Solvation Dynamics of CO₂(g) by Monoethanolamine at the Gas–Liquid Interface: A Molecular Mechanics Approach

Abstract

A classical force field approach was used to characterize the solvation dynamics of high-density CO₂(g) by monoethanolamine (MEA) at the air–liquid interface. Intra- and intermolecular CO₂ and MEA potentials were parameterized according to the energetics calculated at the MP2 and BLYP-D2 levels of theory. The thermodynamic properties of CO₂ and MEA, such as heat capacity and melting point, were consistently predicted using this classical potential. An approximate interfacial simulation for CO₂(g)/MEA(l) was performed to monitor the depletion of the CO₂(g) phase, which was influenced by amino and hydroxyl groups of MEA. There are more intramolecular hydrogen bond interactions notably identified in the interfacial simulation than the case of bulk MEA(l) simulation. The hydroxyl group of MEA was found to more actively approach CO₂ and overpower the amino group to interact with CO₂ at the air–liquid interface. With artificially reducing the dipole moment of the hydroxyl group, CO₂–amino group interaction was enhanced and suppressed CO₂(g) depletion. The hydroxyl group of MEA was concluded to play dual but contradictory roles for CO₂ capture.

1. Introduction

Monoethanolamine (MEA) is extensively used in the electric power industry for directly scrubbing CO₂ from the exhaust gas produced by fossil fuel combustion [1]. MEA nucleophilically attacks the partially positively charged carbon atom of CO₂ with its amino groups, forming the zwitterionic (ZW) intermediate of [HOCH₂CH₂NH₃⁺CO₂⁻], thereby pulling CO₂ from the exhaust gas. This ZW intermediate subsequently stabilizes itself by transferring a proton from its quaternary nitrogen through an intermolecular hydrogen bond (HB) network to other proton acceptors. Such a network can be formed by a combination of the amino or hydroxyl groups of MEA or with other additives that can form HBs. Water and other types of alcoholamines, such as diethanolamines (DEAs) and methyl-diethanolamines (MDEAs), are common additives in MEA solutions to improve overall CO₂ scrubbing efficiency.

MEA is widely produced by controlling the stoichiometry of ethylene oxide reacting with aqueous ammonia. Secondary or tertiary amines can be generated by increasing the ethylene oxide:ammonia ratio. Several studies have performed experimental characterizations [2,3,4,5,6,7,8,9,10,11,12,13,14] and computational investigations [15,16,17] to elucidate and improve CO₂-scrubbing reactivity by using various types of alcoholamines. The engineering of various functionalized ethylene oxides as the starting materials at the industrial scale to synthesize the designed alcoholamines can be challenging. Conway et al. measured the kinetic rate and equilibrium constants of CO₂ (aq) with various sterically hindered alcoholamines and determined iso-butylamine, n-butylamine, and n-propylamine to be the main scrubbers for the CO₂ binding step; none of the three amines contain hydroxyl groups [13]. A hydroxylated compound, 3-amino-1-propanol (MPA), was observed to be approximately 20% less efficient at binding CO₂ (aq) compared with the aforementioned amines. Intuitively, elongating the alcohol groups, such as MPA in respect to and MEA, can improve CO₂ binding affinity. Li and Tsai performed molecular dynamics simulations and stated that an appropriate length of the alkyl chain between the amino and hydroxyl groups can minimize the electronic cancellation effect from the two electron-withdrawing groups [18]. Adding water molecules to alcoholamines can facilitate the deprotonation process [19] and stabilize the ZW intermediates. However, water molecules can also potentially compete for the CO₂ absorption sites by forming intermolecular HBs with the amino groups [18]. If the basicity of alcoholamines is adequately strong, water can be deprotonated to form OH⁻ and protonated alcoholamines, where OH⁻ is known to bind CO₂ efficiently.

Considering the aforementioned observations, the hydroxyl groups of alcoholamines seem to play a trivial role for CO₂ capture. However, because of the use of ethylene oxide as the starting material for synthesizing alcoholamines, the hydroxyl groups are inherently present in CO₂ scrubbers. Compared with the CO₂-binding characteristic of the amino groups, the hydroxyl groups can shuttle protons from the quaternary nitrogen of the ZW intermediates to the CO₂⁻ moieties, as shown by ab initio molecular dynamic simulations of CO₂ in neat MEA liquid [20]. However, this proton transfer capability of the hydroxyl groups is anticipated to be less effective compared with that of water molecules.

