Theses and Dissertations
Permanent URI for this collection
Browse
Browsing Theses and Dissertations by Subject "Adsorption energy"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
Item Embargo Predicting Gas Sensor Materials Using Machine Learning(2026-05-19) Shandukani, Kharavho; Maluta, N. E.; Dima, R. S.; Ranwaha, T. S.Gas sensor devices are employed globally by industries and individuals for the purpose of air quality monitoring and the detection of harmful gases. These sensors are important in the identification and tracking of gases that pose significant threats to human life. The mechanism of these devices is based on the electrode material within the device, which includes its interaction with other gases. Despite the wide range of gas sensors available in the market today, they still face challenges such as sensitivity, selectivity, low-temperature operation, and cost-effectiveness. Addressing these limitations necessitates the exploration of novel materials, a process that is often experimentally time-intensive and costly. To mitigate these challenges, this study proposes an innovative approach to accelerate the discovery of materials for gas sensing applications, which constitutes the focus of this research. This study presents a synergy of density functional theory (DFT) and machine learning (ML) to accelerate the discovery of perovskite oxide-based gas-sensing materials. DFT was employed to calculate the adsorption energy, adsorption distance, and adsorption angle. The study focused on LaCoO3(110) surfaces doped with 20 different transition metals. The data generated from DFT calculations were used to train ML models for classification, support vector classifier (SVC), and for regression (random forest regressor (RFR), gradient boosting regressor (GBR), support vector regressor (SVR), k-nearest neighbors (KNN), decision tree regressor (DTR), ridge regressor (RR), and Lasso regressor (Lasso)). The SVC achieved approximately 90% classification accuracy in distinguishing high and low adsorption materials, while GBR and SVR both yielded the highest regression performance ( 𝑅2 ≈0.97), closely reproducing DFT calculation results for the adsorption energy as the target. Explainable AI framework named SHapley Additive exPlanations (SHAP) analysis identified dopant electronegativity, atomic radius, and electron density as dominant factors influencing adsorption strength. The integration of DFT and ML substantially reduces the computational and experimental screening time, accelerating the identification of promising gas-sensing material candidates.