Artificial Neural Networks for Noise Removal in Data-Sparse Charged Particle Imaging Experiments

Area of Science:

• Physical Chemistry
• Chemical Physics
• Computational Chemistry

Background:

• Charged particle imaging is crucial for studying unimolecular photochemical dynamics.
• Low photon counts often lead to Poissonian noise, hindering quantitative analysis.
• Existing noise reduction methods struggle with extremely low signal levels.

Purpose of the Study:

• To demonstrate the efficacy of artificial neural networks (ANNs) for Poissonian noise removal in low-count charged particle imaging.
• To validate the ANN approach using both simulated and experimental data.
• To assess the potential impact of ANNs on chemical dynamics research.

Main Methods:

• Application of artificial neural networks (ANNs) for noise reduction.
• Testing on simulated and experimental image data from photoion/photoelectron detection.
• Analysis of multiphoton ionization of pyrrole and (S)-camphor.

Main Results:

• ANNs successfully transformed unusable images into statistically reliable data.
• High performance demonstrated, with results showing impressive similarity to benchmark references.
• The method proved effective even with very low signal-to-noise ratios.

Conclusions:

• ANNs offer a powerful solution for noise reduction in low-count charged particle imaging.
• This approach significantly enhances the quantitative analysis capabilities in chemical dynamics.
• The technique holds great promise for studies involving low cross-sections or subtle image features.