XAFS-CTイメージングと機械学習によるゴム―真鍮接着界面における銅硫化反応の三次元可視化―

抄録

Adhesion between rubber and brass-plated steel cords is essential for the durability and reliability of automobile tires. The interface is formed through chemical reactions between sulfur-containing rubber and brass, and copper sulfide species generated at the buried interface play a key role in adhesion and aging behavior. In particular, Cu2S and related non-stoichiometric copper sulfides are associated with favorable adhesion, whereas further chemical changes under thermal and humid conditions can lead to interfacial degradation. However, practical rubber–metal adhesion interfaces are highly heterogeneous and embedded inside opaque rubber composites, making their three-dimensional chemical evolution difficult to observe directly. This article introduces an advanced approach using X-ray absorption fine structure computed tomography（XAFS-CT） combined with machine-learning analysis. XAFS-CT integrates the chemical-state sensitivity of XAFS with the three-dimensional imaging capability of CT, enabling non-destructive visualization of metallic Cu in brass, Cu2S, and CuS in the rubber matrix. A rubber–brass model composite containing numerous dispersed brass particles was designed to provide many spatially isolated adhesion interfaces within a single field of view. This sample design allowed high-throughput analysis of local interfacial reactions while satisfying the X-ray absorption requirements for XAFS-CT. Repeated measurements of the same sample during humid thermal aging further enabled the chemical evolution of individual brass particles to be tracked over aging time. The results showed that metallic Cu was consumed and Cu2S was mainly formed during the early stage of aging. With further aging, CuS formation and Cu2S consumption became more pronounced. Importantly, these reactions did not proceed uniformly; each brass particle exhibited a distinct reaction history depending on its local environment. Machine-learning analysis classified these temporal changes into characteristic reaction types and clarified the statistical distribution of aging pathways. This approach provides a new framework for understanding adhesion aging as a statistical ensemble of heterogeneous local reactions rather than as a single averaged process.