Effects of Moisture Contents on Shale Gas Recovery and CO2 Sequestration

Enhanced recovery of shale gas with CO2 injection has attracted extensive attention as it combines the advantages of improved efficiency of shale gas recovery and reduced greenhouse gas emissions via CO2 geological sequestration. On the other hand, the microscopic mechanism of enhanced shale gas recovery with CO2 injection and the influence of the subsurface water confined in the shale nanopores remain ambiguous. Here, we use grand canonical Monte Carlo (GCMC) simulations to investigate the effect of moisture on the shale gas recovery and CO2 sequestration by calculating the adsorption of CH4 and CO2 in dry and moist kerogen slit pores. Simulation results indicate that water accumulates in the form of clusters in the middle of the kerogen slit pore. Formation of water clusters in kerogen slit pores reduces pore filling by methane molecules, resulting in a decrease in the methane sorption capacity. For the sorption of CH4/CO2 binary mixtures in kerogen slit pores, the CH4 sorption capacity decreases as the moisture content increases, whereas the effect of moisture on CO2 sorption capacity is related to its mole fraction in the CH4/CO2 binary mixture. Furthermore, we propose a reference route for shale gas recovery and find that the pressure drawdown and CO2 injection exhibit different mechanisms for gas recovery. Pressure drawdown mainly extracts the CH4 molecules distributed in the middle of kerogen slit pores, while CO2 injection recovers CH4 molecules from the adsorption layer. When the water content increases, the recovery ratio of the pressure drawdown declines, while that of CO2 injection increases, especially in the first stage of CO2 injection. The CO2 sequestration efficiency is higher under higher water content. These findings provide the theoretical foundation for optimization of the shale gas recovery process, as well as effective CO2 sequestration in depleted gas reservoirs.


INTRODUCTION
Shale gas, which predominantly consists of methane, has become an increasingly important energy source owing to its low emissions, high energy efficiency, and abundant reserves in the world. 1,2 In contrast to the conventional reservoirs, shale rocks have the characteristics of extremely low permeability in the nanodarcy range and low porosity. 3 The development of horizontal and hydraulic fracturing methods has greatly advanced shale gas exploitation in the United States. 4,5 Unfortunately, the widely used depressurization method becomes inefficient and significant amounts remain unrecoverable. 6 Recently, enhanced gas recovery (EGR) with CO 2 injection is considered to be a promising alternative that can not only enhance shale gas productivity but also mitigate the climate change via geological CO 2 sequestration, 7 in which CO 2 captured from power plants is injected into the shale gas reservoirs and sequestrated in a safe and permanent manner. The idea of swapping CO 2 for CH 4 can also be applied to other replacement reactions, such as the recovery of CH 4 from hydrate reservoirs. 8 Kerogen makes up the major organic part of the shale rocks 9 and pores, mainly at the nanoscale, where a large amount of shale gas can be stored. 10 Therefore, it is crucial to understand the adsorption properties of CH 4 and CO 2 in kerogen in an attempt to enhance the efficiency of shale gas recovery and CO 2 sequestration. It is also well known that shale is under moist conditions. 11−14 For example, kerogen from the Kuonamka Formation was reported to have moisture contents ranging from 0.6 to 5%. 11 Plenty of experimental work studied the effects of moisture on pure methane adsorption and suggested that the presence of water occupies the pore volume and thus leads to a sharp decrease in the methane adsorption capacity in shale. 14 −17 On the other hand, relatively few experiments focused on the influence of moisture on the competitive adsorption of CH 4 and CO 2 . Gensterblum et al. 18 investigated the effect of preadsorbed water on the CH 4 and CO 2 adsorption in coals and observed that the CO 2 /CH 4 adsorption ratios at low surface coverage are generally higher in moist coals than in dry conditions, and with the presence of water, the mobility of CO 2 and CH 4 is reduced. Few laboratory studies have been reported on the competitive adsorption of CO 2 and CH 4 in the moist kerogen. Isolation of the kerogen from shale samples with the morphology of the kerogen intact remains challenging for experiments. 9 Moreover, shale rocks contain a large amount of nanoscaled pores, further increasing the difficulties of experimental approaches in probing the adsorption behaviors in shales.
