Study On The Spectral Characteristics And Origin Identification Of Nephrite From Different Producing Areas In China

At present,the research on nephrite mainly focuses on gemological characteristics,structural characteristics,phase composition and coloration mechanism,but there are few studies on the identification methods of nephrite producing areas with different colors.Therefore,in this paper,X-ray powder diffractometer（XRD）,laser Raman spectrometer（Raman）,Fourier transform infrared spectrometer（FTIR）and ultraviolet-visible-near infrared spectrometer（UV-vis）were used to test the ink jade and green jade samples in Xinjiang,Qinghai and Guangxi.Principal component analysis（PCA）was used to extract principal component factors,and PCA-FDA（Principal Component Analysis-Fisher’s Discriminant Analysis）discriminant model and PCA-BP（Principal Component Analysis-Back Propagation Neural Network）neural network model of nephrite from different producing areas were established according to the extracted principal component factors.The discriminant prediction accuracy of different models was compared to explore the better discriminant model of nephrite producing area.The results showed that:（1）XRD diffraction analysis test found that the nephrite from various producing areas is basically consistent with the standard tremolite diffraction peak（PDF#85-0876）.The three strong diffraction peaks of the main diffraction line are d₁₁₀=8.2922,d₃₁₀=3.120,d₁₅₁=2.7225.（2）The Raman spectra of nephrite from Xinjiang and Qinghai show the significant characteristics of tremolite.It is found that the near-infrared spectra of nephrite samples from Qinghai and Xinjiang show that 4190、4321、7176cm^-1 are mainly caused by Mg-OH bending vibration,while 5225cm^-1 shows an obvious adsorption water peak.Nephrite from different origins are composed of tremolite and a small amount of actinolite,with similar chemical composition.The characteristic peaks at 239 nm,663 nm and 942 nm and the wide absorption band centered at 498 nm appear in the UV-Vis-NIR spectra of nephrite from Xinjiang,Qinghai and Guangxi.The sample represents the characteristic peak of chromium ion at 663 nm,which gradually becomes stronger with the deepening of color.（3）The highest accuracy of PCA-FDA discriminant model is 88.9%based on mid-infrared spectroscopy.The highest accuracy of PCA-BP neural network model is100%based on near-infrared spectroscopy.The results show that both neural network models have certain ability to predict the origin of nephrite,and can make a correct judgment on the origin of nephrite to some extent.The best discriminant effect of non-destructive testing is the PCA-BP neural network model based on Raman spectroscopy,and the comprehensive accuracy rate reaches 97.1%.The best discriminant effect is the PCA-BP neural network model based on near-infrared spectral data（KBr tableting）,and the comprehensive accuracy rate is 100%.