Prediction of Arsenic Adsorption on To Metal Organic Frameworks and Adsorption Mechanisms Interpretation by Machine Learning

Metal organic frameworks (MOFs) is a promising adsorbent for the removal of arsenic (As) from wastewater. The removal efficiency of As is influenced by many factors, such as the textural properties of MOFs, adsorption conditions, and As species. Examining all relevant factors through traditional experiments is challenging. In order to predict the As adsorption performance of MOFs and reveal the adsorption mechanisms, four machine learning models were developed to predict adsorption capacities of total, organic, and inorganic As using adsorption conditions, properties of MOFs, and characteristics of different As species as inputs. Results demonstrated that extreme gradient boosting (XGBoost) had the best predictive performance for all predictions (R2 of 0.93–0.96). The further validations indicate the high accuracy of the inorganic As-based XGBoost model. The feature importance analysis showed that the concentration of As, surface area of MOFs, and pH of solution were the three key factors to inorganic As, while those for organic As were the concentration of As, pHpzc value of MOFs, and the state of metal clusters. The formation of coordination complexes between As and MOFs may be the major adsorption mechanisms for both inorganic and organic As. Whereas, electrostatic interaction may have higher effect on organic As adsorption than that on inorganic As adsorption. This work provides a new strategy for evaluating the adsorption of As on MOFs and discovering the underlying decisive factors as well as adsorption mechanisms, both of which facilitate future researches on As wastewater treatment