DOCTORA L T H E S I S
Qingwei Gao Molecular Simulations of Microstructure, Thermodynamics, and Dynamics of Complex Liquid Mixtures at Interfaces and Confined Spaces
Department of Engineering Sciences and Mathematics Division of Energy Science
ISSN 1402-1544 ISBN 978-91-7790-683-4 (print)
ISBN 978-91-7790-684-1 (pdf) Luleå University of Technology 2020
Molecular Simulations of Microstructure, Thermodynamics, and Dynamics of Complex Liquid Mixtures at Interfaces
and Confined Spaces
Qingwei Gao
Energy Engineering
Molecular Simulations of Microstructure, Thermodynamics, and Dynamics of Complex Liquid Mixtures at Interfaces and
Confined Spaces
Qingwei Gao
Energy Engineering Division of Energy Science
Department of Engineering Sciences & Mathematics Luleå University of Technology
SE-971 87 Luleå, Sweden
ABSTRACT
In modern chemical engineering, process intensification is realised by introducing a nano- or micro-interface. The introduction of such an interface breaks the homogeneous nature of the fluid and forms a unique nano- or micro-interfacial structure, which has a significant impact on the properties of fluids. However, the existing traditional chemical engineering theories cannot be used to describe this inhomogeneous behaviour and clarify the underlying intrinsic mechanisms, making it difficult to find control factors which enhance chemical processes. It is necessary to establish theoretical models suitable for liquid–solid systems which can describe the fluid behaviour at interfaces. The key is to accurately recognise the structural as well as thermodynamic and dynamic properties of the complex fluid mixtures at these nano- or micro-interfaces. Previous studies that have been conducted to study the behaviour of simple fluids at interfaces found that, because of the strongly asymmetric interactions, the fluid tends to form layered structures at or close to the interface, which further affects the behaviour of the fluid molecules in close vicinity. However, for complex fluids and/or fluid mixtures, the effects of the interface and the interactions of fluid molecules on the formation of the layered structure and the fluid behaviour above the formed layer have not been elucidated.
In this thesis, to conduct a systematic study, several typical liquid mixtures, which are also important in the modern chemical industry, i.e. immiscible dimethyl carbonate (DMC)/water mixtures with relatively weak van der Waals interactions, miscible aqueous alcohol solutions with strong interactions due to hydrogen bonding (H-bonding), and deep eutectic solvents (DESs) with strong electrostatic interactions, were selected as representatives to construct a complex fluid–solid interface. Molecular dynamics simulation was used as the main tool to quantitatively describe the formation and influence of the adsorption layer on the structures and properties of the fluids at the interface at the molecular level. Additionally, in the future, key parameters will be provided for establishing theoretical nano- or micro-interface models. The main results obtained are summarised as follows:
(1) The local composition and microstructure of DMC/water mixtures in carbon nanotubes (CNTs) were studied. It is found that DMC molecules preferentially get adsorbed
water molecules in the “DMC tube” are more widely dispersed owing to the disordered orientation structure and the destroyed H-bonding structure.
(2) The molecular behaviour of aqueous solutions containing alcohol (i.e. methanol, ethanol, n-propanol, and n-butanol, respectively), confined in a two-dimensional graphene slit, was studied. A distinctly layered structure is formed at the interface, and the alcohol molecules are preferentially adsorbed on the graphene wall. Among the four studied systems, for the n-propanol system, the water molecules on the interfacial adsorption layer have the longest residence time because of the least distortion to the H-bonding network of the water molecules, restricting their motion. The nanophase separation (e.g. separation from water at the interface) of aqueous methanol solutions with stronger intermolecular interactions is less prominent compared to that of DMC/water and other aqueous alcohol solutions.
(3) The wetting behaviour of the deep eutectic choline chloride (ChCl)/urea (1:2) droplets on the ionic model substrate was studied, in which the substrate continuously and linearly changed its state from hydrophobic to hydrophilic. Due to the strong electrostatic interactions between the anions and cations, the neutral urea molecules are stripped out and adsorbed on the interface, forming a stable “new interface”. The orientation and H-bonding structure between the urea molecules in the adsorption layer lead to a difference in hydrophilicity of the “new interface”, further affecting the wetting behaviour of the upper molecules.
A further comparison of the results for the different systems that were studied shows that for a weaker intermolecular interaction in the bulk phase, a clearer separation at the interface of the nanophase is observed. Therefore, for more complex systems (such as aqueous ionic liquids/DESs), it is essential to study their microscopic mechanism in the bulk phase before investigating their interfacial behaviour. Subsequently, the microstructure of ChCl/urea/water (i.e. one type of aqueous DES) was studied. The investigation shows that the hydration strength of chloride ions is higher than that of choline ions, indicating that the anions have a greater impact on the non-ideal behaviour of the mixture, which was further proven by analysing the interaction energy. In addition, the competition between the ion pairs and ionic hydration was suggested as the underlying mechanism for the non-ideal changes in the thermodynamic properties of complex systems with strong electrostatic interactions.
ACKNOWLEDGEMENTS
It is impossible to perform a Ph.D. without the help of others. I am grateful to all the people that have supported me in accomplishing this difficult task. More specifically, I would like to express my great gratitude to my supervisor, Prof. Xiaoyan Ji. During my study at LTU, she has shown me the way to be a good scientist: stay efficient and keep learning. I would like to recognize the invaluable assistance that she provided during my study. Without her persistent help, the completion of this thesis would not be possible. I would like to express my gratitude to the co-supervisor, Prof. Aatto Laaksonen from Stockholm University for his scientific ideas on the research and guidance on the academic and personal level. I would like to express my appreciation to the co-supervisor, Prof. Xiaohua Lu from Nanjing Tech University, for his support and suggestion.
I would like to acknowledge Prof. Yudan Zhu from Nanjing Tech University for her excellent guidance from the beginning of my Ph.D. study. I admire her knowledge of molecular simulation techniques and thermodynamics in general.
I am indebted to Chunyan Ma for her continued help in my life and scientific research.
During the preparation of this thesis, she has spent much time helping me to check the documents needed for my defense. I would like to thank my group members Prof. Yijun Shi, Prof. Liwen Mu, Jingjing Chen, Yunhao Sun, Dr. Yanrong Liu, and Nan Wang for always being there. Without your company and encouragement, I would not have completed my Ph.D. I also would like to thank my colleague and friend Aekjuthon Phounglamcheik and Thamali Rajika Jayawickrama for sharing the working time and spare time with me. I will always remember the happy “fika” and skiing time with you.
I would like to thank Prof. Marcus Öhman for his support and creating such a wonderful working environment in the division. I would like to thank all the colleagues at the Division of Energy Science for the friendly environment and invaluable help. I would like to thank the colleagues and friends at Nanjing Tech University, China, for sharing the knowledge and the joyful moments.
Now I am here at the end of my Ph.D. With the end of this stage, new stages will start.
