Molecular Dynamics Study of the Solution Structure, Clustering, and

Feb 15, 2018 - CO2 sequestration from anthropogenic resources is a challenge to the design of environmental processes at a large scale. Reversible che...
0 downloads 9 Views 3MB Size
Article Cite This: J. Phys. Chem. B XXXX, XXX, XXX−XXX

pubs.acs.org/JPCB

Molecular Dynamics Study of the Solution Structure, Clustering, and Diffusion of Four Aqueous Alkanolamines Sergey M. Melnikov and Matthias Stein* Molecular Simulations and Design Group, Max-Planck-Institut für Dynamik komplexer technischer Systeme, Sandtorstrasse 1, 39106 Magdeburg, Germany S Supporting Information *

ABSTRACT: CO2 sequestration from anthropogenic resources is a challenge to the design of environmental processes at a large scale. Reversible chemical absorption by amine-based solvents is one of the most efficient methods of CO2 removal. Molecular simulation techniques are very useful tools to investigate CO2 binding by aqueous alkanolamine molecules for further technological application. In the present work, we have performed detailed atomistic molecular dynamics simulations of aqueous solutions of three prototype amines: monoethanolamine (MEA) as a standard, 3-aminopropanol (MPA), 2methylaminoethanol (MMEA), and 4-diethylamino-2-butanol (DEAB) as potential novel CO2 absorptive solvents. Solvent densities, radial distribution functions, cluster size distributions, hydrogen-bonding statistics, and diffusion coefficients for a full range of mixture compositions have been obtained. The solvent densities and diffusion coefficients from simulations are in good agreement with those in the experiment. In aqueous solution, MEA, MPA, and MMEA molecules prefer to be fully solvated by water molecules, whereas DEAB molecules tend to self-aggregate. In a range from 30/ 70−50/50 (w/w) alkanolamine/water mixtures, they form a bicontinuous phase (both alkanolamine and water are organized in two mutually percolating clusters). Among the studied aqueous alkanolamine solutions, the diffusion coefficients decrease in the following order MEA > MPA = MMEA > DEAB. With an increase of water content, the diffusion coefficients increase for all studied alkanolamines. The presented results are a first step for process-scale simulation and provide important qualitative and quantitative information for the design and engineering of efficient new CO2 removal processes.

1. INTRODUCTION Chemical absorption by aqueous amines and alkanolamines is considered to be one of the most promising methods for CO2 removal from flue gas1,2 due to its long standing use and efficiency. Monoethanolamine (MEA) is currently the most widely used solvent for this technique. However, its employment is hampered by the high energy required for solvent regeneration, possible degradation of the amine, and problems with equipment corrosion.3,4 These technological difficulties show the demand for advanced, novel solvents with a higher degree of efficiency and more robust molecular properties. The search for new and better solvents can efficiently be supported by molecular simulations, which also help to understand the CO2 capture underlying molecular mechanisms. Especially for cases where experimental molecular characterization is complicated or impossible, computer-aided solvent design is a versatile tool. Indeed, material and processes of CO2 capture are the focus of numerous recent publications. Two recent papers5,6 review the state-of-the-art molecular simulation techniques in this area. Molecular simulations make use of either static quantum mechanical calculations, ab initio molecular dynamics (MD), or classical molecular simulations (molecular dynamics (MD) and Monte Carlo simulations). © XXXX American Chemical Society

Classical MD simulations give important insight into the liquid structure, thermodynamic properties, solubility, viscosities, and dynamic properties of pure solvents and solvent mixtures.7 Recently, computer-aided molecular design was applied for the selection of new solvent candidates for the chemical absorption process of carbon dioxide (CO2) from a gas stream.8 The top candidate compounds were recommended there for further investigations as potential solvents for CO2 capture. We selected a representative subset of these novel compounds for this current MD study: MEA plus three other top candidate molecules from that list. Monoethanolamine (MEA) has been chosen as a standard and the benchmark, 3-aminopropanol (MPA) as an example for a novel primary amine, 2methylaminoethanol (MMEA) as a prototypic secondary amine, and 4-diethylamino butan-2-ol (DEAB) as a representative tertiary amine. The chemical structures of the studied molecules are displayed in Figure 1. There are many comprehensive MD studies of pure MEA9−16 that revealed a wealth of information about Received: October 18, 2017 Revised: February 7, 2018 Published: February 15, 2018 A

