Monireh Khosravi nasab Master of Science nasab_Thesis.… · Monireh Khosravi nasab Master of...

COMPUTATIONAL MODELLING OF ZEOLITE NION EXCHANGE PROPERTIES

Monireh Khosravi nasab Master of Science

Submitted in fulfilment of the requirements for the degree of

Doctor of Philosophy

School of Earth and Atmospheric Sciences

Science and Engineering Faculty

Queensland University of Technology

2020

Computational Modelling of Zeolite N Ion Exchange Properties i

Keywords

Zeolite N, ion exchange, ion selectivity, cation exchange, diffusion, molecular

modelling, molecular mechanics, molecular dynamics, quantum mechanics, density

functional theory, Forcite, DMol3, COMPASS force field, GGA, LDA, PBE, PW91,

geometry optimization, dispersion correction, basis sets, bond length, population

analysis, atomic charge, Mulliken partial charges, hydration energy, ammonium,

monovalent cation, alkaline elements, divalent cation, alkaline earth elements, explicit

water, concentration profile, self-diffusion coefficient, mean square displacement,

radial distribution function, concentration profile, zeolite A, clinoptilolite, heulandite,

natural Australian zeolite, mineralogical characterisation, XRD, XRF, SEM, EDS,

WDS, electron probe microanalysis, EPMA.


Abstract

Zeolites are porous materials with different crystalline structures, pore

topologies and chemical compositions. This diversity provides specific properties that

result in a wide range of industrial applications including oil refining, gas separation,

wastewater treatment and agriculture. Prediction of these properties can enable

economic estimations and enhance the highest performance of these materials in

different industries. Studies have shown that experiments cannot always provide

detailed atomic scale understanding of these mechanisms. However, computational

modelling and simulation techniques are effective tools that help explain the behaviour

of zeolites under different processing conditions at different scales.

Zeolite N is a synthetic zeolite of the EDI framework type with chemical formula

K12Al10Si10O40Cl2.8H2O and is one of the more than 200 known zeolite types; some of

which have been manufactured and used at commercial scale to selectively extract

ammonium ions from wastewaters. Previous experimental laboratory and field data

show that zeolite N has a high capacity for exchange of ions. In this study, the structure

and ion exchange behaviour of synthetic zeolite N in an aqueous environment is

investigated by applying computational techniques based on quantum mechanics and

molecular dynamics simulations.

In this study, we assess the effects of Local Density Approximation (LDA) and

Generalized Gradient Approximation (GGA) DFT models on zeolite structural

parameters and on partial atomic charges of framework atoms. We applied these

functionals with different quality of convergence and SCF tolerances, numerical basis

sets and dispersion correction schemes. Optimized zeolite N structures are evaluated

by comparing the atom positions and framework T-O bond lengths with experimental

data. The obtained Si-O and Al-O bond lengths of optimized structures in this study

are in agreement with previous experimental studies on zeolite N and computational

models of other zeolites. The values of partial atomic charges are sensitive to the

choice of numerical basis sets. Results show that the GGA-PBE functional with DNP-

4.4 basis set and TS dispersion correction scheme is a reliable DFT model to optimize

and establish the structural parameters of zeolite N for further MD simulations.

ii Computational Modelling of Zeolite N Ion Exchange Properties

Molecular dynamics (MD) simulations are used to investigate the hydration

energy and ion exchange properties of zeolite N. The exchange of K+ ions with

univalent ions NH4+, Li+, Na+, Rb+ and Cs+ as well as divalent cations Mg2+ and Ca2+

is investigated under a range of simulation conditions using a three-dimensional

membrane in an electrolyte box containing explicit water molecules. Hydration energy

calculations indicate that zeolite N prefers eight water molecules per cage, which is

consistent with X-ray and neutron diffraction determination of the structure. The

structural arrangements of ions inside zeolite N membranes are investigated by

concentration profiles, ion density maps and radial distributions of atom pairs.

Moreover, the mobilities of ions are estimated by calculating the self-diffusion co-

efficient from the mean square displacement of ions over simulation time. The results

show that the diffusion and exchange of cations are affected by shape and size of

channels controlling the ion exchange flow as well as the nature of cation, ionic size

and charge density. Moreover, the results indicate that the ion exchange by zeolite N

is selective towards NH4+ in preference to other mono- and divalent cations.

The outcomes of this investigation provide qualitative and quantitative insight

into exchange behaviour of zeolite N at atomic scale and the parameters that influence

ion exchange properties. Moreover, the methodology for studying the exchange

behaviour of zeolite N will provide a practical method to predict behaviour of other

zeolites, such as Australian natural zeolites and synthetic zeolites such as zeolites Y,

A and W.

Over the last 20 years, Australian natural zeolites have been investigated for use

in various industrial applications. However, there are few, if any, mineral

characterisation studies on Australian natural zeolites since the early 1990s that use

modern techniques. In this study, a detailed mineralogical analysis was conducted on

zeolite specimens from Avoca and Werris Creek deposits, located in Queensland and

New-South-Wales, respectively, in Australia. The chemical compositions of fine-

grained zeolites were estimated with high accuracy using accepted EPMA protocols

and data reduction methods. The Australian zeolites were identified as magnesium

heulandite-Ca and magnesium clinoptilolite-Ca. This investigation suggests that

Australian zeolites are good candidates for separation and/or exchange processes due

to their microporosity, high thermal stability and presence of Ca and Mg as dominant

extra-framework cations. The chemical composition obtained in this study can be used

Computational Modelling of Zeolite N Ion Exchange Properties iii

to develop models for further simulation studies on ion- exchange behaviour of

Australian zeolites.

This study also shows that combining experiment and computational modelling

can provide a precise understanding of both chemical and physical properties of

zeolites as well as other zeolitic materials. This spatially and compositionally precise

information enables high quality prediction of ion exchange behaviour under real

conditions.


List of Publications

Journal publications:

Khosravi, M., Murthy, V., Mackinnon, I. D. R., 2020, Evaluation of DFT

methods to calculate structure and partial atomic charges for zeolite N,

Computational Materials Science, 171, 109225.

Khosravi, M., Murthy, V., Mackinnon, I. D. R., 2019, The exchange

mechanism of alkaline and alkaline earth elements in zeolite N,

Molecules, 24(20), 3652 (journal cover).

Murthy, V.; Khosravi, M.; Mackinnon, I. D. R., 2018, Molecular

modeling of univalent cation exchange in zeolite N. The Journal of

Physical Chemistry C, 122(20), 10801-10810.

Khosravi, M., Cathey, H. E., Mackinnon, I. D. R., 2020, Detailed

mineralogical investigation of Australian natural zeolite using EPMA,

American Mineralogist, (under revision for re-submission).

Khosravi, M., Murthy, V., Mackinnon, I. D. R., 2020, Molecular

modelling on penetration of monovalent cations into zeolite N

membrane, The Journal of Physical Chemistry C. (in preparation).

Conference presentations:

Murthy, V.; Khosravi, M.; Mackinnon, I. D. R., 2019, Molecular

modelling on penetration of monovalent cations into zeolite N

membrane, 19th International Zeolite Conference, Perth, Australia (Oral

Presentation, Peer-reviewed).

Khosravi, M., Cathey, H. E., Mackinnon, I. D. R., 2019, Detailed

mineralogical investigation of Australian natural zeolite using EPMA,

19th International Zeolite Conference, Perth, Australia (Poster

Presentation, Peer-reviewed).

Khosravi, M., Murthy, V., Mackinnon, I. D. R., 2018, Modelling

hydration and ion exchange for zeolite N, International Conference on


Nanoscience and Nanotechnology (ICONN2018), Wollongong,

Australia, (Poster presentation, Peer-reviewed).

Cathey, H. E., Khosravi, M., Mackinnon, I. D. R., 2018, Towards higher

spatial resolution in quantification of zeolites by field emission electron

probe microanalysis (FE-EPMA), 19th International Microscopy

Congress, Sydney, Australia, (Poster Presentation).

Murthy, V.; Khosravi, M.; Mackinnon, I. D. R., Molecular modeling of

univalent cation exchange in zeolite N, 2017, Associatian of Molecular

Modelers of Australia, Margaret River, Australia (MM2017), (Oral

presentation).


Table of Contents

Keywords .................................................................................................................................. i

Abstract ..................................................................................................................................... i

List of Publications ................................................................................................................... i

Table of Contents ...................................................................................................................... i

List of Figures ......................................................................................................................... iii

List of Tables ......................................................................................................................... vii

List of Abbreviations .............................................................................................................. ix

Statement of Original Authorship ........................................................................................... xi

Acknowledgements ................................................................................................................ xii

Chapter 1: Introduction ...................................................................................... 1

1.1 Background .....................................................................................................................1

1.2 Context ............................................................................................................................3

1.3 Purposes ..........................................................................................................................4

1.4 Significance and Scope ...................................................................................................5

1.5 Thesis Outline .................................................................................................................5

1.6 Refrences ........................................................................................................................9

Chapter 2: Literature Review ........................................................................... 11

2.1 Computational Methods in Zeolite Science ..................................................................11

2.2 Computational Studies of Zeolite Ion-exchange ..........................................................12

2.3 Zeolite N .......................................................................................................................17

2.4 Summary and Implications ...........................................................................................21

2.5 Refrences ......................................................................................................................23

Chapter 3: Methodology .................................................................................... 27

3.1 Computational Chemistry Techniques .........................................................................27

3.2 Quantum Mechanics .....................................................................................................29

3.3 Molecular Mechanics....................................................................................................32

3.4 Refrences ......................................................................................................................41

Chapter 4: Modelling Hydration Behaviour of Zeolite N .............................. 45

4.1 Introduction ..................................................................................................................45

4.2 Methods ........................................................................................................................46

4.3 Results ..........................................................................................................................51

4.4 Discussion .....................................................................................................................57

4.5 Conclusion ....................................................................................................................65

4.6 Refrences ......................................................................................................................67


Chapter 5: Evaluation of DFT Methods to Calculate Structure and Partial Atomic Charges for Zeolite N ................................................................................. 70

5.1 Introduction .................................................................................................................. 70

5.2 Computational and theoretical methods ....................................................................... 73

5.3 Results .......................................................................................................................... 76

5.4 Discussion .................................................................................................................... 80

5.5 Conclusion ................................................................................................................... 86

5.6 Data availability ........................................................................................................... 87

5.7 References .................................................................................................................... 88

Chapter 6: Exchange Mechanism of Alkaline and Alkaline earth Elements in Zeolite N Membranes ............................................................................................... 91

6.1 Introduction .................................................................................................................. 91

6.2 Computational Methods ............................................................................................... 93

6.3 Results .......................................................................................................................... 96

6.4 Discussion .................................................................................................................. 109

6.5 Conclusion ................................................................................................................. 118

6.6 Refrences .................................................................................................................... 120

Chapter 7: Detailed Mineralogical Study on Natural Australian Zeolites . 123

7.1 Introduction ................................................................................................................ 123

7.2 Experimental .............................................................................................................. 125

7.3 Results ........................................................................................................................ 128

7.4 Discussion .................................................................................................................. 142

7.5 Conclusion ................................................................................................................. 147

7.6 Refrences .................................................................................................................... 149

Chapter 8: Conclusions.................................................................................... 155

8.1 Summary .................................................................................................................... 155

8.2 Conclusion ................................................................................................................. 157

8.3 limitations................................................................................................................... 157

8.4 Future Recommendations .......................................................................................... 158

Appendices .............................................................................................................. 161

Computational Modelling of Zeolite N Ion Exchange Properties iii

List of Figures

Figure 1-1 Main applications of zeolites. The blue and grey tetrahedra indicate Si and Al tetrahedra, respectively. from left to right, brown lines represent the cracked hydrocarbon chains, red/violet circles represent exchanging cations and green and red-grey molecules represent N2 and CO2 gas molecules. (This figure has been adapted from Figure 2 of a recent review article by Speybroeck et al2) ............................................ 2

Figure 2-1 ........................................ 15

Figure 2-2 Zeolite N unit cell displayed in polyhedron and atomistic formats to highlight structural relationships oriented at different crystallographic axes. Colours represent, yellow: silicon, pink: aluminium, red: oxygen, white: hydrogen, lilac: potassium and light green: chlorine (derived from Materials Studio programme) ............................................... 19

Figure 3-1 Hierarchy of (a) Time and Length scales and (b) Accuracy of different computational methods. ................................................................ 28

Figure 3-2 Schematic representation of the four key components of molecular mechanics force fields. The green, blue and red balls represent atoms, the black solid lines show covalent bonds and the dashed lines denote non-bond interactions, the green and red atoms indicate a Coulombic interaction and the green and blue atoms a Lennard-Jones interaction. ...... 33

Figure 4-1 Zeolite N (2x2x2) supercell viewed along (a) [001] and (b) [110] crystallographic directions. Yellow represent Si atoms (or silica tetrahedra), pink = Al (or alumina tetrahedra), red = oxygen, white = hydrogen, lilac = potassium ions and light green = chlorides. .................... 47

Figure 4-2 Zeolite N membrane (ZM) in water (water layers, WL1 and WL2). Yellow represents Si atoms, pink = Al, red = oxygen, white = hydrogen, lilac = potassium and light green = chlorine. .............................. 49

Figure 4-3 (a) Variation of HE (ΔUH(Nw)) in a bulk as a function of number of H2O/cage in the ZM (red) and equilibration of H2O/cage in ZM versus total number of H2O added before equilibration (blue). (b) Variation of HE and pressure in ZM as a function of the number of H2O/cage in ZM................................................................................................................ 53

Figure 4-4 Ion retention ratio compared with K+ over 8 ns MD simulations for ZM................................................................................................................ 54

Figure 4-5 Number of water molecules inside ZM over 8 ns of MD simulations. .................................................................................................. 54

Figure 4-6 (a-e) RDFs, g(r), for non-framework ions to framework atoms in zeolite N and (f) for NH4

+ to Ow in the electrolyte. ..................................... 56

Figure 4-7 (a-e) Ion density profiles along the z direction: within ZM denoted by the vertical red dashed lines and in electrolyte solution (on either side of the red dashed lines) after 8 ns MD simulations. (f) and (g) Density field maps for ions in the central cages (magnification of the

iv Computational Modelling of Zeolite N Ion Exchange Properties

region denoted by the green rectangle shown in ZM) of ZM: K+ is left hand panel (f) and M+;is right hand panel (g); the relative intensity in density field maps increases from red to blue. (Yellow represents Si atoms, pink = Al, red = oxygen, white = hydrogen, lilac = potassium and light green = chlorine. ........................................................................... 57

Figure 4-8 Self diffusion co-efficient of ions in ZM .................................................. 64

Figure 4-9 Visualization of hydrogen bonds between NH4+ and O atoms in the

framework at different time steps of simulation time. Other extra framework species (K+, Cl- and H2O) not involved in hydrogen bonding interaction are hidden from view. Yellow: silicon, pink: aluminium, red: oxygen, white: hydrogen, blue: nitrogen and blue dashed line: hydrogen bond ......................................................................... 65

Figure 5-1 a 2x2x2 super cell of Zeolite N along different directions (a) (001), (b) (010) and (c) (100). ................................................................................ 74

Figure 5-2 Electron density profile showing the interaction between potassium cations with chloride anions and framework oxygens in the (110) plane of zeolite N unit cell along z direction. .............................................. 81

Figure 5-3 Comparison of average Si-O and Al-O bond lengths for zeolite N obtained from DFT models in this study with previous experimental and computational studies on zeolite N and other zeolites .......................... 82

Figure 5-4 XRD patterns for zeolite N unit cell containing K and Cl extra framework ions calculated using Reflex ...................................................... 85

Figure 6-1 The illustrations of SI and SII for extra-framework K in zeolite N supercells along (a) [001] and (b) [110] crystallographic directions. The 8-membered ring pore openings in each channel direction are highlighted with green colour. The black dashed lines indicate the interaction of potassium cations in site I and II with framework oxygen atoms. Atoms are coloured as Silicon=yellow, Aluminium=pink, oxygen=red, potassium=purple and chloride=light green. ............................................................................................................ 94

Figure 6-2 The retention ratio of guest to host ions in (a) ZM-001 and (b) ZM-110, the total charge on cations per unit cell of (c) ZM-001 and (d) ZM-110, the retention of total guest cations in (e) ZM-001 and (f) ZM-110, the number of chlorides in each unit cell of (g) ZM-001 and (h) ZM-110, and water molecules per cage of (i) ZM-001 and (j) ZM-110, respectively. ......................................................................................... 99

Figure 6-3 (a) ZM-001 simulation box along z direction and (b-i) ion concentration profiles along z direction after 8.5 ns MD simulations. The two red dashed lines indicate the location of ZM-001 in electrolyte solution. .................................................................................... 102

Figure 6-4 (a) ZM-110 simulation box along z direction and (b-j) ion concentration profiles along z direction after 8.5 ns MD simulations. The two red dashed lines indicate the location of ZM-110 in electrolyte solution ..................................................................................... 103

Computational Modelling of Zeolite N Ion Exchange Properties v

Figure 6-5 Density field maps of Mn guest cations in (a) K+/Li+, (b) K+/Na+, (c) K+/NH4

+, (d) K+/Cs+, (e) K+/ K+, (f) K+/Rb+, (g) K+/Mg2+ and (h) K+/Ca2+ systems retained inside ZM-001 after 8.5 ns MD simulations. .... 104

Figure 6-6 Density field maps of Mn guest cations in (a) K+/Li+, (b) K+/Na+, (c) K+/NH4

+, (d) K+/Cs+, (e) K+/ K+, (f) K+/Rb+, (g) K+/Mg2+ and (h) K+/Ca2+ systems retained inside ZM-110 after 8.5 ns MD simulations ..... 105

Figure 6-7 Self-diffusion coefficients for K+ and Mn guest cations of each exchanging system inside (a) ZM-001 and (b) ZM-110 membranes. The gray lines indicate uncertainties.......................................................... 108

Figure 6-8 Self-diffusion coefficients for K+ and Mn guest cations of each exchanging system in the electrolyte outside (a) ZM-001 and (b) ZM-110 membranes. The gray lines indicate uncertainties. ............................. 109

Figure 7-1 Representative samples of (a) Avoca, QLD and (b) Werris Creek, NSW deposits............................................................................................. 129