I am continuing my path with this belief that I have the support of my family and friends to
LIST OF PUBLICATIONS Publications included in the thesis
Paper Ⅰ
Effect of Surface-induced Dimethyl Carbonate Molecules on the Behavior of Water Confined in Carbon Nanotube
Qingwei Gao, Yumeng Zhang, Aatto Laaksonen, Yudan Zhu, Xiaoyan Ji, Shuangliang Zhao, Yaojia Chen, Xiaohua Lu
Submitted to Chinese Journal of Chemical Engineering, In revision
Paper Ⅱ
Effect of Adsorbed Alcohol Layers on the Behavior of Water Molecules Confined in a Graphene Nanoslit: A Molecular Dynamics Study
Qingwei Gao, Yudan Zhu, Yang Ruan, Yumeng Zhang, Wei Zhu, Xiaohua Lu, Linghong Lu Langmuir, 2017, 33, 11467-11474
Paper Ⅲ
Molecular Insight into Wetting Behavior of Deep Eutectic Solvent Droplets on Ionic Substrates: A Molecular Dynamics Study
Qingwei Gao, Nanhua Wu, Yao Qin, Aatto Laaksonen, Yudan Zhu, Xiaoyan Ji, Xiaohua Lu Journal of Molecular Liquids, 2020, 319, 114298
Paper Ⅳ
Effect of Water Concentration on the Microstructures of Choline Chloride/Urea (1:2)/Water Mixture
Qingwei Gao, Yudan Zhu, Xiaoyan Ji, Wei Zhu, Linghong Lu, Xiaohua Lu Fluid Phase Equilibria, 2018, 470, 134-139
Paper Ⅴ
Physicochemical Properties and Structure of Fluid at Nano-/Micro-interface: Progress in
Green Energy & Environment, 2020, Accepted, https://doi.org/10.1016/j.gee.2020.07.013
Contributions of the author Paper Ⅰ
Qingwei Gao was responsible for planning the work, carrying out the molecular simulations, doing the analysis, and writing the paper together with co-authors.
Paper Ⅱ
Qingwei Gao was responsible for planning the work, carrying out the molecular simulations, doing the analysis, and writing the paper together with co-authors.
Paper Ⅲ
Qingwei Gao was responsible for planning the work, carrying out the molecular simulations, doing the analysis, and writing the paper together with co-authors.
Paper Ⅳ
Qingwei Gao was responsible for planning the work, carrying out the molecular simulations, doing the analysis, and writing the paper together with co-authors.
Paper Ⅴ
Qingwei Gao was responsible for planning the work, reviewing the simulation and experimental studies of liquid–solid system, and writing the paper together with co-authors.
Publications not included in the thesis
Paper Ⅵ
Atomistic Insights into the Effects of Carbonyl Oxygens in Functionalized Graphene Nanopores on Ca2+/Na+ Sieving
Nana Zhao, Jiawei Deng, Yudan Zhu, Yaojia Chen, Yao Qin, Yang Ruan, Yumeng Zhang, Qingwei Gao, Xiaohua Lu
Carbon, 2020, 164, 305-316
Paper Ⅶ
Molecular Insights into the Microstructure of Ethanol/Water Binary Mixtures Confined within Typical 2D Nanoslits: The Role of the Adsorbed Layers Induced by Different Solid Surfaces
Yao Qin, Nana Zhao, Yudan Zhu, Yumeng Zhang, Qingwei Gao, Zhongyang Dai, Yajing You, Xiaohua Lu
Fluid Phase Equilibria, 2020, 509, 112452
Paper Ⅷ
Progress in Molecular-simulation-based Research on the Effects of Interface-induced Fluid Microstructures on Flow Resistance
Yumeng Zhang, Yudan Zhu, Anran Wang, Qingwei Gao, Yao Qin, Yaojia Chen, Xiaohua Lu
Chinese Journal of Chemical Engineering, 2019, 2(6), 1403-1415
Paper Ⅸ
Extra Low Friction Coefficient Caused by the Formation of a Solid-like Layer: A New Lubrication Mechanism Found through Molecular Simulation of the Lubrication of MoS2
Nanoslits
Jiahui Li, Yudan Zhu, Yumeng Zhang, Qingwei Gao, Wei Zhu, Xiaohua Lu, YijunShi Chinese Journal of Chemical Engineering, 2018, 26(12), 2412-2419
Paper Ⅹ
Molecular Dynamics Study of Mg2+/Li+ Separation via Biomimetic Graphene-Based Nanopores: The Role of Dehydration in Second Shell
Yang Ruan, Yudan Zhu, Yumeng Zhang, Qingwei Gao, Xiaohua Lu, Linghong Lu Langmuir, 2016, 32(51), 13778-13786
Paper ⅩI
Preliminary Study on Mechanism of Confined Mass Transfer and Separation: “Secondary Confinement” Effect of Interfacial Adsorption Layer
CONFERENCE CONTRIBUTIONS
The 4th International Conference on Molecular Simulation Shanghai, China, October 23-26, 2016
Poster
Pore Size Effect on Microstructures of Dimethyl Carbonate/Water Mixtures Confined in Nanochannels: A Molecular Dynamics Study
Qingwei Gao, Yudan Zhu, Yang Ruan, Xiaohua Lu, Yumeng Zhang, Jiahui Li
The 8th International Congress on Ionic Liquids (COIL-8) Beijing, China, May 13-17, 2019
Oral presentation with published abstract
Atomistic Insight into the Microstructures of Aqueous Ionic Liquids Qingwei Gao, Yudan zhu, Xiaoyan Ji, Xiaohua Lu
ABBREVIATIONS
AFM Atomic force microscopy
[BMIM][PF6] 1-butyl-3-methylimidazolium hexafluorophosphate
CNT Carbon nanotube
ChCl Choline chloride
[C10C1Im][NTf2] 1-decyl-3-methylimidazolium bis(trifluoromethanesulfonyl) imide
DES Deep eutectic solvent
DMC Dimethyl carbonate
DSC Differential scanning calorimetry
[EMIM][NTf2] 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide [EMIM][BF4] 1-ethyl-3-methylimidazolium tetrafluoroborate
H-bond Hydrogen bond
HBPT Hydrogen bond persistence time
in-situ STM In situ scanning tunnelling microscopy
IL Ionic liquid
LJ Lennard-Jones
MD Molecular Dynamics
NVE Microcanonical ensemble
NPT Isobaric-isothermal ensemble
NVT Canonical ensemble
[P6,6,6,14][BOB] Trihexyltetradecylphosphonium bis(oxalato)borate
RDF Radial distribution function
SFA Surface force apparatus
TABLE OF CONTENT
ABSTRACT ... 1
ACKNOWLEDGEMENTS ... 3
LIST OF PUBLICATIONS ... 5
Publications included in the thesis ... 5
Publications not included in the thesis ... 6
CONFERENCE CONTRIBUTIONS... 8
ABBREVIATIONS ... 9
1. INTRODUCTION ... 13
1.1 Background ... 13
1.2 Interfacial structure ... 14
1.3 Objectives ... 18
2. METHODOLOGY ... 21
2.1 Overview of computational chemistry and molecular modelling ... 21
2.2 Basic principles of classical molecular dynamics simulations ... 22
2.3 Ensembles used in the thesis ... 24
2.4 Analysis methods ... 24
3. RESULTS AND DISCUSSION ... 31
3.1 DMC/water mixture confined in CNTs ... 31
3.2 Aqueous alcohol solutions confined in a graphene nanoslit ... 38
3.3 Deep eutectic solvent droplets on ionic substrates ... 45
3.4 Choline chloride/urea (1:2)/water mixture ... 51
4. CONCLUSIONS ... 56
5. FUTURE DIRECTIONS ... 58
REFERENCES ... 59
1. INTRODUCTION
1.1 Background
In chemistry or chemical engineering, the efficient transformation of materials is often achieved based on the theories of momentum transport, heat (energy) transport, mass transport, and chemical reactions. Using the interaction between microparticles in the bulk fluid, the material and energy regulations in the traditional chemical engineering process can be successfully predicted [1, 2]. However, in modern chemical engineering, process intensification is achieved through interface introduction. Taking membrane separation as an example, it makes full use of the interface to achieve separation and considerably reduces energy usage, which makes it applicable for the next-generation separation in chemical engineering. In particular, in molecular membrane separation, which further extends to the nano/micro scale and involves a multi-phase process, the interface plays a critical role [3-7].
