Molecular simulation study on multicomponent competitive adsorption of CH4, CO2, and H2O in coal Authors: Qun Cheng, Yan Liu, Cen Chen, Can Yang, Shiyun Wei, Ruiying Zhu, Yang Chen, Songze Li Published: 2026 ABSTRACT AbstractCoal exhibits heterogeneous pore networks and chemically diverse surfaces, resulting in complex competitive adsorption among CH4, CO2, and H2O. The underlying molecular mechanisms remain unclear. In this work, molecular simulation methods were applied to investigate the adsorption behavior and interaction characteristics of CH4/CO2/H2O mixtures on two typical coal components (inertinite and vitrinite) under different CO2 enrichment levels, corresponding to gas-phase CO2 mole fractions of 4.8%, 9.1%, and 16.7%. The results demonstrate that CH4 dominates surface occupation in all cases, maintaining 30–70 adsorbed molecules, whereas CO2 adsorption is significantly weaker, remaining below 5 at low loading and increasing to only 10–17 at high loading. This indicates a limited competitive capability of CO2 for adsorption sites. From an interaction perspective, water governs the electrostatic environment, with surface–water Coulombic energies consistently distributed around − 600 to − 750 kJ mol−1. In contrast, CO2–water interactions decrease from − 220 to − 360 kJ mol−1 to − 100 to − 170 kJ mol−1 as CO2 loading increases, reflecting a pronounced screening effect. Meanwhile, direct CO2–surface interactions remain weak (typically − 5 to − 15 kJ mol−1). Overall, CH4 adsorption is primarily controlled by dispersion interactions, while CO2 is constrained by weak surface affinity and reduced hydration strength, resulting in a secondary role in multicomponent competitive adsorption within coal systems.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-50774-x.Keywords: Coal, Competitive adsorption mechanism, CH4/CO2/H2O mixtures, Molecular dynamics simulation, Interfacial interactionsSubject terms: Chemistry, Energy science and technology, Environmental sciences Introduction In recent years, coalbed methane (CBM) has gained growing attention in the context of the global energy transition, especially as the demand for cleaner and more efficient energy sources continues to increase. Unlike conventional gas reservoirs, coal formations present a highly intricate internal structure, composed of multiscale pores and chemically diverse surfaces, which together govern gas storage and migration in a nontrivial manner1–4. Under in-situ conditions, methane rarely exists in isolation; instead, it co-occurs with carbon dioxide and formation water, creating a dynamically interacting multicomponent system. This interplay becomes especially relevant in the context of CO2 injection for enhanced methane recovery and subsurface carbon management, where the relative adsorption preferences of different species dictate both displacement efficiency and storage performance. Although CO2 is generally more strongly attracted to coal matrices than CH4, the presence of water introduces additional complexity by occupying adsorption sites, restructuring interfacial environments, and mediating fluid–solid interactions5–8. Consequently, adsorption behaviors observed in simplified single-component systems often fail to capture the actual processes occurring in coal reservoirs. A detailed understanding at the molecular scale of how CH4, CO2, and H2O compete and coexist within confined coal pores is therefore crucial for improving predictive capability and guiding practical engineering applications9–11. Over the last decade, significant advances have been achieved in elucidating gas adsorption behavior in coal, supported by both experimental studies and molecular simulation approaches. Early efforts largely centered on single-component systems to clarify the intrinsic affinity of coal toward different gases12,13. For instance, Liu et al. utilized grand canonical Monte Carlo (GCMC) simulations to examine CH4 and CO2 adsorption on coal structures containing different functional groups, revealing that CO2 shows a higher adsorption affinity, which is attributed to its larger quadrupole moment and stronger electrostatic interactions with polar surface sites14. Similarly, Zhang et al. incorporated more realistic coal structures and reported that pore size distribution and surface chemistry jointly control adsorption performance, emphasizing the importance of structural heterogeneity. These studies laid the foundation for understanding gas–solid interactions at the molecular scale15. With the increasing interest in CO2-enhanced coalbed methane recovery (CO2-ECBM), attention gradually shifted toward competitive adsorption in binary systems. Hao et al. investigated CH4/CO2 competitive adsorption and found that increasing CO2 concentration significantly suppresses CH4 uptake while enhancing total adsorption, accompanied by coal matrix swelling16. In a related study, Liu et al. combined GCMC and molecular dynamics (MD) simulations and showed that CO2 preferentially occupies low-energy adsorption sites and exhibits stronger binding but lower mobility compared to CH4, which provides a molecular explanation for its displacement advantage17. Collectively, these works have clarified the fundamental mechanisms governing CH4–CO2 competitive adsorption. More recently, the influence of water has attracted growing attention, as moisture is an unavoidable component in coal reservoirs. Zhou et al. examined the ternary CO2/CH4/H2O system and reported that adsorption strength follows the order H2O ≫ CO2 > CH4, with water molecules significantly occupying pore space and reducing gas mobility11. In addition, Lin et al. demonstrated that increasing moisture content leads to a complex, non-linear variation in gas adsorption capacity, while CO2 consistently maintains a competitive advantage over CH418. Other studies have further suggested that water clustering and hydrogen-bond networks can modify local adsorption environments and even influence CO2 affinity under certain conditions19. In parallel with molecular simulation studies, a substantial body of experimental work has also contributed to understanding multicomponent adsorption behavior in coal. For instance, Bustin et al. showed that CO2 exhibits a stronger adsorption affinity toward coal than CH4, and its adsorption can induce significant coal matrix swelling, which in turn affects reservoir permeability20. Likewise, Day et al. experimentally quantified volumetric changes in coal during gas injection and confirmed that CO2 causes more pronounced matrix expansion compared to CH4, highlighting its strong interaction with the coal structure21. To more closely represent reservoir conditions, several studies have extended their focus to competitive adsorption in multicomponent systems. For instance, Ottiger et al. investigated the co-adsorption behavior of CO2, CH4, and N2 on coal and found that CO2 is preferentially adsorbed compared to CH4, highlighting its dominant role in competitive adsorption processes22. Similarly, Du et al. reported that CO2 exhibits stronger and more favorable adsorption than CH4 across different coal ranks, with adsorption selectivity increasing with coal maturity and being governed by pore structure and thermodynamic parameters23. These experimental observations are broadly consistent with simulation results, reinforcing the dominant role of CO2 in competitive adsorption systems. More recently, increasing attention has been directed toward the role of water in modifying adsorption behavior. Levy et al. showed that methane adsorption capacity in coal increases with coal rank but decreases with rising temperature and moisture content, indicating that adsorption behavior is strongly controlled by coal properties and reservoir conditions24. In addition, Azam et al. reported that water confined within coal pores markedly suppresses the adsorption capacities of both CO2 and CH4 and modifies their sorption kinetics, emphasizing the critical role of water–gas interactions under realistic reservoir conditions25. Experimental observations from Mastalerz et al. further indicated that different coal lithotypes exhibit