Several studies have performed computational modeling of the CO₂ capture phenomena at the nanometer and nanosecond scales by using a classical force field approach [21]. Most of these studies primarily focused on predicting the CO₂ transport property and fixation phenomena through physical absorption in various solutions, such as CO₂, H₂O, and a mixture of different types of alcoholamine derivatives. Jhon et al. predicted the nucleophilicities and accessibilities of MEA, 2-methylaminoethanol, and 2-amino-2-methyl-1-propanol, and emphasized that the predicted CO₂-binding reactivity correlates with the accuracy of describing the steric effect of amines by the chosen force field [22]. Turgman-Cohen et al. included the ionic species that resulted from CO₂ chemisorption in the model and qualitatively mimicked the thermogravimetric experiments [23]. Farmahini et al. simulated the separation of CO₂ and CH₄ mixtures in MDEA solution, which was accelerated piperazine [24]. Yu et al. developed a synergistic molecular dynamics model and reported that an optimal ternary mixture outperformed other quaternary or quintuple amine systems in the realm of CO₂ diffusivity [25].

Extensive computational studies have focused on improving CO₂-binding efficiency by using amino groups; however, the roles of the hydroxyl groups of alcoholamines have not been fully characterized. Notably, Du et al. measured the surface vibrational spectrum by using the sum frequency generation technique to address the hydrophobicity of the air–water interface and identified the spectroscopic fingerprint of the dangling hydroxyl groups [26]. Electrostatic interactions resulting from such dangling hydroxyl groups at the air–water interface favored the solvation of anionic ions, as previously reported [27,28,29,30,31,32,33]. Thus, the solvation phenomena of nonpolar CO₂ molecules containing the quadruple moment at the air–alcoholamine interface inspired the present study. Pure CO₂(g) can be considered a hydrophobic phase, whereas scrubber alcoholamines are highly hydrophilic because of the presence of amino and hydroxyl groups. Understanding the dynamic details of removing CO₂ molecules from the hydrophobic environment to be subsequently solvated and absorbed by hydrophilic alcoholamines can provide the underlying microscopic characteristics of these organic scrubbers.

Therefore, we conducted all-atom molecular dynamic simulations to model CO₂ solvation dynamics at the air–liquid interface. We started from the MM3 force field [34,35,36,37,38,39], a classical molecular mechanics force field for hydrocarbons, and modified the required parameters for CO₂ and MEA molecules, then benchmarked the CO₂ potential in the gas phase, MEA(l) potential in the solution phase, and the CO₂–MEA absorption pathway compared with density functional theory (DFT) results. Finally, we monitored the dynamics of the CO₂(g)/MEA(l) interface. Proton transfer from the ZW intermediates for the subsequent chemical absorption step was omitted in this study due to the complexity of parameterizing the reactive force fields. The present study has focused at the dominant physical absorption phenomena of CO₂ molecules at the air–liquid interface.

2. Results and Discussion

2.1. Benchmark Using CO₂ and (CO₂)₂ Models

The intermolecular interactions between CO₂ molecules of three selected characteristic structures—namely, the global minimum, parallel, and T-type (CO₂)₂ structures in Figure 1—were compared at the BLYP-D2, MP2, and molecular mechanics (MM) levels of theory in

Table 1. Comparison of CO₂ monomeric and dimeric models at the BLYP-D2, MP2, and molecular mechanics (MM) levels (kcal/mol).

1 Isomers	Min	Parallel	T-Type	${\mathsf{\nu }}_{\mathbf{\text{CO}}\mathbf{2}}$ (cm−1)
2 BE(BLYP-D2)	−1.06	0.25	−0.77	636, 1302, 2310
BE(BLYP)	0.29	0.76	0.09
BE(disp)	−1.36	−0.50	−0.86
2 BE(MP2)	−1.27	−0.01	−1.07	668, 1310, 2374
BE(HF)	−0.05	0.83	−0.16
BE(disp)	−1.22	−0.84	−0.91
3 Eint(MM)	−1.23	0.16	−0.92	629, 1240, 2374
${\text{Eint}}_{\mathsf{\mu }\mathsf{\mu }}^{\text{MM}}$	−1.00	0.32	−0.70
${\text{Eint}}_{\text{LJ}}^{\text{MM}}$	−0.23	−0.15	−0.22

¹ Graphical representation of the (CO₂)₂ models is presented in Figure 1. ² Binding energy (BE) (BLYP-D2 and MP2) contains the correction of the basis set superposition error by using BLYP-D2 and MP2 optimized structures, respectively. ³ Eint denotes the interaction energy at the MM level, where $\mathsf{\mu }\mathsf{\mu }$ and LJ denote dipole–dipole and Lennard–Jones terms, respectively.

