Rather than storing CO2 in saline aquifers, a competing possibility for sequestering carbon lies in seafloor hydrates, ice-like substances in which gases are trapped within cages of water molecules. One strategy is to store CO2 in solid form as a gas hydrate, or as a pool of liquid CO2 below a cap of its hydrate. CO2 hydrates can also be stored in subsea sediments that are characterized by finegrained silts, mud and clays, usually with very small mean pore diameters. Engineering any technology along these lines requires a fundamental understanding of both the thermodynamics and kinetics of CO2 hydrate formation in bulk as well as in confined geometries associated with the subsea sediment pores. To assess the long-term behavior of the carbon dioxide disposal and geological storage, it is also important to understand the phase behavior of CO2-H2O mixtures over a broad range of thermodynamic conditions as well as in confined geometries. The principles of statistical mechanics, along with molecular-based computer simulations, provide a fundamental link between molecular-level structure and interactions and macroscopic properties that can shed new light on these issues.

Pablo Debenedetti and colleagues are using molecular dynamics (MD), Monte Carlo (MC) and path sampling simulations to study CO2 hydrate phase behavior, the kinetics of hydrate formation and dissociation, and phase behavior of CO2-H2O and CO2-H2O-NaCl. These studies will provide insights important to geological storage of carbon dioxide as well as for the rational design of hydrate-based CO2 storage systems over broad ranges of temperature and pressure.

 

Molecular simulation studies of CO2 hydrate dissociation

To better understand the kinetics of CO2 hydrate dissociation, Debenedetti and colleagues have used a molecular dynamics model to simulate CO2 hydrate in contact with liquid water. A snapshot of the simulation system is shown in Figure 18 . CO2 hydrate has so-called sI hydrate structure, the unit cell of which consists of six large (51262) and two small (512) cages (Xm Yn denotes a cage with m X-sided and n Y-sided polygons). The cage occupancy corresponds to the number of cages filled by guest molecules. Although some studies have suggested that hydrate stability is affected by cage occupancy, no systematic studies have been performed of this important parameter.

Figure 18. Snapshot of CO2 hydrate and water system used in simulations. CO2 molecules are shown in green. Hydrogen bonds between water molecules in the hydrate structure are shown as red dashed lines. Liquid water is shown in blue and water in hydrate structure is shown in red (oxygen) and white (hydrogen).

To investigate this issue, the researchers studied the dissociation behavior of three occupancy cases – 100% occupancy, in which all cages are occupied by CO2 molecules, and two cases of 87% occupancy. The two cases with 87% overall occupancy differ in the distribution of their empty cages — in one case, 50% of the small cages were empty (denoted as 87%S) and in the second case, ~17% of the large cages chosen randomly were kept empty (denoted as 87%L).

The studies reveal two novel aspects of CO2 hydrate dissociation. First, the rate of hydrate dissociation depends not only on the overall occupancy, but also on the cage-specific occupancy. With overall occupancy of 87%, the hydrate structure with empty large cages (87%L) dissociates approximately twice as fast as the structure with empty small cages (87%S) (See Figure 19). Therefore, it is important to consider not only the overall occupancy, but also cage-specific occupancy when modeling the kinetics of hydrate dissociation.

Second, analysis of the dissociation behavior of individual cages reveals that the melting behavior of the cages is dependent on their surrounding hydrate structure. For example, large cages occupied with CO2 molecules have different melting behavior in 87%L and 87%S, melting faster in case of former (See Figure 20). Thus, dissociation of individual cages cannot be explained based on simple distinction between small and large or empty and occupied cages. The surrounding hydrate structure of the cages has to be taken into account. This also suggests that hydrate dissociation is a cooperative phenomenon.

Figure 19. Evolution of energy, E in the hydrate‐water system during the simulation relative to final bulk system energy, <E>e. Increase in the ratio represents melting of the hydrate. Different colors correspond to different occupancy cases. Blue: 100%, Red: 87%S, Magenta: 87%L
Figure 20. Dissociation trajectories of filled large cages (51262 ) in 87%S and 87%L occupancy hydrates at 270 K. Each color represents the dissociation trajectory of an individual cage upon melting. Note differing timescales.

 

Monte Carlo simulations of high-pressure phase equilibria of CO2-H2O mixtures

Another focus of the Debenedetti group has been to assess the ability of state-of-the-art molecular-based computer simulations to predict the phase behavior of CO2-H2O and CO2-H2O-NaCl mixtures over broad ranges of temperature and pressure relevant to carbon capture and storage, as well as CO2-assisted geothermal energy production. The researchers tested five CO2/H2O models (EPM2/TIP4P2005, EPM2/TIP4P, EPM2/SPC, TraPPE/TIP4P2005, and exponential-6 CO2/exponential-6 H2O), focusing on the solubility of CO2 in water, the solubility of water in CO2, and phase diagram of CO2-H2O mixtures at high temperatures. Figure 21 summarizes the comparison between experimental data and predictions from various models studied here.

Reasonable agreement between the experimental data and simulation results for CO2 solubility in water was obtained at low pressures (P < 200 bar) for most of the tested models, but overall the simulation results from most of the models deviate from the experimental curves at higher pressures. Also, the solubility of water in CO2 is not reproduced by most of the models. At high temperatures near the critical point, the experimental data are inconsistent with each other and the simulation results for all the tested models compare poorly with the experimental data.

The analysis suggests that while all of the tested models qualitatively capture the temperature and pressure dependence of solubility, none is capable of reproducing the experimental data over the entire temperature and pressure range. This work is the first comprehensive assessment to date of the state of the art in the molecular modeling of phase behavior in the CO2-H2O system across broad ranges of temperature and pressure.

Figure 21. Comparison of models with experimental data. Left panel: CO2 solubilities in H2O at low temperatures (50° C ≤ T ≤ 250° C). Middle panel: H2O solubilities in CO2 at low temperatures (50° C ≤ T ≤ 250° C). Right panel: Mutual solubilities of CO2/H2O at high temperatures (250° C < T ≤ 350° C). Symbol key: (-∎- ) Takenouchi and Kennedy’s experimental data (Am. J. Sci., 262, 1055, 1964); (-♦-) Todheide and Franck’s experimental data (Berich Bunsen Gesell, 67, 836, 1963); (-●-) Wiebe and Gaddy’s experimental data (J. Amer. Chem. Soc., 61, 315, 1939); (-★- ) Wiebe’s experimental data (Chem. Rev., 29, 475, 1941); ( △ ) exponential‐6 CO2/ exponential‐6 H2O model; (＋) EPM2/TIP4P2005 model; ( ✕ ) TraPPE+TIP4P2005 model; ( O ) EPM2/SPC model; ( ▽ )EPM2/TIP4P model.