significant variations in pore structure and adsorption capacity, with vitrain showing higher surface area and gas uptake than fusain, highlighting the strong control of pore characteristics and coal composition on adsorption behavior26. Although significant progress has been made in both experimental and simulation studies, current understanding of multicomponent adsorption in coal remains incomplete. Many simulations rely on simplified or homogeneous coal models, without explicitly accounting for the structural and chemical differences between maceral components such as vitrinite and inertinite. Experimental studies, while providing valuable macroscopic insights, are inherently limited in resolving molecular-scale interactions and competitive behaviors within confined nanopores. Furthermore, existing research has primarily examined single- and binary-gas systems, leaving the synergistic behavior of CH4–CO2–H2O mixtures largely unresolved. Notably, the role of CO2 loading in regulating competitive adsorption under complex multicomponent environments remains unclear27–30. To overcome these gaps, molecular simulations were conducted to examine the competitive adsorption behavior of CH4, CO2, and H2O on representative vitrinite and inertinite surfaces. By systematically tuning the CO2 concentration, the study elucidates how compositional variation governs interfacial adsorption and interaction mechanisms in multicomponent coal systems. This approach enables a more detailed and realistic understanding of multicomponent interactions in coal systems. Simulation models and methods Simulation models In this study, molecular models of inertinite (INE) and vitrinite (VIT) were constructed to investigate the competitive adsorption behavior of multicomponent species on different coal maceral surfaces (Fig. 1a,b). The molecular structures of inertinite and vitrinite used in this study are constructed as representative coal macromolecular models to capture the essential chemical and structural characteristics of different maceral types. The inertinite model consists of 231 carbon atoms, 229 hydrogen atoms, 23 oxygen atoms, and 1 nitrogen atom, with a total of 484 atoms. The structure is dominated by condensed aromatic ring systems, forming a relatively rigid and highly carbonized framework. The aromatic carbon proportion is high, indicating an advanced degree of coalification. Only a limited number of aliphatic side chains are present, mainly as short alkyl linkages connecting aromatic clusters. The oxygen-containing functional groups are relatively scarce and are primarily distributed as hydroxyl, carbonyl, and ether groups. This results in a comparatively low polarity and a more hydrophobic surface character for inertinite. In contrast, the vitrinite model contains 242 carbon atoms, 258 hydrogen atoms, 35 oxygen atoms, and 2 nitrogen atoms, totaling 537 atoms. Compared with inertinite, vitrinite exhibits a higher proportion of heteroatoms, particularly oxygen, indicating a greater abundance of oxygen-containing functional groups. These groups include hydroxyl, carbonyl, and ether functionalities, which are more widely distributed along both aromatic edges and aliphatic chains. The aromatic structures in vitrinite are less condensed, and the model contains a relatively higher fraction of aliphatic chains and bridge structures, reflecting a lower degree of carbonization. As a result, vitrinite shows stronger polarity and more heterogeneous surface chemistry. Water molecules were represented using a four-site model, illustrated by grey lines (Fig. 1c), to accurately describe their geometric and electrostatic characteristics. Carbon dioxide molecules were modeled using a ball-and-stick representation (Fig. 1d), while methane was described by a united-atom model to simplify its molecular structure without compromising computational efficiency (Fig. 1e). In these structural models, carbon in the coal matrix, hydrogen, oxygen, and nitrogen atoms are depicted in green, white, red, and blue, respectively, whereas the carbon atoms in methane molecules are highlighted in yellow for clear distinction. In this work, inertinite (INE) and vitrinite (VIT) were employed as representative coal molecular models. Each simulation system contained a single coal macromolecule and 400 CH4 molecules. To represent different CO2 enrichment levels under reservoir-relevant conditions, 20, 40, and 80 CO2 molecules were additionally introduced, corresponding to CO2 mole fractions of approximately 4.8%, 9.1%, and 16.7% in the CH4/CO2 gas phase, respectively (Fig. 1f–k). These values represent low, intermediate, and relatively high CO2 exposure scenarios that may occur during CO2 injection and enhanced coalbed methane recovery processes. The remaining space was filled with water molecules according to the density of liquid water under ambient temperature and pressure. All simulations were performed in a cubic box with dimensions of 5 nm × 5 nm × 5 nm. The number of CH4, CO2, and H2O molecules in each system was defined based on this fixed volume, allowing the composition to be interpreted in terms of number density and molar fraction. Specifically, the CO2 loading was systematically varied (20–80 molecules), while the amounts of CH4 and H2O were kept constant to ensure comparability between systems. This setup enables reproducible control of multicomponent composition within a confined nanoscale environment. Molecular structures of (a) inertinite, (b) vitrinite, (c) H2O, (d) CO2, and (e) CH4. Initial structural models of the inertinite system with (f) 20, (g) 40, and (h) 80 CO2 molecules, and of the vitrinite system with (i) 20, (j) 40, and (k) 80 CO2 molecules. Simulation methods To avoid finite-size artifacts and better mimic an infinite system, periodic boundary conditions were applied in all three spatial dimensions (x, y, and z). The intermolecular interactions within the coal macromolecules were described using the OPLS-AA force field31, while methane molecules were modeled with the united-atom variant (OPLS-UA) to improve computational efficiency32. The EPM2 model was adopted to describe CO233, which has been widely applied for capturing its thermodynamic behavior. Water was modeled using the TIP4P-2005 potential34, and molecular geometry was constrained via the SETTLE algorithm to preserve rigidity during the simulations. Van der Waals interactions were described using the Lennard–Jones potential, with parameters combined according to the Lorentz–Berthelot rules. Short-range nonbonded interactions were truncated at 1.20 nm, whereas electrostatic interactions beyond this range were computed using the particle mesh Ewald (PME) approach. A time increment of 2 fs was employed for all simulations. Prior to data collection, the system was first relaxed by energy minimization using the steepest descent method to remove close-contact artifacts35–40. It was then equilibrated for 1 ns under NPT conditions (323 K and 100 bar). The simulation temperature and pressure were set to 323 K and 100 bar, respectively, to represent typical thermodynamic conditions of intermediate-depth coal reservoirs. The selected temperature corresponds to approximately 323.15 K, which falls within the commonly reported range of coal seam temperatures (313.15–333.15 K), while the pressure represents a moderate subsurface condition. It should be noted that, although reservoir conditions may vary, the present study focuses on the underlying molecular interaction mechanisms, which are expectedMP to be qualitatively applicable over a reasonable range of temperature and pressure. Temperature and pressure were maintained using the V-rescale thermostat and Berendsen barostat, with relaxation times of 0.1 ps and 2 ps, respectively. During this stage, pressure coupling was treated anisotropically, allowing independent fluctuations along each dimension of the simulation cell. Subsequently, production simulations were extended to 20 ns under identical thermodynamic conditions. A Nosé–Hoover thermostat in combination with a Parrinello–Rahman barostat was employed to achieve more accurate regulation of temperature and pressure, with