Molecular simulations have been successfully applied to interfacial and colloid science, 19 including adsorption, 20−22 wetting, 23,24 and surfactants. 25 Much research in recent years has made significant progress in understanding the competitive adsorption of CH 4 and CO 2 in kerogens using grand canonical Monte Carlo (GCMC) simulations. 26−30 Huang et al. 29 investigated the effect of moisture on the CO 2 /CH 4 competitive adsorption in kerogen matrix and found that the increased moisture content leads to a decrease in adsorption capacity, and the CO 2 /CH 4 adsorption selectivity first decreases and then increases. However, most of the work focused on the adsorption inside the kerogen matrix, in which the pores are isolated and not well connected, and the pore size is only a few angstroms. Recently, Wang et al. 31 studied the adsorption of CH 4 /CO 2 mixtures in moist kerogen using a 2 nm wide slit-pore model and stated that the effect of moisture content on adsorption selectivity is not obvious for shale kerogen slit nanopores. A more realistic kerogen pore structure is used in their study, but the water molecules are placed inside the kerogen matrix rather than within the slit pores. The interaction between water and the adsorbate gas, which is mainly located in the slit pores, is weakened due to the separation. Therefore, the moisture effects on the adsorption of CH 4 and CO 2 are still not fully understood. In addition, the dependence of adsorption selectivity on pressure and pore size is extensively used in the previous research. 29 −32 To some extent, it can imply the relative adsorption affinity of CH 4 and CO 2 to the adsorbent but fails to reflect the shale gas recovery process with CO 2 injection, namely, CO 2 huff-n-puff process, consisting of three stages: CO 2 injection, soaking, and production. 6 CO 2 is first injected into the production well (huff) and then the well is shut in and soaks for a period; finally comes the production (puff). 6 In this work, we aim to reveal the effects of moisture on the competitive adsorption of CH 4 /CO 2 binary mixtures in kerogen using GCMC simulations. Kerogen slit-pore models of different pore widths are constructed to represent the pore structure in shale. 33,34 First, we analyze the adsorption of pure methane in kerogen slit pores with and without the presence of moisture under a wide range of pressures from 10 to 60 MPa.
Then, we discuss the moisture effects on the adsorption of CH 4 /CO 2 binary mixtures in detail. Finally, we simulate the enhanced shale gas recovery by CO 2 huff-and-puff and compare the mechanisms of the pressure drawdown and CO 2 injection processes on shale gas recovery. The effects of moisture on the shale gas recovery and CO 2 sequestration are also revealed.

METHODS
2.1. Molecular Models. Kerogen can be classified into four types according to the depositional origin. 35,36 Type II kerogen, typically derived from marine sediments, is selected in this work for its abundance in shale deposits, as well as good potential for generating oil/gas. 37−39 The kerogen molecular unit (type II-C) used in the simulations was proposed by Ungerer et al. 40 on the basis of experimental data from the Duvernay shale formation. 41 Besides, type II-D kerogen has also been used to study the shale gas sorption behavior. 42 Molecular dynamics (MD) simulations are conducted in the canonical ensemble (NVT) and isobaric−isothermal ensemble (NPT) using the large-scale atomic/molecular massively parallel simulator (LAMMPS) package 43 to generate kerogen matrixes and slit pores. The Dreiding force field 44 is employed to describe the kerogen properties. First, the initial structure of the kerogen macromolecule is relaxed individually. Then, 12 relaxed kerogen units are randomly placed in a simulation box of 100 × 100 × 100 Å 3 . The final configuration of the kerogen matrix was created through a series of annealing procedures, as reported in Collell et al. 45 The density of the simulated kerogen matrix is 1.22 ± 0.02 g/cm 3 , which is within the range of the experimental value of mature shales (1.18−1.35 g/ cm 3 ). 46 The pore size distribution of the kerogen matrix is presented in Figure S1, calculated by the method of sphere insertion proposed by Bhattacharya and Gubbins. 47 Finally, we extend the simulation box in the Z-direction to create a kerogen slit nanopore, as shown in Figure 1. Note that we use the slit-shaped pore model as slit pores are very common in shale formations. 33,34 The pore width W is defined as the distance between the rightmost atom in the left slab and the leftmost atom in the right slab in the Z-direction. Two pore widths of 2 and 4 nm are constructed to investigate the pore size effects.