DOI: 10.1021/acs.jpcb.7b10322 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

obtained with parameters published by López-Rendón et al. We chose the OPLS-AA forcefield, which well reproduced OCCNtorsion conformations and provided good results for both density and diffusion coefficients. The OPLS-AA forcefield also offers the possibility of investigating such a diverse class of amine molecules (from primary to tertiary amines) with linear and branched structures. Water molecules are represented by the charge/extended (SPC/E) model24 because this model was shown to better reproduce experimental diffusion coefficients.25 Long-range electrostatic interactions were treated with the particle-mesh Ewald algorithm. Nonbonded interactions were modeled with a 12−6 Lennard-Jones potential. A cutoff radius of 1.4 nm was used for all interactions. Lennard-Jones parameters for unlike interactions were calculated using the geometric average rule. 2.2. Simulations Setup. MD simulations were carried out with Gromacs 5.1.226 at a temperature of 298 K. The following sequence of simulation steps was performed for each alkanolamine type and aqueous mixtures. Initially, an isothermal−isobaric (NPT) ensemble 2 ns simulation run (preceded by 500 ps equilibration) was performed to define the solvent density and simulation box size for the next simulation stage. A Berendsen thermostat and barostat,27 with coupling parameters of 0.2 and 2 ps, respectively, were used to maintain temperature and pressure. Then, a canonical (NVT) ensemble 20 ns simulation run (preceded by a 1 ns equilibration run) was carried out. A Nosé−Hoover thermostat28 with a coupling constant of 0.4 ps was used for temperature control. The equations of motion were integrated with a time step of 1.0 fs for both ensembles. Simulation trajectories were saved every 1 ps for further data analysis. Simulations were carried out for solvent ratios of 0/100 (fully diluted), 7.5/92.5 (6/94 for the case of MEA), 30/70, 50/50, 80/20, and 100/0 (pure case) (w/w) alkanolamine/water for the four studied molecule types. For MEA, additionally two compositions at 59/41 and 93/07 (w/w) MEA/water were simulated. The amount of molecules and simulation box sizes for all studied molecular ensembles are provided in the Supporting Information. Simulation box sizes were determined from preliminary simulations and checked for an absence of finite size effects. For the sake of convenience, the studied mixture compositions expressed in mass fractions are related in Table 1 with corresponding molar fractions and concentrations. 2.3. Data Analysis. The cluster size distribution was calculated by the algorithm proposed by Geiger et al.,29 taking periodic boundary conditions into account. A formation of hydrogen bond (HB) was defined when satisfying two geometrical criteria: a distance between HB acceptor and donor of the hydrogen atom of less than 0.35 nm and angle between the vector pointing from donor X to acceptor Y atom

Figure 1. Chemical structures of the studied alkanolamine molecules.

thermodynamics, liquid structure, and dynamic properties of liquid MEA. However, only a few simulations dealt with aqueous solutions of MEA.15,17,18 The liquid structures of aqueous MEA solutions at different mixture compositions were investigated by Gubskaya17 and Orozco,18 diffusion coefficients were calculated by Gubskaya17 and da Silva,15 the conformer distribution for MEA at infinite dilution in water by da Silva,15 and a good agreement for mixed solvent densities was obtained.18 In this work, we investigate structural and dynamic properties, hydrogen-bond formation, and clustering behavior of aqueous MEA in comparison with those of the three novel alkanolamine compounds for CO2 chemical absorption. To our knowledge, this is the first such complete MD study of molecular structures, liquid properties, and diffusion coefficients for MPA, MMEA, and DEAB molecules in pure and mixed aqueous solutions at different molecular compositions. In this work, we develop an understanding how the studied compounds are solvated in aqueous solution and how the liquid structure and mobility of solvent molecules change with a variation of the water content.