corresponding coupling times of 2 ps and 4 ps. In contrast to equilibration, pressure was applied isotropically during the production phase. In this study, adsorption is defined based on a geometric and local environment criterion. A CH4 molecule is considered to be in a bulk-like (non-adsorbed) state if the number of surrounding water oxygen atoms within 5.5 Å is not less than 15, while the number of nearby gas molecules within 4.5 Å does not exceed 2. Otherwise, the molecule is considered to be located near the coal surface and classified as adsorbed. All simulations were carried out using the GROMACS 5.0.7 simulation package. Results and discussion The evolution of the simulated structure over time In the inertinite (INE) system, distinct evolution behaviors were observed under different CO2 concentrations. At a relatively low CO2 loading (20 molecules), CH4 molecules rapidly aggregated and formed a well-defined spherical bubble within approximately 5 ns (Fig. 2a). The INE macromolecule preferentially adsorbed at the gas–liquid interface, with its main framework closely conforming to the surface of the CH4 bubble. It is worth noting that a quantitative criterion was adopted to distinguish dissolved and aggregated CH4 molecules based on their local coordination environments. Specifically, a CH4 molecule is considered to be in the dissolved state when at least 15 water oxygen atoms are located within 5.5 Å of the carbon atom, and no more than 2 gas molecules are found within 4.5 Å. Otherwise, the molecule is classified as participating in a gas-rich cluster. Based on this definition, the observed structures correspond to nanoscale CH4 aggregation rather than macroscopic bubbles. A fraction of CO2 molecules was also found at the bubble interface, while the remaining molecules were dispersed in the aqueous phase. As the simulation progressed, CO2 molecules gradually migrated toward the CH4-rich region, and by around 15 ns, nearly all CO2 molecules had been captured by the bubble, leading to a relatively stable configuration. The structural features at 20 ns showed minimal deviation from those observed at 15 ns, indicating that the system had reached a quasi-equilibrium state. When the CO2 concentration increased to 40 molecules (Fig. 2b), the morphology of the CH4 aggregation changed significantly. Instead of forming a compact spherical bubble, CH4 molecules assembled into a more elongated, cylindrical-like structure. This configuration exhibited a larger spatial extent but a relatively lower local density. Despite the higher number of CO2 molecules, the expanded CH4 domain enabled more efficient capture of CO2 at an early stage, with a substantial portion of CO2 molecules being associated with the CH4-rich region as early as 5 ns. In contrast to the low-CO2 case, the INE structure no longer tightly adhered to the bubble surface but adopted a more extended and less compact conformation. Even after 20 ns of simulation, a small fraction of CO2 molecules remained distributed in the bulk water phase, indicating incomplete localization. At a higher CO2 loading (80 molecules) (Fig. 2c), the interaction between CO2, CH4, and the INE surface became more pronounced. A larger number of CO2 molecules were rapidly incorporated into the CH4-rich region within the initial 5 ns. Notably, CO2 molecules exhibited a stronger tendency to accumulate near the INE surface rather than being uniformly distributed around the bubble. This preferential localization became increasingly evident as the simulation proceeded, suggesting that higher CO2 concentrations enhance interfacial interactions and promote the coupling between gas aggregation and surface adsorption processes41–45. Time evolution of simulation configurations over 20 ns for the INE system containing (a) 20, (b) 40, and (c) 80 CO2 molecules, and for the VIT system containing (d) 20, (e) 40, and (f) 80 CO2 molecules. For the vitrinite (VIT) system, the overall evolution of CH4 aggregation and the adsorption behavior of CO2 exhibit trends broadly consistent with those observed in the INE system. At low CO2 concentration, CH4 molecules preferentially assemble into a compact spherical bubble, with the VIT macromolecule closely adhering to the gas–liquid interface in a well-fitted configuration. As the simulation proceeds, CO2 molecules gradually migrate toward the CH4-rich region and become increasingly localized at the bubble interface, eventually being incorporated into the aggregated phase. These observations suggest that the formation of CH4-rich domains plays a crucial role in facilitating CO2 accumulation and adsorption. Moreover, increasing CO2 concentration has a pronounced impact on the morphology of the CH4 aggregates. At intermediate and high CO2 loadings, the CH4 bubble undergoes a structural transition from a spherical to a more elongated, cylindrical-like configuration, indicating a concentration-dependent restructuring of the gas phase. In addition, the adsorption of CO2 on the coal surface is found to be primarily governed by the carbonaceous framework of the VIT macromolecule. The distribution and arrangement of carbon atoms provide the dominant interaction sites, thereby influencing the spatial localization of CO2 molecules and their affinity toward the coal surface (Fig. 2d–f). The time-dependent evolution of two-dimensional density distributions of CO2 and CH4 was examined in both INE and VIT systems to reveal their spatial organization and interaction characteristics. In the INE system containing 20 CO2 molecules, the spatial distribution of CO2 exhibits a dispersed pattern throughout the simulation. During the initial stage (0–3 ns) (Fig. 3a), CO2 molecules appear as isolated spots with no evident clustering behavior. As the system evolves into the intermediate stage (3–10 ns) (Fig. 3b), this discrete distribution becomes less pronounced, accompanied by a gradual spreading tendency. In the later stage (10–20 ns) (Fig. 3c), CO2 molecules are further dispersed, indicating a continuous redistribution within the aqueous phase rather than aggregation. In contrast, CH4 molecules display a markedly different behavior. Even within the early stage (0–3 ns) (Fig. 3d), CH4 already exhibits a clear aggregation tendency, forming elongated, band-like structures with a density gradient decreasing from the central region toward the periphery. As time progresses, this aggregation becomes increasingly pronounced, with a more compact and dense core developing at the center, reflecting the progressive coalescence of CH4 molecules into a stable, high-density domain (Fig. 3e,f). It must be noted that the apparent discrepancy between the spherical CH4 aggregates observed in three-dimensional configurations and the band-like features in the two-dimensional density maps can be attributed to projection and averaging effects. Specifically, the density maps represent spatially averaged distributions along a given direction, which can transform three-dimensional cluster geometries into elongated high-density regions. In addition, interfacial interactions with coal surfaces and periodic boundary conditions further distort the ideal spherical morphology, leading to anisotropic distributions in the projected density profiles. With increasing CO2 concentration, a more pronounced dispersed distribution is observed when the number of CO2 molecules reaches 40. In the early stage (Fig. S1a), the point-like distribution of CO2 becomes more evident compared to the case with 20 molecules, and such features remain observable even at later simulation times. However, despite the increased concentration, no clear aggregation of CO2 is detected, indicating that the molecules remain predominantly dispersed within the system. The spatial distribution of CH4 under this condition follows a pattern similar to that described previously, with aggregation behavior persisting and showing no fundamental change in its overall morphology (Fig. S1b,c). At the highest CO2 loading (80 molecules), a distinctly different evolution pattern is observed. During the initial stage (0–3 ns) (Fig. S2a), CO2 molecules begin to assemble into a loosely