Figure 7-2 (a) Polarised light microscope image showing the mineralogical distribution in Avoca thin section, (b) BSE image of pink layer of Avoca sample showing mineral diversity and (c) BSE image of dark pink layer of Avoca sample showing the mineral diversity in this layer (red circles on the images represent the position of EPMA point analysis) ..................................................................................................... 129

Figure 7-3 (a) Polarised light microscope image showing the mineralogical distribution in Werris Creek thin section, (b) BSE image of general matrix of Werris Creek sample showing mineral diversity and (c) BSE image of dark red layer of Werris Creek sample showing the mineral diversity in this layer (red circles on the images represent the position of EPMA point analysis) ............................................................................ 130

Figure 7-4 XRD patterns of representative zeolite samples from Avoca and Werris Creek (Werris Creek graph superimposed and offset upwards). Corundum was added as internal standard to micronized samples. ........... 133

Figure 7-5 The proportion of mineral phases in Avoca and Werris Creek samples determined by XRD quantitative analysis, compared with previous XRD studies by Flood et al.8 ....................................................... 133

Figure 7-6 SEM image of (a) Avoca and (b) Werris Creek samples showing surface morphologies and different types of amorphous and microcrystalline phases, including platy/tabular-shaped heulandite crystals, fibrous mordenite crystals, bulky feldspar laths and sheets of clay minerals. ............................................................................................. 133

Figure 7-7 WDS map showing stoichiometric proportion of (a) Si:Al and the element wt% for (b) Ca, (c) Mg, (d) K and (e) Fe in Avoca sample. ........ 136

Figure 7-8 WDS map showing stoichiometric proportion of (a) Si:Al and the element wt% for (b) Ca, (c) Fe and (d) K in Werris Creek sample ........... 136

Figure 7-9 WDS map image showing the relation of (a) Si element wt% and (b) Fe element wt% in the dark brown layer of Werris Creek sample. ........... 137

Figure 7-10 Ternary diagrams demonstrating (a) variation in major cation compositions for clinoptilolite and heulandite (green circles represents

vi Computational Modelling of Zeolite N Ion Exchange Properties

the Flood and Taylor (1991)4 and (b) feldspar diversity. Compositions were obtained by EPMA. ........................................................................... 138

Figure 7-11 TG, DTG and DSC curves of (a) Avoca and (b) Werris Creek zeolite samples. .......................................................................................... 139

Figure 7-12 Characteristic N2 adsorption/desorption isotherms of (a) Avoca with 1-2mm particle size and (b) Avoca micronized particles as well as differential pore size distribution of (c) Avoca samples with 1-2mm particle size and (d) Avoca micronized particle size.. ............................... 141

Computational Modelling of Zeolite N Ion Exchange Properties vii

List of Tables

Table 2-1 Examples of computational and theoretical methods applied to zeolite science. This table is adapted from Table 1 of the recent review article by Speybroeck et al2. ........................................................................ 13

Table 2-2 Summary of previous investigation of ion-exchange property of zeolites using computational chemistry methods ........................................ 18

Table 4-1 Partial charge and COMPASS force field atom types used on all atoms ............................................................................................................ 48

Table 4-2 Hydration energy of zeolite N; a) Bulk and b) ZM .................................. 52

Table 4-3 Number of ions in ZM with K+, NH4+, Na+, Rb+ and Cs+: before and

after MD simulation. .................................................................................... 55

Table 4-4 The nearest neighbour distance (Å) between univalent ions (M) with atoms in the zeolite N framework, estimated from RDFs. .......................... 55

Table 4-5 Experimental data on zeolite N ion exchange selectivity in mixed cation solutions25 .......................................................................................... 58

Table 4-6 Self diffusion co-efficient (D) of ions and water molecules in the Zeolite membrane (ZM) and solution calculated from MD simulation for 9ns, at 298K............................................................................................ 63

Table 5-1 DFT models employed in this study with different convergence quality, approximation, functional, basis sets, dispersion corrections and thermal smearing parameters ................................................................ 75

Table 5-2 Arithmetic mean of atomic displacements (d-av (Å)) and maximum atomic displacement (d-max (Å)) obtained by COMPSTRU programme47. The Codes are explained in Table 5-1. ................................. 77

Table 5-3 Bond distances ( in Å) between framework Si/Al atoms with oxygen atoms derived from DFT calculations and the calculated mean absolute deviation (MAD) of computational bond length from experimental data. The Codes are explained in Table 5-1. .......................... 78

Table 5-4 Calculated Mulliken atomic charges of zeolite N framework atoms derived from DFT calculations .................................................................... 80

Table 5-5 Diversity of atomic charge of framework atoms of zeolites with Si/Al=1 ......................................................................................................... 84

Table 5-6 Refined positional parameters of zeolite N structure obtained from experiment17 and DFT calculation and calculated atomic displacement ..... 86

Table 6-1 Number of initial and retained ions in K-ZM without guest cations, ZM-001 and ZM-110 membranes as well as their comparison with previous study. The potassium retained in Site I and Site II after 8.5ns simulations are presented as a percentage. ................................................ 100

Table 6-2 The nearest distances of Mn ions into framework oxygen (O-Mn), silicon (Si-Mn) and aluminium (Al-Mn) atoms, chloride ions (Cl-Mn)

viii Computational Modelling of Zeolite N Ion Exchange Properties

and oxygen of water molecules (Ow-Mn), inside membranes and their comparison with previous study. ............................................................... 106

Table 6-3 Self-diffusion coefficient of ions inside ZM-001 and ZM-110 membranes and outside in the electrolyte solution. ................................... 107

Table 7-1 Cation Exchange Capacity of different particle size of zeolite samples ....................................................................................................... 131

Table 7-2 Bulk chemical composition of zeolite samples according to the XRF analysis, presented as wt %oxides ............................................................. 132

Table 7-3 Summary of zeolite cationic compositions as determined using EPMA for different layers of Avoca and Werris Creek samples ............... 139

Table 7-4 Mass loss of samples at different temperature ranges (all values are in %) ........................................................................................................... 140

Table 7-5 Parameters obtained from N2 adsorption/desorption isotherms for Avoca and NSW samples ........................................................................... 142

Table 7-6 Average weight % oxide in zeolite composition obtained by EPMA point analysis following data reduction and quality control protocol of Campbell et al (2016) 41. ............................................................................ 145

Table 7-7 Zeolite Formula (normalized to 72 oxygen atoms) obtained by EPMA point analysis following data reduction and quality control protocol of Campbell et al (2016) 41. ......................................................... 145

Computational Modelling of Zeolite N Ion Exchange Properties ix

List of Abbreviations

Silicon Si pico second ps

Aluminium Al nano second ns

Oxygen O Centigrade degree oC

Hydrogen H Angstrom Å

Lithium Li kilovolt kV

Sodium Na Potential Hydrogen pH

Potassium K International Zeolite Association IZA

Rubidium Rb Aluminophosphate ALPO

Caesium Cs Silicoaluminophosphate SAPO

Magnesium Mg Acid Mine Dainage AMD

Calcium Ca High Performance Computers HPC

Titanium Ti Personal computers PCs

Vanadium V Edingtonite EDI

Chromium Cr Periodic Building Units PBU

Manganese Mn X-ray Fluorescence analysis XRF

Iron Fe X-ray Diffractometer Analysis XRD

Chlorine Cl Scanning Electron Microscopy SEM

Phosphorene P Energy Dispersive Spectroscopy EDS

Fluorine F Thermal Gravimetric analysis TG

Gallium Ga Differential Thermal Gravimetric DTG

Germanium Ge Brunaur-Emmet-Teller model BET

Lead Pb Barret-Joyner-Halenda model BJH

Cadmium Cd Wavelength-Dispersive

spectroscopy WDS

Hydroxyl group OH Electron Probe Micro-Analysis EPMA

Water molecules W Coal Seam Gas CSG

Ammonium NH4+ Back Scattered Electron image BSE

Oxygen atoms of water

molecules Ow New South Wales NSW

Hydrogen atoms of water

molecules Hw Queensland QLD

Nitrogen gas N2 Molecular Mechanics MM

ammonium NH4+ Quantum Mechanics QM

gram g Monte Carlo MC

x Computational Modelling of Zeolite N Ion Exchange Properties

Kilogram Kg Molecular Dynamics MD

Milligram per litter mg/L Crystallographic Information File CIF

Giga Pascal GPa Density Functional Theory DFT

Milliequivalent per Litter Meq/L Residential Time Distribution RTD

Mole per Litter Mol/L Radial Distribution Function RDF

Kilo Volt kV Mean Square Displacement MSD

Milli-Ampere mA Self-Diffusion Coefficient D

Kelvin K Density functional dispersion

correction

DFT-D

Generalised Gradient

Approximation GGA

Protein Consistent Force Field PCFF

Local Density Approximation LDA Consistent Force Field CFF

Tkatchenko and Scheffler

scheme

TS Universal Force Field UFF

Ortmann, Bechstedt, and

Schmidt scheme

OBS Consistent Valance Force Field CVFF

Double Numerical plus d-

functions

DND Symmetry Constrained Intensity

Bonding Search method

SCIBS

Double Numerical plus

polarization

DNP Quantitative computed

tomography

QCT

Hydration Energy HE Cambridge Serial Total Energy

Package

CASTEP

Particle-Particle Particle-Mesh PPPM Vienna Ab initio simulation package VASP

Nose-Hoover-Langevin NHL Frozen Density Embedding Theory FDET

Condensed-phase Optimized

Molecular Potentials for

Atomistic Simulation Studies

COMPASS

our own n-layered integrated

molecular orbital and molecular

mechanics

ONIOM

Computational Modelling of Zeolite N Ion Exchange Properties xi

Statement of Original Authorship

The work contained in this thesis has not been previously submitted to meet

requirements for an award at this or any other higher education institution. To the best

of my knowledge and belief, the thesis contains no material previously published or

written by another person except where due reference is made.

Signature:

Date: 22 May 2020

QUT Verified Signature

xii Computational Modelling of Zeolite N Ion Exchange Properties

Acknowledgements

principal supervisor Prof. Ian Mackinnon, for his insightful guidance and help

on both research and life throughout my PhD journey. Also, I acknowledge his

financial support towards the IZC19 conference. I extend my gratitude to my associate

supervisors Dr. Vinuthaa Murthy and Prof. Graeme Millar for their significant help on

computation and experiment parts of my PhD research study.

I acknowledge laboratory members of the Central Analytical Research Facilities

in the Institute for Future Environments of QUT for their help and company during my

candidature. My appreciation also goes to the High-Performance Computing (HPC)

group at QUT for providing the pioneering computing resources and help on software

maintenance.

I am grateful for the Institute for Future Environments Scholarship granted and

the Higher Degree Research Tuition Fee Sponsorship provided by Queensland

University of Technology.

Finally, I thank my family for in my life as

well as all my friends for their support during my PhD.

Computational Modelling of Zeolite N Ion Exchange Properties 1

Chapter 1: Introduction

This chapter introduces the background (section 1.1) and the context (section

1.2) of the research as well as the purpose (section 1.3), and the scope (section 1.4) of

this research. Finally, an outline of the following chapters in this thesis is presented in

section (1.5).

1.1 BACKGROUND

Zeolites are crystalline structures with a framework that consists of SiO4 and

AlO4 tetrahedra linked together by bridging oxygen atoms. The presence of aluminium

imposes a negative charge to the framework which is compensated by inclusion of

cationic species. These cationic species can be any of the alkali, alkaline earth,

lanthanide, transition metal and organic cations that are situated in available extra-

framework positions inside the channels and cages of the zeolite structure. Water

molecules are a complementary part of zeolite structures; these occupy lower

symmetry sites inside the zeolite pores and surround the extra-framework elements1.

Presently, more than 200 framework types of zeolites have been approved by the

International Zeolite Association (http://www.iza-structure.org/databases/) This

number includes more than 60 types of natural zeolites which are found in deposits

worldwide (http://www.izaonline.org/natural/default.htm). However, most zeolites

that are used in industry are synthesized under hydrothermal conditions.

In a typical synthesis process, an aqueous gel consisting of SiO2, Al2O3 and

suitable desired cations crystallise at high pH and under medium pressure and

temperature2. Other members of the zeolite family, called zeotype materials, are ALPO

(Aluminophosphate) and SAPO (Silicoaluminophosphate) that are formed from PO4-

AlO4 and SiO4-AlO4-PO4 tetrahedra, respectively. Moreover, frameworks that have

other elements, such as Ti, V, Cr, Ga, Ge, Fe and Mn, instead of Si and Al are

considered zeotype materials2.The abundance, different crystalline structure, pore

topology diversity and variation in the chemical composition provide these materials

with special properties including molecular sieving, adsorption, catalysis, thermal

stability and ion exchange. Each of these properties, or combinations of properties,

result in a wide range of industrial applications.

2 Computational Modelling of Zeolite N Ion Exchange Properties

Zeolites are excellent catalysts due to Bronsted and Lewis acid sites that occur

in the framework and extra-framework locations, respectively3. These acidic sites are

critical enablers of efficient catalytic transformations used in many industries4. Zeolite

acidity has resulted in significant use in the petrochemical industry5-7 for cracking long

chain hydrocarbons, isomerisation and for the synthesis of hydrocarbons8. In addition,

zeolites are commonly used as molecular sieves because of their microporous

structure. The porous framework is built from cages of bonded SiO2, Al2O3 (or, as

noted above, other combinations of oxides) connected together by channels of various

shapes. This diversity of structure and structure types provides zeolites with a capacity

to separate molecules of different sizes and shapes9. The extensive use of zeolites in

the gas separation industry is based on this feature10. Moreover, the presence of extra-

framework cations, loosely bonded to the zeolite framework provides the ion-

exchange property also widely used in industry1. Specific zeolites, such as synthetic

zeolite N and natural clinoptilolite, are effective in reducing, or eliminating,

contaminants from aqueous solutions like ammonium11, heavy metals12, inorganic

anions, organic compounds, dyes and humic substances13. Removal of metals from

acid mine drainage (AMD), removal of ions from landfill leachate14 and horticultural

applications (soil amendments15, fertiliser additives16) are other key uses of zeolites

with respect to their ion-exchange properties. Figure 1-1 provides a schematic that

generically describes the three fundamental chemical processes that give rise to these

three primary applications of zeolites.

Figure 1-1 Main applications of zeolites. The blue and grey tetrahedrals indicate Si and Al tetrahedra, respectively. from left to right, brown lines represent the cracked hydrocarbon chains, red/violet circles represent exchanging cations and green and red-grey molecules represent N2 and CO2 gas

molecules. (This figure has been adapted from Figure 2 of a recent review article by Speybroeck et al2)


The precise chemical and physical behaviour of zeolites during catalysis,

adsorption, gas separation and ion-exchange processes are challenging to determine at

an atomic scale. This challenge is, in part, due to the fact that many processes occur

inside the pores of zeolites and, in part, because direct methods to probe within these

pores are limited2. For example, to understand an adsorption process, knowledge of

the adsorbed reactant, its diffusion to the active sites, the conversion process on those

sites and finally, the desorption of that reactant are required. The rates of reactions

during adsorption and formation of the final product are affected by each of the above-

mentioned steps. The nature of the adsorbed molecules and those elements situated

inside the zeolite pores, make it difficult to acquire information at atomic level by

using experimental techniques. In many cases, macro- or micro-scale experimental

data allow for inferences on the most likely chemical or physical processes occurring

within zeolite pores. However, computational simulations on zeolite structures and

their behaviour have shown that detailed atomic-scale models can often accurately

reflect the macro- or micro-scale properties experimentally determined for a number

of industrially important zeolites. Speybroeck et al2 provides a comprehensive review

on the application of computational chemistry in the field of zeolite science and

technology. This review indicates the focus of computational studies mostly on

catalysis and gas separation applications of zeolites and less attention on ion-exchange

application.

1.2 CONTEXT

In this research project, the focus is on investigating synthetic zeolite N and its

ion-exchange behaviour by applying computational simulations and modelling. We try

to explain qualitatively why zeolite N has better exchange reactions with some cations

in comparison to other elements. In addition, this project will also quantitatively

predict exchange behaviour of zeolite N for the cations of interest under conditions of

interest by developing molecular models.

This research is conducted from a computational perspective and focuses on

molecular modelling and simulations that can develop and expand our understanding

of the exchange behaviour of zeolite N at a molecular scale. Thus, while this work will

focus on computational techniques and outcomes, the results will be evaluated in light

of experimental data on Zeolite N.


1.2.1 Hypothesis

In this research, we assume that ion-exchange behaviour of zeolite N in aqueous

environments can be predicted by applying computational techniques based on

quantum mechanics and molecular dynamic simulations.

1.3 PURPOSES

1.3.1 Aim

This research project focuses on a qualitative and quantitative study of the ion

exchange behaviour of zeolite N at atomic scale by applying a combination of

computational simulations and comparison with experiments. This fundamental study

will focus on ion exchange in aqueous solutions containing cations in mixed

assemblages such as Li+, Na+, K+, Ca2+ and NH4+ and other mixtures of mono- and di-

valent ions. Zeolite N is known, by previous experiments, to exchange some of these

cations at high capacities. This work aims to understand the underlying atomic scale

factors that influence the ion exchange property of zeolite N and to develop an

approach to evaluate relative selectivity for exchange of different ions in solution.

1.3.2 Research Objectives

The main objectives of this research include:

Study the hydration behaviour of zeolite N and estimate its effects on ion-

exchange processes (chapter 4).

Study the structure of zeolite N using computational techniques and to define

the atomic charges and other atom parameters using comprehensive DFT

calculations (chapter 5).

Perform molecular dynamics simulations based on quantum mechanics

descriptions of electron distributions to investigate the ion-exchange dynamics

of zeolite N. The ion exchange behaviour will be evaluated in aqueous

solutions to model a hydrophilic environment (chapter 4).

Compare the exchange mechanism of univalent and divalent cations with

zeolite N to evaluate the influences of atomic number and ion charge on

exchange dynamics (chapters 4 and 6).


Investigate the impacts of various parameters on ion-exchange behaviour of

zeolite N (chapters 4 and 6).

Compare computational outcomes with the existing experimental data in order

to evaluate the validity of the computational models (chapters 4, 5 and 6).

