Author

Kuijun Li

Date of Graduation

2016

Document Type

Dissertation

Degree Type

PhD

College

Statler College of Engineering and Mineral Resources

Department

Mechanical and Aerospace Engineering

Committee Chair

David S Mebane

Committee Co-Chair

Brian J Anderson

Committee Member

Fernando V Lima

Committee Member

Xingbo Liu

Committee Member

David C Miller

Abstract

The mechanism of CO2 uptake in silica-supported, polyethylenimine (PEI)-impregnated sorbents under dry and humid conditions has been investigated in a new chemical model based on quantum chemical calculations and a comparison of reaction-diffusion models with experimental data. First-principles quantum chemical calculations showed that the zwitterion, which is not stable in the gas phase even when considering a polar environment similar to water, can be stabilized by nearby water and amine molecules. Stabilized zwitterions serve as diffusive intermediates to transport adsorbed CO2 into the PEI bulk. Compared to amine-stabilized zwitterions, water-stabilized zwitterions are more numerous and have a lower activation energy for deprononation. The formation of bicarbonate is not favored because of the high energy of formation and activation energy.;A reaction-diffusion model based on this new chemistry has been developed and implemented using a finite volume method. The model treats transport only in the bulk phase of PEI, as this is much slower than gas-phase diffusion. The simulation results were compared with a set of experimental data from thermogravimetric (TGA) analysis. This model reproduced the experimental adsorption behaviors both qualitatively and quantitatively under both dry and humid conditions.;Bayesian calibration using quantum chemistry calculated priors for physical parameters was employed to connect the model with experimental data. Uncertainties from both the model parameters and model form discrepancy from the adsorption model were quantified in the form of a joint sample-based distribution obtained through Markov chain Monte Carlo (MCMC) sampling. Dynamic discrepancy functions in the Bayesian Smoothing Splines (BSS-ANOVA) framework were included in the equilibrium constants and diffusion coefficients. Results of calibration showed that the proposed approach is capable of replicating the experimental behavior, both qualitatively and quantitatively: The TGA data was well covered by the calibration results. The calibration results were then incorporated within a process-scale model (of a bubbling fluidized bed adsorber) by propagating the posterior distribution.

Share

COinS