. Geometric constraints were imposed for optimizing the parallel and T-type structures. The intermolecular interactions (Eint) characterized at the MM level were compared using the binding energies (BE) of the BLYP-D2 or MP2 levels. For the global minimum structure of (CO₂)₂, Eint was predicted to be −1.23 kcal/mol versus −1.06 and −1.27 kcal/mol at the BLYP-D2 and MP2 levels, respectively. The parallel (CO₂)₂ structure with C–C constrained at 4.0 Å was the most repulsive one, where the MM level (0.16 kcal/mol) qualitatively reproduced the interaction, as predicted at the BLYP-D2 and MP2 levels (0.25 and −0.01 kcal/mol). T-type (CO₂)₂ was consistently predicted to interact less strongly than the global minimum at all three levels. The current nonpolarizable MM approach dominantly contributed through dipole–dipole (μμ) interactions for intermolecular CO₂ interactions, but van der Waals (vdW) interactions primarily contributed to the BE of BLYP-D2 and MP2 levels. Using the functional forms of classical electrostatic and Lennard–Jones (LJ) potentials to distinguish the electrostatic and vdW contributions have been reported as challenging in the case of water model development [40,41,42,43]. The success of such distinction is mainly subject to the selected functional form for the classical representation of vdW interactions. Thus, further improvement of current CO₂ potential is in progress and not included in the scope of this study. Nonetheless, the total intermolecular CO₂ interaction calculated by MM still remained a reasonable agreement with the quantum mechanical (QM) results. Finally, vibrational frequencies of the CO₂ monomer calculated at the MM level were consistent with those calculated at the BLYP-D2 and MP2 levels.

2.2. Phase Transition of (CO₂)₁₃

The phase transition temperature of (CO₂)₁₃ clusters was characterized to benchmark the thermodynamic property prediction by using the current CO₂ potential. Maillet et al. reported the melting transition of (CO₂)₁₃ at 95 K [44]. Liu and Jordan observed the solid–solid phase transition at approximately 90 K [45] by using a two-body model developed by Murthy et al. [46]. Murthy potential is a two-body potential using a rigid monomer, 5-point electrostatic sites, and atomic 6–12 Lennard–Jones terms. Comparably, the current CO₂ model showed vibrational capability, with atomic 6–12 Lennard–Jones terms and bond-centered dipole–dipole (μμ) interactions. The calculated heat capacity curve from 88 K to 97 K of (CO₂)₁₃ clusters is shown in Figure 2. Each canonical ensemble (NVT) simulation was pre-equilibrated for 2.5 ns and sampled for 100 ns at a 0.5-fs time step. A solid–solid phase transition was identified at 91 K, followed by cluster evaporation beyond 97 K. The heat capacity was calculated using Equation 1.

$${\mathrm{C}}_{\text{NVT}}\left(\mathrm{T}\right)=\frac{\langle {\mathrm{U}}^2\rangle -{\langle \mathrm{U}\rangle }^2}{{\mathrm{N}}_{\mathrm{A}} {\mathrm{k}}_{\mathrm{B}} {\mathrm{T}}^2}$$

(1)

where U is the sampled potential energy of (CO₂)₁₃, N_A is the Avogadro constant, k_B is the Boltzmann constant, 〈U²〉 and 〈U〉² denote the mean of U² and the square of the average U, respectively.

2.3. Benchmark of the MEA Potential

Three stable geometries are identified for the monomeric MEA at MP2/aug-cc-pVTZ with the polarizable continuum model (PCM) as shown in Figure S1. One of the cis-form MEA monomers containing an intramolecular hydrogen bond (intraHB) interaction between OH···NH₂ is predicted to be the lowest energy at MP2 level. The energetics between these three stable geometries are calculated to be deviated by a trivial difference at both MP2 and MM levels, as summarized at Table S1 in Supplementary Materials. Two hydrogen-bonded geometries, denoted as NHN and OHO, were selected for comparing the current MEA potential at the MM level with BLYP-D2, as shown in Figure 3. Constrained optimizations were conducted to converge the minimum geometry searching with the XH···X angle frozen at 180°. The intermolecular interactions between two MEA molecules are summarized in

Table 2. Comparison of intermolecular interactions between the BLYP-D2 and MM levels (kcal/mol).

