Accelerating process design for direct air capture: Bayesian optimization of temperature-vacuum swing adsorption with fiber sorbents

• A dual-bed TVSA system for DAC was refined using NbOFFIVE-1-Ni/CA fiber sorbents. • Bayesian Optimization was employed to optimize CO 2 capture capacity and OPEX. • Higher flow rates increase CO 2 capture but also elevate blower energy consumption. • Most desorption occurs within 10 minutes, limiting the need for prolonged desorption. • DAC performance curves guide optimal trade-offs between OPEX and capture amount. Solid sorbent-based direct air capture (DAC) is a key carbon dioxide removal strategy, yet its process optimization remains challenging due to the need to simultaneously address multiple, high-dimensional objectives. These include maximizing CO 2 capture capacity, minimizing operating costs, and reducing overall carbon emissions across the system's life cycle. In this study, we experimentally present a refined approach for optimizing a dual-bed temperature and vacuum swing adsorption (TVSA) cycle using a fiber sorbent based on the metal-organic framework NbOFFIVE-1-Ni, aiming to understand and reconcile the competing performance objectives. To achieve this, a structured exploration of the critical operational variables was conducted, encompassing adsorption flow rate, adsorption time, desorption temperature, and desorption time. Recognizing the complexity and interrelated nature of the performance metrics, we adopted Bayesian Optimization, a powerful data-driven method, to iteratively identify operating conditions that maximize CO 2 capture efficiency while minimizing operational expenditure (OPEX). Extensive cycle-level testing and performance assessment produced a DAC performance profile characterized by distinct Pareto fronts, which delineate the inherent trade-offs between energy consumption and capture efficiency. These insights enabled the determination of optimal operating conditions. Notably, the lab-scale dual bed system achieved a capture capacity of 21.76 mol CO 2 per year (0.52g-CO 2 /g-sorbent per day), supporting its feasibility for large-scale, cost-effective, and environmentally responsible DAC deployment.