Evaluation of the standard normal variate method for Laser-Induced Breakdown Spectroscopy data treatment applied to the discrimination of painting layers

• Evaluation of SNV correction efficiency on LIBS spectra in case of laser fluctuations
• Evaluation of SNV efficiency on semi-quantification of trace elements in lead standards with LIBS calibration curves
• Evaluation of SNV efficiency on stratigraphic study with metallic and paint layers as found in cultural heritage
• Representation of spectra corrected with SNV for in depth analysis in 3D color images

Nowadays, Laser-Induced Breakdown Spectroscopy (LIBS) is frequently used for in situ analyses to identify pigments from mural paintings. Nonetheless, in situ analyses require a robust instrumentation in order to face to hard experimental conditions. This may imply variation of fluencies and thus inducing variation of LIBS signal, which degrades spectra and then results. Usually, to overcome these experimental errors, LIBS signal is processed. Signal processing methods most commonly used are the baseline subtraction and the normalization by using a spectral line. However, the latter suggests that this chosen element is a constant component of the material, which may not be the case in paint layers organized in stratigraphic layers. For this reason, it is sometimes difficult to apply this normalization. In this study, another normalization will be carried out to throw off these signal variations. Standard normal variate (SNV) is a normalization designed for these conditions. It is sometimes implemented in Diffuse Reflectance Infrared Fourier Transform Spectroscopy and in Raman Spectroscopy but rarely in LIBS. The SNV transformation is not newly applied on LIBS data, but for the first time the effect of SNV on LIBS spectra was evaluated in details (energy of laser, shot by shot, quantification). The aim of this paper is the quick visualization of the different layers of a stratigraphic painting sample by simple data representations (3D or 2D) after SNV normalization.In this investigation, we showed the potential power of SNV transformation to overcome undesired LIBS signal variations but also its limit of application. This method appears as a promising way to normalize LIBS data, which may be interesting for in-situ depth analyses.