In fact, using a liquid–solid interface to enhance process performance is a common subject for the new generation of chemical engineering processes, and other typical examples include those of confined catalysis [8], crystal growth, nano lubrication [9, 10], electrochemistry [11, 12], colloidal systems, and membrane separation [13].
When introducing an interface, especially at the nano/micro scale, owing to the strong and asymmetric interactions between the fluid and solid surface, the properties of fluids are completely different from those in the bulk phase, resulting in innovative process performance. Studies have been extensively conducted to investigate the interfacial properties, from which abnormal phenomena have been observed. For instance, Fujimori et al. [14] observed that sulphur formed one-dimensional chains and exhibited metal phase transition at the atmospheric pressure in a 0.6 nm CNT using the advanced techniques such as high-resolution transmission electron microscopy and synchrotron X-ray diffraction. In comparison, bulk sulphur requires an extremely high pressure (490 GPa) to become metallic.
Similar interface-induced phase transition processes have also been found in many other studies, such as KI crystals confined in single-walled carbon nanohorns [15], organic liquid octamethylcyclotetrasiloxane [16] and toluene [17] confined between two mica surfaces. By designing the catalyst interface, Jiao et al. [18] converted synthesis gas to light olefins with surprisingly high selectivity, breaking the theoretical limit of the Fischer-Tropsch synthesis.
Robeson upper bound [19]. In our previous work [20, 21], we also found an abnormal phenomenon in CO2 absorption with an ionic liquid (IL) supported on the substrate. The CO2
absorption capacity increased with decreasing IL thickness, and it could be even 10 times higher in the bulk phase.
Using liquid–solid interfaces to enhance process performance is a unique advantageous of modern chemical engineering. The introduction of an interface leads to unusual properties and extraordinary phenomena which facilitates super-high process performance, making it essential to focus on the fluid–solid interface. In the conventional process, the fluid behaviour is homogeneous and isotropic, and the properties depend on temperature, pressure, composition, etc. However, the introduction of an interface breaks the nature of the homogeneous fluid itself and forms a unique interfacial structure, which significantly affects the fluid properties and thus process performance. Unravelling the molecular structure at the liquid–solid interface is, therefore, of great significance to understand the physical phenomena in interfacial systems [22], while investigations of such systems in practical applications remain challenging. In advanced chemical engineering, the studied systems are complex and far from the “ideal” or “model” case. Many factors affect the molecular structure of the fluid on the interface, such as miscibility or ability to form a eutectic mixture, and forces varying from van der Waals, hydrogen-bonding (H-bonding), and electrostatic forces are at play. Moreover, the solid material presents sophisticated structures and properties such as roughness, electrostatic effects, and chemical heterogeneities.
1.2 Interfacial structure
The interfacial structure can be investigated both experimentally and theoretically. In general, it is a challenge to obtain a nano/microscopic structure for fluids at the solid surface through conventional experimental measurements. In recent years, many advanced experimental techniques have been developed, making it possible to experimentally study the interfacial structure of fluids. The most common analysis techniques are differential scanning calorimetry, surface force apparatus, atomic force microscopy (AFM), in situ Raman spectroscopy [23, 24], and in situ scanning tunnelling microscopy (in situ STM). The research progress has been summarised in the review articles [25-28], and several important systems are briefly introduced below.
Ma et al. [29] combined a qPlus-based non-contact AFM/STM system to investigate the interfacial structure of water on Au (111) surface at the atomic level (see Figure 1.1). A 2D
approximately 2.5 Å. Within each half-layer, three high-lying oxygen atoms are surrounded by either a lower contrast oxygen atom or the core of the hexamer, which can be viewed as a stand-alone 2D crystal ice. Bampoulis et al. [30] studied the structure and dynamics of four aqueous alcohol solutions confined between graphene and mica surfaces at the molecular level using AFM and revealed that there is a nanophase separation between alcohol and water. The alcohol molecules prefer to adsorb on top of the ice layers on mica in direct contact with the graphene surface, resulting in alcohol-rich islands. Griffin et al. [31]
measured the density distribution of [C10C1Im][NTf2] at the mica surface using a surface force balance. Layered distributions are observed at the interface, and the density oscillates in a period of 23 Å. Similar work also shows the distribution of ILs at the interfaces. For instance, [EMIM][NTf2] at a single-crystal gold electrode [32], [EMIM][NTf2] and [EMIM][BF4] at the mica surface [33].
Figure 1.1 AFM characterisation of the 2D bilayer ice on Au (110) surface. Reproduced from Ref.
[29] with permission from 2020 Nature.
In addition to the experimental characterisation, molecular simulations are also effective approaches to investigate structural changes, in which statistical mechanics-based theories are used to reproduce the behaviour of molecules. Material and chemical scientists usually perform molecular simulations to gain a molecular-level understanding of materials for better designs and applications of nanomaterials. To achieve molecular understanding of the anomalous interfacial phenomenon involved in modern chemical engineering, molecular simulation has become an indispensable method to explore the structure of fluid at the interface [34, 35]. Previous intensively conducted studies have been summarised in these
layer structure will affect the behaviour of the upper fluid molecules. Below are some important studies that are highlighted as examples.
Wang et al. [38] chose an ionic model substrate as the solid surface to study the wetting behaviour of water droplets. The opposite charges were symmetrically distributed on the substrate surface, making the overall surface charge-neutral. In their work, several charge values (0–1.0 e) were chosen, transforming the substrate gradually from hydrophobic to hydrophilic. A stable interfacial adsorption water layer was observed in all cases, explaining the unexpected phenomenon of “water does not wet a water monolayer” at room temperature.