Methane molecules are simulated using TraPPE force field, and the united-atom model is applied. 48 The carbon dioxide molecule is treated as a rigid and linear structure with TraPPE-EH force field, where the C−O bond length and O−C−O bond angle are fixed as 1.16 Å and 180°, respectively. 49 The SPC/E model 50  (1)  where r ij is the distance between atoms i and j; ε ij and σ ij represent the LJ potential well depth and the zero-potential distance, respectively; q is the charge of atoms; and k e = 8.988 × 10 9 N·m 2 ·C −2 is the electrostatic constant. The force field parameters are listed in Table 1.
Lorentz−Bertherlot mixing rules 51 are adopted to calculate interactions between unlike atoms. A cutoff distance of 14 Å is employed for short-range Lennard-Jones interactions, and analytical tail corrections are applied. 52 As our system has a finite length along the Z-direction, the conventional three-dimensional Ewald summation technique is not valid for the calculation of the long-range electrostatic interactions. Here, we use the same approach as for the three-dimensional Ewald summation, but an empty space is inserted between periodic replicas to avoid the artificial influence from the periodic images in the Z-direction. 52,53 Tests are carried out to ensure the length of the empty space is long enough that the artificial effects could be eliminated. The entire simulation box including the empty space is shown in Figure S2. During the simulations, the molecules are only allowed to move within the kerogen slit pore, but not into the vacuum.

Simulation Details.
GCMC simulations are carried out in the grand canonical ensemble (μVT) to investigate the adsorption of CH 4 and CO 2 in kerogen slit pores. In the GCMC simulations, gases inside the kerogen slit pores are assumed to be in equilibrium with an external bulk reservoir under the same temperature and chemical potentials. The equilibration process is achieved by performing insertion, deletion, and translation moves for the gas molecules. For CO 2 and H 2 O molecules, rotational moves are also applied. The chemical potentials are obtained by Widom's insertion method 54,55 using Monte Carlo simulations in the NVT ensemble, where the fluids are simulated in bulk phase without confinement. The calculated chemical potential values are verified by μVT simulations compared to those of NIST database 56 ( Figure S3). The bulk densities of CO 2 and CH 4 mixtures at a given pressure and temperature are calculated from the Peng−Robinson equation of state (PR-EOS). 57,58 Bulk densities calculated by PR-EOS have been verified by comparing μVT simulations, as shown in Figure S4. MCCCS Towhee, a Monte Carlo molecular simulation code, is utilized in all of the GCMC simulations. 59 During the GCMC simulations, the kerogen slit is kept rigid, but the water molecules are allowed to move. The number of water molecules is kept constant, based on the assumption that the water molecules remain in the pore during the shale gas recovery process. Besides, the water content is reported as the volumetric mass density of water, ρ H 2 O ave , in the kerogen slit pore with the unit of g/cm 3 . The pressure in the simulations denotes the pressure of the external bulk reservoir, which is in equilibrium with the confined system.

RESULTS AND DISCUSSION
In this section, we first investigate the adsorption of pure CH 4 and CH 4 /CO 2 binary mixtures in kerogen slit pores, and the moisture effects are analyzed in detail. Then, we discuss the practical implications for the shale gas recovery and CO 2 sequestration during the CO 2 huff-and-puff process.