defined spherical distribution, characterized by a relatively low density in the central region and a higher density toward the periphery. As the simulation progresses to 3–10 ns (Fig. S2b), the peripheral regions gradually become more compact, indicating that dispersed CO2 molecules in the surrounding phase are continuously migrating toward and integrating into the growing cluster. In the later stage (10–20 ns) (Fig. S2c), a more stable spherical structure is established, maintaining the characteristic of a dense outer shell and a less populated core. The location of this spherical domain is inferred to coincide with the position of the INE macromolecule, suggesting a strong coupling between CO2 aggregation and the coal surface46–51. In contrast, CH4 also exhibits a spherical aggregation pattern under this condition, but with an opposite density distribution. At 0–3 ns, CH4 remains relatively loosely distributed; however, with increasing simulation time, the density within the central region gradually increases, forming a more compact core, while the density at the outer region becomes comparatively lower (Fig. S2d–f). Two-dimensional density contour maps of CO2 molecules in the INE system with 20 CO2 molecules during (a) 0–3 ns, (b) 3–10 ns, and (c) 10–20 ns, respectively. And two-dimensional density contour maps of CH4 in the INE system with 20 CO2 molecules during (d) 0–3 ns, (e) 3–10 ns, and (f) 10–20 ns, respectively. In the VIT system with 20 CO2 molecules, the early-stage distribution (0–3 ns) exhibits a trend similar to that observed in the INE system (Fig. 4a), where CO2 molecules are dispersed in a point-like manner without noticeable aggregation. As the simulation proceeds, this discrete pattern gradually diminishes, indicating that CO2 molecules tend to diffuse throughout the system rather than forming clusters. By the later stage (10–20 ns), CO2 molecules are widely distributed across the simulation domain, and distinct localized features become difficult to identify (Fig. 4b,c). In contrast, CH4 displays a different evolution behavior. At the initial stage (0–3 ns, Fig. 4d), CH4 molecules form a roughly spherical distribution, characterized by a higher density in the central region and a gradual decrease toward the periphery. With increasing simulation time, this spherical configuration progressively transforms into a more elongated, band-like structure, while maintaining a similar density gradient with a dense core and less populated edges (Fig. 4e). During the later stage (10–20 ns), CH4 molecules further concentrate toward the central region of the elongated structure, leading to an expansion of the surrounding low-density regions (Fig. 4f). When the number of CO2 molecules increases to 40, the point-like distribution becomes more pronounced, although it is still insufficient to induce noticeable aggregation. As the simulation proceeds, this discrete distribution gradually weakens; however, compared to the system containing 20 CO2 molecules, the spatial heterogeneity remains more evident throughout the simulation (Fig. S3a–c). In contrast, CH4 molecules exhibit a distinct band-like distribution as early as 0–3 ns, although the structure is initially relatively loose (Fig. S3d). During the intermediate stage (3–10 ns), CH4 molecules progressively accumulate toward the central region, eventually forming a more compact elongated structure characterized by a high-density core and a lower-density periphery (Fig. S3e–f). At a higher CO2 loading of 80 molecules, the distribution behavior changes markedly. In the early stage (0–3 ns), CO2 molecules organize into a ring-like structure, displaying a characteristic density profile with a depleted center and an enriched outer region (Fig. S4a). As time evolves to 3–10 ns, this ring-like arrangement becomes distorted, transforming into an irregular and spatially heterogeneous cluster (Fig. S4b). In the later stage (10–20 ns), the structure partially recovers into a quasi-ring configuration, although the density distribution remains uneven (Fig. S4c). Meanwhile, CH4 aggregation also undergoes significant morphological changes under this condition. The initially formed spherical cluster deforms into an ellipsoidal structure during 3–10 ns, closely resembling the spatial distribution of CO2, suggesting a high degree of overlap between the two species (Fig. S4d–e). This ellipsoidal configuration persists into the later stage (10–20 ns), indicating a stable co-localization of CH4 and CO2 within the system (Fig. S4f). Two-dimensional density contour maps of CO2 molecules in the VIT system with 20 CO2 molecules during (a) 0–3 ns, (b) 3–10 ns, and (c) 10–20 ns, respectively. And two-dimensional density contour maps of CH4 in the VIT system with 20 CO2 molecules during (d) 0–3 ns, (e) 3–10 ns, and (f) 10–20 ns, respectively. Interaction and diffusion properties Radial distribution functions (RDFs) were employed to characterize the local structural organization and interaction tendencies between the coal matrices (INE and VIT) and CO2, CH4, and H2O molecules. Across the three CO2 loadings, the RDF profiles collectively illustrate a gradual yet distinct transition in competitive adsorption behavior on the INE surface, involving not only changes in interaction strength but also redistribution of species at the molecular scale. At low CO2 concentration (20 molecules) (Fig. 5a), CO2 exhibits limited interaction with the coal matrix, with the first discernible peaks appearing at ~ 3.2–3.4 Å and relatively low intensities (g(r) ≈ 0.95–1.0). A slight preference toward carboxyl groups over the carbon backbone is observed, indicating that polar sites initially dominate CO2 adsorption. In contrast, CH4 shows a broader and more stable correlation with the aromatic carbon structure, maintaining g(r) > 1.0 from ~ 3.3 Å up to beyond 5.0 Å, reflecting a typical hydrophobic association. Its interaction with carboxyl groups indicates a reduced local density compared to the reference state (g(r) < 1.0 in the first shell). Meanwhile, water strongly accumulates around polar functional groups, with a sharp rise starting at ~ 2.7 Å and reaching ~ 0.8 at longer distances, forming a relatively stable hydration shell that partially hinders CO2 access to these sites. When the CO2 content increases to 40 molecules (Fig. 5b), both the intensity and spatial extent of correlations are significantly enhanced. CO2 begins to approach the surface at shorter distances (~ 2.7–3.0 Å), and its interaction with carboxyl groups strengthens (g(r) ≈ 1.1 at ~ 3.25 Å), while its association with the carbon backbone becomes more pronounced, reaching g(r) ≈ 1.4–1.6 around ~ 3.5–4.0 Å. The persistence of elevated g(r) values (> 1.3) over a wider range suggests the development of a CO2-enriched interfacial region. Under these conditions, CH4 still preferentially interacts with the carbon backbone, maintaining g(r) ≈ 1.0–1.3 between ~ 3.5 and 5.0 Å, but its spatial distribution becomes less concentrated due to partial competition from CO2. Water continues to favor carboxyl groups, with similar peak positions (~ 2.7–3.0 Å) and magnitudes (~ 0.8), although the slightly enhanced CO2–carboxyl interaction indicates the onset of competitive disruption of the hydration layer. At high CO2 loading (80 molecules) (Fig. 5c), the adsorption environment undergoes a more substantial reorganization. CO2 exhibits strong and sustained correlations with both the carbon backbone and functional groups. The CO2–carboxyl peak remains located at ~ 3.25–3.45 Å with g(r) ≈ 1.0–1.2, while the CO2–carbon backbone interaction becomes dominant, with a broader and higher peak (~ 3.75–4.05 Å, g(r) ≈ 1.5–1.7) and extended correlations persisting up to ~ 6–7 Å. This indicates not only stronger adsorption but also the formation of a dense, quasi-layered CO2 distribution near the coal surface, likely driven by combined surface affinity and intermolecular packing effects. In this regime, CH4 interactions are clearly suppressed: its correlation with the carbon framework weakens (g(r) ≈ 0.8–1.0), and its already weak association with carboxyl groups remains below unity, suggesting