Detailed structural analysis revealed that a considerable number of dangling OH bonds remains in the first adsorbed layer, further affecting the H-bonding structure between water molecules beyond the first layer. There is no room for the water molecules beyond the first layer to form a hydrogen bond (H-bond) with those in the first layer. This greatly reduces the interaction between water molecules in the first layer and those beyond the first layer, resulting in the water monolayer being unfavourable for further growth of interfacial ice.
Phan et al. [39] studied the structure as well as the adsorption behaviour and dynamics of water-ethanol mixtures confined in alumina pores. Due to stronger interaction between the water molecules and the alumina surface, water is always preferentially adsorbed in the alumina nanopores, regardless of the water concentration. Analysis of the residence autocorrelation function and mean square displacement suggests that water diffuses faster than ethanol in the narrow pores, making alumina a promising material for the permselective membranes. They also found that the CH2-CH3 bond of the ethanol molecule is preferentially parallel to the surface, and the OH group prefers to form H-bonds with the surface, which may also affect the mobility of the molecules. The interfacial structure of [BMIM][PF6] in a 2.8-nm-wide graphene slit was systematically investigated by Dai et al. [40]. According to the density distribution, ILs were divided into three distinct regions, compact layer (com- layer), subcompact layer (sub-layer), and central layer (cen-layer). Analysing of orientational order parameter distribution for the imidazolium rings shows that in the com-layer region, the imidazolium rings of ILs prefer “parallel” and “tilted standing” orientations. The strong π-π interaction between the imidazolium ring and graphene surface plays a dominant role in this orientational structure and enhances the layering structure [41, 42]. In the sub-layer and cen-layer regions, a part of the [BMIM] imidazolium ring has a preferred “tilted standing”
orientation. They also found that the ILs in the com-layer have longer H-bond lifetimes and cation-anion pair dissociation times compared to the other two layered regions, which means
In summary, the understanding of fluid structure at the liquid–solid interface has advanced tremendously over the past years owing to the routine availabilities of real-time, high-resolution imaging techniques, and theoretical research. However:
1. Advanced experimental techniques can only be used to explore the distribution and properties of the first adsorption layer molecules.
2. Most of the theoretical work (such as molecular simulations) is performed on pure systems or mixtures, but only on the structure of the first layer or nanophase separation phenomenon at the interface.
3. The interfacial adsorption layer may play an important role in the properties of the upper fluid (Figure 1.2). The hypothesis proposed in this thesis is based on the analysis of the available work, which has not been previously highlighted.
Figure 1.2 Schematic representation of the physical model extracted from three systems (water, aqueous alcohol, and IL) at the interface. The adsorbed molecular layer can be regarded as a new
interface, which may affect the structure and properties of other fluid molecules.
However, owing to the complexity of both the solid surface and fluid molecules in current systems, the understanding of the interfacial interaction mechanism for complex fluids and fluid mixtures is still inadequate. Further quantitative and systematic investigation of the influence of the nano- or micro-interface on the structure and properties of complex fluids and fluid mixtures, with the focus on the role of the interfacial adsorption layer, is essential. Because the current experimental techniques cannot fulfil these requirements, molecular simulation will be a highly effective method to complement current techniques
1.3 Objectives
The overall objective of this thesis was to perform a systematic study to clarify how the interfacial adsorption layer affects the structure and properties of the upper fluids in complex systems, determining the properties and process performance (Figure 1.3). Molecular simulations were used, and three liquid mixtures ((1) immiscible dimethyl carbonate (DMC)/water with van der Waals interactions; (2) miscible aqueous alcohols with H-bonding;
(3) deep eutectic solvent (DES), that is, solvent forming from two salts with electrostatic interactions) involved in the advanced chemical industry, were selected as representatives in constructing the complex fluid–solid systems. The objectives were as follows:
1. To evaluate the contribution of the interfacial adsorption layer in immiscible mixtures with weak interactions in a DMC/water mixture under different confinements.
2. To study the influence of the molecular layer structure at the interface on the upper molecules in miscible mixtures with H-bonding.
3. To investigate the effect of the interfacial adsorption layer at the interface with varying hydrophilicity on the upper molecules in the DES system with electrostatic interactions.
4. To explore the feature structure of choline chloride (ChCl)/urea/water in the bulk phase, which is the premise of extending aqueous solutions with novel liquid materials (ILs, DESs) to the interface.
The overall research ideas, work, and goals are summarised in Figure 1.3. In view of the unique phenomena of the interface intensification process in chemical engineering, we studied complex liquid–solid interfaces at the nano/micro scale. Chapter 1 begins with a brief introduction to the background of the topic, the significance, and the relevant research status;
the research objectives were also introduced. In Chapter 2, the methods used in this thesis are described. The results of the four studied systems are presented in Chapter 3. Finally, the main conclusions are summarised in Chapter 4, and future work is stated in Chapter 5.
Figure 1.3 The framework of this thesis.
2. METHODOLOGY
2.1 Overview of computational chemistry and molecular modelling
Figure 2.1 summarises the theoretical calculation methods for different time and length scales. It includes (1) ab initio methods based on quantum mechanics at the electronic scale, and (2) semi-empirical methods which fall between the fully electronic and atomistic levels of description. (3) Molecular simulation methods based on molecular mechanics at atomistic and molecular scales, (4) mesoscale simulation based on coarse-graining methods, and (5) a continuum simulation method based on macroscopic conservation laws.
Figure 2.1 Simulation methods with different time scales and length scales [43].
Over the past two decades, our ability to model physical and chemical processes at the atomic level has increased rapidly, mainly due to the following reasons: the improvement of theoretical methods based on quantum mechanics or statistical mechanics (such as ab initio and classical density functional theory), improvement of the intermolecular force field, rapid growth of computer speed and memory, and more efficient algorithms. Unlike the electronic or ab initio scale, protons and electrons are not explicitly modelled in classical molecular simulations, and atoms or molecules interact with each other through a force field or intermolecular potential energy, which can greatly reduce the computing demands. Both mesoscale and continuum simulation methods will lose most of the atomic or molecular scale information, which is very important for studying the behaviour and properties of the
equilibrium, as no dynamics are considered. MD has the great advantage that the dynamic behaviour is readily calculated, making it possible to directly observe diffusive, convective, and other modes of motion at the molecular level.
It is well known that the experimental phenomena are the result of interactions of complex structures and forces at different scales, and it is difficult to decouple the factors and accurately analyse the microscopic phenomena by simple experimental characterisation.
MD simulations have been proposed as a powerful tool to study the microstructural changes of fluids at the molecular scale and have attracted increasing attention from industry and academia. Nowadays, MD simulations have penetrated the research scope of many disciplines (e.g. physics, mechanics, and biology) and have become an indispensable tool [34, 35]. In this thesis, MD simulations were chosen as the main methods for conducting the research.
