Highly sensitive Fourier-transform coherent anti-Stokes Raman scattering spectroscopy via genetic algorithm pulse shaping

We report highly sensitive Fourier-transform coherent anti-Stokes Raman scattering spectroscopy enabled by genetic algorithm (GA) pulse shaping for adaptive dispersion compensation. We show that the non-resonant four-wave mixing signal from water can be used as a fitness indicator for successful GA training. This method allows GA adaptation to sample measurement conditions and offers significantly improved performance compared to training using second-harmonic generation from a nonlinear crystal in place of the sample. Results include a 3 × improvement to peak signal-to-noise ratio for 2-propanol measurement, as well as a 10 × improvement to peak intensities from the high-throughput measurement of polystyrene microbeads under flow.