3.1. Adsorption of Pure CH 4 in Kerogen. 3.1.1. Dry Condition. We present methane density distributions in various nanopores under dry conditions at 338.15 K in Figure  2. For all of the pressures, two strong methane adsorption layers are formed near the kerogen surfaces, and the methane density within the adsorption layer increases with pressure. Under higher pressures (over 10 MPa), the CH 4 −CH 4 interaction becomes stronger and a weak second adsorption layer can be observed. In the middle of kerogen slit pores, the methane density is higher in the smaller pore (W = 2 nm) due to the strong fluid−surface interactions. As the pore width increases (W = 4 nm), the fluid−surface interaction becomes weaker. As a result, the methane density in the middle of the kerogen slit pore reaches bulk.  . For all of the pressure conditions ranging from 10 to 60 MPa, water molecules form clusters in the middle of the kerogen slit pore, unlike in clay nanopores. 60 This is because the kerogen is hydrophobic, while the clay is hydrophilic. Similar phenomena were also reported on graphene and montmorillonite. 61 The effects of water on the methane density distribution in 2 nm kerogen slit pores under different pressures are presented in Figure 4. In general, the CH 4 density in the middle of the kerogen slit pore decreases significantly as water molecules are predominantly distributed in the middle of the pore in the form of clusters. The methane density distribution in the  presence of water is not symmetric because the water cluster is not located exactly at the center of the slit pore. For methane within the adsorption layer, the density also slightly decreases. From the water density distribution (as shown in Figure S5), some water molecules are distributed near the kerogen surface, occupying the adsorption sites on the surface and resulting in the decrease of methane in the adsorption layer. The average density of methane stored in the kerogen slit pore can be given as where ⟨N⟩ is the ensemble averaged number of methane molecules in kerogen slit pores, V denotes the volume of the slit pore, and N A is the Avogadro constant. We only consider the methane molecules in slit pores, excluding those inside the kerogen matrix. Figure 5 shows the total uptake of CH 4 at 338.15 K in kerogen slit pores of different pore widths with and without moisture. Statistical uncertainties on the sorption isotherms are examined, and the errors bars are less than the symbols. For both pore widths, the amount of adsorbed methane decreases monotonically with increasing water content. As shown in Figure 3, the water molecules are distributed in the middle of the slit pore in the form of clusters and occupy a fraction of the pore volume. Thus, they impede the pore filling of methane molecules in the middle of the pore. By comparing the CH 4 total uptakes in different pores ( Figure  5), the reduction in the average density of methane is similar for 2 and 4 nm kerogen slit pores, when they have the same water volumetric density (ρ H 2 O ave 0.186 g/cm 3 ) rather than the same surface density. It demonstrates that the volumetric density of water may be more appropriate to define the water content in kerogen to quantify the effect of moisture on gas adsorption. CH 4 excess sorption isotherms in 2 and 4 nm dry kerogen slit pores are also calculated using effective pore volume from helium adsorption, 62 and the results are shown in Figure S6. It first reaches a maximum at around 15 MPa and then decreases with increasing pressure. Excess sorption in 2 nm kerogen slit pores is larger than that in 4 nm kerogen slit pores due to the stronger fluid−wall interaction.  Figure 6. The density distributions of both CO 2 and CH 4 have two peaks near the kerogen walls, but the density of CO 2 is much higher than that of CH 4 , as CO 2 has a stronger affinity to kerogen than CH 4 . 63 In the middle of 2 nm kerogen pores, both the densities of CO 2 and CH 4 are higher than the bulk density due to the strong fluid−surface interactions. When W = 4 nm, the fluid−surface interaction becomes weaker. Therefore, the densities in the middle of pores approach the    bulk density. The density distributions of pure methane in kerogen slit pores under the same pressure are also shown in Figure 6 for comparison. With the competitive adsorption of CO 2 , the density of methane decreases significantly, especially on the adsorption layer. Figure 7 displays the total uptake of CH 4 −CO 2 binary mixtures of different compositions in the dry kerogen slit pores of different pore widths at 338.15 K and over a wide range of pressures from 10 to 60 MPa. Similar trends can be observed for CH 4 in the kerogen slit pores of different pore widths that the average density increases quickly with pressure at the beginning and gradually reaches a plateau. The average density of CH 4 in kerogen slit pores increases as its mole fraction in the mixtures increases. For CO 2 molecules, sorption increases gradually with the increasing pressure. Similar to CH 4 , as the mole fraction of CO 2 in the binary mixtures increases, an increase in the corresponding CO 2 adsorption can be observed. In the equimolar mixtures, the CO 2 average density in kerogen slit pores is much higher than that of CH 4 .

