OPTIMIZING BAND GAP AND PHOTODEGRADATION PERFORMANCE FROM DOPED ZNO WITH MACHINE LEARNING APPLICATION

• Main Author: Amaldi, Haris
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/71739
• Institution: Institut Teknologi Bandung

Synthetic textile dyes are one of the types of pollutants that are classified as Persistent Organic Pollutants (POPs) because they cannot be degraded by conventional waste treatment. For this case, photocatalytic-based Advanced Oxidation Processes (AOPs) is an attractive alternative solution to POPs because it is using sunlight to degrade synthetic dyes into CO2 and H2O. One of widely used photocatalysts is zinc oxide (ZnO) due to its good oxidation, good photocatalytic properties, and appropriate band gap. However, ZnO has a wide band gap (3.2 eV) that can only use the UV light, which make up 7% of the sunlight spectrum. Adding the dopant for ZnO is one of the methods to decrease its band gap and improve its capability to use the sunlight. The band gap and photocatalytic performance of ZnO depend on the amount of dopant. Hence, it is necessary to optimize the amount of dopant so that ZnO can use the sunlight optimally. In this case, machine learning can be an appropriate tool to optimize the doping concentration as scientists had previously used the same method to optimize another photocatalyst. This research is studying the use of machine learning to optimize dopant concentration on ZnO. This is done to get the lowest band gap, the best photodegradation performance of ZnO, and the accuracy level of machine learning. In this research, methylene blue (MB) is chosen because of its wide application in textile coloring and its lethality to microorganisms. Optimizing is done by varying the type of dopant (Fe and Al), light source (visible, UV, sunlight), and degradation time using data from previous research. The machine learning algorithm uses Python language. From this research, we know that this machine learning can predict ZnO band gap accurately with average accuracy at 97.5%. However, this machine learning cannot accurately predict the performance of ZnO in MB photodegradation. The accuracy of the three models ranges from 72.22 – 84.87%. The first recommendation for future research includes conducting a laboratory experiment to test band gap in real life. Secondly, photodegradation data can be added to increase the accuracy of machine learning. Lastly, the database can be updated with results from other types of textile dyes.