2.2 Basic principles of classical molecular dynamics simulations
The molecular model is composed of empirical functions or force fields, which are used to describe the interaction between molecules (U) [44, 45]. The force fields define the ways of assigning functional forms and parameters to describe inter- and intra-molecular interactions. We can conceptually divide the force field into two parts: “Bonded” and “Non- bonded”:
UtotalEbondedEnon-bonded (2.1) 2.2.1 Non-bonded contribution (Enon-bonded):
The “non-bonded” part includes van der Waals and electrostatic interactions. The van der Waals interaction function has many forms, from simple quadratic forms to Morse functions, Lennard-Jones (LJ) potentials, and others. In this thesis, the LJ 12-6 function combined with the Coulomb term was used to describe the non-bonded interactions (Equation 2.2).
12 6
( ) 4
4
ij ij i j
ij ij
ij ij r ij
U r q q
r r r
(2.2)
where 𝑟 is the distance between atoms i and j, 𝑞 and 𝑞 are the charges of atoms i and j, respectively, 𝜀 is the dielectric constant, 𝜀 and 𝜎 are the energy and size parameters, respectively, between atoms i and j.
2.2.2 Bonded contributions (Ebonded):
Bonded interactions are not only exclusive pair interactions but also include 3- and 4- body interactions. There are bond stretching (2-body), angle bending (3-body), and dihedral angle (4-body, including improper torsion) interactions. In this thesis, the bond stretching between two covalently bonded atoms i and j (Ubond-stretch) is represented by a harmonic potential:
bond-stretch B 2 1-2 pairs
K ( ij eq)
U
b b (2.3) where 𝑏 is the bond length between atoms i and j, 𝑏 is the equilibrium bond length, and K is the force constant.The bond-angle vibration between a triplet of atoms i - j - k (Uangle-bend) was also represented by a harmonic potential on the angle 𝜃 :
angle-bend θ 2
angles
K ( ijk eq)
U
(2.4) where 𝜃 is the equilibrium bond angle, and K is the force constant.Improper dihedrals can be used to keep planar groups (e.g. aromatic rings) planar, or to prevent molecules from flipping over to their mirror images. The harmonic potential was used to describe the improper torsion angle motion (Uimproper):
2
improper ω
impropers
K ( eq)
U
(2.5) where is the improper dihedral angle, 𝜔 is the equilibrium improper dihedral angle, and K is the force constant.The normal dihedral interaction (Udihedral) can be described by the following potential, which is also used in the OPLS force field:
dihedral
1 2 3 4
1 K (1 cos( )) K (1 cos(2 )) K (1 cos(3 )) K (1 cos(4 ))
U 2 (2.6)
where K , K , K , and K are the constants, and is the dihedral angle.
After defining and building the system, the distances between particles and the forces acting on each atom can be calculated based on the initial coordinates and velocities that are selected. Thereafter, the coordinates and velocities can be determined by the equations of motion. The trajectory is the coordinates and velocities for a complete dynamic run. After
2.3 Ensembles used in the thesis
In a conventional MD simulation, the number of particles N, volume V, and total energy E are constants of motion. Thus, MD simulations measure the (time) averages in an ensemble that is very similar to the microcanonical, namely, the NVE ensemble. Currently, it is common to carry out MD simulations in ensembles other than NVE in practise, for example, isobaric-isothermal (NPT) and canonical (NVT) ensembles. The development of different ensembles is because most experiments are carried out under controlled pressure and temperature conditions. In the NPT ensemble, the number of particles N, pressure P, and temperature T are kept constant, and it is used initially to obtain a reasonable density.
Production simulations are usually performed in the NVT ensemble with the volume taken from the NPT simulations.
In this thesis, simulations were performed using an isobaric-isothermal (NPT) ensemble or canonical (NVT) ensemble, wherein the Nosé–Hoover thermostat was selected to control the temperature in the NVT, while the Berendsen pressure coupling was used to regulate the pressure of the simulation system in NPT. More specifically, the simulations of DMC/water were performed in the NVT ensemble, wherein the temperature was controlled by the velocity-rescaling thermostat; the Nosé-Hoover thermostat was used in the aqueous alcohol and ChCl/urea (1:2) interface systems. For the bulk ChCl/urea (1:2)/water, the velocity-rescaling thermostat was selected to control the temperature, and the Berendsen pressure coupling was used to regulate the pressure of the simulation system in the NPT ensemble.
2.4 Analysis methods
Both static and dynamic properties can be analysed based on the trajectories produced from MD simulations, such as the local composition, phase behaviour, and rheology of liquids. In addition, microstructures, such as H-bonds and orientation, can also be obtained by time-averaging, which is equal to ensemble averaging if the simulation is performed long enough for all the phase space to be visited (ergodic hypothesis). The analysis methods used in this thesis are as follows.
2.4.1 Radial distribution function
In statistical mechanics, the radial distribution function (RDF) (also called pair correlation function) is used to describe how the particle density changes as a function of the
distance from the reference atom in multi-particle systems (atoms, molecules...), and the RDF 𝑔 𝑟 between particles of A and B is defined in the following way:
AB
( ) B( )
B local
g r r
(2.7) where B( )r is the particle density of type B at a distance r around particle A, and
B local
refers to the bulk-mean density of type B.
2.4.2 Hydrogen bond
The microstructure of H-bonding networks is considered as an important and effective indicator for aqueous solutions [46, 47]. In this thesis, the geometric criteria [48] (Figure 2.2) was adopted to determine the formation of H-bonds among water, alcohol, and urea molecules. The three criteria that were considered to determine whether two molecules (one acts as a donor and the other as an acceptor) form a H-bond are listed below.
(1) The distance between the oxygen atom of an acceptor molecule and that of a donor molecule (ROO in Figure 2.2) must be shorter than 3.5 Å.
(2) The distance between the oxygen atom of an acceptor and the hydrogen atom of a donor (ROH in Figure 2.2) must be shorter than 2.5 Å.
(3) The angle of H-O···O (the first two atoms H-O belong to a donor molecule and the third oxygen atom belongs to an acceptor molecule) ( in Figure 2.2) must be less than a threshold value of 30°.
A H-bond can be formed between the two molecules only if the three criteria are met simultaneously.
Figure 2.2 Geometric criteria of H-bond formation.
2.4.3 Hydrogen bonding network of water molecules
Distribution of H-bonds per water molecule can provide insight into the microstructures of water molecules. It was also proposed as the common method used for analysing the microstructure of H-bonding networks in this thesis [49]. Distribution of different H-bond microstructures can also suggest different mobility of water molecules [50, 51].