Investigate the physical and mineralogical properties of Australian natural

zeolites and determine their chemical composition for further ion-exchange

simulations (chapter 7).

1.4 SIGNIFICANCE AND SCOPE

Synthetic zeolite N is one of more than 200 types of zeolites with properties that

enable industrial use. Its use in industry will depend, in large part, on a specific

property such as ion-exchange capacity or combinations of properties such as particle

size, selectivity for a particular ion or thermal stability. Therefore, there is a need to

reliably predict the properties of this zeolite for key applications, in particular, for

selective exchange of important ions that commonly occur in solution in our

environment (e.g. NH4+, Li+, Na+, K+, Mg2+ and Ca2+). The capacity to predict this and

other properties (e.g. behaviour with change of pH in solution) enables evaluation of

potential technical and economic performance for use in different industries, including

the petroleum, wastewater treatment and agricultural industries.

For this purpose, studies have shown that experiments cannot address the

detailed mechanics of ion exchange due to limited detail on the chemical and physical

mechanisms that occur during ion exchange processes. However, computational

techniques can accurately develop and expand our understanding of the ion exchange

behaviour of zeolite N for ions of interest under conditions from atomic scales to meso-

scales. Ultimately, an in-depth understanding of both theory and experiment for this

zeolite will allow the first principles design of new zeolite forms or, preferably, the

prediction of their behaviour under actual conditions.

1.5 THESIS OUTLINE

In the following, an outline of chapters in this thesis is provided:

Chapter 1: includes an introduction to the research background and

research context. This chapter also describes the purposes and

significance of this research followed by the outline of the study.


Chapter 2: contains a comprehensive review on the literature available

in the field. Initially, the application of computational chemistry

techniques in the field of zeolite science is covered and then molecular

modelling studies of ion-exchange properties of zeolites. Then, the

reasons supporting the choice of zeolite N for this study are explained.

Finally, the significance and implications of research in the current

literature is reviewed.

Chapter 3: represents an introduction on available molecular modelling

techniques, followed by a brief description of quantum mechanics

calculations and molecular dynamics simulations, as well as applied

molecular modelling techniques used in this study. This chapter includes

model development, simulation settings and outcome analyses.

Chapter 4: studies show that water molecules play an important role in

the ion-exchange process of zeolites17, 18. Therefore, this research started

with investigating the behaviour of water molecules in zeolite N. In this

chapter, the hydration behaviour of zeolite N is investigated with

molecular dynamics calculations and validated with experimental data

provided in the literature. Then, molecular dynamics simulations are

conducted to study the exchange of monovalent cations (NH4+, Na+, K+,

Rb+ and Cs+) in zeolite N. The outcomes are compared with available

experimental data for NH4+ and Na+.

Chapter 5: The partial atomic charges employed in Chapter 4 models

were obtained from previous studies on zeolite LTA19 which has a similar

Si/Al ratio to zeolite N. However, zeolite N contains two different

crystallographic sites for framework Si and Al atoms and contains

different extra-framework cations compared to zeolite A (or zeolite

LTA)19, 20. Therefore, in this chapter a range of Density Functional

Theory (DFT) models are evaluated to optimize structures of zeolite N

with and without extra-framework ions. The effects of DFT models on

zeolite N structural parameters and partial atomic charges of framework

atoms are assessed. Structural parameters and partial charges are

compared with experimental and computational studies on zeolite N and

other zeolites. The structural parameters and partial atomic charges


obtained for zeolite N by the most reliable DFT model in this study are

used in further MD simulations.

Chapter 6: presents results from a comprehensive simulation of ion-

exchange mechanism within zeolite N membrane for monovalent (NH4+,

Li+, Na+, K+, Rb+ and Cs+) and divalent (Mg2+ and Ca2+) cations. The

framework of zeolite N consists of two channels along different

crystallographic directions with different size and shape. Therefore, the

exchanges of cations within zeolite N membranes are investigated along

different crystallographic directions. Moreover, the effects of different

partial charges for framework atoms on the dynamic behaviour of zeolite

N are investigated. Finally, the ion selectivity for zeolite N is predicted

based on simulation outcomes.

Chapter 7: the outcomes of previous chapters on zeolite N ion-exchange

behaviour indicate that molecular modelling studies can enhance our

understanding of mechanisms that are difficult to explain experimentally.

For example, these simulation techniques may then be employed to study

the ion-exchange behaviour of natural Australian clinoptilolite.

Developing a reliable model for simulating the exchange behaviour of a

specific zeolite requires a precise understanding of its chemical

composition, Si/Al ratio as well as type and amount of extra-framework

cations. However, clinoptilolite shows variable chemical compositions

in nature and, consequently, shows variable experimental exchange

behaviour when obtained from different deposits. Therefore, to develop

viable models for simulation studies, the detailed measurement of

composition, with high spatial resolution, of two natural Australian

clinoptilolites using electron probe microanalysis (EPMA) was obtained.

Consequently, this chapter includes a comprehensive mineralogical

analysis to determine physical properties, thermal behaviour, porosity

and mineral composition.

Chapter 8: summarises and concludes the major outcomes of this

research. Moreover, the recognised limitations in the research process are

described. Regarding the utilisation of the outcomes of this study in the


field of research and addressing limitations, several options are suggested

for future studies.


1.6 REFRENCES

1. Jacobs, P.; Flanigen, E. M.; Jansen, J.; van Bekkum, H., Introduction to zeolite science and practice. Elsevier: 2001; Vol. 137. 2. Speybroeck, V. V.; Hemelsoet, K.; Joos, L.; Waroquier, M.; Bell, R. G.; Catlow, C. R. A., Advances in theory and their application within the field of zeolite chemistry. Chemical Society Reviews 2015, 44, 7015–7430. 3. Corma, A., From microporous to mesoporous molecular sieve materials and their use in catalysis. Chemical Reviews 1997, 97 (6), 2373-2420. 4. Martínez, C.; Corma, A., Inorganic molecular sieves: Preparation, modification and industrial application in catalytic processes. Coordination Chemistry Reviews 2011, 255 (13–14), 1558-1580. 5. Primo, A.; Garcia, H., Zeolites as catalysts in oil refining. Chemical Society Reviews 2014, 43 (22), 7548-7561. 6. Vermeiren, W.; Gilson, J.-P., Impact of zeolites on the petroleum and petrochemical industry. Topics in Catalysis 2009, 52 (9), 1131-1161. 7. Degnan, T. F., The implications of the fundamentals of shape selectivity for the development of catalysts for the petroleum and petrochemical industries. Journal of Catalysis 2003, 216 (1), 32-46. 8. De Vos, D. E.; Dams, M.; Sels, B. F.; Jacobs, P. A., Ordered mesoporous and microporous molecular sieves functionalized with transition metal complexes as catalysts for selective organic transformations. Chemical Reviews 2002, 102 (10), 3615-3640. 9. Auerbach, S. M.; Carrado, K. A.; Dutta, P. K., Handbook of zeolite science and technology. CRC press: 2003. 10. Smit, B.; Maesen, T. L. M., Towards a molecular understanding of shape selectivity. Nature 2008, 451 (7179), 671-678. 11. Mackinnon, I. D. R.; Barr, K.; Miller, E.; Hunter, S.; pinel, T., Nutrient Removal from waste water using high performance materials. Water Science and Technology 2003, 47, 101-107. 12. Fu, F.; Wang, Q., Removal of heavy metals ions from wastewaters: A review. Journal of Environmental Management 2011, 92, 407-418. 13. Wnag, S.; Peng, Y., Naturalzeolites as effective adsorbents in water and wastewater treatments. Chemical Engineering Journal 2010, 156, 11-24. 14. Delkash, M.; Ebrazi Bakhshayesh, B.; Kazemian, H., Using zeolitic adsorbents to cleanup special wastewater streams: A review. Microporous and Mesoporous Materials 2015, 214, 224-241. 15. Zwingmann, N.; Singh, B.; Mackinnon, I. D. R.; Gilkes, R. J., Zeolite from alkali modified kaolin increases NH4+ retention by sandy soil: Column experiments. Applied Clay Science 2009, 46 (1), 7-12. 16. Zwingmann, N.; Mackinnon, I. D. R.; Gilkes, R. J., Use of a zeolite synthesised from alkali treated kaolin as a K fertiliser: Glasshouse experiments on leaching and uptake of K by wheat plants in sandy soil. Applied Clay Science 2011, 53 (4), 684-690. 17. Pissis, P.; Daoukaki-Diamanti, D., Dielectric studies of molecular mobility in hydrated zeolites. Journal of Physics and Chemistry of Solids 1993, 54 (6), 701-709. 18. Maurin, G.; Bell, R. G.; Devautour, S.; Henn, F.; Giuntini, J. C., Modeling the Effect of Hydration in Zeolite Na+-Mordenite. Journal of Physical Chemistry B 2004, 108 (12), 3739-3745.


19. Salmas, R. E.; Demir, B.; Yıldırım, E.; Sirkecioğlu, A.; Yurtsever, M.; Ahunbay, M. G., Silver–Sodium Ion Exchange Dynamics in LTA Zeolite Membranes. Journal of Physical Chemistry C 2013, 117, 1663. 20. Christensen, A. N.; Fjellvag, H., Crystal structure determination of zeolite N from synchrotron X-ray powder diffraction data. Acta Chemica Scandinavica 1997, 51, 969-973.


Chapter 2: Literature Review

Since the introduction of the term ‘zeolite’ by Axel F. Cronstedt in 1756, many

thousands of books, articles and conference papers have been published. These works

cover all aspects of zeolite science and technology from synthesis to applications.

Researchers from a variety of disciplines have deployed experimental and

computational techniques available at the time to develop insight and knowledge of

these extraordinary materials. This review is focused on the literature that can explain

the existing gap in zeolite science related to the use of computational chemistry for

studying the ion-exchange behaviour of synthetic zeolite N and related Al-Si zeolites.

This chapter begins with an introduction to the application of computational

chemistry techniques in zeolite science (section 2.1) and reviews the literature on

computational studies of ion-exchange in zeolites (section 2.2). Section 2.3 introduces

zeolite N and the motives for selecting zeolite N for this study. Finally, section 2.4

highlights the significance and implication of the study.

2.1 COMPUTATIONAL METHODS IN ZEOLITE SCIENCE

Zeolites are a favoured topic for researchers because of the wide variety of

industrial applications of these materials. In addition, over the years new structures

have been developed and the structures of existing zeolites have been optimised to

serve a particular problem in practice1. As this review will show, the use of

computational techniques to design zeolitic materials, at the molecular scale, for

specific applications has been well demonstrated. Computational chemistry and

materials design have progressed rapidly and effectively not only due to novel methods

and algorithms but also to rapid increases in the speed and memory capacity of modern

computers. A large number of methods have been applied for modelling and

simulation of zeolites2. These methods have solved many challenging issues in zeolite

science and today, modelling has found a predictive role in this field 2.

Advancement of HPC resources has influenced progress in all aspects of

computational modelling3, especially with regard to the theory of zeolite chemistry. In

addition, advances in applied methodologies have resulted in greater accuracy in

simulations. The combination of significant improvements in computational methods,


along with improved software and hardware performance has allowed more complex

structures, and systems containing these structures, to be modelled by computational

chemistry4.

Nowadays, computational and theoretical techniques are used routinely in many

fields of zeolite science2, 5 not only to model but also to predict the structure and

function of zeolites. For example, even the dissolution and growth of a zeolite during

synthesis can be elucidated by modelling methods. Modelling has been an effective

tool responsible for the success of zeolitic materials in catalysis. Finally, sorption and

diffusion investigations have greatly benefited from molecular modelling methods.

While the many successes of computational modelling in the field of zeolites are too

numerous to elucidate in this review, Table 2-1 presents representative examples of

applications of computational techniques2 in this field.

A review of the available literature shows that in the past decades researchers

have applied computational modelling and simulation techniques predominantly for

catalysis and gas separation applications of zeolites. Other applications, such as ion-

exchange use in industrial applications, while undertaken experimentally by many

researchers, have had significantly less focus.

2.2 COMPUTATIONAL STUDIES OF ZEOLITE ION-EXCHANGE

Ion-exchange is one property of zeolites that is commonly exploited for

industrial applications. However, the diffusion of water molecules and ions, as well as

their correlation and interaction within a particular zeolite structure are not well

understood at an atomic scale despite many experimental results and macroscopic

mathematical models to describe the phenomena6. Nevertheless, we know from

previous studies that computational modelling and simulation techniques are an

effective means to explain experimental results2. In the following review, specific

examples that will guide the way to an understanding of zeolite N behaviour are

outlined.


Table 2-1 Examples of computational and theoretical methods applied to zeolite science. This table is

adapted from Table 1 of the recent review article by Speybroeck et al2.

Field of study Issues Computational techniques

Zeolite synthesis

Condensation of small silica

clusters in gas phase

Molecular Mechanics and DFT

calculation

Surface structure and crystal growth Periodic DFT calculation and Monte

Carlo simulation

Structural

Modelling

Hypothetical zeolite structure

Symmetry Constrained Intensity

Bonding Search method (SCIBS) and

tiling theory with structure

optimization using force fields

External surface of zeolites Interatomic Potential Method

Spectroscopy

Scaling methods for frequencies of

infrared spectroscopy

Large cluster models, periodic DFT

calculations and Molecular Dynamic

simulations with periodic DFT codes

Studying aluminophosphates and

structure-spectrum relationship in

NMR spectroscopy

Periodic DFT calculations

Studying dye molecules in zeolites

using optical spectroscopy

Frozen Density Embedding Theory

(FDET) and first principles Molecular

Dynamics simulation

Zeolite

applications

Gas separation and screening of

materials

Grand Canonical Monte Carlo

simulation with force fields

Alkane adsorption and product

selectivity in alkane cracking

Periodic DFT-D, cluster models,

ONIOM, first principles Molecular

Dynamics and QCT simulations

Redox/Oxidation studies in metal

exchanged zeolites for NH3-SCR

Periodic DFT, extended cluster models

and first principle Molecular dynamics

simulations

Diffusion in zeolites

Molecular Dynamic simulations using

force fields and kinetic Monte Carlo

simulations


2.2.1 Ion-exchange theory

The ion-exchange property of zeolites is primarily due to substitution of Si by

Al atoms in the zeolite framework. This replacement induces on the zeolite framework

a negative charge that is balanced by alkaline and alkaline-earth cations. These cations

are situated in extra-framework sites inside the cages and channels of the zeolite

structure, loosely bonded to water molecules or bridging oxygen and can be exchanged

by other cations under suitable chemical conditions7. For example, in the presence of

variable concentrations of aqueous solutions over a range of temperatures, pH

conditions or pressures, the ion exchange performance of a particular zeolite is

determined by key crystallographic parameters.

Of the various parameters that affect the ion-exchange behaviour of zeolites, the

most significant factor is the Si/Al ratio because the amount and nature of extra-

framework cations depends on the Al content in the zeolite framework. Increasing

aluminium content causes a more negative charge and a higher ion-exchange capacity

of the zeolite. The shape and size of the microporous structure of the zeolite is the

second important factor. When the ionic radii of targeted cations are close to the pore

size of the zeolite, the exchange capacity for that cation is considerable and viable8-10.

Other parameters that play important roles in the ion selectivity of a zeolite include the

amount of water molecules inside the zeolite structure, the hydration energy of the

cations, the position of the extra-framework cations, the nature of the exchangeable

cations in the solution, as well as the particle size, uniformity, purity and stability of

the zeolite material8, 10.

In general, under different chemical conditions, there is an order of ion

selectivity for zeolite structures. The following rules generally govern the types of ions

that can be exchanged within aluminosilicate zeolite structures7:

1) Silica rich frameworks prefer large monovalent cations while zeolites with

Si/Al ~ 1 prefer small multivalent cations;

2) Cations such as Li+ and Mg2+, with high heats of hydration tend not to be

exchanged easily at the ambient temperature normally used to define exchange

isotherms; 1q1

3) Zeolites often do not prefer transition metal cations for ion exchange; these

cations primarily situate on the external surfaces of zeolites.


In theory, the ion-exchange efficiency of zeolites is described by analysing exchange

isotherms. These isotherms are graphical indications of relative selectivity of zeolites

to the cations7. For example, assume that we have a K-rich zeolite and a solution

containing K+ and NH4+ ions. If the zeolite has high selectivity for NH4

+ at the desired

pH, pressure and temperature, the isotherm will be an upward convex curve (red line

in Fig. 2-1); and if it has high selectivity for K+ under those condition, the isotherm

will follow a downward concave curve (green line in Fig. 2-1). The isotherm will show

a straight line when there is no difference between free energy of NH4+ ions in the

solution with K+ ions in the zeolite (black line in Fig. 2-1). Nevertheless, in practice

exchange isotherms do not exactly match these theoretical shapes.

Figure 2-1

Moreover, understanding the diffusion coefficient of exchangeable cations is

important in ion exchange processes11. In this study, the mobility of cations within the

zeolite is influenced by a number of processes, which must be considered:

1) Energy barriers that cations experience while encountering with zeolite

framework atoms;

2) Competition between different cations as they diffuse through the same

framework;

3) Energy barriers occur due to the self-exchange of the individual cations;

4) Self-diffusion process of water molecules within zeolite framework and their

movements as hydration shell of cations diffusing through the zeolite

framework.

However, experimentally measuring diffusion coefficients is complicated due to

not only the various physical and chemical conditions under which diffusion occurs,


but also the range of scales over which diffusion may be observed can be affected by

experimental techniques. For example, by changing the scale of experimental

techniques from macroscopic to mesoscopic and microscopic, the diffusion

coefficients may differ by orders of magnitude12.

In order to design cation exchange materials with the highest performance, the

mechanism, isotherms and parameters of the exchange processes should be understood

and predictable. This understanding is difficult to achieve by experiment alone.

Therefore, applying computational techniques can make a difference to explore

unsolved experimental observations.

2.2.2 Ion-exchange modelling and simulation studies

Modelling of ion-exchange experiments can provide insight on detailed

mechanisms within the pore structure of the material as well as on circumstances that

are difficult to interpret via laboratory experiment. For example, if the cation

concentration is low in both solution and the zeolite structure, it may be difficult to

gain sufficiently accurate analytical data to construct a reliable exchange isotherm7.

Other conditions may also be amenable to computation more readily than experiment.