Isomers 3	NHN	OHO
1 BE(BLYP-D2)	−4.62	−6.90
1 BE(BLYP)	−1.35	−3.50
1 BE(disp)	−3.28	−3.32
1 ${\text{Eint}}_{\mathsf{\mu }\mathsf{\mu }}^{\text{MM}}$	−1.85	−3.61
1 ${\text{Eint}}_{\text{VDM}}^{\text{MM}}$	−1.88	−1.40
2 ${\text{Eint}}_{\text{Total}}^{\text{MM}}$	−4.16	−6.15
2 ${\text{Eint}}_{\mathsf{\mu }\mathsf{\mu }}^{\text{MM}}$	−2.17	−4.33
2 ${\text{Eint}}_{\text{VDM}}^{\text{MM}}$	−1.99	−1.82

¹ Constrained optimized geometry at the BLYP-D2/aug-cc-pVTZ level. ² Constrained optimized geometry at the MM level. ³ The same notation is adopted as that in Table 1.

. The binding energy of OHO was predicted to be stronger than that of NHN at −6.90 and −6.15 kcal/mol for the BLYP-D2 and MM levels, respectively. NHN interactions were estimated to be −4.62 and −4.16 kcal/mol for the BLYP-D2 and MM levels, respectively. Even by using the BLYP-D2-optimized geometry, the MM level predicted acceptable interactions compared with the MM-optimized geometry, as shown by ¹Eint versus ²Eint in Table 2. This consistency indicated strong similarity between the two constrained optimized geometries at the BLYP-D2 and MM levels. However, weak contribution to the intermolecular binding from the vdW interactions was still in presence for the current MEA potential. Phase transition of the (MEA)₁₂₈ model was simulated to investigate the thermodynamic property prediction by using the current MEA model. Eight NVT simulations were performed at 280–420 K, with a density of 1.01 g/cm³. Each run was pre-equilibrated and sampled for 3 ns to predict the corresponding heat capacity. A phase transition was identified at approximately 320 K as shown in Figure 4, and the experimental freezing point of MEA(l) was 283 K [47]. Thus, the current MEA potential slightly overestimated the melting property of MEA.

2.4. CO₂ Binding Using MEA

Constrained optimizations in the gas phase along the C_CO2–N_MEA distance were calculated at both the BLYP-D2 and MM levels along the CO₂-binding pathway. The relative energetics considering each physical absorption minimum structure at both levels is summarized in Figure 5. CO₂ was physically absorbed at a distance of approximately 2.9 Å from N_MEA at the BLYP-D2/aug-cc-pVTZ level and approximately 2.7 Å for the case predicted at the MM level. Close agreement was observed between the two theory levels for the C_CO2–N_MEA distance (r_CN) ranging from 2.8 to 2.2 Å. At r_CN = 2.0 Å, MM showed more than 2.5 kcal/mol deviated from BLYP-D2. This deviation exponentially increased as r_CN increased beyond 2.0 Å. Natural population analysis was performed to estimate the fragment charge of the amino group of MEA (q_NH2) and CO₂ (q_CO2) along the absorption pathway, as plotted at the secondary y-axis in Figure 5. At r_CN = 2.2–2.8 Å, q_NH2 and q_CO2 slightly varied. However, considerable increases and decreases were observed for q_NH2 and q_CO2 beyond r_CN = 2.0 Å, respectively, suggesting a substantial charge transfer effect as CO₂ approached the amino group of MEA. The current nonpolarizable potential did not consider such electronic redistribution, regardless of the reparameterization of the balance between electrostatic and vdW interactions in this approach. Thus, marked deviations were anticipated between the MM and BLYP-D2 levels. Nonetheless, the accuracy of the current potential was reasonable at r_CN > 2.0 Å.