effective displacement from favorable adsorption regions. Water still preferentially associates with polar groups, but its RDF profile becomes smoother and less structured, with g(r) stabilizing around ~ 0.7–0.85, implying that the hydration network is partially disrupted by CO2 occupation. This indicates that the evolution of CO₂–water interaction with increasing concentration is not uniform, but results from the interplay between electrostatic screening and enhanced short-range contacts. Radial distribution functions (RDFs) among different components in the INE system with (a) 20, (b) 40, and (c) 80 CO2 molecules and in the VIT system with (d) 20, (e) 40, and (f) 80 CO2 molecules. Overall, the system evolves from a water-dominated and weakly competitive regime at low CO2 concentration to a strongly competitive and CO2-dominated adsorption environment at high loading. CO2 progressively strengthens its interaction with both polar and nonpolar sites, expands its spatial influence near the surface, and ultimately suppresses CH4 adsorption while perturbing the interfacial water structure. In the VIT system at low CO2 loading (20 molecules) (Fig. 5d), the interfacial structure is already highly heterogeneous, with CO2 showing strong and selective affinity for polar sites. No direct contacts are observed below ~ 2.5 Å, but a sharp CO2–carboxyl peak emerges at ~ 2.65 Å with g(r) ≈ 1.80, indicating localized and strong binding to oxygen-containing groups. This feature is more pronounced than in the INE system, suggesting more accessible or reactive polar sites in VIT. In contrast, CO2–carbon backbone interactions develop over a broader range (~ 3.6–4.4 Å) with g(r) ≈ 1.5–1.8 and remain above ~ 1.3 up to ~ 7–8 Å, reflecting a combination of dispersion-driven adsorption and the formation of an extended CO2-enriched interfacial region. CH4 primarily associates with the carbon backbone, with RDF values of ~ 0.7–1.3 between ~ 3.5 and 5.0 Å, indicating stable but relatively delocalized hydrophobic interactions. Unlike INE, CH4 also shows partial accessibility to polar sites, with a noticeable CH4–carboxyl peak (~ 1.5 at ~ 4.15 Å), suggesting weaker steric constraints or greater surface heterogeneity. Water preferentially occupies carboxyl groups, with RDF values increasing from ~ 2.3 Å to ~ 0.6–0.75 beyond ~ 4 Å, consistently exceeding water–carbon correlations (< 0.55). However, its overall intensity is still lower than that of CO2–carboxyl interactions, implying that CO2 can effectively compete for polar adsorption sites even at low loading. At intermediate CO2 loading (40 molecules) (Fig. 5e), both the strength and spatial extent of CO2 adsorption increase significantly. CO2 begins to interact with the surface at slightly shorter distances (~ 2.6–2.8 Å) and exhibits multiple strong peaks associated with polar sites, reaching g(r) ≈ 1.8 at ~ 3.25 Å and further increasing to ~ 2.0–2.3 in the 4.0–4.5 Å region. Meanwhile, CO2–carbon backbone correlations also intensify, with a broad peak around ~ 3.8–4.1 Å (g(r) ≈ 1.4–1.6) and sustained values above ~ 1.3 up to ~ 7–8 Å. CH4 retains its preference for the carbon backbone, but its RDF becomes weaker and more fluctuating (typically ~ 0.7–1.3 over 3.5–5.5 Å), indicating partial displacement by CO2. Occasional CH4–carboxyl peaks (~ 1.2–1.3) suggest transient access to polar sites, but overall affinity remains limited. Water continues to favor carboxyl groups, with RDF values reaching ~ 0.7–0.85, yet the buildup becomes less sharp at short range, implying that CO2 increasingly perturbs the local hydrogen-bond network and competes for these sites. At high CO2 loading (80 molecules) (Fig. 5f), the system transitions to a dense and cooperative adsorption regime. CO2 exhibits strong and persistent correlations with both surface types, with CO2–carbon backbone RDF values reaching ~ 1.8–2.0 at ~ 3.8–4.6 Å and remaining above ~ 1.5 up to ~ 7–8 Å, indicative of multilayer-like accumulation or clustering. CO2–carboxyl interactions remain comparably strong (g(r) ≈ 1.5–1.9. Under these conditions, CH4 adsorption is clearly suppressed: the CH4–carbon backbone RDF decreases in intensity and becomes broader (~ 1.0–1.6), while CH4–carboxyl correlations remain weak and highly fluctuating (< ~ 1.2), confirming effective competitive exclusion. Water still preferentially associates with carboxyl groups, maintaining RDF values of ~ 0.7–0.85 at longer distances, but the more gradual increase indicates partial disruption of the hydration structure. Overall, increasing CO2 loading drives a clear evolution from localized, site-specific adsorption to a dense, cooperative interfacial structure. CO2 progressively dominates both polar and nonpolar regions, displaces CH4 from favorable adsorption domains, and weakens the integrity of the interfacial water network, reflecting a coupled effect of competitive adsorption and crowding-induced restructuring. It should be noted that, although coordination number analysis based on RDF integration can provide additional quantitative insight, the present study focuses on the relative changes in RDF peak position and intensity, which are sufficient to characterize the variation of intermolecular interactions and local structure in this dynamic multicomponent system. To quantify the mobility of gas molecules under varying CO2 loadings, the mean square displacement (MSD) of CO2 and CH4 was systematically evaluated in both INE and VIT systems. This analysis provides direct insight into how competitive adsorption and interfacial interactions influence molecular diffusion behavior. The MSD profiles in the INE system exhibit a clear time-dependent growth for both CO2 and CH4, with an overall quasi-linear regime after 2 ns, indicating normal diffusive behavior. Notably, CH4 consistently shows higher MSD values than CO2 at all loadings, suggesting intrinsically faster mobility. For instance, at 10 ns, the MSD of CH4 reaches 724, 716, and 843 for 20, 40, and 80 CO2 cases, respectively, whereas CO2 only attains 506, 501, and 496. This corresponds to an approximate 1.4–1.7 times enhancement in CH4 mobility relative to CO2, reflecting weaker interactions or lower steric constraints for CH4 in the INE matrix. Increasing CO2 loading induces a non-negligible impact on both species. For CO2, the MSD at 20 ns increases from 898 (20 CO2) to 937 (40 CO2) and further to 1107 (80 CO2), indicating that higher CO2 concentration promotes its own diffusion, likely due to collective effects and reduced confinement. This corresponds to an increase of 23% from low to high loading. A similar trend is observed for CH4, where the MSD rises from 1417 (20 CO2) to 1291 (40 CO2) and up to 1685 (80 CO2). Although the 40 CO2 case shows a slight suppression compared to 20 CO2 (9% decrease), the 80 CO2 system significantly enhances CH4 mobility (19% increase relative to 20 CO2), suggesting a non-monotonic dependence governed by competitive occupancy. From a dynamic perspective, the slope of MSD curves (proportional to diffusion coefficient) further confirms this trend. Taking the linear region (10–20 ns), the effective diffusion rate of CO2 increases progressively with loading, whereas CH4 exhibits a minimum at intermediate CO2 concentration and a maximum at high loading. This implies that moderate CO2 content may hinder CH4 transport due to competitive adsorption or transient blocking, while excessive CO2 likely alters the microenvironment (e.g., enlarging free volume or disrupting local ordering), thereby facilitating CH4 migration (Fig. 6a). Mean square displacement (MSD) of CO2 and CH4 at different CO2 concentrations in the (a) INE/(b) VIT system. The MSD evolution in the VIT system shows a well-defined transition to a linear regime after 2 ns, indicating stable Fickian diffusion for both gases. Similar to the INE case, CH4 consistently exhibits higher MSD values than CO2 across all loadings, confirming its superior mobility. Quantitatively, at 10 ns, CH4 reaches 743, 752, and 736 for 20, 40, and 80 CO2 systems, respectively, while CO2 remains at 445, 556, and 481. This corresponds to a 1.3–1.7 times higher mobility of CH4. The