Abstract: A combined Raman and electrochemical study of natural orpiment (As2S3), an arsenic sulfide pigment, was used to assess the quick formation of oxidized species such as arsenic oxide (As2O3) upon exposing the pigment to 405 nm or 532 nm monochromatic light while simultaneously recording the Raman spectra of the exposed sample. During this process, a distortion of the main band at 355 cm−1, associated with the stretching of the AsS3/2 pyramids of natural orpiment, was observed as well as an increased intensity of the 359 cm−1 band, corresponding to covalent AsAs bonds in natural orpiment. The distortion was accompanied by an overall decrease of the global Raman signal for natural orpiment, which could be explained by a loss in the crystal structure. The same phenomena were recorded in reference natural orpiment model paint samples stored for a long time together with verdigris (Cu(OH)2·(CH3COO)2·5H2O) and minium (Pb3O4) paints, the latter two appearing darkened on their sides closest to the orpiment sample as well as in several historical samples containing natural orpiment mixed with various blue pigments. By SEM-EDX and XRPD analysis, respectively on loose material and cast thin-sections of model paint samples, the darkening was identified as dark sulfide species such as chalcocite (Cu2S) and galena (PbS), suggesting the release of volatile sulfide or related species by the natural orpiment paint. XANES analyses of paint samples presenting AsAs bond increase indicated the presence of sulfur species most likely identified as organosulfur compounds formed upon the AsAs bond formation and explained the darkening of the Cu- and Pb-based pigments. To the best of our knowledge, this article reports for the first time the light-induced formation of AsAs bonds in natural orpiment used as an artists' pigment and objectively demonstrates the incompatibility between orpiment and (arsenic) sulfide-sensitive pigments.

Abstract: The added value of supervised Machine Learning (ML) methods to determine the Absolute Configuration (AC) of compounds from their Vibrational Circular Dichroism (VCD) spectra was explored. Among all ML methods considered, Random Forest (RF) and Feedforward Neural Network (FNN) yield the best performance for identification of the AC. At its best, FNN allows near-perfect AC determination, with accuracy of prediction up to 0.995, while RF combines good predictive accuracy (up to 0.940) with the ability to identify the spectral areas important for the identification of the AC. No loss in performance of either model is observed as long as the spectral sampling interval used does not exceed the spectral bandwidth. Increasing the sampling interval proves to be the best method to lower the dimensionality of the input data, thereby decreasing the computational cost associated with the training of the models.

Abstract: The use of a detailed Monte Carlo simulation code for X-ray fluorescence spectrometers is demonstrated for calculating the outcome of X-ray scattering experiments in the incident energy range 40-80 keV. The code was validated by comparisons of experimental and simulated spectral distributions in the case of thick, homogeneous samples in which multiple photon scattering occurs with high probability. The experimental spectral distributions were collected at beamline BW5 of HASYLAB, Germany, where a highly energetic, monochromatic synchrotron beam is available. With respect to heterogeneous samples, the code was employed to evaluate the use of Rayleigh and Compton scatter signals for obtaining three dimensional information on the sample dark matrix composition.