2.5. CO₂(g) Dissolution by MEA(l)

A conceptual interfacial model using high-density CO₂(g) surrounded by MEA(l) was characterized to study the solvation dynamics of CO₂ in MEA(l). CO₂(g) represented by a (CO₂)₄₄ cluster model was equilibrated at 400 K for 4 ns under NVT ensemble. A repulsive spherical potential (r = 9.1257 Å) was applied, and the sphere density was fixed at 1.01 g/cm³. The final atomic coordinates and velocities of the (CO₂)₄₄ cluster were adopted for the subsequent CO₂(g)/MEA(l) simulation. A 4 ns NVT simulation of (MEA)₈₆₄ under the periodic boundary condition, T = 400 K, and density = 1.01 g/cm³ was conducted for creating the MEA(l) model. The MEA molecules located within r = 9.5 Å were removed from the final structure of (MEA)₈₆₄ simulation, and an (MEA)₈₀₄ model was obtained by adopting the corresponding atomic velocities. Finally, the initial condition of the CO₂(g)/MEA(l) simulation was generated by combining the (CO₂)₄₄ cluster and (MEA)₈₀₄ model by using the same cell size as that of the (MEA)₈₆₄ simulation, as shown in Figure 6.

A 250 ps simulation using the isoenthalpic-isobaric ensemble (NPH) was subsequently conducted for the interfacial CO₂(g)/MEA(l) model. The external pressure of the NPH ensemble was 7415 atm, according to the average value of the previous (MEA)₈₆₄ NVT simulation. Radial distribution functions, g(r), of C_CO2–C_CO2, C_CO2–N_NH2, and C_CO2–O_OH, denoted as $g_{CcCc}^{orig}\left(r\right), g_{CcN}^{orig}\left(r\right), and g_{CcO}^{orig}\left(r\right)$, were plotted considering the simulation time, as shown in Figure 7. Multiple g(r) results were analyzed for various time periods (0, 20, 40, 60, 80, 100, 120, 150, 200, and 250 ps), as shown in the legends. In Figure 7a, $g_{CcCc}\left(r\right)$ substantially increased for the $r_{CcCc}>16$ Å region after 250 ps simulation, and this increase represented the depletion kinetics of high-density CO₂(g). The profound $g_{CcCc}^{orig}\left(r=3.7 \AA \right)$ indicated the presence of a CO₂ clustering effect. Such an effect is considered to diminish with the increasing simulation time because of the stronger interaction of CO₂–MEA than of CO₂–CO₂. Between $g_{CcN}^{orig}\left(r\right)$ and $g_{CcO}^{orig}\left(r\right)$, the probability of CO₂ binding was higher by the hydroxyl group than by amino groups of MEA. Notably, such a difference cannot be attributed to the binding energy of CO₂ to the amino and hydroxyl groups of MEA because the intermolecular interaction of CO₂ with the binding motif at the MM level was 5.50 and 5.05 kcal/mol, respectively. According to 250 ps simulation, the hydroxyl groups contributed more to the depletion of the (CO₂)₄₄ cluster and suppression of the interactions between the CO₂ and amino groups.

A control experiment using 250 ps CO₂(g)/MEA(l) simulation was performed to clarify the role of the hydroxyl groups of MEA, in which ${\mu }_{OH}$ was scaled down by 90% and other parameters were the same as those in prior unscaled simulation. The resulting g(r) of C_CO2–C_CO2, C_CO2–N_NH2, and C_CO2–N_OH, denoted as $g_{CcCc}^{sc}\left(r\right)$, $g_{CcN}^{sc}\left(r\right)$, and $g_{CcO}^{sc}\left(r\right)$, are summarized in Figure 7d–f. Figure 7d shows that the $g_{CcCc}^{sc}\left(3.7\right)$ peak was substantially higher than $g_{CcCc}^{orig}\left(3.7\right)$ at the end of 250 ps, suggesting the presence of a stronger CO₂ clustering effect. In other words, CO₂(g) depletion was suppressed because of weak ${\mu }_{OH}$••${\mu }_{C=O}$ interactions. The physical absorption of CO₂ by the amino groups was markedly enhanced, as observed in $g_{CcN}^{sc}\left(2.8\right)$ in Figure 7e. The probability of CO₂ being physically absorbed by the hydroxyl groups was observed at further $r_{CcO}$ distance by $g_{CcO}^{sc}\left(3.0\right)$ compared with $g_{CcO}^{orig}\left(2.7\right)$ in Figure 7c of the unscaled case. The time evolution of $g_{CcCc}^{orig}\left(3.65\right)$ and $g_{CcCc}^{sc}\left(3.65\right)$ was compared with consideration of the simulation time, as shown in Figure 8, and CO₂(g) depletion was substantially enhanced because of the introduction of the reduced ${\mu }_{OH}$ of MEA molecules.

