Publications

Hydrogen is a natural gas that can be used as a low-carbon renewable energy source. Ultramafic rocks that have undergone serpentinization serve as source rocks for hydrogen. Hydrogen forms as a result of the oxidation of iron (Fe²⁺) in olivine and orthopyroxene minerals to (Fe³⁺) through reactions with fluids.

In this process, in addition to physical parameters like fluid presence, temperature, and pressure, the abundance of olivine and orthopyroxene minerals is a key parameter in the formation of hydrogen source rocks. The higher the Fe content in an ultramafic rock—represented by ultramafic minerals such as olivine and pyroxene—the greater the likelihood that the rock will undergo high-grade serpentinization, where olivine and pyroxene are serpentinized into serpentine, thus contributing to the formation of a serpentine-rich rock and generating hydrogen. Therefore, an efficient method is needed to determine and quantify the abundance of olivine and pyroxene in ultramafic rocks to identify high-grade serpentinization.
This research employs an Artificial Neural Network (ANN) to classify the degree of serpentinization in thin sections of ultramafic rocks. In this study, the ANN is used for supervised learning, where the input consists of image features extracted from thin section images of the rocks. These features could include pixel intensity values, texture patterns, and mineralogical characteristics that are indicative of the serpentinization grade. The ANN is trained on a dataset of labeled images, with each image associated with a known serpentinization grade. During training, the network learns to map these image features to the corresponding grades of serpentinization. Once trained, the ANN can classify new images of ultramafic rock thin sections, providing an automated and efficient means of assessing the degree of serpentinization and identifying potential hydrogen source rocks.