gap slightly narrows at higher CO2 loading, implying that CO2 mobility becomes more competitive under crowded conditions. CO2 diffusion in the VIT system shows a clear dependence on loading. At 20 ns, the MSD increases from 790 (20 CO2) to 1364 (40 CO2), and further to 972 (80 CO2). Compared to the lowest loading, the 40 CO2 case exhibits a substantial enhancement (73%), while the 80 CO2 system shows a moderate increase (23%). This non-monotonic behavior suggests that intermediate loading provides optimal conditions for CO2 transport, possibly due to balanced intermolecular spacing and minimized clustering, whereas excessive CO2 may introduce local crowding that partially offsets mobility gains. For CH4, the trend is more complex. At 20 ns, the MSD rises from 1423 (20 CO2) to 1416 (40 CO2) and then decreases to 1362 (80 CO2). Although the variation is not drastic, a gradual suppression (4% from low to high loading) is evident. This indicates that increasing CO2 concentration imposes a mild hindrance on CH4 diffusion, likely due to competitive occupation of free volume and transient blocking effects. Unlike the INE system, no enhancement of CH4 mobility is observed at high CO2 loading, highlighting a key difference in microenvironment response between the two systems. From the slope perspective (10–20 ns), CO2 diffusion is most pronounced at intermediate loading, whereas CH4 diffusion slightly declines with increasing CO2 content. This implies that CO2–CO2 interactions in VIT may facilitate cooperative transport up to a threshold, beyond which steric effects dominate. Meanwhile, CH4 transport appears more sensitive to spatial restriction and lacks the cooperative enhancement observed in INE (Fig. 6b). To elucidate the intermolecular interaction characteristics in the INE and VIT systems, the Lennard–Jones (LJ) potential was employed to quantify the non-bonded interactions among different components. Across CO2 loadings of 20, 40, and 80 molecules, the Lennard–Jones interactions in the INE system exhibit a consistent hierarchy with clear equilibration behavior. In all cases, the CH4–water interaction is the dominant contribution, rapidly decreasing in magnitude during the first ~ 2 ns and stabilizing around − 900 to − 1000 kJ mol−1, indicating a fast reorganization of the water network and a strong preference for methane solvation in the aqueous phase. The INE–CH4 and INE–water interactions constitute the second tier, both generally distributed within − 300 to − 500 kJ mol−1 after equilibration. Their comparable magnitudes suggest that INE simultaneously interacts with methane and water, acting as an interfacial regulator (Fig. 7a). With increasing CO2 loading (Fig. 7b,c), these interactions show slightly enhanced fluctuations, implying more dynamic competition at the interface. In contrast, CO2-related interactions are consistently weaker but show a clear dependence on concentration. The INE–CO2 interaction remains minimal (typically within − 50 to − 130 kJ·mol−1), confirming the limited affinity of CO2 for the INE phase. Meanwhile, CO2–water interactions increase from roughly − 100 kJ mol−1 at low loading to about − 400 to − 500 kJ mol−1 at high loading, indicating progressively enhanced solvation of CO2 in water. The CO2–CH4 interaction remains moderate (generally − 150 to − 400 kJ mol−1) and does not exhibit a strong systematic trend, suggesting limited mixing between the two gas species. From a temporal perspective, all interaction energies reach a quasi-equilibrium state within 2–3 ns, after which only moderate thermal fluctuations are observed. Overall, the interaction strength follows the order: CH4–water ≫ INE–CH4 ≈ INE–water > CO2–water > CO2–CH4 > INE–CO2. Increasing CO2 loading does not change this hierarchy but enhances CO2–water interactions and interfacial fluctuations, highlighting the growing role of CO2 in modulating the local structure while remaining less strongly bound than methane. In the INE system, the Lennard–Jones (LJ) interactions among different components at CO2 loadings of (a) 20, (b) 40, and (c) 80. In the VIT system, the Lennard–Jones (LJ) interactions among different components at CO2 loadings of (d) 20, (e) 40, and (f) 80. The LJ interaction analysis across different CO2 loadings (20, 40, and 80 molecules) in the VIT system reveals a consistent energetic hierarchy, accompanied by a rapid initial relaxation followed by long-term dynamic stability. In all cases, interaction energies remain negative throughout the simulation, confirming that dispersion interactions dominate the system behavior, while their magnitudes define the preferential association pathways. A common feature across all systems is the overwhelming dominance of CH4–water interactions. At low CO2 loading (20 CO2) (Fig. 7d), this interaction starts from an extremely strong value (− 3000 kJ mol−1) and rapidly relaxes within ~ 1 ns to a stable range of − 900 to − 1100 kJ·mol−1. A similar trend is observed at higher CO2 contents (40 and 80 CO2), where the equilibrated values remain in a comparable window (− 850 to − 1100 kJ·mol−1), indicating that methane–water interactions are robust and largely insensitive to CO2 concentration. This behavior reflects fast restructuring of the aqueous network and persistent solvation of methane, establishing CH4–water as the primary stabilizing interaction in the system. The second tier is formed by VIT–fluid interactions, particularly VIT–CH4 and VIT–water. Quantitatively, VIT–CH4 interactions remain relatively stable across all cases, typically in the range of − 300 to − 450 kJ·mol−1 (average − 350 kJ·mol−1), suggesting a sustained affinity that favors methane accumulation near the VIT interface. VIT–water interactions show comparable magnitudes (− 250 to − 460 kJ·mol−1) but consistently exhibit larger fluctuations, implying a more dynamic hydration layer where water molecules continuously exchange at the interface. This competitive interaction indicates that methane has a slight but consistent advantage over water in accessing VIT surface sites. CO2-related interactions occupy the lower tier and show clear concentration dependence. At 20 CO2, CO2–water interactions are relatively weak (− 130 to − 230 kJ·mol−1), but increase significantly at higher loadings, reaching − 250 to − 420 kJ·mol−1 (40 CO2) and further strengthening to − 550 to − 700 kJ·mol−1 (80 CO2). This trend indicates enhanced CO2 solvation with increasing concentration. In contrast, CO2–CH4 interactions remain moderate and relatively stable across all systems, generally within − 60 to − 120 kJ·mol−1 (20 CO2) and increasing to − 150 to − 400 kJ·mol−1 at higher loadings, suggesting persistent but non-dominant mixing between gas species without clear phase separation. Notably, VIT–CO2 interactions are consistently the weakest across all conditions. Even with increasing CO2 concentration, they remain limited (approximately − 5 to − 60 kJ·mol−1 at 20 CO2, − 10 to − 100 kJ·mol−1 at 40 CO2, and − 50 to − 130 kJ·mol−1 at 80 CO2), indicating minimal direct affinity between CO2 and the VIT surface. This quantitatively confirms that CO2 is largely excluded from interfacial adsorption and preferentially resides in the bulk aqueous phase. Overall, a unified interaction hierarchy can be established as: CH4–water ≫ VIT–CH4 ≈ VIT–water > CO2–water ≥ CO2–CH4 ≫ VIT–CO2. This pattern highlights that system stability is governed by water-mediated interactions, with methane strongly stabilized by both water and the VIT interface, whereas CO2 primarily participates in bulk-phase solvation with weak interfacial involvement (Fig. 7e,f). Coulombic interactions provide key insight into the role of electrostatics in INE and VIT systems, particularly for polar species and interfacial regions. Analyzing these interactions helps clarify how charge effects influence component distribution and competitive behavior beyond dispersion forces. Across all CO2 loadings (20, 40, and 80 molecules) (Fig. 8a–c), the Coulombic interactions in the INE system exhibit a consistent hierarchy dominated by water-mediated electrostatics. The INE–water interaction is always the largest