Moist Condition.
Sorption of CH 4 /CO 2 binary mixtures in the moist kerogen slit pores is also investigated using GCMC simulations. As shown in Figure 8, water clusters are formed in the kerogen slit pores with the sorption of CH 4 / CO 2 binary mixtures under pressures ranging from 10 to 60 MPa as pure CH 4 sorption with moisture. When there is only   methane sorption in the moist kerogen slit pore, the pore surface is occupied by methane molecules. With the CO 2 cosorption, the adsorption sites on the kerogen surface are mainly covered by CO 2 molecules. Besides, the solubility of methane in water is much lower than that of CO 2 . 64 As a result, CH 4 sorption in the moist condition is significantly reduced due to the cosorption of CO 2 . Effects of moisture contents on the total uptake of CH 4 / CO 2 binary mixtures are illustrated in Figure 9. Similar to the trends in dry conditions, the average densities of both CH 4 and CO 2 in moist kerogen slit pores increase when their mole fraction increases. As the moisture content increases, the methane average density decreases significantly, independent of its mole fraction in the mixtures (Figure 9a). However, the effect of moisture content on CO 2 adsorption is associated with the mole fraction of CO 2 in the binary mixtures. For the case of low CO 2 bulk mole fraction in mixtures (y CO 2 = 0.25), a slight increase in CO 2 adsorption can be found in the moist condition, while the opposite is true at high CO 2 mole fraction (y CO 2 = 0.75). As stated in the previous section, the presence of moisture fills up the pore volume in kerogen slit pores, which results in the reduction in CH 4 sorption. With the CO 2 cosorption, the water−CH 4 binary interaction turns into the water−CH 4 −CO 2 ternary interaction. Since the water−CO 2 interaction is much stronger than the water−CH 4 interaction, the decrease in accessible pore volumes for CH 4 can partially be occupied by the more favorable CO 2 sorption. Besides, within the region near the kerogen surface, CO 2 exhibits a higher affinity to kerogen compared to CH 4 . Therefore, the CO 2 has a competitive advantage over CH 4 both in the middle of the kerogen slit pore and near the surface, which are shown in the snapshots (Figure 8) and density profiles ( Figure S7). When the bulk mole fraction of CH 4 is relatively high, the decrease in methane sorption caused by the presence of water is pronounced. When the mole fraction of CO 2 is higher, both CO 2 and CH 4 need to compete with water for adsorption sites or space. The occupation of water in the kerogen slit pores limits the sorption of CO 2 and CH 4 , thereby diminishing their sorption capacities. The sorption selectivity of CO 2 over CH 4 with different moisture contents is presented in Figure S8 to characterize the preferential adsorption. For the moisture contents considered in this work, the CO 2 /CH 4 adsorption selectivity increases with increasing moisture content.