There are four types of H-bonding networks for water molecules, that is, F1, F2, F3, and F4, where Fn denotes a different number of water molecules that form the n number of H- bonds with the surrounding water molecules. Figure 2.3 illustrates the characteristic microstructures of H-bonding networks for F1, F2, F3, and F4, respectively, in which the H- bond between the water molecules was determined according to the geometrical criteria described in section 2.4.2. If most of the water molecules exist in the form of F1, it means that the mobility of water molecules is relatively free and less affected by the surrounding water. On the contrary, if F3 and F4 are the majority, the water molecules are bound strongly, even as strong as that of ice, which therefore results in a poor mobility.
Figure 2.3 Schematic diagram of the H-bonding network.
2.4.4 Hydration number of ions and ion pairs
In order to investigate the microstructural changes in the ChCl/urea (1:2)/water mixtures in the bulk phase, the average hydration numbers of cations and anions were analysed based on the RDFs. Specifically, the average hydration numbers were obtained by integrating the corresponding RDF over a radius range spanning from the onset of the primary correlation peak up to the first minimum. For instance, if the distance between the chloride anion and the oxygen atom of a water molecule is shorter than the first minimum value of the anion- water RDF curve, this water molecule is called the hydrated molecule with respect to the chloride anion. A similar definition was used by Hammond et al. to study the liquid structure of pure ChCl/urea [52].
structure than that of the chloride anion, it is difficult to accurately obtain the hydration number based on the RDF between the centre of mass for choline and the oxygen atom of water molecules. To solve this problem, the RDFs of each site in the choline cation were calculated first. Subsequently, if a water molecule is within any site of the hydration shell, the water molecule is considered as the hydrated molecule of the choline cation, rather than simply adding the number of water molecules in each hydration shell. Based on this modified method, the overlapped water molecules can be eliminated (Figure 2.4), and the hydration number of the choline cation can thus be obtained accurately.
Figure 2.4 Illustration of choline cation hydration shell.
In addition to the hydrated ions, ion pairs can also be found in many aqueous solutions, such as aqueous ChCl/urea (1:2) studied in this thesis. Ion-pairing plays a crucial role in the microstructure and properties of systems in contact with water [53]. In this thesis, the number of ion pairs was calculated based on the cation-anion RDFs. If the distance between a pair of anions and cations is shorter than the position of the first peak valley of the RDF, it is considered that an ion pair is formed.
2.4.5 Residence time
Residence time is extensively used to reflect the mobility of molecules in a particular region in simulations, which can be obtained by integrating the residence correlation function [54-56]. The microstructure of the water molecules on the preferential adsorption layer of alcohol molecules is one of the focuses of this thesis. In general, the density profiles of the water molecules confined in the two adsorbed alcohol layers are axisymmetric (Figure 2.5).
density profile of the water molecules.
Figure 2.5 Snapshots of equilibrium configurations (side views) for aqueous propanol in contact with the graphene slit.
The normalised residence autocorrelation function (CR(t)) is calculated as
𝐶 𝑡 〈〈 〉〉 (2.8) where Ow(t) represents the state of the oxygen atom of a water molecule at time t (ns) and Ow(0) is the state at time origin (t = 0). Ow(t) = 1 represents an atom that belongs to a particular layer; otherwise, Ow(t) = 0. The angular brackets denote ensemble averages. CR(t) decays from 1.0 to 0.0 with time evolution, indicating that the water molecules move into and out of the particular layer.
According to Equation 2.9, the mean residence time () was calculated to quantitatively assess the mobility of the water molecules on the preferential adsorption layer by integrating the values of the residence autocorrelation functions:
𝜏 𝐶 𝑡 𝑑𝑡 (2.9) 2.4.6 Interaction Energy
To understand the evolution of the microstructure, the intermolecular interaction energies among different species were analysed. This method has been widely used in the qualitative comparison of liquid–solid interactions [57, 58]. In this thesis, the interaction energy (Einter) was defined including both electrostatic and van der Waals interactions, which were modelled with Coulomb and LJ potential functions, respectively.
2.4.7 Calculation of the contact angle
For the ChCl/urea (1:2) interface systems, the contact angles of the droplets on the ionic
wettability. Following steps were used to obtain the contact angles,
Step 1, the boundary density counter for a given droplet was determined using the method reported by Ghalami [59] and Giovambattista [60], that is, it was determined at the position where the average density over the entire production trajectory was 0.2 g·cm−3.
Step 2, the points on the boundary were fitted into a curve using an iterative least- squares estimation.
Step 3, the contact angle of the droplet was calculated with Equation 2.10.
cos 𝛽 1 (2.10) where h and R are the geometry parameters of the droplet, h refers to the height of the droplet, and R is the radius of the fitted circle. A schematic diagram is shown in Figure 2.6.
Figure 2.6 Schematic diagram for calculating the contact angle.
3. RESULTS AND DISCUSSION
To systematically study the microstructures of fluids at the interface to clarify how the interfacial adsorption layer affects the structure and properties of the upper fluids, three liquid mixtures were selected as representatives, considering the interaction strength. They are immiscible DMC/water, with relatively weak van der Waals interactions, miscible aqueous alcohol solutions with H-bonding (methanol, ethanol, propanol, and butanol with different polarities), and ChCl/urea with electrostatic interaction (a DES formed from two salts). These specific fluids were chosen also because of their importance in practical applications, as described below in each subsection. In addition, aqueous ChCl/urea in the bulk phase was studied to explore the microstructure of the system before exploring its interfacial behaviour.
3.1 DMC/water mixture confined in CNTs
DMC is an important and environmentally friendly chemical [61] with many advantages, such as low haze point, low toxicity, and fast biodegradability. In recent years, its use has increased rapidly as a carbonylation agent to replace lethal phosgene and prepare polycarbonate and carbamate polymers, as well as a substitute for dimethyl sulphate and methyl halide in methylation reactions. In addition, many studies have been conducted to explore the applications of DMC in other fields, such as a solvent in lithium-ion batteries [62, 63] and oxidants in internal combustion engine fuel [64]. Because trace water in DMC will have a great impact on its performance, especially in the application of fuel oil and batteries [65, 66], DMC needs to be further purified after common separation. Due to the high energy demand with traditional separation methods, the molecular separation membrane has been proposed as one of the most promising technologies [3, 67-69], among which laminated graphene and CNTs have been widely developed as membrane materials [70-77]. However, the molecular mechanisms under nanoconfinement are still unclear, which leads to the preparation of molecular separation membranes based on our experience with intensive experimental explorations and lack of theoretical guidance.
In this context, MD simulation was used to study the behaviour of DMC/water mixtures confined in CNTs with various chiral parameters ((14,14), (16,16), (18,18), (20,20)), which were named (14,14)_DMC, (16,16)_DMC, (18,18)_DMC, and (20,20)_DMC, respectively.
We mainly focussed on exploring the effects of the adsorption layer on the microstructure,
The distribution and microstructure of the H-bonding network in the system were further analysed and compared, deepening the molecular understanding of the fluid behaviour and properties of the DMC/water mixture at the nano- or micro-interface.
