Interfacial Behavior from the Atomic Blueprint: Machine Learning-Guided Design of Spatially Functionalized a-SiO2 Surfaces

Abstract

a-Quartz surfaces functionalized with hydroxyl and methyl groups provide a versatile platform for controlling interfacial properties critical to applications such as catalysis, protective coatings, and energy conversion. The arrangement of these functional groups strongly influences interfacial interactions at solid-liquid interfaces, highlighting their relevance to colloid and interface science. However, conventional models often treat surface functionalization as spatially homogeneous, overlooking the atomic-scale organization of surface groups. We hypothesize that this spatial distribution, beyond overall composition, plays a decisive role in governing surface stability and interfacial behavior. To test this hypothesis, we employ a multi-scale simulation workflow combining density functional theory, ab initio molecular dynamics (AIMD), and machine-learned force fields (MLFFs). This approach allows us to explore a range of spatial patterns of OH/CH3 functionalization on the a-quartz (0001) surface. We evaluate the impact of spatial arrangements on mixing energy, hydrogen bonding networks, and vibrational properties with high accuracy and robustness. Our results reveal that spatial patterning strongly influences surface stability and interfacial structure. A thermodynamically favored unpaired configuration emerges near 67 % CH3 substitution, where isolated OH groups form secondary hydrogen bonds through reorientation toward subsurface oxygen atoms. This rearrangement induces a characteristic blue shift in OH stretching frequencies, indicating weaker H-bonding. These effects are absent in clustered arrangements. By establishing a clear link between functional group patterning and interfacial behavior, our work uncovers the underlying mechanisms to guide and accelerate the rational design of silica-based materials and coatings, directly relevant to colloid and interface science.