STUDY OF NANO PARTICLE LINEAR AGGREGATE CONFIGURATION USING MACHINE LEARNING ON THE SCATTERING CROSS SECTION SPECTRUM

• Main Author: Innayah, Nazla
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/84797
• Institution: Institut Teknologi Bandung

The development of nanoscale light technology has now been widely utilized, both in terms of health such as helping to detect cancer and in the field of light manipulation such as wave guidance. In this nanoscale technology, one of the components used is the nanoparticle system. Of the various types of geometric structures, spherical geometry is a geometry that is relatively easy to synthesize and study. However, in the synthesis of spherical nanoparticles, aggregation sometimes occurs, for example chain aggregation or clustered aggregation. However, for a particular application, one form of aggregation is more desirable than the others. Therefore, a fast and easy mechanism is needed, where after synthesis the aggregation that occurs can be determined. One way to identify nanoparticle systems and their applications is through light scattering in a particular system. In this thesis, the scattering response to spherical geometry in the form of a scattering cross section is studied using an electromagnetic computational method in the form of a surface integral equation. In the first part, the case of one ball to five balls arranged linearly is reviewed. It was found that the scattering cross section is highly dependent on the configuration of the scattering geometry, the direction of incidence and the polarization of the illumination wave. Depending on the type of configuration and direction and polarization of the illumination, a case with a scattering cross-section with two peaks with quite significant differences in value is obtained, and another case is obtained with a scattering cross-section shape that has a wide distribution with a peak value. In the second part of this thesis, a machine learning system is built to study the scattering case by stating a linear aggregate system of nano spheres which is divided into three aggregation classifications. The first classification is the case of non-aggregated nano spheres, the second classification is that there are some nano spheres that aggregate and some that do not, and the third classification is that all nano spheres aggregate. The machine learning system is trained using part of the dataset in the form of information about the aggregate configuration and its scattering crosssection spectrum. Part of the dataset that is owned and not used for Training is used as a test case for the machine learning that is created. There are six algorithms used in this machine learning, namely Logistic Regression (LR), Linear Discriminant Analysis (LDA), K-nearest neighbors Classifier (KNN), Decision Tree Classifier (CART), Gaussian Naive Bayes (NB), and Support Vector Machine (SVM). The results of machine learning on the case of testing three ball data using trained models for one, two, four, and five balls show the test data accuracy value on SVM of 0.91 and CART of 0.90 with the training data accuracy value on SVM of 0.67 and CART of 0.52. Keywords: nanoparticle aggregates, scattering cross-section, machine learning, and surface integral equations