Abstract

Carbon capture and sequestration (CCS) offers an effective approach to mitigate global climate change by storing CO₂ in subsurface formations. However, uncertainties in flow dynamics, geochemical reaction rates, and monitoring hinder reliable estimation of long-term CO₂ storage. This dissertation develops a suite of experimental and modeling studies to quantify CO₂ trapping in subsurface environments. Chapter 1 presents a Bayesian-optimized machine learning model trained on high-fidelity two-phase flow simulations under vertical-equilibrium assumptions. Latin-hypercube sampling provides a broad parameter space for rapid Monte Carlo-based uncertainty quantification (UQ) and Sobol-index sensitivity analysis of CO₂ trapping mechanisms, including mineral trapping, in the Johansen deep saline aquifer case. Chapters 2–4 explore mineral trapping and geochemical behavior in natural stilbite. Batch reactor experiments quantify the extent of mineral carbonation by stilbite at 60 °C. Mixed flow-through reactor tests at 60 °C and 120 °C measure far-from-equilibrium dissolution rates, revealing a U-shaped pH dependence for Si release and nearly pH-independent Ca release that accelerates by an order of magnitude at higher temperatures. These results confirm rapid ion exchange and carbonate formation across acidic to alkaline fluids. Chapter 5 validates a co-injection tracer method in a coiled-tubing reactor packed with crushed basalt. Paired non-reactive tracer and CO₂ slugs yield breakthrough curves whose separation directly quantifies CO₂ retention, achieving 90–95% trapping under laboratory conditions. Taken together, these standalone studies—machine-learning surrogates, dissolution kinetics, carbonation experiments, and tracer-based monitoring—provide robust, scalable tools for predicting and verifying long-term CO₂ mineral sequestration.