These conditions will generally involve zeolite structures with complex, tortuous or

multiple pore structures and sizes13-15 as well as properties of exchanging ions,

including ionic size, charge density, hydrated ionic size and free energy of solvation16.

In ion-exchange processes, there is always a possibility that exchangeable

cations in solution cannot reach the exact targeted extra-framework sites in the zeolite

pores. During ion-exchange, to remove an extra-framework cation from its site, the

coordination energy between the oxygen of the framework and surrounding water

molecules must be overcome. In principle, this coordination shell must be disrupted

and a new shell should be constructed with other framework and water molecules. On

the other hand, the exchangeable cations in solution must rebuild a hydration shell with

water molecules inside the zeolite pores and then develop a new coordination shell

with the bridging oxygen atoms of the zeolite framework. These complex re-

arrangements occur through connected mechanisms which are difficult to visualise6

but which can be calculated and, at specific time intervals, shown as “snapshots” as

the computation proceeds17, 18.


Nevertheless, recent studies using computational techniques on zeolites

demonstrate that molecular modelling can explain the zeolite structure and its ion

exchange property accurately2 19. In addition, modern techniques allow a quantitative

and qualitative understanding of the chemical and physical mechanisms that occur

during the exchange process6, 20. Examples include studies on the cation distribution

in faujasite14, the interaction between zeolite A structure and water molecules21, and

the influence of water molecules on ion selectivity in niobate molecular sieves22. Table

2-2 presents a summary of molecular modelling studies related to ion-exchange

behaviour of zeolites.

Computational models have focused on key zeolites such as zeolites A, Y,

faujasite and mordenite with fewer detailed studies on clinoptilolite. In this study,

attention will focus on an industrially significant zeolite – zeolite N – that has been

manufactured and used at commercial scale to selectively extract ammonium ions from

wastewaters.

2.3 ZEOLITE N

The potassium-rich zeolite N, with the general formula K12Al10Si10O40Cl2.5H2O

was initially synthesised by Barrer et al.23 in 1953. Later, Christensen and Fjellvag24,

25 determined the crystal structure of zeolite N using high resolution X-ray and neutron

diffraction data. The structure of zeolite N is orthorhombic with space group I22224, 26.

In earlier work, Barrer et al.27-29 using less accurate diffraction methods, proposed that

the structure is tetragonal. Zeolite N is in the EDI framework group and is considered

a fibrous zeolite built from chains consisting of one dimensional Periodic Building

Units (PBU). These tetrahedral PBUs consist of 5T units (T can be Si or Al) connected

together by bridging oxygen atoms along the a and b axes, translated along the c axis

to make connected channels. The intersection of these channels makes eight-ring pores

in this zeolite.

The porous structure of zeolite N is constructed from a channel network with the

minimum pore diameter 3.6 Å as shown in Figure 2-2. The two main channels run

along the [001] and [110] directions and exhibits two different eight-membered rings.

The intersection of these channels makes a cage with 6.3 Å width hosting extra-

framework potassium and chloride ions as well as eight water molecules.


Table 2-2 Summary of previous investigation of ion-exchange property of zeolites using

computational chemistry methods

Zeolite type Computational method Application

Low-Al gmelinite30 ab initio MD Investigation of Na+/NH4+ position and dynamics as a function of hydration degree

Zeolite A21 Classical minimization techniques Embedded cluster methods

Investigation of water molecules location and their effects on the position and stability of extra-framework cations

Na-Mordenite31 MC and MD calculations Investigation of the effects of Si/Al ratios and water content on statics and dynamics of extra-framework cations

Na-A32 Classical MD Qualitative ion-exchange study of Li+ and Ca2+ cations

ZK-433 Classical MD Separation of water molecules from NaCl solution

Hydrated zeolite A6 MD using a sophisticated empirical potential function Studying the Ca-Na exchange

Titanosilicate Na-ETS-1017

DFT calculations for estimating partial charges MD using Universal force field (UFF)

Investigating the exchange process of bivalent heavy metals (Pb2+, Cd2+ and Cu2+)

Mordenite8 First principles DFT calculations Studying the mechanism of cation selectivity of zeolites for Cs+ cations

Clinoptilolite34 ONIOM DFT/MM Periodic DFT

Examining and optimizing the extra-framework cations sites inside the pores

LTA18 MD simulations Investigating water flow through the zeolite framework

LTA13 Periodic DFT calculations MD using Consistent Valence Force Field (CVFF)

Clarifying the defects of ion concentration, crystal thickness and temperature parameters on the Na+/Ag+ exchange rate

Zeolite Y, Mordenite, Zeolite A and ZK-49

Semi-grand canonical MC simulation

Evaluating the cation selectivity and exchange isotherm by considering the effect of Si/Al ratio

Al-modified clinoptilolite35 DFT calculations Investigating the arsenic mobility under

anhydrous and hydrated conditions

Zeolite Y14 DFT calculations MC and MD simulations Predicting the Na+/NH4+ exchange isotherms

Faujasite/Zeolite Y36 Semi-grand canonical MC simulations

Calculating ion exchange isotherms between Na+ and Li+, K+, Cs+ and Rb+

Chabazite37 Periodic ab initio Selectivity of H, Li+, Na+ and K+

Clinoptilolite38 MD and grand canonical MC simulations

Study the dehydration and exchange of Na+, K+ and Cs+ cations


DON, CFI, BEC, MFI, LTA and ERI framework types39

MD simulations Study the transport of vanadium and oxovanadium ions

LTA, FAU, LTN, THO, NAT and EDI40 MD simulations Study the separation of Cu2+, Cd2+ and Pb2+

cations from solution

Figure 2-2 Zeolite N unit cell displayed in polyhedron and atomistic formats to highlight structural relationships oriented at different crystallographic axes. Colours represent, yellow: silicon, pink:

aluminium, red: oxygen, white: hydrogen, lilac: potassium and light green: chlorine (derived from Materials Studio programme)

Several parameters make zeolite N an interesting candidate for molecular

dynamic simulations. On the one hand, considering the end-member (un-exchanged)

composition of zeolite N, the Si/Al ratio is equal to 1. Therefore, the zeolite can

accommodate a high ion-exchange capacity because, as mentioned above, a low Si/Al

ratio is favourable for ion-exchange applications. On the other hand, the pore size (3.6-

6.8 Å) of synthetic zeolite N is sufficient for ion movements and the Al-Si framework

has low tortuosity. That is, compared to other zeolite framework types, the

predominant eight membered channel along the c axis provides an unimpeded path for

ions to transfer or transport to exchangeable sites inside the pores. Moreover, the extra-


framework cation of zeolite N, potassium, is located at two different sites inside the

pores. The K1is located at Site I (SI) at the centre of eight-member rings constructing

channels along the [001] and K2 at site II (SII) is located at the other eight-membered

rings that form a channel along [110]. Hence, these cations can be exchanged due to

their accessible positions and weak electrostatic bonds to water molecules and

framework atoms.

Zeolite N is not a natural zeolite, but it can be synthesised relatively easily.

Barrer and Marcilly28 (1970), Barrer and Munday29 (1971) and Barrer et al.27 (1968)

reported that zeolite N or K-F(Cl), generally, can be formed from zeolite Na-X

contacted with a K rich solution by a static hydrothermal process at temperatures

between 200 oC -300 °C after 7 days. Christensen and Fjellvag24 (1997) synthesised

zeolite N using zeolite 4A as starting material under similar hydrothermal conditions.

More recently, Mackinnon et al.41 (2010) produced zeolite N, at manufacturing scale,

at lower temperatures between 60 oC -100 °C and in shorter times from clay minerals

such as kaolinites and montmorillonites as sources of silicon and aluminium. These

reactions produced zeolite N in less than 20 hours in a continuous stirred reactor using

potassic and potassic-sodic solutions. Later, these reaction conditions were improved

further with zeolite N produced from kaolin42 under hydrothermal conditions in less

than two hours. Sengyang et al43 , recently, synthetised zeolite N from metakaolinite

after 24 hours under hydrothermal conditions at 175 °C. They found the CEC of 590

meq/100 g for zeolite N synthesised by this method.

The high capacity of zeolite N for selective ion exchange applications, compared

with competitive natural zeolites, has been verified by experimental studies.

Mackinnon et al. (2003) used MesoLite (a previous commercial name of zeolite N) to

remove ammonia from return side streams of wastewater treatment plants. In this

demonstration, Mackinnon et al.44 indicated that the material has a robust potential for

ammonia removal of up to 90% with an inlet ammonium concentration ranging

between 700 mg/L and 900 mg/L. Thornton et al.45 (2007) investigated the exchange

isotherms of zeolite N under different conditions of solution pH and concentration,

contact time and presence of competitive cations. These ion exchange investigations44,

45 reported 45-55g NH4+-Nkg-1 ammonium exchange capacity for synthetic zeolite N.

In comparison, the natural zeolite, clinoptilolite which has been used extensively for


ammonium removal applications, shows a much lower exchange capacity for

ammonium, in the range of 0.94-21.52g NH4+-Nkg-1 45.

In other applications of zeolite N, Zwingmann et al.46 (2009) demonstrated that

adding small amounts (0.4%) of zeolite N to sandy soils increased NH4+ retention

capability effectively. In controlled glasshouse trials, Zwingmann et al. 46 showed that

the performance of zeolite N is 11 times higher than natural zeolite clinoptilolite under

the same conditions. In addition, in this work Zwingmann et al. 46 suggested that NH4+-

zeolite N - produced by exchanging the extra-framework K cations with NH4+ ions -

can be used as a slow release fertiliser46, 47.

These experimental laboratory and field data studied the exchange isotherms of

zeolite N and showed that zeolite N has a high capacity for exchange of ions. However,

the ion-exchange mechanism of zeolite N is still unknown and thus, is an appropriate

candidate for computational modelling studies.

2.4 SUMMARY AND IMPLICATIONS

This literature review covered two main aspects in this study:

i. Computational chemistry techniques that are the significant part of this review

and include both quantum computational methods and atomistic simulation

approaches that are focused on:

Application of computational techniques for studying the structure,

property and application of zeolites

Investigating the ion-exchange behaviour of zeolites using

computational methods

ii. The ion-exchange experiments and applications focusing on:

Experimental investigation of zeolite N

Application of ion-exchange ability of zeolite N in industry, for

instance agricultural and horticultural applications and wastewater

treatment.

This review shows that computational modelling and simulation techniques are

effective implementations to describe experimental outcomes, and during past

decades, researchers have applied computational chemistry methods to improve


zeolite science and technology. However, most studies have concentrated on catalysis

and gas separation applications of zeolites, while considerably fewer computational

studies have focused on ion exchange behaviour.

Recent studies reveal that aspects of zeolite ion-exchange behaviour can be

accurately explained at a molecular or atomic level by applying computational

chemistry techniques. These computational studies pay attention to special zeolites

including zeolite A, Y, mordenite and clinoptilolite. Given that there are more than

200 framework types of zeolites, many of which are of industrial importance, there is

good opportunity to pursue computational studies on ion-exchange mechanisms of

other zeolites that experimentally show high ion selectivity.

For this study, zeolite N has been chosen, because experimental investigations

present valuable data on the ion-exchange capability of zeolite N. These data serve as

a verifiable template from which to understand, at an atomic or molecular scale, the

detailed mechanism of exchangeable ions in this structure and provide great

encouragement that computational modelling on this zeolite will be invaluable. Hence,

this research aims to provide a better qualitative and quantitative understanding of the

structure of zeolite N, describing the ion-exchange mechanisms and prediction of its

ion exchange behaviour in aqueous environments under conditions relevant to

practical use.


2.5 REFRENCES

1. Guo, P.; Shin, J.; Greenaway, A. G.; Min, J. G.; Su, J.; Choi, H. J.; Liu, L.; Cox, P. A.; Hong, S. B.; Wright, P. A.; Zou, X., A zeolite family with expanding structural complexity and embedded isoreticular structures. Nature 2015, 524 (7563), 74-78. 2. Speybroeck, V. V.; Hemelsoet, K.; Joos, L.; Waroquier, M.; Bell, R. G.; Catlow, C. R. A., Advances in theory and their application within the field of zeolite chemistry. Chemical Society Reviews 2015, 44, 7015–7430. 3. Woodley, S. M.; Catlow, C. R. A., High performance computing in the chemistry of materials. Physical Chemistry Chemical Physics 2014, 16 (39), 21001-21001. 4. De Jong, W. A.; Bylaska, E.; Govind, N.; Janssen, C. L.; Kowalski, K.; Müller, T.; Nielsen, I. M.; van Dam, H. J.; Veryazov, V.; Lindh, R., Utilizing high performance computing for chemistry: parallel computational chemistry. Physical Chemistry Chemical Physics 2010, 12 (26), 6896-6920. 5. Smit, B.; Maesen, T. L. M., Towards a molecular understanding of shape selectivity. Nature 2008, 451 (7179), 671-678. 6. Suffritti, G. B.; Demontis, P.; Gul´ın-Gonz´alez, J.; Sale, R., Ca-Na cation exchange in zeolite A: A microscopic approach using molecular dynamics simulations. IL Nouvo Cimento 2008, 123, 10-11. 7. Jacobs, P.; Flanigen, E. M.; Jansen, J.; van Bekkum, H., Introduction to zeolite science and practice. Elsevier: 2001; Vol. 137. 8. Nakamura, H.; Okumura, M.; Machida, M., First-Principles Calculation Study of Mechanism of Cation Adsorption Selectivity of Zeolites: A Guideline for Effective Removal of Radioactive Cesium. Journal of the Physical Society of Japan 2012, 82 (2), 023801. 9. Nakamura, H.; Okumura, M.; Machida, M., Monte Carlo simulation studies of cation selectivity in ion exchange of zeolites. RSC ADVANCES 2014, 4 (95), 52757-52761. 10. Cooney, E. L.; Booker, N. A.; Shallcross, D. C.; Stevens, G. W., Ammonia Removal from Wastewaters Using Natural Australian Zeolite. I. Characterization of the Zeolite. Separation Science and Technology 1999, 34, 2307–2327. 11. Smit, B.; Maesen, T. L. M., Molecular simulations of zeolites: adsorption, diffusion, and shape selectivity. Chemical Reviews 2008, 108, 4125–4184. 12. Kärger, J.; Ruthven, D., Diffusion in zeolites. Handbook of Zeolite Science and Technology 1992, 341. 13. Ekhteiari Salmas, R.; Demir, B.; Yıldırım, E.; Sirkecioğlu, A.; Yurtsever, M.; Ahunbay, M. G., Silver–Sodium Ion Exchange Dynamics in LTA Zeolite Membranes. The Journal of Physical Chemistry C 2013, 117 (4), 1663-1671. 14. Wang, L.; Sun, H., Prediction of Na+/NH4+Exchange in Faujasite Zeolite by Molecular Dynamics Simulation and Thermodynamic Integration Method. The Journal of Physical Chemistry C 2013, 117 (27), 14051-14060. 15. Krishna, R.; van Baten, J. M., A molecular dynamics investigation of the diffusion characteristics of cavity-type zeolites with 8-ring windows. Microporous and Mesoporous Materials 2011, 137 (1-3), 83-91. 16. Hinkle, K. R.; Jameson, C. J.; Murad, S., Using Molecular Simulations To Develop Reliable Design Tools and Correlations for Engineering Applications of Aqueous Electrolyte Solutions. Journal of Chemical & Engineering Data 2016, 61 (4), 1578-1584.


17. Nalaparaju, A.; Hu, Z. Q.; Zhao, X. S.; Jiang, J. W., Exchange of heavy metal ions in titanosilicate Na-ETS-10 membrane from molecular dynamics simulations. Journal of Membrane Science 2009, 335 (1-2), 89-95. 18. Turgman-Cohen, S.; Araque, J. C.; Hoek, E. M.; Escobedo, F. A., Molecular dynamics of equilibrium and pressure-driven transport properties of water through LTA-type zeolites. Langmuir 2013, 29 (40), 12389-99. 19. Salmas, R. E.; Demir, B.; Yıldırım, E.; Sirkecioğlu, A.; Yurtsever, M.; Ahunbay, M. G., Silver–Sodium Ion Exchange Dynamics in LTA Zeolite Membranes. Journal of Physical Chemistry C 2013, 117, 1663. 20. Catlow, C.; van Santen, R.; Smit, B., Dynamic monte carlo simulations of diffusion and reactions in zeolites. Computer modelling of microporous materials 2004, 109. 21. Higgins, F. M.; de Leeuw, N. H.; Parker, S. C., Modelling the effect of water on cation exchange in zeolite A. Journal of Materials Chemistry 2002, 12 (1), 124-131. 22. Nenoff, T. M.; Ockwig, N. W.; Cygan, R. T.; Alam, T. M.; Leung, K.; Pless, J. D.; Xu, H.; Hartl, M. A.; Daemen, L. L., Role of water in the ion selectivity of niobate-based octahedral molecular sieves. The Journal of Physical Chemistry C 2007, 111 (35), 13212-13221. 23. Barrer, R.; Hinds, L.; White, E., 299. The hydrothermal chemistry of silicates. Part III. Reactions of analcite and leucite. Journal of the Chemical Society (Resumed) 1953, 1466-1475. 24. Christensen, A. N.; Fjellvag, H., Crystal structure determination of zeolite N from synchrotron X-ray powder diffraction data. Acta Chemica Scandinavica 1997, 51, 969-973. 25. Christensen, A. N.; Fjellvag, H., nuetron powder diferaction study of the dehydration of zeolite N. Acta Chemica Scandinavica 1999, 53, 85-89. 26. Baerlocher, C.; McCusker, L. B.; Olson, D. H., Atlas of zeolite framework types. Elsevier: 2007. 27. Barrer, R.; Cole, J.; Sticher, H., Chemistry of soil minerals. Part V. Low temperature hydrothermal transformations of kaolinite. Journal of the Chemical Society A: Inorganic, Physical, Theoretical 1968, 2475-2485. 28. Barrer, R.; Marcilly, C., Hydrothermal chemistry of silicates. Part XV. Synthesis and nature of some salt-bearing aluminosilicates. Journal of the Chemical Society A: Inorganic, Physical, Theoretical 1970, 2735-2745. 29. Barrer, R.; Munday, B., Cation exchange in the synthetic zeolite KF. Journal of the Chemical Society A: Inorganic, Physical, Theoretical 1971, 2914-2921. 30. Benco, L.; Demuth, T.; Hafner, J.; Hutschka, F., Ab initio molecular dynamics simulation of hydration and ionexchange processes in low Al-zeolites. Microporous and Mesoporous Materials 2001, 42, 1-19. 31. Maurin, G.; Bell, R. G.; Senet, P.; Devautour-Vinot, S., Static and Dynamic Properties of the Nonframework Cations in Na-Mordenites Zeolite. Molecular Simulation 2004, 30 (9), 587-592. 32. Murad, S.; Jia, W.; Krishnamurthy, M., Ion-exchange of monovalent and bivalent cations with NaA zeolite membranes : a molecular dynamics study. Molecular Physics 2004, 102 (19-20), 2103-2112. 33. Lin, J.; Murad, S., A computer simulation study of the separation of aqueous solutions using thin zeolite membranes. Molecular Physics 2001, 99 (14), 1175-1181. 34. Uzunova, E. L.; Mikosch, H., Cation site preference in zeolite clinoptilolite: A density functional study. Microporous and Mesoporous Materials 2013, 177, 113-119.