The geometries of MEA molecules containing the intramolecular hydrogen bond interactions are analyzed for the 2 ns bulk MEA simulation and the first 200 ps of the CO₂(g)/MEA(l) simulation. Two hundred snapshots are recorded for each case where 10 MEA molecules are selected and monitored out of these recorded snapshots. The histograms of the intramolecular N_NH2-H_OH distance (r_intraNH) are recorded in Figure S2 (10 × 200 data points are included at each figure). By comparing the recorded data points of r_intraNH < 2.2 Å, the CO₂(g)/MEA(l) simulation contains substantially more intraHB geometries during the first 200 ps over the case of bulk MEA simulation. The less hydrophilic CO₂ phase could stabilize the presence of intraHB MEA orientations, while the bulk MEA liquid could favor the intermolecular hydrogen bond interactions.

3. Experimental Section

Molecular mechanics (MM) calculations were performed using Tinker version 7.1 [48]. Force field parameters for MEA and CO₂ adopted the relevant fragments defined in MM3 force field, followed by comprehensive reparameterization routines in order to maximize the agreement with the quantum mechanical (QM) calculations at MP2 or DFT levels for the hydrogen-bonded complexes of (MEA)₂ and the CO₂••MEA physically absorbed structure, as discussed below. MM3 was originally designed for describing hydrocarbons and thus provided a good starting point for the subsequent parameterizations [34,35,36,37,38,39]. BLYP functional [49,50] augmented with Grimme’s D2 dispersion correction [51], being reported to appropriately describe CO₂–alcoholamine interactions by using a triple-zeta quality basis set, was adopted for the DFT calculations [52]. The CO₂ parameters took into account molecular stretching and bending motions, interatomic Lennard–Jones interactions, and bond-centered dipole–dipole interactions. The interatomic van der Waals (vdW) interactions were described using the arithmetic and geometric combining rules for radius and epsilon, respectively. The cutoff values of vdW and dipole–dipole interactions were set at 10.0 and 12.0 Å, respectively. Berendsen thermostat was used for the molecular dynamic simulations with a coupling time of 0.1 ps [53]. All electronic structure calculations in this study were performed using the GAUSSIAN 09 package [54]. MP2 and BLYP-D2calculations used the aug-cc-pVTZ basis set.

4. Conclusions

In the present study, a nonpolarizable MM approach was reported to characterize CO₂(g) dissolution in MEA(l). The current CO₂ and MEA potentials showed reasonable agreement with the MP2 and BLYP-D2 levels of theory for describing the intramolecular vibration and intermolecular interactions of CO₂, (CO₂)₂, and (MEA)₂. Heat capacity simulations by using (CO₂)₁₃ clusters reproduced the solid–solid phase transition at approximately 90 K, as reported in previous studies. The melting transition predicted using the (MEA)₁₂₈ model was approximately 320 K, in which the experimental freezing potential of pure MEA was 283 K. Along the CO₂ absorption pathway by MEA, the MM approach reproduced the energetics of the BLYP-D2 level if CO₂ was bound at a distance of approximately 2.0 Å from the amino group of MEA. The absence of polarizable terms in the current potential could not describe the charge transfer effect, which was included in the BLYP-D2 level of theory, because CO₂ was absorbed by MEA at a short C_CO2 to N_MEA distance (<2.0 Å).

At the gas–liquid interface represented by the (CO₂)₄₄ and (MEA)₈₀₄ models, CO₂(g) was almost completely depleted within 250 ps. There are notably more intramolecular hydrogen bond interactions identified than the case of bulk MEA simulation. The hydroxyl groups of MEA more actively approached CO₂ at the interfacial region and suppressed the absorption probability of the NH groups. The artificial reduction of the dipole moment of the hydroxyl bonds enhanced CO₂–amino group interactions, but the depletion of CO₂ gas phase was reduced as well. The hydroxyl groups of MEA were identified to play contradictory roles in the interfacial CO₂ capture process. The current findings could provide additional perspectives for the synthetic design of molecular CO₂ scrubbers.