|
| | Abstract | | Because mineral textures of rocks as well as their mineral compositions have quantitative information on their formation histories, it is important to develop a technique of the mineral determination and quantitative texture description. To map minerals of pillow and massive basalts of the Juan de Fuca plate, chemical compositional maps on thin sections of the basaltic samples from Integrated Ocean Drilling Program Expedition 301 were constructed via Electron Probe Micro-Analyzer, EPMA. From the chemical compositional maps via EPMA, the mineral map has been estimated by considering the fraction of chemical components of each mineral. For the lithology which contains several kinds of minerals, however, we need to determine mineral map by considering several chemical compositions simultaneously. Therefore we should develop a method to compare the chemical components on multi-dimensional domain. Herein, mineral maps were constructed via an unsupervised neural network called as Self-Organizing Map from the several compositional maps. The classified mineral maps reveal mineralogical and texture differences between pillow basalt and massive flow basalt. Using the classified mineral maps, the fraction of bulk mineral components could be estimated. Furthermore, the digitalized data of the mineral map enable us to quantify size distribution of phenocryst and its fractions. |
|
|
|
| |
|