35. Awuah, J. B.; Dzade, N. Y.; Tia, R.; Adei, E.; Kwakye-Awuah, B.; Catlow, R. A.; De Leeuw, N. H., A density functional theory study of arsenic immobilization by the Al(iii)-modified zeolite clinoptilolite. Physical Chemistry Chemical Physics 2016, 18 (16), 11297-11305. 36. Jeffroy, M.; Boutin, A.; Fuchs, A. H., Understanding the equilibrium ion exchange properties in faujasite zeolite from Monte Carlo simulations. The Journal of Physical Chemistry B 2011, 115 (50), 15059-15066. 37. Civalleri, B.; Ferrari, A. M.; Llunell, M.; Orlando, R.; Mérawa, M.; Ugliengo, P., Cation selectivity in alkali-exchanged chabazite: An ab Initio Periodic Study. Chemistry of Materials 2003, 15 (21), 3996-4004. 38. Johnson, M.; O’Connor, D.; Barnes, P.; Catlow, C. R. A.; Owens, S. L.; Sankar, G.; Bell, R.; Teat, S. J.; Stephenson, R., Cation Exchange, Dehydration, and Calcination in Clinoptilolite: In Situ X-ray Diffraction and Computer Modelin. Journal of Physical Chemistry B 2003, 107, 942-951. 39. Hinkle, K. R.; Jameson, C. J.; Murad, S., Transport of Vanadium and Oxovanadium Ions Across Zeolite Membranes: A Molecular Dynamics Study. The Journal of Physical Chemistry C 2014, 118 (41), 23803-23810. 40. Khanmohammadi, H.; Bayati, B.; Rahbar-Shahrouzi, J.; Babaluo, A.-A.; Ghorbani, A., Molecular simulation of the ion exchange behavior of Cu2+, Cd2+ and Pb2+ ions on different zeolites exchanged with sodium. Journal of Environmental Chemical Engineering 2019, 7 (3), 103040. 41. Mackinnon, I. D. R.; Millar, G. J.; Stolz, W., Low temperature synthesis of zeolite N from kaolinites and montmorillonites. Applied Clay Science 2010, 48 (4), 622-630. 42. Mackinnon, I. D. R.; Millar, G. J.; Stolz, W., Hydrothermal syntheses of zeolite N from kaolin. Applied Clay Science 2012, 58, 1-7. 43. Sengyang, P.; Rangsriwatananon, K.; Chaisena, A., Preparation of zeolite Nfrom metakaolinite by hydrothermal method. Journal of Ceramic Processing Research 2015, 16, 111-116. 44. Mackinnon, I. D. R.; Barr, K.; Miller, E.; Hunter, S.; pinel, T., Nutrient Removal from waste water using high performance materials. Water Science and Technology 2003, 47, 101-107. 45. Thornton, A.; Pearce, P.; Parsons, S. A., Ammonium removal from solution using ion exchange on to MesoLite, an equilibrium study. J Hazard Mater 2007, 147 (3), 883-9. 46. Zwingmann, N.; Singh, B.; Mackinnon, I. D. R.; Gilkes, R. J., Zeolite from alkali modified kaolin increases NH4+ retention by sandy soil: Column experiments. Applied Clay Science 2009, 46 (1), 7-12. 47. Zwingmann, N.; Mackinnon, I. D. R.; Gilkes, R. J., Use of a zeolite synthesised from alkali treated kaolin as a K fertiliser: Glasshouse experiments on leaching and uptake of K by wheat plants in sandy soil. Applied Clay Science 2011, 53 (4), 684-690.


Chapter 3: Methodology

In this chapter, methods used to understand the exchange behaviour of zeolite N

are outlined. Section 3.1 introduces different molecular modelling approaches of

computational chemistry. Sections 3.2 and 3.3 discuss the computational techniques

based on molecular mechanics or quantum mechanics principles and molecular

dynamics simulation methods employed to accomplish the objectives of this research.

The detailed simulation methodologies and settings related to the particular work are

presented in the following chapters.

3.1 COMPUTATIONAL CHEMISTRY TECHNIQUES

In computational chemistry, the behaviour of a molecular system is investigated

and predicted by mathematically calculating the interactions between all available

atoms then coupling these atoms using not only quantum mechanics but also

minimisation, simulation, molecular mechanics and all other computational techniques

that then link molecular level interactions to experimentally accessible macroscopic

quantities. The internal energy (and forces, etc.) for a given set of nuclear positions are

computed using molecular modelling techniques, that are based on the laws of physics

and chemistry and can be Molecular Mechanics (MM) or Quantum Mechanics (QM)1,

2. Then, the nuclear potential energy surface is explored using classical and statistical

mechanics, for example, Monte Carlo (MC) and Molecular Dynamics methods (MD)3,

4 .

MC methods randomly sample the surface of potential energy and try to

converge to the proper Boltzmann-weighted distribution to achieve values about the

minimum free energy of the ensemble of states2 5. MC is a statistical method1, and

thus, does not represent behaviour of a molecular system over time which means that

dynamic information on the system is difficult to obtain. However, by numerically

solving Newton’s equations of motion in the MD method, the velocities and positions

of molecules over time can be computed1 2. By applying these MM methods, the

macroscopic behaviour of large systems can be determined and modelled over long

time scales.


In contrast, quantum chemistry is grounded on the laws of quantum mechanics

which explicitly consider the behaviour of electrons2. Thus, it is possible to calculate

in detail the chemical bonds and atomic interactions of a specific structure or molecular

arrangement by computational methods that are often called ab initio methods.

Nevertheless, the high computational cost of these techniques results in applications

that are limited to smaller systems or to specific locations in a crystal structure such as

a bonding site. QM calculations can be employed for both MC and MD simulations

where the energies of the whole system at each time step of simulations are calculated

by QM. Ab initio Molecular Dynamics is an example of these types of simulations1.

As mentioned in Chapter 2, computational chemistry methods can describe

physical and chemical processes in a wide range of zeolites and their applications.

These techniques are usually categorised according to the size of the chemical system

that they are able to simulate and the attainable accuracy 6 as shown schematically in

Fig.3-1. Fig. 3-1a describes the relationship between physical parameters, length scale

against time scale for three fundamental computational methods. Fig 3-1b shows the

complementary alignment of molecule or structure size (in number of atoms) versus

computational accuracy for these same three computational methods. Thus, for a

molecular system, it is critical to adopt appropriate length and time scales for the

modelled system aligned with application of an appropriate method and accuracy.

Higher accuracy of models requires substantially more computational time that may

restrict calculations to smaller sized systems. Thus, there is not a formulaic approach

to implement a modelled solution but rather a balance between dimensional and

energetic accuracy, speed of computation, and structure size is required to obtain

useful and predictable outcomes from such techniques.

Figure 3-1 Hierarchy of (a) Time and Length scales and (b) Accuracy of different computational methods.


In this study, an in-depth understanding of charge distribution, bonding, ion

exchange and diffusivity of ions in the zeolite N system will be obtained by using a

combination of quantum mechanical and molecular mechanical simulations from

electronic to macroscopic scales over a range of length and time scales. A theoretical

introduction to these simulation methods is given in the following sections.

3.2 QUANTUM MECHANICS

Quantum mechanics can provide insight into a wide range of properties of a

molecular system. These properties not only include structural and thermodynamic

values but also properties dependent on the electronic distribution in a system. Based

on the laws of quantum mechanics, the energy state and other related properties of a

many-electron system are acquired by solving the time-independent Schrödinger

equation2:

Equation 3-1

This equation implicitly assumes the Born-Oppenheimer approximation, where

the H is a Hamiltonian, associated with the kinetic and potential energies of an

electronic system and Ψ is the many-electron wave function that describes the ground

state of a system. In quantum mechanics, having determined the ground state

wavefunction, Ψ, by solution of equation 3-1, the ground state electronic energy,

electron density and any other ground state property can be calculated2.

There are two main quantum mechanical calculation methods ab initio and semi-

empirical methods. Ab initio is a Latin phrase that means “from first principles”. In

this method, only real physical constants are used in quantum mechanical calculations

and the Schrödinger equation can not be solved exactly for many-bodied systems and

a hierarchy of approximations provides a hierarchy of ab initio methods, which, in

principle, can yield essentially exact solutions2. In contrast, semi-empirical methods

use experimentally derived parameters or neglect some terms to simplify the equations

and reduce computational costs. Ab initio methods deliver more accurate properties of

some molecular systems compared to semi-empirical methods, since they are not

limited to experimental approximations. However, practical ab initio methods have

limitations that lead to systematic errors, while semi-empirical methods are fitted


against experiment and can be very accurate for special molecular systems. There is

another method called Density Functional Theory that investigates the structural,

magnetic and electronic properties of a many-body system. In this study, the structure

and properties of zeolite N in its ground state are determined using Density Functional

Theory (DFT) as described in the following section.

3.2.1 Density Functional Theory (DFT)

Density Functional Theory (DFT) is an electronic structure method based on the

electron density distribution of atoms7. The advantage of DFT calculations is that it

considers the three-dimensional distribution of electrons which is much simpler to

converge than the molecular orbitals that need to be described in molecular orbital-

based ab initio methods with a many electron wave function used in Schrodinger’s

equation8 . In addition, the simple formalism of DFT results in significantly reduced

computational costs compared with some wave function calculations8, 9 and increased

calculation accuracy2, 7.

In the DFT method, the total energy E(ρ) of a system with charge density ρ is

calculated from Equation 3-2:

Equation 3-2

Where, EK(ρ) is the kinetic energy, EC(ρ) is the Coulombic energy associated

with electron-electron interactions and EXC(ρ) is associated with a combination of

exchange and correlation energies and also is a function of density. The kinetic energy

EK(ρ) can be obtained from Kohn-Sham wave functions of the system or from charge

density for special systems. Moreover, the electrostatic energy EC(ρ) can be calculated

by the sum of the electron-nucleus attractions and electron-electron repulsion.

However, the exchange-correlation energy EXC(ρ) is derived from local densities, ρ,

which is assumed to be that of a homogeneous gas. There are various approximations

available with different approaches and accuracies10, including Local Density

Approximation (LDA) 11, 12, Generalised Gradient Approximation (GGA) 12, 13, Meta-

Functionals (meta-GGA functionals)14 and Hybrid functionals15, 16. In this study, the

first two approximations are used to define the structural properties that aligns with

experimentally-determined crystallographic data on zeolite N17.


Local Density Approximation (LDA)

In this method, the exchange-correlation energy is only derived from local

densities This method assumes electrons in a homogeneous electron gas model to

calculate exchange-correlation energy. This method provides results with lower

computational costs, but with a lower level of accuracy11, 12. The VWN12 and PWC13

are two examples of many LDA functionals based on parameterisation.

Generalised Gradient Approximation (GGA)

Unlike LDA, which is based on uniform density that results in the overestimation

of exchange-correlation energy, the GGA method takes account of heterogeneous

electron density for the exchange-correlation energy. These gradient corrections result

in more accurate computational estimations12, 13. The PW9113 and PBE14 are

commonly used GGA functionals.

3.2.2 DFT calculation on zeolite N structure

In a solid structure, whether crystalline or amorphous, the charges on atoms in

the structure greatly influence general properties of the structure. The partial charges

on atoms within the zeolite N structure influence the magnitude of electrostatic

interactions both within the structure itself and involving extra-framework ions and

water molecules. In this study, structural properties and partial atomic charges of

zeolite N are determined by calculating electron density, electro-statistics and

population analysis using various DFT quantum mechanical methods. The details of

DFT models, parameters and settings are presented in Chapter 5. The DMol3 software

module is employed for DFT quantum mechanical calculations and analyses.

DMol3

In this study, all DFT calculations and analyses were conducted using DMol3 18,

19 (DMol3; Accelrys Inc.: San Diego, CA, 2016.) code in Accelrys Materials Studio

software package 2017. Using density functional theory (DFT), this code allows

prediction of structural, electronic, electrostatic, energetic and thermodynamic

properties and geometry optimization of structures. This module offers modelling of

three-dimensional organic and inorganic materials, periodic systems, solids and

surfaces with high accuracy and reasonably low computational cost.


3.3 MOLECULAR MECHANICS

In molecular modelling, the properties of very large systems cannot be described

by quantum mechanical calculations. Quantum mechanical methods consider the

electrons in a calculation of the energy and properties of a system, so these calculations

are computationally time-consuming for large molecular systems. However, molecular

mechanics methods, calculate the energy based on nuclear positions rather than

electronic motions. Thus, molecular mechanics methods are able to simulate systems

with a significant number of atoms and calculate their structure and properties at

reasonable levels of accuracy and lower computational cost compared to quantum

mechanical methods.

3.3.1 Force Fields

In molecular mechanics methods, the potential energy U can be described by

four key contributions: bond stretching, angle bending, bond torsion, and non-bonded

interactions (Coulombic and Lennard-Jones):

Equation 3-3

Where, kb, kθ and kφ are bond, angle and torsion constants, respectively, b and θ

are immediate bond length and angle, respectively, while b0 and θ0 are the initial bond

length and angle. The multiplicity and phase of dihedral angle φ are represented by n

and β in the third component. In the last component, ɛ is the depth of the potential

energy well that represents the energy of van der Waals interactions, σ is the distance

at which the van der Waal’s potential is a minimum, rij is the distance between atom

pairs i and j, qi is the partial charge of atom i and ke is dielectric coulomb constant.

Figure 3-2 schematically represents these four potentials.

All parameters and constants in equation 3-3 as well as torsions and out of plane

interactions are mandatory for developing a molecular mechanics simulation using

force fields. For example, defining the partial charges of atoms and Lennard-Jones

parameters of all atom pairs in the system is crucial for a molecular mechanics

simulation. The parameters of each component in equation 3-3 can be determined from


experiments or approximated via obtained results for bonds, angles, dihedral and

partial charges from quantum mechanics calculations2.

Figure 3-2 Schematic representation of the four key components of molecular mechanics force fields. The green, blue and red balls represent atoms, the black solid lines show covalent bonds and the

dashed lines denote non-bond interactions, the green and red atoms indicate a Coulombic interaction and the green and blue atoms a Lennard-Jones interaction.

In general, force fields take two forms including rule-based force fields and

parametrised fitted force fields, which include two generations. The first generation

uses simple potential energy functional forms in equation 3-3. The AMBER20,

CHARM21, OPLS22 and GROMAS23 force fields are in this category. The second

generation uses higher-order anharmonic potentials and cross-coupling functional

forms for intermolecular interactions. The UFF24 PCFF25 and COMPASS26 force fields

are some examples from this generation. The key difference between two generations

relates to description of valance functionals, including diagonal and off-diagonal

cross-coupling functionals, and non-bond interaction functionals.

Choosing the appropriate force field is critical because the accuracy of the

selected force field affects the reliability of simulation outcomes27. In order to select

the most suitable force field, the following factors should be considered:

Are all of the atoms characterised by the force field?

Are there any similar studies for comparison?

What is the quality of the calculated force field? (with respect to energy

expressions and parameters)

Are the force field results validated with trustworthy data?

In this study, to simulate the ion exchange behaviour of zeolite N in an aqueous

environment we need a force field that is suitable for condensed phases and able to


describe the solid-liquid interactions. Therefore, Condensed-phase Optimised

Molecular Potential for Atomic Simulation Studies (COMPASS) force field26 is

implemented for all molecular mechanics simulations including, geometry

optimisations, energy minimisations and molecular dynamics calculations.

COMPASS Force Field

Condensed-phase Optimised Molecular Potential for Atomic Simulation Studies

(COMPASS) force field26 is a second generation force field derived from ab initio

data. The functional form in equation 3-4 is used in this force field26.

Equation 3-4

Where, γ is the out-of-plane angle and the rest of parameters are same in equation

3-3. In this force field, the Lennard-Jones parameters for unlike atom pairs are different

from the L-J- 9-6, ɛ and σ, of like atom pairs. For unlike atom pairs the off-diagonal

parameters are calculated from equations 3-5 and 3-626.


Equation 3-5

Equation 3-6

In this force field the parametrization is separated into two stages. In the initial

stage, partial charges and valance parameters are derived by fitting to ab initio

potential energy surfaces. Then in the next stage, the force field parameters are

optimised to reach good agreement with experimental data. The COMPASS force field

uses different models to describe different systems. The CFF valance model26, the

ionic model and a semi-ionic model are used to describe the covalent, ionic and metal

oxide systems. The validations of this force field for zeolites28, 29 were conducted based

on energy minimization on crystals. Moreover, this force field has been used

successfully to predict the molecular structure, properties and behaviour of organic,

inorganic and polymeric materials30, 31.

3.3.2 Energy Minimization

The stability of a structure can be calculated by determining the minimum

potential energy for a particular configuration of atoms. The calculation of a minimum

potential energy configuration is described in computational modelling as geometry

optimization or energy minimization.

In this study, energy minimization is used to remove the stress from the initial

configuration of the system and bring it to a relaxed condition before molecular

dynamics simulations. The Quasi-Newton minimization method32, 33 is implemented

in all geometry optimizations prior to MD simulations described in following chapters.