contribution, remaining strongly negative in the range of − 700 to − 1020 kJ·mol−1, with typical averages around − 800 to − 850 kJ·mol−1. This persistent magnitude indicates that the interfacial hydration structure governs the overall electrostatic stability, with only moderate fluctuations after rapid equilibration (1 ns), suggesting a dynamically stable hydrogen-bond network. As CO2 concentration increases, the most notable change occurs in the CO2–water interaction. At low loading (20 CO2), it remains relatively weak (− 80 to − 180 kJ·mol−1), but strengthens to − 180 to − 300 kJ·mol−1 at 40 CO2 and further to − 300 to − 750 kJ·mol−1 at 80 CO2, with an average approaching − 430 kJ·mol−1. This progressive enhancement indicates that CO2 increasingly participates in the water network, likely through more frequent contacts and collective polarization effects, reflecting a transition toward denser hydration structures at high loading. The INE–CO2 interaction, although negligible at low concentration (within ± 10 kJ·mol−1), becomes slightly more attractive at higher loading (typically − 5 to − 20 kJ·mol−1 at 80 CO2), suggesting a weak but non-negligible tendency for CO2 to associate with polar sites on the INE surface. In contrast, all CH4-related Coulombic terms remain effectively zero across all cases, confirming that methane does not contribute to electrostatic interactions and remains governed by nonpolar interactions. Overall, increasing CO2 loading does not alter the dominant role of water but significantly amplifies CO2–water electrostatic coupling and slightly enhances INE–CO2 affinity. The system therefore evolves from a water-dominated electrostatic environment with minor CO2 involvement to a more strongly coupled polar network where CO2 actively participates, while CH4 remains electrostatically inert. In the INE system, the Coulomb interactions among different components at CO2 loadings of (a) 20, (b) 40, and (c) 80. In the VIT system, the Coulomb interactions among different components at CO2 loadings of (d) 20, (e) 40, and (f) 80. The Coulombic interactions in the VIT systems consistently show a clear hierarchy dominated by water, with limited contribution from CO2 and negligible involvement of CH4. For all CO2 loadings (Fig. 8d–f), the VIT–water interaction remains the strongest term, generally ranging from − 500 to − 780 kJ·mol−1 and stabilizing around − 600 to − 700 kJ·mol−1, indicating a persistent and stable hydration shell at the interface. In contrast, VIT–CO2 interactions are weak (typically within − 5 to − 10 kJ·mol−1, occasionally reaching − 25 kJ·mol−1), reflecting only transient and non-persistent electrostatic association. The CO2–water interaction represents the secondary contribution but shows a clear dependence on loading. At lower concentrations, it is relatively strong (down to − 300 kJ·mol−1), while at higher loadings (e.g., 80 CO2), it weakens significantly (typically − 60 to − 195 kJ·mol−1, averaging − 125 kJ·mol−1), suggesting that intermolecular screening and local crowding reduce effective CO2 hydration. Despite this attenuation, CO2 does not establish strong direct electrostatic coupling with the VIT surface. All CH4-related Coulombic interactions remain essentially zero throughout, confirming its nonpolar nature and negligible electrostatic role. Dynamically, the systems reach equilibrium rapidly (within 1–2 ns), after which all interactions fluctuate within narrow ranges without systematic drift. Overall, the electrostatic behavior can be summarized as VIT–water ≫ CO2–water ≫ VIT–CO2, with increasing CO2 concentration primarily weakening CO2–water coupling rather than enhancing interfacial electrostatic interactions. It should be noted that the dominant role of water in the electrostatic environment is mainly reflected by its strong Coulombic interactions with the coal surface. Although hydrogen bond analysis could provide additional structural details of the water network, the present conclusions are primarily supported by the energetic contributions and interaction characteristics observed in the simulations. Competitive adsorption properties The dynamic evolution of CO2 and CH4 adsorption on mineral surfaces is examined across different CO2 loadings. This comparison highlights how adsorption behavior varies between INE and VIT interfaces as the CO2 concentration increases. In the INE system, adsorption behavior is consistently dominated by CH4, while CO2 shows limited surface affinity even as its loading increases. At 20 CO2 (Fig. 9a), the number of adsorbed CO2 molecules remains very low (0–5, average 2–3), indicating only sporadic attachment to the surface. In contrast, CH4 adsorption is much higher, fluctuating between 33 and 73 with an average around 55–60, demonstrating that the INE surface strongly favors methane occupation. When the CO2 loading increases to 40 (Fig. 9b), CO2 adsorption rises moderately to 2–11 (average 6–7), but CH4 still occupies most surface sites (39–69, average 50–55), suggesting that competitive adsorption exists but remains weak. At 80 CO2 (Fig. 9c), a further increase in CO2 adsorption is observed (5–17, average 11–12), accompanied by a noticeable reduction in CH4 adsorption (19–61, average 40–45). This trend indicates that higher CO2 concentration enhances its access to the surface and partially displaces methane. However, CH4 still maintains a substantial presence, confirming that CO2 cannot fully outcompete CH4 for adsorption sites on INE. From a dynamic perspective, all systems reach a quasi-equilibrium state within 2–3 ns, after which adsorption numbers fluctuate within relatively narrow ranges without systematic drift. Overall, although increasing CO2 loading promotes its adsorption and weakens CH4 coverage to some extent, the INE surface retains a clear preference for CH4, reflecting stronger methane–surface interactions and limited competitive capability of CO2. Evolution of the number of adsorbed CO2 and CH4 molecules on the INE surface over time with CO2 loadings of (a) 20, (b) 40, and (c) 80, respectively. Corresponding evolution on the VIT surface with CO2 loadings of (d) 20, (e) 40, and (f) 80. In the VIT system, adsorption behavior still shows a clear preference for CH4, but the competitive effect of CO2 is more pronounced than in the INE case. At 20 CO2 (Fig. 9d), the number of adsorbed CO2 molecules remains low, fluctuating between 1 and 4 with an average of 2–3, indicating weak surface affinity. Meanwhile, CH4 adsorption is dominant, ranging from 32 to 74 with an average around 55–60, suggesting that the VIT surface is initially occupied primarily by methane. When the CO2 loading increases to 40 (Fig. 9e), CO2 adsorption rises moderately to 1–11 (average 4–5), while CH4 adsorption decreases slightly to 36–70 (average 50–55). This indicates the onset of competitive adsorption, where CO2 begins to occupy part of the available surface sites but still cannot rival CH4. At the highest loading (80 CO2) (Fig. 9f), CO2 adsorption increases further to 4–17 (average 10–12), accompanied by a more noticeable reduction in CH4 adsorption (16–59, average 40–45). Compared with the INE system, this decline in CH4 coverage is more evident, implying that CO2 exhibits a relatively stronger competitive capability on the VIT surface. From a dynamic perspective, adsorption reaches a quasi-steady state within 2 ns for all cases, after which both CO2 and CH4 fluctuate within limited ranges without long-term drift. Overall, although CH4 remains the dominant adsorbate, increasing CO2 concentration significantly enhances its surface occupation and weakens methane adsorption more effectively on VIT than on INE, reflecting a comparatively higher affinity of CO2 for the VIT interface. The temporal evolution of water adsorption on INE/VIT surfaces provides direct insight into interfacial hydration and stability. In the INE system, the number of adsorbed water molecules shows a clear dependence on both time and CO2 loading, reflecting dynamic restructuring of the interfacial hydration layer. At 20 CO2, water adsorption starts at a high level (230–250 within the first 3 ns) and then gradually