Sorbent Performance Estimation
Adsorption equilibrium information is a fundamental resource to evaluate
the performance of sorbent candidates for application at process
scale.36 We estimated CO2 and
N2 single-component adsorption isotherms by Grand
Canonical Monte Carlo (GCMC) simulations37,38 in
UiO-66 and MIL-101(Cr). This technique has been extensively used to
estimate the adsorption properties of CO2 and similar
species in a range of porous sorbent materials.39-44GCMC simulations were carried out at a range of operating conditions for
which we performed design calculations, i.e., sub-ambient temperatures
ranging from 213 to 273 K. The details of molecular modeling are
provided in the Supporting Information S4.1.1. We then employed Ideal
Adsorbed Solution Theory (IAST)45 to predict the bulk
0.14:0.86 CO2:N2 mixture adsorption
equilibrium. IAST predicts the mixture equilibrium from single-component
adsorption isotherms under the assumption that an ideal solution is
formed by the adsorbed phase.46-48 H2O
was not considered in mixture equilibrium assuming only a trace amount
remains in a feed into the separation unit after pretreatment of the
flue gas. The mixture isotherms were used to estimate sorbent
performance via the full-order model outlined below for fiber sorbents
with heat management. For zeolite 13X, we applied the extended dual-site
Langmuir (DSL) model reported by Haghpanah et al.19(Table S10) for CO2:N2 mixture
equilibrium evaluation.
A fixed-bed packed with thermally-modulated fiber
composites31,32 was used for modeling of a pressure
swing adsorption (PSA) cycle. A complete account of details describing
this model can be found in Rubiera Landa et al49 and
the Supplemental Information S4.1.2. The cycle employed for the VPSA
process consists of four steps: 1. light product pressurization, 2.
adsorption, 3. co-current blowdown, and 4. counter current evacuation.
To compute Pareto frontiers efficiently the VPSA-cycle simulation is
treated as a black-box and the multi-objective optimization algorithm
‘Surrogate Optimization of Computationally Expensive Multiobjective
Problems’ (SOCEMO), developed by Müller50 was
implemented in Matlab.51 This optimizer allows
computation of Pareto frontiers at a fraction of the computational cost
required for other commonly applied genetic
algorithms.52 The Pareto frontier values (purity,
recovery, system operating conditions) were integrated into the Aspen
Plus process model to simulate performance of different adsorbent and
cycle design choices.19