3.3.3 Molecular Dynamics Simulations

Molecular dynamics (MD) simulations based on molecular mechanics

calculations allows simulation of the time-dependent behavior of a large molecular

system in long time duration with shorter time computation compared to quantum

mechanical calculations. MD simulations are categorized into three different

approaches: ab initio MD34, classical MD and coarse grain MD35.

Classical MD models the nuclear interaction of particles. Coarse grain MD

considers a single particle as a representation of a cluster of atoms for calculating the


empirical potential. In ab initio MD, the atomic forces are derived from first-principle

quantum mechanics calculations.

The choice between these three MD methods depends on the desired level of

information from simulating the molecular system. The classical and coarse grain MD

methods provide results with lowest accuracy and least amount of information of the

behavior of molecular system. However, ab initio MD, using electronic structure-

based parameters, provides the highest level of detail and of accuracy for chemical

behavior of a physical system.

Molecular dynamics simulations generate trajectories that describe the positions

and velocities of the particles (with specific mass) in a finite system over time.

Equation 3-7 represents Newton’s second law that provides the trajectories:

Equation 3-7

Where, Fi is the force on the ith particle of the system with mi mass at xi position

and U is potential energy obtained from Eq. 3-4. In order to simulate the properties of

a molecular system with initial coordinates and velocities, MD simulation repeats the

following steps for a specified simulation time:

a) Calculating the potential energy of bond length, angle, torsion and non-bond

interactions as denoted in Eq. 3-4 for each atom at current time step t0,

b) Calculating the force vectors on each atom as derivatives of potential energy

at t0 from Eq. 3-7,

c) Computing the acceleration vectors from Eq. 3-7 for t0,

d) Computing and updating the coordinates and velocity vectors for each atom

for the next time step (t+δt).

The positions (xi), velocities (vi) and accelerations (ai) of atoms in a molecular

system can then be predicted for the next time step (t+δt) of MD simulation. The time

step δt significantly depends on the integration method of Eq. 3-7. The velocity-Verlet

algorithm36 is the most commonly used algorithm in MD simulations. This algorithm

is used as a three-step procedure as in the equations below:

Equation 3-8


Equation 3-9

Equation 3-10

In this study, we implement the velocity-Verlet algorithm for all MD

simulations.

3.3.4 Equilibrium Ensembles

The thermodynamic state in a MD simulation is kept constant using different

equilibrium ensembles based on the preserved thermodynamic parameter. The number

of particles is not constant in all ensembles, for example as in grand canonical

calculations. The notation of ensembles represents the variables considered constant

during the time period of the MD simulations. The various thermodynamic ensembles

are summarised below:

NVE: is a micro-canonical (constant internal energy, E) ensemble

characterisation related to adiabatic systems with unrestricted

temperature and pressure.

NPT: is an isobaric-isothermal ensemble applied for periodic systems.

NPH: is an isoenthalpic-isobaric ensemble applied on periodic systems.

NVT: is a canonical ensemble

The temperature and pressure in NPT, NPH and NVT ensembles are controlled

using various thermostats and barostats, including Velocity Scaling37, Nose38-40 41,

Andersen42, Berendsen43 and NHL44, 45 thermostats and Parrinello46, Andersen42,

Berendsen43 and Souza-Martins47 barostats. In this study, the dynamical behaviour of

zeolite N configurations are simulated using the NVT ensemble with the NHL

thermostat.

3.3.5 MD Simulation Procedure

In this section, the general procedure for MD simulations is introduced and the

detailed simulation settings are provided in each related chapter.


Software

In this study, the Forcite module in the Materials Studio suite of programmes48,

licenced from Biovia and implemented on the QUT High Performance Computer

(HPC) system is used to optimize the geometry of zeolite N conformations prior to the

MD simulations on the targeted zeolite N systems. Forcite is a module that allows

study of a wide range of molecular systems using various molecular mechanics tools.

This module implements several force fields to approximate the potential energy

surface.

Model Construction

In molecular modelling, the starting structure of a molecular system that

represents initial atomic positions can be imported from crystallography or nuclear

magnetic resonance (NMR) experimental studies or built in software programs. In this

study, experimental data on the structure of zeolite N in the form of CIF files from

previous work17, 49 is used for initial model construction. The partial atomic charges

for zeolite N framework atoms obtained from DFT calculations are used and described

in Chapters 4 and 6. The designed simulation models are described in-detail in

Chapters 4 and 6.

Boundary Conditions

The type of boundary conditions are identified before simulation setups. In this

study, a periodic boundary condition is applied in all three directions, x, y and z.

Atomic and lattice constraints are used to satisfy the MD simulation targets.

Simulation Setup

For geometry optimisation setup prior to MD simulations, the minimisation

algorithm, convergence criteria quality and the number of iteration steps are

determined.

For a dynamic setup, the quality of calculations, thermodynamic ensemble,

thermostats, barostats, desired temperature and pressure, the total simulation time,

time steps and number of steps are identified.

Energy Calculation Setup

The desired force field to calculate the potential energy of the system is

identified. The summation methods for the potential energy of electrostatic and van

der Waals interactions and preferred cut-offs are determined.


Simulation Run

Conducting a MD simulation can be divided into three steps:

Geometry optimization, that minimizes the energy of the system to the

accepted level prior to MD simulations.

Equilibration MD simulations, that conducts the MD calculations on

optimised molecular system for a few ps to deliver an equilibrium

condition for temperature or pressure.

Production MD simulation, that performs the MD simulation for

desired simulation time (usually many ns) in order to collect the

trajectories for further analysis.

3.3.6 Analysis

The Forcite module provides a three-dimensional atomistic trajectory containing

all atomic positions at each time step of the MD simulation run. The analysis function

in Forcite allows extraction of the dynamics, statistics and structural information of

the system simulated from the trajectories.

Dynamic analysis, provides information about time-dependent

properties, including, mean square displacement, velocity correlation

function, velocity and temperature profiles, etc.

Statistical analysis, represents evolution of properties frame-by-frame,

for example, temperature, pressure, kinetic energy, potential component

energy, density, etc.

Structural analysis, provides information about evolution and

distribution of structural parameters, such as length, angle, torsion,

atomic concentration, radial distribution function, etc.

3.3.7 Limitations of MD simulations

A variety of sources cause errors in the results of MD simulations. These sources

include inaccuracy of interatomic potentials (e.g. applied force field and the

summation methods for calculating the electrostatic and van der Waals interactions),

length and time scale of simulations, the constructed model and statistical

uncertainties. The examples of calculating some of these errors are included in Chapter


6. Moreover, the limitations of molecular modellings conducted in this thesis are

reviewed in Chapter 8.


3.4 REFRENCES

1. Smit, B.; Maesen, T. L. M., Molecular simulations of zeolites: adsorption, diffusion, and shape selectivity. Chemical Reviews 2008, 108, 4125–4184. 2. Lewars, E. G., Computational chemistry: introduction to the theory and applications of molecular and quantum mechanics. Springer: 2016. 3. Alder, B. J.; Wainwright, T. E., Studies in molecular dynamics. I. General method. The journal of chemical physics 1959, 31 (2), 459-466. 4. Rahman, A., Correlations in the Motion of Atoms in Liquid Argon. Physical Review 1964, 136 (2A), A405-A411. 5. Metropolis, N.; Rosenbluth, A. W.; Rosenbluth, M. N.; Teller, A. H.; Teller, E., Equation of state calculations by fast computing machines. The journal of chemical physics 1953, 21 (6), 1087-1092. 6. Pidko, E. A.; Hensen, E. J. M., Chapter 3 - Computational Chemistry of Zeolite Catalysis A2 - Sels, Bert F. In Zeolites and Zeolite-Like Materials, Kustov, L. M., Ed. Elsevier: Amsterdam, 2016; pp 111-135. 7. Tsuneda, T., Density functional theory in quantum chemistry. Springer: 2014. 8. Kohn, W., Nobel Lecture: Electronic structure of matter—wave functions and density functionals. Reviews of Modern Physics 1999, 71 (5), 1253. 9. Cohen, A. J.; Mori-Sánchez, P.; Yang, W., Insights into current limitations of density functional theory. Science 2008, 321 (5890), 792-794. 10. Cohen, A. J.; Mori-Sánchez, P.; Yang, W., Challenges for density functional theory. Chemical Reviews 2011, 112 (1), 289-320. 11. Vosko, S. H.; Wilk, L.; Nusair, M., Accurate spin-dependent electron liquid correlation energies for local spin density calculations: a critical analysis. Canadian Journal of physics 1980, 58 (8), 1200-1211. 12. Perdew, J. P.; Wang, Y., Pair-distribution function and its coupling-constant average for the spin-polarized electron gas. Physical Review B 1992, 46 (20), 12947. 13. Perdew, J. P.; Burke, K.; Ernzerhof, M., Generalized gradient approximation made simple. Physical review letters 1996, 77 (18), 3865. 14. Tao, J.; Perdew, J. P.; Staroverov, V. N.; Scuseria, G. E., Climbing the Density Functional Ladder: Nonempirical Meta-Generalized Gradient Approximation Designed for Molecules and Solids. Physical review letters 2003, 91, 146401. 15. Axel D, B., Density‐functional thermochemistry. III. The role of exact exchange. The journal of chemical physics 1993, 98 (7), 5648-5652. 16. Stephens, P. J.; Devlin, F. J.; Chabalowski, C. F.; Frisch, M. J., Ab Initio Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density Functional Force Fields. The Journal of Physical Chemistry 1994, 98 (45), 11623-11627. 17. Christensen, A. N.; Fjellvag, H., Crystal structure determination of zeolite N from synchrotron X-ray powder diffraction data. Acta Chemica Scandinavica 1997, 51, 969-973. 18. Delley, B., An all-electron numerical method for solving the local density functional for polyatomic molecules. The journal of chemical physics 1990, 92, 508. 19. Delley, B., From molecules to solids with the DMol3 approach. The journal of chemical physics 2000, 113, 7756. 20. Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P. A., A Second Generation Force Field for the Simulation of Proteins, Nucleic Acids, and


Organic Molecules. Journal of the American chemical society 1996, 118 (9), 2309-2309. 21. MacKerell, A. D.; Bashford, D.; Bellott, M.; Dunbrack, R. L.; Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.; Joseph-McCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, F. T. K.; Mattos, C.; Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E.; Roux, B.; Schlenkrich, M.; Smith, J. C.; Stote, R.; Straub, J.; Watanabe, M.; Wiórkiewicz-Kuczera, J.; Yin, D.; Karplus, M., All-Atom Empirical Potential for Molecular Modeling and Dynamics Studies of Proteins. The Journal of Physical Chemistry B 1998, 102 (18), 3586-3616. 22. Jorgensen, W. L.; Maxwell, D. S.; Tirado-Rives, J., Development and Testing of the OPLS All-Atom Force Field on Conformational Energetics and Properties of Organic Liquids. Journal of the American chemical society 1996, 118 (45), 11225-11236. 23. Schuler, L. D.; Daura, X.; Gunsteren, W. F. v., An improved GROMOS96 force field for aliphatic hydrocarbons in the condensed phase. Journal of computational chemistry 2001, 22 (11), 1205-1218. 24. Rappé, A. K.; Casewit, C. J.; K. S., C.; Goddard, W. A.; Skiff, W. M., UFF, a full periodic table force field for molecular mechanics and molecular dynamics simulations. Journal of the American chemical society 1992, 114, 10024. 25. Sun, H.; Mumby, S. J.; Maple, J. R.; Hagler, A. T., An ab initio CFF93 all-atom forcefield for polycarbonates. Journal of the American Society 1994, 116, 2978-2987. 26. Sun, H., COMPASS: an ab initio force-field optimized for condensed-phase applications overview with details on alkane and benzene compounds. The Journal of Physical Chemistry B 1998, 102 (38), 7338-7364. 27. McDaniel, J. G.; Schmidt, J. R., Next-Generation Force Fields from Symmetry-Adapted Perturbation Theory. Annual Review of Physical Chemistry 2016, 67, 467-88. 28. Hill, J. R.; Sauer, J., Molecular mechanics potential for silica and zeolite catalysts based on ab initio calculations. 1. Dense and microporous silica. The Journal of Physical Chemistry 1994, 98 (4), 1238-1244. 29. Hill, J.-R.; Sauer, J., Molecular Mechanics Potential for Silica and Zeolite Catalysts Based on ab Initio Calculations. 2. Aluminosilicates. The Journal of Physical Chemistry 1995, 99 (23), 9536-9550. 30. Zhang, H.; Ouyang, D.; Murthy, V.; Wong, Y.; Xu, Z.; Smith, S. C., Hydrotalcite intercalated siRNA: computational characterization of the interlayer environment. Pharmaceutics 2012, 4 (2), 296-313. 31. Murthy, V.; Smith, H. D.; Zhang, H.; Smith, S. C., Molecular Modeling of Hydrotalcite Structure Intercalated with Transition Metal Oxide Anions: CrO42–and VO43–. The Journal of Physical Chemistry A 2011, 115 (46), 13673-13683. 32. Dai, Y.-H., Convergence Properties of the BFGS Algoritm. SIAM Journal on Optimization 2002, 13 (3), 693-701. 33. J. E. Dennis, J.; Moré, J. J., Quasi-Newton Methods, Motivation and Theory. SIAM Review 1977, 19 (1), 46-89. 34. Marx, D.; Hutter, J., Ab initio molecular dynamics : basic theory and advanced methods. Cambridge University Press: Cambridge, UNITED KINGDOM, 2009. 35. Balbuena, P. B.; Seminario, J. M., Molecular dynamics : from classical to quantum methods. Elsevier Science & Technology: Oxford, Netherlands, 1999. 36. Verlet, L., Computer" experiments" on classical fluids. I. Thermodynamical properties of Lennard-Jones molecules. Physical Review 1967, 159 (1), 98.


37. Bussi, G.; Donadio, D.; Parrinello, M., Canonical sampling through velocity rescaling. The journal of chemical physics 2007, 126 (1), 014101. 38. Nosé, S., A molecular dynamics method for simulations in the canonical ensemble. Molecular Physics 1984, 52 (2), 255-268. 39. Nosé, S., A unified formulation of the constant temperature molecular dynamics methods. The journal of chemical physics 1984, 81 (1), 511-519. 40. Shuichi, N., Constant temperature molecular dynamics methods. Progress of Theoretical Physics Supplement 1991, 103, 1-46. 41. Hoover, W. G., Canonical dynamics: Equilibrium phase-space distributions. Physical Review A 1985, 31 (3), 1695. 42. Andersen, H. C. J. T. J. o. c. p., Molecular dynamics simulations at constant pressure and/or temperature. The journal of chemical physics 1980, 72 (4), 2384-2393. 43. Berendsen, H. J.; Postma, J. v.; van Gunsteren, W. F.; DiNola, A.; Haak, J., Molecular dynamics with coupling to an external bath. The journal of chemical physics 1984, 81 (8), 3684-3690. 44. Samoletov, A. A.; Dettmann, C. P.; Chaplain, M. A., Thermostats for “slow” configurational modes. Journal of Statistical Physics 2007, 128 (6), 1321-1336. 45. Leimkuhler, B.; Noorizadeh, E.; Penrose, O., Comparing the efficiencies of stochastic isothermal molecular dynamics methods. Journal of Statistical Physics 2011, 143 (5), 921-942. 46. Martyna, G. J.; Tobias, D. J.; Klein, M. L., Constant pressure molecular dynamics algorithms. The journal of chemical physics 1994, 101 (5), 4177-4189. 47. Souza, I.; Martins, J., Metric tensor as the dynamical variable for variable-cell-shape molecular dynamics. Physical Review B 1997, 55 (14), 8733. 48. BIOVIA, D. S. Materials Studio, 18.1; San Diego: Dassault Systèmes: San Diego, USA, 2018. 49. Ekhteiari Salmas, R.; Demir, B.; Yıldırım, E.; Sirkecioğlu, A.; Yurtsever, M.; Ahunbay, M. G., Silver–Sodium Ion Exchange Dynamics in LTA Zeolite Membranes. The Journal of Physical Chemistry C 2013, 117 (4), 1663-1671.


Chapter 4: Modelling Hydration Behaviour of Zeolite N

In this chapter molecular dynamics (MD) simulations are used to investigate the

hydration energy and ion exchange properties of synthetic Zeolite N. Section 4.1

contains an introduction on molecular modelling methods and their application to

study the ion exchange property of zeolites. Section 4.2 describes the methodology for

investigating the exchange of K+ ions with univalent ions such as NH4+, Na+, Rb+ and

Cs+ under a range of simulation conditions using a three dimensional membrane in an

electrolyte box containing explicit water molecules. The obtained results from the

hydration energy calculations and ion exchange simulations are represented in section

4.3 and followed by a comprehensive discussion in section 4.4. Moreover, this section

evaluates the validity of simulations by comparing the results with available

experiments. Finally, section 4.5 provides outcomes from this study.

4.1 INTRODUCTION

Zeolites are widely used in industrial processes including catalysis1-3, gas

separation4-6 and ion exchange7, 8. Early applications relied on systematic experiments

based on knowledge of the framework structure and composition9, 10 to validate or infer

likely success with catalysis, separation or exchange functionality. With the advent of

powerful computational tools, including molecular and atomistic modelling of

complex structures, structural and compositional data are used to accurately predict

zeolite behaviour under specific conditions11, 12 . For example, the dynamics of

exchange between Na+ and Ag+ ions in synthetic Zeolite A (LTA-type) in contact with

an electrolyte solution calculated with Density Functional Theory (DFT) and

Molecular Dynamics (MD) simulations conform closely with experimental data over

a range of temperatures and concentration gradients13. Other simulations have

successfully explored the cation selectivity in Zeolite A 14 and flow of water through

Zeolite A membranes with application of pressure for hydrophilic and hydrophobic

surfaces15. For more complex structures, DFT calculations have been used to

determine the site preferences for Ca2+ and K+ in the natural zeolite clinoptilolite16.

DFT methods have also been used to estimate the stability of the clinoptilolite structure


after various levels of de-alumination by acid treatment17, 18. In these cases, the

dominant exchangeable cation within the zeolite interstices is sodium.

These successes stimulated our interest in evaluating the exchange kinetics of

univalent ions in the potassium-rich zeolite N. The structure of zeolite N determined

by Christensen and Fjellvag19, 20 is orthorhombic (Space Group I222), classified as an

EDI framework structure with end-member composition K12Cl2 [Al10Si10O40] (H2O)5.