2. DETAILS OF SIMULATIONS 2.1. Choice of Forcefield. To find an apt forcefield to model the alkanolamine molecules, several preparatory MD simulations with pure MEA at various temperatures were performed. We tested four forcefields: the two parameter sets proposed by López-Rendón et al.10 and by da Silva et al.,15 the AUA4 forcefield,12 and the all-atom optimized potentials for liquid simulations (OPLS-AA) forcefield.19 Simulated densities, torsional angle distributions of the OCCN-torsion, and diffusion coefficients (data can be found in the Supporting Information) were compared to those in available experimental data20−22 and previous simulations.23 The comparison shows that pure MEA solvent densities and conformations of the OCCN-torsion are better reproduced by the AUA4 forcefield, whereas diffusion coefficients closest to the experiment are

Table 1. Molar Fractions and Concentrations of Studied Mixture Compositions MEA

MPA

MMEA

DEAB

composition, % wt of alkanolamines

molar fraction

concent, mol/L

molar fraction

concent, mol/L

molar fraction

concent, mol/L

molar fraction

concent, mol/L

100 80 50 30 7.5 6

1.0 0.542 0.228 0.112

17.02 13.71 8.48 5.03

1.0 0.490 0.194 0.093 0.019

13.52 10.95 6.85 4.06 1.00

1.0 0.490 0.194 0.093 0.019

12.38 10.36 6.67 4.01 1.00

1.0 0.332 0.110 0.051 0.010

6.06 5.06 3.31 2.04 0.52

0.019

1.00 B

DOI: 10.1021/acs.jpcb.7b10322 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

The solvent densities determined from our simulations reproduce well the experimental density−composition relation for the MEA case. For MPA and MMEA, the MD simulation results are in good agreement with those of the experiment. Altogether, the general agreement with experimental data provides confidence in the reliability of parameters and further simulation results in this work. 3.3. Solvent and Solvation Structure. Radial distribution functions (RDFs) are commonly used in molecular simulations as a primary means to present the liquid structure at the atomistic level. RDFs of aqueous MEA (including those of the pure case) for selected atom pairs are presented in Figure 3. Amine MEA−MEA interactions are presented to the left and at the right hand side are MEA−water and water−water interactions. We, here, display RDFs only for atom pairs that are relevant to determine HB interactions and indicate the relative spatial orientations of molecules. Throughout the article, the symbol “OW” denotes the oxygen atom of a water molecule; the other “O” symbol designates atoms of the alkanolamine, here MEA. The subscript “N” designates the carbon atoms closest to the nitrogen atoms in MEA, and the subscript “CC” denotes the center-of-mass of the C−C bond of MEA. The RDF of the N−O pair was calculated only for intermolecular interactions. At first, we note that RDFs, describing the pair interactions between nitrogen and oxygen atoms of MEA and water (Figure 3a−c,e,f), are very similar. The curves display a sharp maximum at an interval between r = 0.28−0.31 nm and smaller yet distinguishable peaks at r = 0.46−0.58 nm. These sharp maxima indicate a strong propensity for HB formation between the corresponding amino and hydroxyl groups of MEA and water. The exact appearance of the maxima depends on the actual atom pair type. For an O−O pair, it occurs at 0.29 nm, for an N−N pair at 0.31 nm, and for an N−O pair at about 0.305 nm. The amplitude and form of the RDF maxima show how strong the HB formation between a particular atom pair is. For the pure case of MEA, the O−O, N−N, and N−O HB interactions occur equally often. However, at an increase of water content, the N−N interaction is more suppressed than that of the O−O pair. For the 6/94 (w/w) MEA/water (see dotted curves in Figure 3a,b), the maximum for the O−O pair has an amplitude of 1.07 versus 0.71 for the N−N pair. This suggests that the N− OW pair has a larger ability for HB formation than the N−N pair; at a large water fraction, this effect becomes even more distinct. For the MEA−water interactions (Figure 3e,f), the excess of the O−OW interactions over those of N−OW at large water fraction is less pronounced. The second feature of the RDFs (Figure 3a−c,e,f) is the twofold peak at an interval of r = 0.46−0.58 nm, which does not originate from the second solvation. In fact, these peaks

and the XH-bond vector of the donor of