing a solution to the electromagnetic wave scattering inverse problem. A real-time visualization of autocorrelation function was achieved for femtosecond spectroscopy of phase-change materials. The development of parallel computational algorithms has been carried out by graphics processing unit (GPU) computing using Compute Unit Device Architecture (CUDATM) with an "NVIDIA Tesla K80" GPU, and the Compute Unified Fast Fourier Transform (CUFFT) parallel programming library. The parallel implementation of phase-retrieval algorithms provides ~30x speed-up and efficiency.