Potassium is known to exchange with NH4+ and Na+ while the structure shows high

selectivity towards NH4+ over other cations including divalent ions7, 21. Synthesis of

this zeolite has been undertaken at both ambient and hydrothermal conditions from a

range of source aluminosilicates such as kaolin22-24, montmorillonite22 meta-kaolin24,

25 and zeolite23. A detailed understanding of zeolite N ion exchange properties is of

interest as it shows high potential for a range of applications that require control of

nitrogen-rich nutrients in the environment6, 8.

MD simulation is widely recognized as a robust tool that can provide atomic

level insight into the distribution, exchange and mobility of ions and water in a zeolite

framework. MD simulations have been conducted to study ion exchange dynamics by

Salmas et al.13, Murad et al.26, 27 and Nalaparaju et al.28 on several zeolite membranes.

While Murad et al.26, 27 have used explicit water molecules on NaA and Na-ETS-10

membranes, Salmas et al.13 and Nalaparaju et al.28 have used implicit solvent methods

with a dielectric continuum model for ionic solutions on LTA and Na-ETS-10

membranes, respectively to reduce computational costs. In this study, we report ion

exchange processes in hydrated zeolite N with explicit water molecules. We also

develop a model for a slab of zeolite N containing eight unit cells with explicit

representations of water within the structure and on either side of the slab. This

approach allows an evaluation of hydration energy and the relative diffusion of ions

and water into and out of the framework structure to the surrounding electrolyte. We

use non-equilibrium MD calculations to investigate the transport of water molecules

and ions within, into and out of a thin membrane of zeolite N driven by chemical

potentials on either side of the framework slab.

4.2 METHODS

The initial structural parameters are from a synchrotron X-ray study of Zeolite

N by Christensen and Fjellvag19. The unit cell structure orthorhombic


distortion with lattice parameters a =9.9041 Å, b = 9.8860 Å and c =13.0900 Å. The

zeolite N framework is similar to edingtonite in which SiO4 and AlO4 tetrahedra form

an ordered framework19, 20 as shown in Figure 4-1. The zeolite N framework is

constructed of α-cages by the sharing of eight-membered rings in which the corner

sites are alternatively occupied by Si and Al atoms. This arrangement results in

complete Si/Al ordering with Si/Al = 1. Ion exchange processes in hydrated zeolite N

with explicit water molecules are evaluated by MD simulations with an ab initio force-

field and DFT generated partial charges on all atoms.

Figure 4-1 Zeolite N (2x2x2) supercell viewed along (a) [001] and (b) [110] crystallographic

directions. Yellow represent Si atoms (or silica tetrahedra), pink = Al (or alumina tetrahedra), red = oxygen, white = hydrogen, lilac = potassium ions and light green = chlorides.

4.2.1 Geometry Optimisation and Ionic Charge

MD simulations are performed using Forcite in Materials Studio (MS) 8.1 and

MS 201729, 30. The COMPASS Force Field31, which is a general ab initio-derived force


field, is used for all geometry optimizations and MD simulations. COMPASS is a

general force field that consolidates parameters for organic and inorganic materials.

COMPASS force field is assigned to all atoms in the zeolite N framework, extra

framework ions and water molecules. The flexible SPC water model32 incorporated in

COMPASS is used for water molecules.

Partial charges on all atoms of zeolite N are calculated by periodic DFT methods.

Initially, calculations are performed with extra-framework ions (e.g. K+, Cl-) and water

molecules removed from the structure. With this configuration, the total charge of the

unit cell framework is set to -10 electrons and geometry optimization is obtained using

the GGA/PW91 functional33. This calculation includes an all-electron, double

numerical basis set with d functions (DND 3.5, comparable to a Gaussian 6–31G* basis

set) on all non-hydrogen atoms. The SCF convergence criterion is set at an energy

change of 10-5 Hartree (Ha). The convergence criterion for geometry optimisations are,

for maximum energy, force and displacement convergence, 2x10-5 Ha, 0.004 Ha/Å and

5 x10-3 Å, respectively. Charge calculations are performed with the DMol3 program

in Materials Studio (MS) 8.129, 30. The geometry optimized unit cell is cleaved and

capped with –OH groups on both surfaces along the (001) plane with a vacuum slab

of 5 Å. The cell is optimized by DFT using the constraints given above to obtain the

partial charges on the surface of O and H atoms. The atomic charges are determined

with the Mulliken algorithm of population analysis for DMol3 calculations. Then, the

defined partial charges by DFT calculation were used by incorporation into the

COMPASS force field atom types. The charges and force field atom types in this study

are given in Table 4-1.

Table 4-1 Partial charge and COMPASS force field atom types used on all atoms

Atom COMPASS force field atom type Charge Zeolite N Si si4z 1.54 Al al4z 1.11 Obulk o2z -0.912 Osurface o2* -0.701 Hsurface h1o 0.245 N in NH4

+ n4+ -0.783 H in NH4

+ h14 0.446 H in H2O h1o 0.41 O in H2O o2* -0.82 K+ k+ 1 Na+ na+ 1 Rb+ rb+ 1 Cs+ cs+ 1


4.2.2 Molecular Dynamics of Zeolite N membrane

The hydration energy and ion exchange process parameters for zeolite N while

submerged in water and/or electrolyte solution, are determined for a membrane using

the following approach. A 2×2×2 supercell is created, cleaved along the (001) plane

(with a vacuum slab of 20 Å on either side). Therefore, the size of simulation box is

19.8×19.7×68.2 Å3 with 2524 atoms. Terminal Al and Si atoms are then capped with

–OH groups. Thus, a zeolite N membrane as shown in Figure 4-2 and labelled as “ZM”

is generated. Initially, water molecules with a density of 1g/cm3 are added to the Water

Layer 1 (WL1) and Water Layer 2 (WL2) regions of the model. In subsequent

calculations, the numbers of water molecules and ions are varied to create a chemical

potential inside and outside of the zeolite membrane. For all atoms, calculated partial

charges determined by the method described in section 4.2.1 and listed in Table 4-1

are used.

Figure 4-2 Zeolite N membrane (ZM) in water (water layers, WL1 and WL2). Yellow represents Si atoms, pink = Al, red = oxygen, white = hydrogen, lilac = potassium and light green = chlorine.

Prior to MD simulations, geometry optimization of the zeolite membrane in

water is carried out. Electrostatic interactions are calculated by the Particle-Particle

Particle-Mesh method (PPPM)34 and van der Waals forces are determined by the

Ewald summation method35 with a cut-off distance of 12 Å, similar to previous

studies13, and minimizations carried out by a Quasi-Newton procedure36. Periodic

boundary conditions are applied in three dimensions so that the simulation cell is

effectively repeated infinitely in each direction. Initially, the zeolite framework is held

rigid to allow the extra-framework species (ions and water molecules) to vary and to

optimize to a minimum with respect to each other. These optimized structures are then

used as the starting configurations for MD simulations, performed in the NVT-


ensemble (constant-volume/constant-temperature) where framework and extra-

framework atoms are released. Since all atoms in each system are completely free to

move during these simulations, use of the constant-volume model with a fixed cell

shape does not introduce significant limitations to the resulting zeolite structure, nor

to the dynamics and energetics of water and ion exchange. The structure of the

framework is found to be quite stable during simulations (the change in bond distances

are less than 0.05 Å). The small changes to the structure of the zeolite framework

during simulation are shown in Figure 1 in Appendix A.

MD simulations performed in an NVT-ensemble at 298 K with a time step of 1.0

fs is used for all simulations. Analysis of these simulations reveals that equilibrium

values for the thermodynamic parameters are generally achieved within the first 20 ps

using an Andersen thermostat37. An MD simulation of 30 ps using the Andersen

thermostat is initiated and then followed by 500 ps and 8 ns simulations with the Nosé-

Hoover-Langevin (NHL) thermostat38, 39 for different hydration states, n, of the

system.

4.2.3 Hydration Energy

The hydration energy, ΔUH is a measure of the preferred hydration state(s) for

the zeolite framework and is defined by the equation 4.-1:

Equation 4-1

where, Nw is the number of water molecules and U(Nw) and U(0) are the total

potential energies of the system with Nw and zero water molecules, respectively.

For the calculation of ΔUH, we have used a 2x2x2 super cell for the bulk crystal

and a 2x2x2 slab for ZM. Initially, simulations of 2 ns are run with extra framework

ions (e.g. K+, Cl-) originally present in the crystal structure but water molecules are

removed from the structure. Water molecules are then randomly added to the center of

the supercell incrementally (for both bulk and ZM) and allowed to equilibrate by

running simulations of the same duration. We have calculated the hydration energy of

the ZM by creating a vacuum of 15 Å on either side of the ZM slab for water molecules

to flow out. Water molecules are added to the central section of the membrane

incrementally and allowed to transfer out to the surrounding water box over 2 ns

periods for each simulation.


4.2.4 Ion Exchange

The zeolite N membrane consists of 80 Al and Si atoms, 128 water molecules (8

H2O /cage), 96 K+ ions and 16 Cl- ions with a charge of -80e on the framework. To

create a chemical potential between the inside and outside of the membrane and to

check the retention of K+ ions by the membrane (“ZM” in Figure 4-2), an extra 80 K+

ions (to create an initial ratio of 1:1 between the original cations and the exchanging

cations) are placed inside the membrane and 40 Cl- ions are equally placed in the water

column on either side of the membrane. This computational step is designated as

“zeoliteN_K+/K+” in subsequent sections. Similarly, exchange of other univalent ions

is simulated by randomly placing cations such as NH4+, Na+, Rb+, or Cs+ inside the

ZM. These exchange calculations are designated as zeoliteN_K+/NH4+,

zeoliteN_K+/Na+, zeoliteN_K+/Rb+ and zeoliteN_K+/Cs+, respectively. The exchange

of ions between the zeolite N membrane and solution is determined after 8 ns of NVT

simulation. This method allows evaluation of cation preference(s) within the zeolite

N membrane. NVT-ensemble MD simulations of 1 ns duration, including an initial 30

ps time for equilibration, are undertaken to calculate the radial distribution functions

(RDF), mean square displacement (MSD), self-diffusion coefficient(s), and

concentration profiles of ions inside and outside the zeolite N membrane.

4.3 RESULTS

We present results from a priori calculations based on the potassium-rich end-

member composition for zeolite N as determined via both X-ray and neutron

diffraction studies19, 20. Experimental data suggest that other partially-exchanged

compositions are possible for zeolite N25 but are not considered in this work. This

fundamental study focuses on the hydration behaviour of zeolite N and the diffusion

of ions and counter-ions in a 3D membrane.

4.3.1 Hydration

A 2x2x2 zeolite N bulk and ZM contain 16 cages of aluminosilicate framework

with Si/Al = 1.0. The hydration energy (HE) of zeolite N obtained using equation 4-1

is calculated for different hydration levels by stepwise increase of water molecules into

the model as listed in Table 4-2a. The HE, number of water molecules retained by the

ZM, and pressure obtained from simulations undertaken with incremental additions of

water molecules to the ZM are shown in Table 4-2b.


Table 4-2 Hydration energy of zeolite N; a) Bulk and b) ZM

a) Zeolite N Bulk b) Zeolite N membrane Total H2O

loaded

H2O\ cage

HE (kcal\mol)

Total H2O

loaded

H2O retained

H2O \ cage

HE (kcals\mol)

Pressure

(GPa)

0 0 0 0 0 -2.67 64 4 -19.41 64 64 4 -19.54 -2.43 80 5 -18.06 128 99 6.2 -15.38 -1.19 96 6 -16.73 160 129 8.1 -13.04 -0.16

112 7 -14.69 192 128 8.0 -12.49 -0.34 128 8 -13.62 256 142 8.9 -10.46 0.15 144 9 -12.10 318 156 9.8 -11.31 0.12 160 10 -9.94 352 169 10.6 -11.29 0.10 176 11 -8.02 384 174 10.9 -6.74 2.84 192 12 -6.13

A plot of calculated HE for different amounts of water molecules per cage in

zeolite N is shown in Figure 4-3. In Figure 4-3a, the HE of bulk, is denoted by the red

line as the number of water molecules per cage increases. The blue line indicates the

number of water molecules retained per cage (after equilibration) in the ZM versus the

total number of H2O molecules added to the ZM initially (before equilibration). The

HE increases linearly with the number of water molecules per cage, with a slight dip

as the number of water molecules in zeolite N bulk approaches eight H2O/cage. Figure

4-3b shows the variation of HE and pressure with the number of water molecules per

cage in the ZM. Additional calculations related to structural behaviour and mobility of

water molecules, potassium and chlorine ions during hydration are provided in

Appendix A.

4.3.2 Ion exchange

The number and percentage retention of ions and water molecules inside the

zeolite N membrane are summarized in Table 4-3 and shown in Figures 4-4 and 4-5.

The data in Table 4-3 show time points of the simulation at 500 ps, 1 ns, 5 ns and 8 ns.

For all simulations, ion exchange is at an equilibrium condition after 1 ns, as

demonstrated by the limited exchange of ions at subsequent time points (Table 4-3).

Minor fluctuations in the number of ions and water molecules inside and outside ZM

indicate that the system is in dynamic equilibrium.


Figure 4-3 (a) Variation of HE (ΔUH(Nw)) in a bulk as a function of number of H2O/cage in the ZM (red) and equilibration of H2O/cage in ZM versus total number of H2O added before equilibration (blue). (b) Variation of HE and pressure in ZM as a function of the number of H2O/cage in ZM.

As shown in Table 4-3, at 8 ns for the ZeoliteN_K+/K+ system, 49.4 % of the K+

ions loaded into the membrane are retained and 50.6% are released to the solution

outside the membrane. Similarly, in the mixed Zeolite N_ K+/M+ systems, Na+, Rb+

and Cs+ show 62.5%, 60% and 60% retention respectively, while NH4+ ions exhibit

the highest retention of 68.8%. The retention ratio is determined from the number of

ions within the membrane at 8 ns simulation divided by the total number of ions in the

system. These values are listed in Table 4-3 and plotted over simulation times in Figure


4-4. The ratios of the exchanging ions (M+) to K+ ions inside the zeolite membrane

after 8 ns are 2.0 for NH4+/K+, 1.57 for Na+/K+, 1.56 for Rb+/K+, and 1.40 for Cs+/K+.

Figure 4-4 Ion retention ratio compared with K+ over 8 ns MD simulations for ZM.

The number of water molecules retained within the membrane over 8 ns are

plotted in Figure 4-5. The number of water molecules per cage in the zeoliteN_K+/K+

system fluctuates between 7.5 to 8 water molecules/cage with an average of 7.7

H2O/cage. ZeoliteN_K+/Na+ shows the highest average water content per cage of 8.3

H2O/cage. Calculations on zeoliteN_K+/Rb+ and zeoliteN_K+/Cs+ membranes show

an average of 6.4 and 6.2 H2O/cage, respectively. Simulation of the exchange of

ammonium ion with potassium (zeoliteN_K+/NH4+) shows that the membrane holds

7.3 H2O/cage.

Figure 4-5 Number of water molecules inside ZM over 8 ns of MD simulations.


Table 4-3 Number of ions in ZM with K+, NH4+, Na+, Rb+ and Cs+: before and after MD simulation.

ZeoliteN_K+/K+ ZeoliteN_K+/NH4+ ZeoliteN_K+/Na+ ZeoliteN_K+/Rb+ +/Cs+ZeoliteN_K

Time Ions ZM % ret Ions ZM %

ret Ions ZM % ret Ions ZM %

ret Ions ZM % ret

0 K+ 176 K+ 96 K+ 96 K+ 96 K+ 96

NH4 + 80 Na+ 80 Rb+ 80 Cs+ 80

500 ps K+ 88 50 K+ 36 37.5 K+ 40 41.7 K+ 37 38.5 K+ 39 40.6

NH4 + 59 73.8 Na+ 52 65.0 Rb+ 48 60.0 Cs+ 53 66.3

1 ns K+ 86 48.9 K+ 34 35.4 K+ 39 40.6 K+ 36 37.5 K+ 39 40.6

NH4 + 55 68.8 Na+ 49 61.3 Rb+ 51 63.8 Cs+ 52 65.0

5 ns K+ 84 47.7 K+ 34 35.4 K+ 39 40.6 K+ 36 37.5 K+ 37 38.5

NH4 + 55 68.8 Na+ 50 62.5 Rb+ 49 61.3 Cs+ 48 60.0

8 ns K+ 87 49.4 K+ 33 34.4 K+ 39 40.6 K+ 37 38.5 K+ 41 42.7

NH4 + 55 68.8 Na+ 50 62.5 Rb+ 48 60.0 Cs+ 48 60.0

4.3.3 Ion Distributions

This computational approach to evaluation of zeolite membrane behaviour

allows determination of the relative distribution of ions within the framework and in

the regions surrounding the framework. These tools include (a) calculated radial

distribution functions (RDFs) to determine the bonding characteristics of framework

and non-framework atoms and (b) measuring the positions of atoms within the

membrane after exchange simulations as described above. The RDFs for exchanged

ions with framework atoms O, Si and Al are shown in Figure 4-6a-e. In these figures,

the first peak for each framework atom represents the nearest neighbour distance of

the exchanged ion(s) to the framework atoms. These nearest neighbour distances are

compiled in Table 4-4. For comparison, Figure 4-6f also shows the RDF for

ammonium ions in the electrolyte surrounding the membrane.

Table 4-4 The nearest neighbour distance (Å) between univalent ions (M) with atoms in the zeolite N

framework, estimated from RDFs.

Ion (M) M-O M-Al M-Si

NH4+ 1.43 2.39 2.39

Na+ 2.07 2.79 2.83

K+ 2.49 3.25 3.39

Rb+ 2.65 3.47 3.45

Cs+ 3.03 3.59 3.59


The proportion of ions preferentially retained within the zeolite N framework in

comparison to those released into solution after 8 ns of MD simulation are determined

using concentration profiles or ionic density profiles. Figure 4-7 shows the number

distribution and the density field maps of ions along the z direction within the

membrane and in the solution outside the membrane. The number of ions at specific

locations within the membrane along the z axis are shown in Figure 4-7a-d for each

simulation of univalent ion exchange. These plots also track the number and species

of ion(s) present in the surrounding electrolyte after 8 ns of simulated exchange

reaction. The relative positions of each ion inside the framework obtained after

simulation of exchange reaction(s) for 8 ns are visualized through density field maps

and are shown in Figures 4-7f-g.

