2.4. Spectral data processing
The FTIT-ATR and LIBS spectra were pre-processed with baseline correction and a Savitzky–Golay smoothing filter (Savitzky & Golay, 1964), to reduce the baseline float/noise and improve the signal-to-noise ratio. Matlab software (Mathworks, USA) was used to smooth the data and for all other statistical analyses performed.
The 2DCOS analysis was performed using the method established by Noda and Ozaki (Noda, 2018; Noda & Ozaki, 2005). The intensity of the synchronous 2D correlation spectrum, Ф (λ1, λ2), represents the simultaneous or coincidental changes in the spectral intensity at wavenumbers λ1 and λ2 with an external perturbation. The peaks on the diagonal line are called auto peaks, while the peaks on the off-diagonal line are called crosspeaks. The asynchronous spectrum, Ψ (λ1, λ2), represents the out-of-phase or sequential changes in spectral intensities with external perturbation. There is no diagonal peak, but there is a cross peak in off-diagonal areas (Noda, 2018). The red and blue colors on the map represent positive and negative intensities, respectively.
In this study, different salt contents were applied as the external perturbation for the FTIR spectra, while the different vertical detecting points were used as external perturbations for the LIBS spectra. In our study, the spectral synchronous and asynchronous correlation maps were generated from the analysis of the ATR and LIBS spectra, respectively.