3.1.1 Adsorption DMC layer in CNTs
The distribution of molecules plays an important role in the understanding of the behaviour of complex mixtures at the nano- or micro-interface, and thus, the two- dimensional density distributions of DMC and water molecules in the X-Y plane were analysed. In the analysis, the carbonyl oxygen atoms of DMC were selected as the characteristic points, and the water molecules were represented by their oxygen atoms.As shown in Figure 3.1, nanophase separation of the mixture occurs under all four different nanoconfinements. DMC molecules always preferentially adsorb on the inner wall of CNT, forming a special structure, the so-called “DMC tube”, and the corresponding “diameters”
are around 1.060, 1.316, 1.578, and 1.752 nm, respectively.
It is also found that, except for (20,20)_DMC, there are light-grey ring areas in the other three types of tubes, and these areas gradually disappear with the increase in diameter, which indicates that in addition to the adsorption layer, there is a small amount of DMC molecules in the central region of the tube. This thesis only focussed on the influence of the adsorption layer on the distribution and microstructure of inner water molecules; thus, the trace amount of free DMC molecules was ignored for further analysis.
Figure 3.1 (a) Sectional view of the equilibrium configuration. (b) X-Y planar number density distributions of DMC and water molecules confined in the CNTs with different diameters.
3.1.2 Distributions of water molecules confined in “DMC tubes” and CNTs
To investigate the difference between the effect of the “DMC tube” on the inner water molecules and that of CNTs in a pure water system, the behaviour of the water molecules in the four plain CNTs was studied, where the diameter of each plain CNT is equivalent to that of the “DMC tubes”, i.e. (8,8), (10,10), (12,12), and (14,14). These four cases were represented by the labels (8,8)_CNT_H2O, (10,10)_CNT_H2O, (12,12)_CNT_H2O, and (14,14)_CNT_H2O. The two-dimensional density distributions of the water molecules in the X-Y plane in the “DMC tube” and CNT with a similar diameter were analysed and compared to qualitatively understand the effect of the preferential adsorption of the DMC layer on the behaviour and properties of the water molecules (Figure 3.2). The water molecules are stratified in the “DMC tube” due to the influence of the adsorption DMC layer, from one layer of water molecules in (14,14)_DMC_H2O to three layers of water molecules in (20,20)_DMC_H2O. The inner water layer regions in the “DMC tube” are also bright, indicating that the distribution of water molecules is more uniform and dispersed. Unlike in the “DMC tube”, there is only one particularly bright layer near the surface when the water molecules are confined in the CNTs. Except for the case of (8,8)_CNT_H2O with only one layer, the water molecules in other CNTs can be divided into two regions: the boundary adsorption region and the central dispersion region. The lightness of the outermost water layer in plain CNTs is brighter than in other regions, indicating the concentrated distribution of water molecules. The comparison clearly reveals that the adsorption DMC layer can be considered as a new interface for the water molecules, that is, the adsorption DMC layer results in a secondary confinement of the inner water molecules. The adsorption DMC layer shows a more notable effect on the local compositions of the water molecules compared to the CNT itself, and the water molecules inside the “DMC tube” are more uniform and dispersed.
Figure 3.2 X-Y planar number density distributions of the water molecules confined in the CNTs with different diameters.
3.1.3 Detailed microstructural analysis of water molecules
To further reveal the influence of the DMC layer on the water molecules at the molecular level, a detailed microstructure analysis was carried out. The characteristic orientation distributions of the water molecules in the “DMC tube” and CNTs were analysed and compared (Figure 3.3). It was found that a characteristic angle θ was formed by the dipole moment of water molecules and the axial direction of the CNTs (Figure 3.3(a)), and the narrower the distribution of θ, the more ordered the orientation structure. It is found that the angular distribution of the water molecules in (8,8)_CNT_H2O has two characteristic peaks at 44° and 139°, respectively. In the case of (14,14)_CNT_H2O, with the increase in diameter, the characteristic peaks become gradually weakened and the orientation structure is more uniform, which indicates that the water molecules become disordered and closer to the water molecules in the bulk phase. Similar phenomena have been found in our previous work simulating water molecules in CNTs with different chiral parameters [78]. In contrast, for the
“DMC tubes”, except in the case of (14,14)_DMC_H2O which shows a wide characteristic peak at 100°, no obvious characteristic peak is observed for other cases with larger pore sizes, that is, the orientation distribution of the water molecules becomes more disordered in the
“DMC tubes”.
Figure 3.3 (a) Definition of angle θ for the dipole moment of the water molecule and the axial direction of CNT; Distributions of angle θ of the water molecules confined in CNT (b) and “DMC tube” (c) with
different diameters.
After a preliminary understanding of how the “DMC tube” affects the orientation of the water molecules, the work was continued to analyse the average number of H-bonds between the confined water molecules in CNT and DMC. The results are shown in Figure 3.4, where the dotted line represents the average number of H-bonds between the water molecules in the bulk phase. The diameters of CNTs in Figure 3.4(a) are similar to those of the “DMC tubes”
in Figure 3.4(b). It can be seen that, although the pore sizes are almost the same for these two systems, the number of H-bonds between the water molecules confined in the “DMC tubes” is significantly less than that in CNTs. The average H-bonds of the water molecules confined in the plain CNTs increase gradually and approach the bulk value, while those for the water molecules in the “DMC tube” increase rapidly from a very low value and finally reach the value close to that of the bulk phase. This may be due to the destruction of the water structure at the DMC interface. It is worth noting that the average number of H-bonds between water molecules in (14,14)_DMC_H2O is only 0.81. Therefore, it can be inferred that the water molecules may transfer in the form of molecular chains in the “DMC tube”
with a size of 1.060 nm [79]. This may be useful for designing membrane materials with high water permeability.
Figure 3.4 Average H-bonds between the water molecules confined in (a) CNT and (b) DMC with different diameters.
The H-bond distribution Fn of the confined water molecules was also calculated to gain a deeper understanding of the microstructure of water molecules in these two systems, as it has been reported that the mobility of water molecules is affected by the H-bond distribution [50, 51]. As described in the previous section, Fn represents the number of water molecules that form n-amount of H-bonds with the surrounding water molecules. For water in the bulk phase, the H-bond distribution is mainly concentrated in F3 (33.59%) and F4 (20.23%). As shown in Figure 3.5, the H-bonding network of the water molecules in CNTs mainly exists in the form of F3, which is close to that of the bulk phase. Moreover, increasing the diameter of CNTs has a slight effect on the H-bonding distribution.
For the water molecules confined in the “DMC tubes”, the proportion of F1 for the water molecules in (14,14)_DMC is as high as 69.92%, and the proportions of F3 and F4 are only 3.73 and 0.13%, respectively, further indicating and confirming the inference that the water molecules in this pore may transport in the form of molecular chains (Figure 2.3). In addition, with the increase in diameter, the proportions of F2, F3, and F4 in the system increased significantly and finally approached those for the bulk water molecules. All these observations are consistent with previous results obtained on the basis of H-bonding.
