A Two-Criterion Framework for Reaction-Set Reduction in Geochemical Simulation of CO2 Mineral Trapping in Basaltic Formations: Integrating Mineral Trapping Efficiency and pH Buffer Capacity

Simulating CO2 mineral trapping in basaltic formations requires geochemical reaction networks, but including all candidate reactions is often computationally prohibitive, and no established criterion exists for selecting a reduced yet representative set. We propose a two-stage reduction framework, motivated by the insight that trapping efficiency and pH buffer capacity are orthogonal criteria: a set that maximizes one is not guaranteed to satisfy the other. Stage A combines K-means clustering, Bayesian optimization, and a genetic algorithm to select reactions maximizing a trapping-efficiency score (J-score, integrating reaction extent, saturation state, and ionic contribution). Stage B independently verifies whether the selected set preserves pH buffer capacity (B-score), using CMG-GEM flow-coupled simulations validated against laboratory experiments. This two-stage check proved necessary: a set selected by J-score alone failed to converge in flow simulation due to excessive mineral precipitation, confirming that trapping efficiency alone is insufficient and identifying Anorthite as the dominant pH-buffering phase. A minimal two-reaction set (Anorthite and Dolomite) achieved a 94% reduction in computational cost under the present single-block column model, while maintaining pH fidelity against benchmark data; extrapolation of this efficiency gain to field-scale, multi-cell simulation remains a direction for future work. This framework reframes reaction-set selection from single-objective optimization to constrained design. The results are demonstrated for a single basaltic rock sample (Basalt Sample A); the approach is potentially transferable to reducing reaction-set uncertainty in other CO2-basalt mineralization systems, subject to lithology-specific recalibration.