3.3. Implications for Shale Gas Recovery and CO 2 Storage. As outlined in the Introduction, this work aims to unravel the CH 4 recovery mechanisms under moist condition during the pressure drawdown process and CO 2 injection to investigate the recovery and sequestration efficiency. The simplified recovery process consists of two pressure drawdowns and two CO 2 injections, as illustrated in Figure 10. The initial pressure of the target shale gas reservoir is assumed to be 30 MPa, which is within the typical pressure range of realistic reservoir conditions. The CH 4 sorption amount in the kerogen slit pores is obtained by GCMC simulations. Then, the shale gas recovery is initiated by the primary pressure drawdown process, and the reservoir pressure is reduced to 20 MPa, while the amount of CH 4 residing in the kerogen slit pores is calculated via μVT simulation. Subsequently, CO 2 is injected into the shale gas reservoirs. During this process, we assume that the pore volume in the fractures (external bulk reservoir) remains the same; therefore, the CH 4 density in the bulk phase of CH 4 /CO 2 mixtures is the same as that in pure CH 4 . The resulting CH 4 /CO 2 mixtures in the kerogen slit pores are determined by the chemical potentials of CH 4 /CO 2 mixtures, which are obtained by NVT simulations. After the system has reached equilibrium, a pressure depletion process is applied  again, while the bulk CH 4 /CO 2 composition stays the same. The dynamics and transport processes are not considered. 65 This process achieves one cycle of CO 2 huff-n-puff in enhanced gas recovery. 6,66 Consequently, a second CO 2 huffn-puff process is operated. Evolution in the composition of fluids in the kerogen slit pores during the shale gas recovery process is shown in Figure  11. During the pressure drawdown, the average CH 4 density in the kerogen slit pores decreases continuously, and with the injection of CO 2 , the CH 4 molecules are further released. The CO 2 molecules are sequestrated in the kerogen slit pores after injection. The effects of moisture content are also shown in Figure 11; as the moisture content increases, the average density of methane at the initial stage is reduced from 13.2 mmol/cm 3 in the dry condition to 7.68 mmol/cm 3 with an average water density of 0.372 g/cm 3 , but the recovery ratio of CH 4 after the whole recovery process is increased from 58.5 to 70.1%.
To compare the efficiency of the shale gas recovery and CO 2 sequestration during every intermediate process and quantify the effects of moisture contents, we introduce two parameters: CH 4 recovery ratio and CO 2 sequestration ratio, taking the initial pressure (30 MPa) of the reservoir as the reference. The CH 4 recovery ratio is defined as the number of CH 4 molecules released during a single process relative to the initial amount in the kerogen slit pore under 30 MPa with/without moist. The CO 2 sequestration ratio is defined as the number of CO 2 molecules sequestrated during the CO 2 injection process relative to its adsorption amount in the kerogen slit pore under 30 MPa, namely, the maximum amount of CO 2 sequestration

Langmuir
Article in a 30 MPa reservoir with/without moist. The CO 2 sequestration ratio can serve as a parameter for indicating the extent of CO 2 sequestration. The two parameters are plotted as a percentage in Figure 12 under different moisture contents. As can be seen from Figure 12a, as the moisture content increases, the CH 4 recovery ratio of the pressure drawdown processes decreases, while that of CO 2 injection processes increases, which means in the moist conditions, the performance of the CO 2 injection method can be even better, while the efficiency of the pressure drop method is lowered. In the dry conditions (0 g/cm 3 ), the first pressure drawdown process accounts for the largest proportion of the total amount of shale gas recovered, while in the moist conditions, the CH 4 recovery ratio of the first CO 2 injection process is the largest over the whole recovery process. To clarify the role of CO 2 injection in the recovery process, we calculate the recovery ratio of the process with the pressure drawdown only from 30 to 20 MPa. The recovery ratios between the two processes are compared ( Figure S9). With the combination of CO 2 injection, recovery ratios are enhanced by 2.4, 3.1, and 5.1 times for water densities of 0, 0.186, and 0.372 g/cm 3 , respectively. For the CO 2 sequestration during the CO 2 injection process (Figure 12b), the CO 2 sequestration ratio rises with increasing moisture content, especially in the first stage. A considerable CO 2 sequestration ratio of 63.4% in total can be achieved in the moist kerogen slit pores at an average water density of 0.372 g/cm 3 .