decreases, reaching 90–120 by 20 ns. The average value is around 170–190, indicating a relatively strong but progressively weakened hydration layer. For 40 CO2, the behavior is more fluctuating, with values ranging from 86 to 263 and an average of 160–180. A noticeable drop occurs around 10–12 ns (86–119), followed by partial recovery, suggesting intermittent disruption and reorganization of water structure due to increased CO2 presence. At the highest loading (80 CO2), water adsorption is initially high (240–270) but exhibits larger fluctuations and an overall downward trend, stabilizing around 110–150 after 15 ns, with an average of 180–200 in the early stage and 130–150 in the later stage. Compared with lower loadings, the fluctuation amplitude is larger (up to 170 molecules), indicating a more unstable hydration environment. This behavior suggests that increasing CO2 concentration increasingly perturbs the water network near the INE surface, reducing long-term water retention. From a dynamic perspective, all systems undergo significant rearrangement within the first 5–10 ns, after which the number of adsorbed water molecules gradually approaches a quasi-steady state, albeit with noticeable fluctuations. Overall, while the INE surface maintains substantial water adsorption under all conditions, increasing CO2 loading weakens the stability of the hydration layer and enhances temporal variability, reflecting stronger competition between CO2 and water for interfacial space (Fig. 10a). Evolution of the number of adsorbed water molecules on the (a) INE/(b) VIT surface. In the VIT system, water adsorption exhibits a more sensitive response to CO2 loading, with distinct temporal evolution patterns. At 20 CO2, the number of adsorbed water molecules decreases markedly over time, from 200 to 215 in the initial stage to 60–120 after 15–20 ns, with an overall average around 140–160. This pronounced decline indicates a progressive depletion of the hydration layer, suggesting relatively weaker water retention on the VIT surface. For 40 CO2, water adsorption remains comparatively stable at a higher level, fluctuating within 150–220 for most of the simulation, with an average of 180–200. The absence of a clear downward trend implies a more persistent hydration environment under intermediate CO2 loading. At 80 CO2, water adsorption becomes more fluctuating and slightly reduced overall, ranging from 99 to 212 in the early stage and stabilizing around 100–150 in the later stage, with an average of 140–160. Compared with the 20 CO2 case, the decrease is less monotonic but still reflects a weakened hydration layer under high CO2 concentration. Notably, transient increases (e.g., 212 at 11 ns) suggest dynamic restructuring of interfacial water rather than steady depletion. From a dynamic standpoint, all systems undergo significant rearrangement within the first 5–10 ns, after which the adsorption levels fluctuate around quasi-equilibrium values. Overall, while water remains a major interfacial component, increasing CO2 loading on VIT leads to enhanced fluctuations and moderate reduction in water coverage, indicating that CO2 more effectively perturbs the hydration structure compared to the INE system (Fig. 10b). Although the conventional adsorption selectivity coefficient requires well-defined bulk-phase compositions, which are not strictly applicable in the present confined simulation system, the ratio of adsorbed CH4 to CO2 still provides a useful indicator of preferential adsorption behavior. The results show that CH4 consistently exhibits higher adsorption capacity than CO2 across all studied conditions, confirming its dominant occupation of adsorption sites. Moreover, with increasing CO2 loading, the relative increase of CO2 adsorption remains limited, further indicating its weaker competitive ability compared to CH4, particularly under hydrated conditions. These findings are consistent with the interaction energy analysis, where CH4 adsorption is mainly governed by dispersion interactions, while CO2 is constrained by weaker surface affinity and hydration effects. Conclusion This work systematically elucidates the coupled behavior of hydrate decomposition, gas release, and interfacial transport under varying CO2 loadings and surface chemistries by integrating molecular simulations. Across both INE and VIT systems, a consistent electrostatic hierarchy is observed, where water dominates the Coulombic interactions (typically − 600 to − 750 kJ mol−1), followed by CO2–water interactions, while all CH4-related electrostatic contributions remain negligible. Notably, increasing CO2 concentration does not proportionally enhance its electrostatic participation; instead, it induces screening and local crowding effects, weakening CO2–water coupling (e.g., from − 220 to − 360 kJ mol−1 at moderate loading to − 100 to − 180 kJ·mol-1 in denser environments) and maintaining water as the primary stabilizing component. From an adsorption perspective, CO2 exhibits limited but concentration-dependent affinity toward solid surfaces. In both INE and VIT systems, CO2 adsorption remains low at 20 molecules (generally < 5), but increases to 5–10 at 40 molecules and up to 10–17 at 80 molecules, indicating enhanced surface occupation under higher loading. In contrast, CH4 consistently shows significantly higher adsorption (typically 30–70 molecules), demonstrating its stronger accumulation tendency near the interface despite its electrostatic inactivity. This highlights a decoupling between adsorption behavior and Coulombic interactions, where CH4 adsorption is governed primarily by van der Waals interactions rather than polarity. Comparatively, VIT surfaces exhibit slightly more dynamic but still weak CO2 affinity, while INE systems show relatively more stable adsorption patterns. In all cases, the systems reach a quasi-equilibrium state rapidly (within 1–2 ns), after which both interaction energies and adsorption numbers fluctuate within narrow ranges without systematic drift. Overall, the results reveal that interfacial water structure plays a decisive role in governing system stability, while CO2 contributes in a secondary, concentration-limited manner, and CH4 behavior is dominated by nonpolar interactions. These findings provide molecular-level insight into how interfacial chemistry and gas composition jointly regulate hydrate-related processes and gas transport efficiency. It should be noted that the observed preferential adsorption of CH4 over CO2 differs from the commonly reported trend in dry, single-component systems, where CO2 typically exhibits stronger adsorption affinity. This discrepancy can be attributed to the hydrated multicomponent conditions considered in this study. Specifically, water molecules strongly occupy adsorption sites and form a stable hydration layer on the coal surface, which screens electrostatic interactions and reduces the accessibility of polar sites for CO2. As a result, CO2 adsorption is significantly constrained, while CH4, dominated by dispersion interactions, maintains relatively stable adsorption behavior. In addition, the simplified coal model and confined simulation environment may also influence the relative adsorption behavior. Therefore, the conclusions of this work are primarily applicable to hydrated multicomponent systems and should not be directly extrapolated to dry conditions. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Author contributions Qun Cheng, Methodology, investigation, writing—original draft; Yan Liu, Formal analysis, data curation; Cen Chen, Conceptualization, writing—original draft, funding acquisition; Can Yang, Methodology, investigation; Shiyun Wei, Methodology, investigation; Ruiying Zhu, Formal analysis, data curation; Yang Chen, Formal analysis, data curation; Songze Li, Writing—review and editing. Funding Project supported by the National Natural Science Foundation of China (No. 42272175), Hubei Provincial Natural Science Foundation of China (No. 2025AFA106), National Science and Technology Major Project (No. 2025ZD1400807). Data availability The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author.