Uncovering Key Factors in Graphene Aerogel-Based Electrocatalysts for Sustainable Hydrogen Production: An Unsupervised Machine Learning Approach.

Abstract

The application of principal component analysis (PCA) as an unsupervised learning method has been used in uncovering correlations among diverse features of aerogel-based electrocatalysts. This analytical approach facilitates a comprehensive exploration of catalytic activity, revealing intricate relationships with various physical and electrochemical properties. The first two principal components (PCs), collectively capturing nearly 70% of the total variance, attested the reliability and efficacy of PCA in unveiling meaningful patterns. This study challenges the conventional understanding that a material's reactivity is solely dictated by the quantity of catalyst loaded. Instead, it unveils a complex perspective, highlighting that reactivity is intricately influenced by the material's overall design and structure. The PCA bi-plot uncovers correlations between pH and Tafel slope, suggesting an interdependence between these variables and providing valuable insights into the complex interactions among physical and electrochemical properties. Tafel slope stands to be positively correlated with PC1 and PC2, showing an evident positive correlation with the pH. These findings showed that the pH can have a positive correlation with the Tafel slope, however, it does not necessarily reflect a direct positive correlation with the overpotential. The impact of pH on current density (j)and Tafel slope underscores the importance of adjusting pH to lower overpotential effectively, enhancing catalytic activity. Surface area (from 30 to 533 m2 g-1) emerges as a key physical property, inclusively inverse correlation with overpotential, indicating its direct role in lowering overpotential and increasing catalytic activity. The introduction of PC3, in conjunction with PC1, enriches the analysis by revealing consistent trends despite a slightly lower variance (60%). This reinforces the robustness of PCA in delineating distinct characteristics of graphene aerogels, affirming their potential implications in diverse electrocatalytic applications. In summary, PCA proves to be a valuable tool for unraveling complex relationships within aerogel-based electrocatalysts, extending insights beyond catalytic sites to emphasize the broader spectrum of material properties. This approach enhances comprehension of dataset intricacies and holds promise for guiding the development of more effective and versatile electrocatalytic materials.