The structural and electronic properties of anionic SinCn clusters with n = 1-6 were investigated. Ab initio Monte Carlo (MC) simulations were used to search for anionic silicon carbide (SiC) clusters corresponding to the local min- ima. The anionic SiC clusters corresponding to the global minimum were then calculated using density functional theory (DFT). Neutral and cationic SinCn clusters with n = 1-6 were simulated using ab initio MC simulations and DFT calculations. The relative energies of the neutral, cationic, and anionic SiC clusters were determined. The numbers of isomers of the anionic SiC clusters were 5 (Si2C2), 12 (Si3C3), 20 (Si4C4), 31 (Si5C5), and 20 (Si6C6). The atomization energies, second differences of the energies, average bond lengths and angles of the anionic SiC clusters, and adiabatic and vertical electron affinities were calculated. The Bader and Mulliken charges of the anionic SiC clusters were also analyzed.

In this study, a machine learning-based redox reaction judgment program was developed to help students under- stand and assess the nature of redox reactions. Students directly generated information on various chemical equations based on the electron transfer model, oxidation number change model, and Goodstein model, and created a classification model using a decision tree algorithm based on this information. When machine learning showed errors in classification, the decision tree was used to identify the cause of the error, allowing students to correct their misconceptions independently, and the process of correcting misconceptions was analyzed. The study was conducted as an online class for 18 sessions for third-year high school students, and the program's effectiveness was evaluated by analyzing students' academic achievement, attitudes toward using redox reaction models, and perceptions of AI-based classes. The results showed that the redox reaction judgment program using decision tree classification model was effective in improving students' academic achievement and correcting misconceptions, with particularly notable results in the oxidation number change model and the Goodstein model. As the study progressed, the percentage of students who responded ‘I cannot explain’ decreased, and they demonstrated an understanding of the advantages, disadvantages, and limitations of the models, choosing appropriate models to interpret chemical reactions according to the situation. Furthermore, students modified their approach from interpreting redox reactions based on simple memorization or calculation to developing an attitude of understanding chemical bonds, structures, and electronegativity, interpreting them from a process perspective. The immediate feedback from AI contributed to improving the learning experience by helping students correct errors and automate their schemas, and students showed a positive attitude toward using AI tools in their chemistry learning. These results suggest that the program developed in this study contributed to the cultivation of basic digital and artificial intelligence literacy. Therefore, the need for chemistry classes using artificial intelligence tools that students can easily handle is emphasized, and teachers should understand and utilize the possibilities and limitations of artificial intelligence in a balanced manner as guides and facilitators of learning.