Search

The advent of laser-based optical tools featuring few-cycle pulses with durations of less than a hundred femtoseconds in the late 1980s enabled scientists to initiate and observe the evolution of chemical reactions. This powerful approach combined the interactions of light and matter and unleashed an unprecedented metrology concept that tracks the interactions of atoms and molecules in their natural timescales. Electron wavepacket dynamics take place in the attosecond range, a thousand times faster than molecules. In optical terms, such durations typically last less than the half-cycle duration of optical fields. Consequently, the investigation of such electronic processes necessitates measurement techniques capable of resolving the oscillations of the electric field of light. The primary objective of this thesis is to develop and advance novel field characterisation techniques based on photoconductive sampling. The first portion of this thesis addresses broadband field characterisation based on nonlinear photoconductive sampling. A theoretical analysis of current formation and localisation in solids is presented, prompting the fabrication of a heterostructured sample with the aim of enhancing the magnitude of the signal obtained from the measurement technique. A thorough proof-of-principle experiment is performed, whereby a significant enhancement in signal magnitude is established. As a consequence of signal improvement, the heterostructured sample reaches the desired stability regime earlier than its traditional bulk counterparts. Moreover, the performance of the heterostructured sample for field characterisation is compared to fused silica and benchmarked against the well-established technique of electro-optic sampling. These results pave the way towards field sampling in low pulse energy systems. The following section details broadband field characterisation based on linear photoconductive sampling by employing tailored pulses from a waveform synthesiser. Visible-ultraviolet pulses are utilised to inject carriers in a common semiconductive material (gallium phosphide), enabling the complete characterisation of a mid-infrared test field. Furthermore, the technique is validated against electrooptic sampling. When compared to electro-optic sampling, the response function of linear photoconductive sampling is concerned with the intensity envelope of the gating field, relaxing the strict requisites on the temporal phase of the gate. The demonstrated results represent a significant achievement in extending field sampling techniques beyond 100 THz and towards the visible range. Finally, a machine learning-based algorithm for denoising waveforms obtained from a laboratory setting is developed and implemented. The algorithm is based on a one-dimensional convolutional neural network, ideal for processing data presented on an evenly spaced grid. The model is compared with well-established methodologies, namely denoising via the fast Fourier transform and wavelet analysis and exhibits excellent performance, extending the repertoire of tools typically used for combating noise. The field characterisation methodologies presented in this thesis pave the way towards accessible and cost-effective field sampling techniques, enabling researchers to study field-induced electron dynamics in matter and usher in ultrafast optoelectronic signal processing towards the PHz range. In general, the field characterisation techniques presented occupy a small footprint, and the measurements take place in ambient air conditions, facilitating their integration in existing experimental infrastructures. With the aid of AI-accelerator chips, the machine learning tool developed in this thesis can be implemented during laboratory measurements as a concurrent denoising technique.