Figure 3.5 The percentage of water molecules confined in (a) CNT and (b) “DMC tubes” with n (n = 1, 2, 3, 4) H-bonds (Fn).
3.1.4 Summary
Based on the systematic study of the local compositions and microstructures of DMC/water mixture in CNTs with four diameters, it is found that DMC molecules preferentially adsorb on the inner surface of CNT and form a “DMC tube” structure, and a new secondary confinement structure is formed for the water molecules. The distribution, orientation structure, and H-bonding network of the water molecules confined in the “DMC tube” were compared with those in CNTs of similar size. Owing to the influence of the DMC adsorption layer, the distribution of the water molecules in the tube is more uniform and dispersed. In particular, when the mixture is confined in the (14,14)_CNT case, 69.92% of the water molecules have less than one H-bond, and the water molecules may transmit in the pulse-like form, similar to the water molecules in the plain (6,6)_CNT [79]. This may be important for designing high-performance membranes for DMC/water separation.
Based on these analyses, it acquired a basic understanding of the molecular mechanism for immiscible mixtures with relatively week interaction (DMC/water mixture) under nanoconfinement and identified the influence of the preferential adsorbed DMC molecules.
To further explore the influence of the interface on the miscible system and to gain a deeper understanding of the role of interfacial molecules, the interfacial behaviour of aqueous alcohol solutions was studied in the following subsection.
3.2 Aqueous alcohol solutions confined in a graphene nanoslit
The separation of the alcohol/water mixtures under confinement is significant in the fields of life science, medicine, and chemical engineering [80, 81], and other liquid-liquid separation scientific research. However, the highly efficient separation of alcohol/water is still a challenge, and understanding the microstructures of the mixtures is required to develop new technologies. Graphene is extensively studied as a membrane material for liquid-liquid separation because of its unique atomic-scale thickness and extraordinary performance, and it has shown a significant potential to achieve separation at the molecular scale [82-85].
Recent studies [86] found that when graphene is used as a membrane material, the ethanol molecules are preferentially adsorbed on the inner surface of the pore wall and form an adsorption ethanol layer under 2D nanoconfinement, affecting the interfacial friction of fluids and thus separation performance. However, the dynamic properties of intermediate water, which is particularly important for membrane separation, has not received sufficient attention. This raises the question to whether this adsorbed alcohol layer affects the structure and mobility of the water molecules that cover the alcohol layer.
To evaluate the effect of the alcohol adsorption layer on the structure and mobility of the water molecules, molecular simulations were performed to investigate four types of alcohol/water binary mixtures confined under a 20 Å graphene slit. An interlayer space of 20 Å was selected in the study because it was proven to facilitate the formation of the preferential alcohol adsorption layer, and it also ensured that every kind of alcohol molecule could form an adsorption layer for better comparison. In addition, it is a typical size for a nanoslit in graphene-based membranes and could be used in many applications (e.g. water purification, pharmaceutical industry, and fuel separation), where a high-performance separation of large molecules and small waste molecules is required [70, 86]. The chosen alcohols were methanol, ethanol, 1-propanol, and 1-butanol, and their mixtures with H2O were represented by the specific labels, namely, Me_W, Et_W, Pro_W, and Bu_W, respectively. The various polarities of these four alcohols lead to different interaction strengths between the alcohol and water molecules. Greater polarities of the alcohol molecule showed weaker intermolecular interaction in the system.
Firstly, density distributions of the molecules nanoconfined in the graphene slit were analysed. Then, the residence times of the water molecules covering the adsorbed alcohol layers were calculated to evaluate the effect of the interfacial adsorption layer on the
dynamics of upper water molecules. Finally, the detailed network microstructures of H-bonds were analysed to elucidate the underlying mechanisms at the molecular scale.
3.2.1 Density distributions
The density profiles of the alcohol molecules in the direction that is perpendicular to the graphene surface were analysed to quantitatively describe the preferred distribution of molecules in the confined slit. For each type of alcohol molecule, the density profiles of methyl and hydroxyl groups were analysed. As shown in Figure 3.6(a–d), two obvious peaks in both density profiles are concentrated near the slit walls for all the studied cases, indicating that all four types of alcohol molecules can form an almost discrete layer near the graphene wall, that is, nanophase separation occurs. The peak positions of the methyl groups are closer to the graphene walls than those of the hydroxyl groups. These observations reflect the preferential orientations of the alcohol molecules, that is, the methyl group faces the graphene wall, whereas the hydroxyl group is oriented towards the inner pore. Ren et al. [87] employed MD simulations to study the dehydration behaviour of ethanol/water mixtures confined in two hydrophobic plates and observed similar preferential orientations.
For all the alcohols studied, the distance between the methyl group and the graphene wall is 2.95 Å, which is closer to the surface compared with that between the hydroxyl group and the graphene wall. This indicates that the methyl group may show less of an effect on the water molecules. Meanwhile, the distance between the peaks of methyl and hydroxyl groups becomes longer as the length of the alcohol chain increases, which may result in a smaller confined space for the water molecules. The distances between the peaks of methyl groups and water molecules are 2.55, 3.30, 3.60, and 3.75 Å, respectively.
The spatial distribution of the confined water molecules was represented by the density distribution of the oxygen atom in the water molecules and their Y-Z planar density distribution. Although a certain amount of water molecules (see insets in Figure 3.6) also exist in the alcohol layer (first adsorption layer), especially for methanol, this phenomenon rapidly disappears to a negligible degree with the increasing length in the alcohol chain.
Therefore, the adsorption layer in the studied cases is almost pure alcohol, and the consideration of the water molecules in this first-layer adsorption can be neglected for the time being.
To evaluate the influence of the preferentially adsorbed alcohol layer on the behaviour of water molecules covering it, the water molecules that were distributed near the adsorbed alcohol layer were further analysed. As shown in the insets of Figure 3.6(a–d), the peaks of the water molecules are heterogeneously but symmetrically distributed, suggesting that the alcohol molecules have a confined effect on the water molecules. This can also be reflected by the 2D density maps, where the brighterregion represents the higher number density of the water molecules. In addition, based on the distribution width of water molecules, it was found that, with the increase in the alcohol chain length, the distribution area of water gradually decreases and the confined space becomes smaller.
Comparing the results of these four alcohol systems shows that no obvious peak of water molecules exists in the slit centre for Pro_W, while for the other three cases, there is always a peak. To further understand the influence of alcohols on upper water, in addition to the distribution, dynamic properties and microstructure analysis were conducted.
3.2.2 Residence time of water molecules on the adsorption alcohol layer
Residence time is extensively used to reflect the mobility of molecules in a particular region in simulations [54-56]. To investigate how the adsorption alcohol layer affects the mobility of the upper molecules, the residence time of the water layer above the adsorption alcohol layer was set as the research focus. According to the positions of the first two valleys
