UC Berkeley

Increasing atmospheric CO₂ levels are of concern due to potential links

to negative environmental impacts such as increasing global temperatures and

ocean acidification. Strategies for reducing CO₂ emissions involve both

reducing the use of fossil fuels a primary energy sources and employing

processes to capture CO₂ from gas streams incidental to energy

generation and store in deep geologic features. The one of the main targets

for CO₂ separation are post-combustion gas streams at electricity

generating plants, which represent a large fraction of the CO₂ emitted.

While current process technology (amine scrubbing) could be scaled to the

accomplish the task, it is a relatively inefficient process and would

substantially reduce the efficiency of electricity generation.

Adsorption-based processes have the potential to reduce the parasitic load on

generating plants by reducing the amount and quality of heat diverted from

generating cycle.

Adsorption processes involve the use of solid porous materials with large

internal surface areas to separate components of a gas mixture. One or more

components will preferentially adsorb and be enriched in the adsorbed mixture,

which can then be desorbed in a separate part of the process. While a variety

of adsorbents have wide application in industry, there is a need to identify

the most efficient materials for this process to ensure its economic

viability. Some materials of interest are zeolites and metal-organic

frameworks, and recent experimental and theoretical work have identified

thousands of possible new materials. To evaluate this large range of

materials, molecular simulation techniques useful for quickly generating

thermodynamic data and understanding the molecular-level mechanisms

responsible for selectivity.

This work details efforts addressing several aspects using Monte Carlo

simulations to evaluate different materials for CO₂ separations. One

aspect is the proper description of how mixtures adsorb in materials with

heterogeneous surfaces where components can competitively adsorb at spatially

distinct sites. By applying ideal adsorbed solution theory (IAST) to separate

Langmuir sites, it is possible to improve predictions of mixture adsorption

isotherms compared to applying IAST to the whole isotherm. One critical

element of applying IAST accurately is ensuring the saturation loadings of

different components in a mixture are estimated as accurately as possible.

Another part addresses how to apply simulation techniques to millions of

related structures simultaneously and evaluate them with a simple model of

generating plant performance. Using GPU-accelerated Monte Carlo simulations, a

database containing thousands of hypothetical zeolite structures was screened

for CO₂/N₂ separations, and many structures were identified that

potentially would have a lower energy penalty to operate than the standard

amine scrubbing process.

Next, a method for fitting parameters for a classical force field from ab

initio calculations was developed and used to predict the adsorption of

CO₂ in Mg-MOF-74, a promising MOF material for CO₂ separations.

Using a modified Buckingham potential with an additional r^-5 attractive

term was able to describe the enhanced interaction between CO₂ molecules

and the coordinative-unsaturated Mg atoms. Finally, the adsorption of water in

zeolite 13X was studied, showing the strong effect it has on the co-adsorption

of CO₂. Rearrangement of sodium cations in the zeolite pores was

important for predicting the correct isotherms, and at the highest water

loadings, some sodium cations are removed from the pore walls become

coordinated by water closer to the center of the pore. This rearrangement

may explain the the steep elbow of the isotherm.