To explain the different effects of moisture contents on the pressure drop method and CO 2 injection method, we further investigate the CH 4 recovery mechanisms of the two methods. The evolution of the CH 4 density distributions during the gas recovery process with different moisture contents is plotted in Figure 13. After the pressure drawdown, the CH 4 density in the middle of the kerogen slit pore is reduced, while after the CO 2 injection, the decrease in CH 4 density mainly occurs in the adsorption layer near the kerogen surface, indicating different mechanisms in shale gas recovery. In other words, the pressure drawdown releases the free gas in the middle of pores, while the CO 2 injection recovers CH 4 from the adsorption layers. In the presence of water in kerogen slit pores ( Figure  13b,c), since the water molecules accumulate as clusters in the middle of kerogen slit pores, the moisture content negatively affects the performance of the pressure drawdown. Meanwhile, the adverse effect of moisture on the CO 2 injection method is negligible, as the water−CO 2 interaction is much stronger than the water−CH 4 interaction. Hence, the presence of water could further improve the efficiency of the CO 2 injection method. Generally, in the moist conditions, the CO 2 injection behaves better, while the efficiency of the pressure drawdown is lower.

CONCLUSIONS
In this work, the effects of moisture on the adsorption of pure CH 4 and CH 4 /CO 2 binary mixtures in kerogen slit pores are investigated using GCMC simulations. The shale gas recovery mechanisms of the pressure drawdown and CO 2 injection are studied, and the moisture effects on the shale gas recovery and CO 2 sequestration efficiency are further explored.
Our simulation results show that for the sorption of pure CH 4 in the moist conditions, water clusters are formed in the middle of a kerogen slit pore and the clusters occupy the volume in the slit pore, thereby impeding the pore filling of CH 4 . Regarding the sorption of CH 4 /CO 2 binary mixture, the CH 4 sorption capacity in the mixture decreases with increasing moisture content, while the effect of moisture on CO 2 adsorption capacity depends on the mole fraction of CO 2 in the mixture due to the ternary interactions among CH 4 , CO 2 , and H 2 O. During the shale gas recovery process, the pressure drawdown and CO 2 injection present different mechanisms. The pressure drawdown releases CH 4 molecules in the middle of kerogen slit pores, while CO 2 injection can release CH 4 molecules in the adsorption layer. As the water content increases, the recovery ratio of the pressure drawdown declines, while that of CO 2 injection increases, especially in the first stage of CO 2 injection. Finally, the CO 2 sequestration efficiency is higher under moist conditions. This work provides important insights into the effects of moisture content on gas adsorption in kerogen. As the subsurface water in the shale formations is saline, further work is planned to examine the effects of saline water on the recovery process. Besides, the recovery process introduced in this work can be extended to the optimization of the shale gas recovery process. For example, adjusting the CO 2 injection pressure and depletion pressure may lead to optimal shale gas recovery efficiency or CO 2 sequestration efficiency. The recovery process can also be applied to other displacement processes in confined systems.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.langmuir.9b00862.
Chemical potentials of pure methane and mixtures (Table S1); additional details of bulk fluids (Table S2); pore size distribution of the kerogen matrix ( Figure S1); entire simulation box ( Figure S2); densities of bulk CH 4 ( Figure S3); validation of the PR-EOS ( Figure S4); water density distributions ( Figure S5); CH 4 excess adsorption isotherms ( Figure S6); density distributions Figure 13. Evolution of the CH 4 density distributions inside the 2 nm kerogen slit pores during the gas recovery process with varying moisture contents: (a) 0 g/cm 3 , dry condition (b) 0.186 g/cm 3 , and (c) 0.372 g/cm 3 . The CO 2 density distribution during the CO 2 injection process is presented in Figure S10.