Figure 4-6 (a-e) RDFs, g(r), for non-framework ions to framework atoms in zeolite N and (f) for NH4+

to Ow in the electrolyte.


Figure 4-7 (a-e) Ion density profiles along the z direction: within ZM denoted by the vertical red dashed lines and in electrolyte solution (on either side of the red dashed lines) after 8 ns MD

simulations. (f) and (g) Density field maps for ions in the central cages (magnification of the region denoted by the green rectangle shown in ZM) of ZM: K+ is left hand panel (f) and M+;is right hand

panel (g); the relative intensity in density field maps increases from red to blue. (Yellow represents Si atoms, pink = Al, red = oxygen, white = hydrogen, lilac = potassium and light green = chlorine.

4.4 DISCUSSION

Experimental data on the ion exchange performance of zeolite N is documented

in the patent literature25 as well as in articles describing applications such as

wastewater treatment7, 21 and agronomy6, 8. These data show that the Cation Exchange

Capacity (CEC) for zeolite N powders ranges between 450 and 503 meq/100g for Si:Al


= 1.0 depending on synthesis conditions25. Experimental data show that the potassic

form of zeolite N prefers univalent ions over divalent ions in a multi-element aqueous

solution6, 25. For example, Table 4-5 provides examples of loading data for zeolite N25

for three ammonium concentrations in solutions with 50 mg/L Ca2+ and 20 mg/L Mg2+.

Similar preference for ammonium ion is observed when zeolite N is equilibrated with

solutions containing both 2,000 mg/L Na+ and 100 mg/L Ca+2 ions25.

Table 4-5 Experimental data on zeolite N ion exchange selectivity in mixed cation solutions25

NH4+ in starting

solution (mg/L) NH4

+ loading (meq/100g)

Ca+2 Loading (meq/100g)

Mg+2 Loading (meq/100g)

Solution 1 30 104 10 4 Solution 2 200 347 25 0 Solution 3 1000 444 18 0

For this study, we focus on univalent cation exchange in an electrolyte solution

containing Cl- as the counter-ion. This format is similar to experimental data which

also utilizes Cl- as the counter-ion in 200 mL aqueous solution with addition of 0.2g

of zeolite N equilibrated for periods of 1–2 hours25. These exchange reactions are pH

dependent as noted by Thornton et al.7 who demonstrate that maximum loading of

ammonium occurs in the pH range 6–7. For this study, we assume pH ~ 7 for MD

simulations.

Ion exchange models for zeolite structures developed from an interest in the

molecular behavior of sodic zeolites such ZK-440 and zeolite 4A41 (also identified as

Na-LTA13) under osmosis or reverse osmosis conditions. The basis for modelling a

zeolite membrane established in this early work, led to detailed comparison of ion

exchange by divalent and univalent ions26 as well as between Ag+ and Na+ in Na-

LTA13. In a study of supercritical and subcritical electrolyte solutions exchanged with

Na-LTA, Murad et al.26 demonstrate that for Na+ it is energetically favourable to

diffuse to the outside of a membrane. Salmas et al.13 show that for Na-Ag exchange in

LTA, the driving force for exchange is strongly influenced by electrolyte

concentration. In this work, we invoke a chemical potential between the membrane

and the electrolyte by inserting additional ions (e.g. K+ or Na+ etc) and water molecules

into the middle of ZM. We then allow the simulation to reach an equilibrium condition

over nine nanoseconds (including the 8 ns of production simulations and 1ns

simulation for calculating the diffusion of ions).


As noted by Salmas et al.13, an implicit or explicit water model for simulation of

LTA ion exchange shows no difference in outcome. In this work we utilize an explicit

water model, as it helps to determine the exact hydration state34 and the involvement

of water molecules during ion exchange. The framework charge is also an important

influence on ion exchange properties of aluminosilicates42. Using DFT, we calculate

the partial ionic charge on framework atoms for zeolite N as shown in Table 4-1, and

note that these values are similar to that determined by Salmas et al.13 for LTA zeolite.

4.4.1 Hydration

Ion exchange in zeolites is predominantly in an aqueous environment and, as

such, the atomic-scale dynamics of extra-framework cations should be considered in

the context of a hydrated system or membrane. Water molecules within zeolite

channels significantly affect the location of extra-framework cations in the structure20

and consequently, the ion-exchange properties of a zeolite. For example, strongly

hydrated extra-framework cations show reduced tendency to exchange structural

positions with other cations. In addition, the mobility of water molecules controls the

motion of the exchangeable cations and, as a result, controls the performance of ion-

exchange and diffusion processes42, 43.

Water has two roles in zeolites: (a) completing the coordination of available

cations inside the zeolite channels which increases their mobility and (b) minimising

the electrostatic repulsion of the bridging oxygen in the zeolite framework44, 45.

Moreover, the Al content in the zeolite framework controls the amount of adsorbed

water, because by decreasing the Si/Al ratio, the hydrophilicity of zeolite increases46.

Understanding the influence of water on behaviour of zeolites or, in other words, the

hydrophilicity of zeolites, can be enhanced by comparing the differences between

dehydrated and hydrated states of a zeolite system.

As shown in Table 4-2 and Figure 4-3a, incremental increases in the number of

water molecules to progressively hydrated states in ZM are observed after 2 ns

simulation. The plot in Figure 4-3a which combines the results of two different sets of

simulations i.e. bulk and ZM provides significant information. The HE curve for the

bulk (red line) increases linearly with the number of water molecules per cage,

exhibiting a slight dip as the number of water molecules in the cages approaches eight

H2O/cage and indicates that it is energetically favourable. This, in tandem with the

results of the ZM indicated in the blue line, shows a constant number of water


molecules retained per cage in the ZM over two increments of H2O molecules added

to the ZM. This condition also corresponds with eight water molecules per cage for

the zeolite N membrane. This outcome of eight water molecules per cage is consistent

with the experimental data for zeolite N based on X-ray and neutron diffraction

studies19, 20. This result validates our zeolite N membrane model and the methodology

developed to simulate water behavior in this material.

For the membrane system, the calculated pressure shows a similar trend as that

for change in hydration energy, HE, shown in Figure 4-3. For example, Table 4-3b

(Figure 4-3b) shows the change in HE as well as calculated pressure as water

molecules are added to the membrane. In this system, pressure increases in concert

with HE and when the membrane achieves a preferred number of water molecules (i.e.

eight per cage), pressure also plateaus at an equilibrium level up to addition of 384

water molecules (i.e. 11 per cage). At this stage and with further addition of water,

pressure increases rapidly and implies a maximum operating condition for a zeolite N

membrane.

4.4.2 Ion Exchange

Using this model for ZM with surrounding electrolyte, the exchange selectivity

by univalent ions for the zeolite N framework can be readily determined. Data in Table

4-3 provide a clear quantitative guide at specific points in time for the

retention/inclusion of ions relative to K+ in ZM. In general, NH4+ uptake is stronger

during initial phases of simulation (up to 500 ps) and reaches a stable condition more

rapidly than other ions. Figure 4.4 plots the ratio of exchanged ions to K+ within ZM

for the 8 ns simulation and clearly shows that NH4+ ion is preferred by zeolite N. Other

ions such as Na+, Rb+ and Cs+ will exchange but at lower levels (or rates). As

anticipated, the ratio of K+ within ZM before and after simulation remains relatively

constant, albeit the influence of H2O molecules is implied by the ~4% variation in ratio

of K+/K+ over time. The relative retention ratios shown in Figure 4-4 are consistent

with experimental data over a wide range of ion concentrations in aqueous solutions.

These general attributes of univalent ion exchange in zeolite N are determined by local

atomic bonding and interactions with the aluminosilicate framework.

Figure 4-5 shows the number of water molecules within ZM during univalent

ion exchange reactions up to 8 ns of simulation. The relative amounts of water

molecules follow the sequence Na+ > K+ > NH4+ > Rb+ > Cs+ with the highest water


contents for K+/Na+ exchange. This sequence is consistent with the relative increase in

ionic radius for these univalent ions. The water content in ZM with each ion exchange

reaction is highly variable and ranges from the average by 3%–4% for K+, up to 8%

for Rb+, with trends suggesting lower content over time for Na+ and K+ exchanges.

These variations in water content are difficult to interpret albeit the influence of cation

hydration spheres may be implicated.

4.4.3 Ion Localisation

Figures 4-6 and 4-7 provide three different measures of the locality of ions after

ion exchange simulations. For example, radial distribution functions (RDFs) measure

the intensity of distances between specific atom pairs. For zeolite N, the relative

distances between framework and non-framework atoms provides an average measure

of localized bonding influences within and around structural cages and are shown in

Figure 4-6 for each ion exchange simulation. Ion distribution plots are shown in Figure

4-7. These plots track the proportion of each ion along the z axis direction within ZM

as well as within the electrolyte and are shown in the centre panel of Figure 4-7. Ion

density maps are shown on the left and right hand panels of Figure 4.7 and provide

detailed density distribution(s) of K+ and M+ within the cage(s) of the zeolite structure

in the membrane.

The position of first peak obtained from RDFs for each of the simulated ion

exchange models in Figure 4-6 show that the values of first peak for O–K+, Al–K+ and

Si–K+ distances are 2.49 Å, 3.25 Å and 3.39 Å, respectively, for the calculated zeolite

N structure containing additional K+ ions. As expected, these values compare well with

the relative bond lengths calculated from X-ray diffraction data by Christensen and

Fjellvag19. This outcome validates the model and approach used to simulate zeolite N.

As shown in Table 4-4, with increased ionic size of cations, M+, exchanged into the

structure, O–M+ distances increase.

For the ZeoliteN_K+/NH4+ simulation (Figure 4-6e), the first peak at 1.43 Å

corresponds to the HNH4+ to framework O distance. This distance is smaller than the

NH4+ hydrogen bonded to the oxygen atom in water (Ow) distance of 1.75 Å shown in

Figure 4-6f for the electrolyte. Comparison of peak heights, that is, the function g(r)

in Figure 4-6, shows substantial variations in O–M+ and Al–M+ pairs in different

simulations. The peak heights of Al–K+, Al–Rb+ and Al–Cs+ are higher than the O–

M+ peaks and increase from K+ < Rb+ < Cs+. In contrast, the peak heights for O–Na+


and O–NH4+ are higher than Al–M+ peaks. This comparison shows that significant

populations of Na+ and NH4+ ions reside near the O atoms in the framework, while K+,

Rb+ and Cs+ mostly reside in the centre of zeolite cages. Moreover, the higher values

of g(r) for O–Na+ and O–NH4+ represent nearest neighbour distances, indicating that

Na+ and NH4+ ions have stronger interactions with the zeolite N framework compared

to other M+ ions. For K+, Rb+ and Cs+, the highest values for maxima in g(r) are not

nearest neighbour distances. These relative RDF peak intensities indicate that a higher

proportion of K+, Rb+ and Cs+ ions inside ZM show weaker interactions with

framework atoms.

Ion density profiles, shown in Figure 4-7 for each of the simulations, provide an

indication of locational preferences along the z axis direction for these univalent ions.

For example, Figure 4-7 (centre panel) shows that for all simulations, K+ ions are

predominantly located in the middle of the membrane. This outcome is not unexpected

given that the simulation method places K+ (and M+) at the centre of ZM at time t = 0.

In contrast, Na+, Cs+ and Rb+ ions tend to show higher densities closer to the end(s),

or just outside, of ZM. NH4+ ions show an even distribution and density across the

breadth of the membrane which suggests that these ions prefer the internal cages of

ZM.

Close inspection of the individual cages, as shown in Figure 4-7g, shows that

Na+ ions are positioned closer to framework atoms, while NH4+ ions are located at the

edges and the middle of the cages. Both Rb+ and Cs+ are predominantly located in the

middle of the cages. Nevertheless, the position of K+ ions after these exchange

simulations (Figure 4-7g) shows slight variations depending on the exchangeable ion.

With addition of excess K+ only, these ions seem to occupy additional sites within the

cage. For Na+ exchange, the remaining K+ ion shows highest density within the centre

of the cage while for NH4+ exchange, the K+ ion is slightly off-centre of the cage. For

both Rb+ and Cs+ exchange, the remaining K+ ions are displaced from the centre of the

cage. These relative displacements of ions within the cages of ZM influence the rates

of diffusion of ions and their propensity for moderation by the presence of water

molecules.

4.4.4 Ion Diffusion

We calculate the relative mobility of ions in the membrane and in solution using

self-diffusion coefficients (D) calculated from the mean square displacements (MSD)


of ions and water molecules. These calculated D values provide an estimate of the

average rate of transfer of ions within the medium using our computational methods

outlined earlier. A comparison of these values for each of the ion exchange simulations

is presented in Table 4-6.

Table 4-6 Self diffusion co-efficient (D) of ions and water molecules in the Zeolite membrane (ZM) and solution calculated from MD simulation for 9ns, at 298K.

Self-diffusion co-efficient, D (cm2/s)

Zeolite N Ions/W Ions inside ZM Ions in electrolyte

K+/K+ K+ 1.51x10-9 1.07x10-5

water 1.62x10-9 2.46x10-5

K+/NH4+ K+ 1.76x10-8 1.62x10-5

NH4 + 8.32x10-9 7.39x10-6 water 3.54x10-8 3.03x10-5

K+/Na+ K+ 1.13x10-10 1.76x10-5

Na + 3.33x10-12 1.27x10-5 water 1.08x10-10 2.76x10-5

K+/Rb+ K+ 4.57x10-10 1.29x10-5

Rb + 3.49x10-10 1.19x10-5 water 1.13x10-9 1.80x10-5

K+/Cs+ K+ 1.80x10-10 1.71x10-5

Cs + 1.08x10-10 1.44x10-5 water 5.90x10-10 2.83x10-5

The average self-diffusion coefficient for water molecules in the electrolyte is

2.58 × 10-5 cm2/s. This value is similar to the experimentally determined self-diffusion

coefficient of water at 298K that ranges between 2.1 × 10-5 cm2/s and 2.7 × 10-5 cm2/s

for chloride solutions47. The D values for ions in the electrolyte are in the range of 1.1

× 10-5 to 1.7 × 10-5 cm2/s (average 1.42 × 10-5 cm2/s), except for NH4+ ions. The NH4

+

ions have the lowest diffusion coefficient at 7.39 × 10-6 cm2/s.

The D value for K+ ions inside ZM are larger than all other exchanging ions

evaluated in these simulations. For example, Rb+ is 1.3x, Cs+ is 1.7x, NH4+ is 2.1x and

Na+ is 34x slower than K+ ions in the membrane. However, as shown in Figure 4-8,

NH4+ ions in the Zeolite N_K+/NH4

+ exchange simulation within ZM show the highest

D values compared to all other exchanged ions in these simulations. Consistent with

earlier data shown in Section 3.0, the value of D for Na+ ions within ZM are the lowest

of all univalent ions. This outcome, as well as inferences from ion distribution profiles,


confirms that zeolite N membranes show exchange selectivity for specific univalent

ions in the series NH4+ > Rb+ > Cs+ > Na+. While experimental data are not available

for Rb+ or Cs+ exchange with K+, these models confirm experimental data obtained for

NH4+ and Na+.

Finally, the high preference and selectivity of NH4+ ions in Zeolite N can be

attributed to hydrogen bonding of the NH4+ ions with the oxygen atoms in the

framework. The four hydrogen atoms in NH4+ ions tend to form hydrogen bonds with

three oxygen atoms of the framework and one oxygen atom of the water molecule, as

seen in Figure 4-9. On close inspection of hydrogen bond interactions of NH4+ ions

during the course of the simulation, hydrogen bonds of the ions break and reform with

different oxygens in the framework, frequently migrating across adjacent oxygens and

water molecules that ease the movement of ammonium ions within the zeolite N

membranes. This behaviour explains not only the high mobility (diffusion coefficient

value) of the NH4+ ions but also the rapid exchange of K+ ions with NH4

+ ions. In

contrast, due to the absence of hydrogen bonding for Na+ ions, though held close to

the framework by electrostatic interactions with the framework, they exhibit low

mobility and low exchange rates.

Figure 4-8 Self diffusion co-efficient of ions in ZM

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

K+/K+ K+/NH4 + K+/Na + K+/Rb + K+/Cs +

Self

diffu

sion

co-e

ffici

ent (

cm2/

s)

D of ions inside ZM

K+

ion


Figure 4-9 Visualization of hydrogen bonds between NH4

+ and O atoms in the framework at different time steps of simulation time. Other extra framework species (K+, Cl- and H2O) not involved in

hydrogen bonding interaction are hidden from view. Yellow: silicon, pink: aluminium, red: oxygen, white: hydrogen, blue: nitrogen and blue dashed line: hydrogen bond

4.5 CONCLUSION

The outcomes in this chapter address the defined objectives for studying the

hydration and exchange behaviour of zeolite N. For example, MD calculations on

potassic zeolite N demonstrate that computational modelling of hydration and of ion

exchange simulates, in general, experimental outcomes for NH4+ and Na+ exchange

with K+ 48. Simulation of ion-exchange with Rb+ and Cs+ also shows that partial

exchange of K+ in the zeolite N structure is likely to occur in practice. The

computational method, which includes explicit water molecules in a membrane and in

the electrolyte, provides time- and location-dependent data on the relative efficacy of

univalent ion exchange in zeolite N.

We demonstrate that zeolite N prefers K+ exchange with NH4+ ions through a

high retention ratio (i.e. NH4+/K+ = 2.0) and NH4

+ shows the highest value for diffusion

coefficient of the univalent ions evaluated. Simulations for 8 ns show that ~70% of the

K+ ions are exchanged by NH4+. Other ions, such as Na+, Rb+ and Cs+, also partially

exchange but with significantly lower values for diffusion coefficients. Of significance

is the very low, by a factor of 34 times, diffusion coefficient for Na+ compared with

K+ determined by these simulations. This slow rate, in combination with the tendency

for Na+ to locate in close proximity to framework oxygen within the zeolite cage,

suggests a strong interaction with the aluminosilicate framework. Unlike with NH4+,

an exchange interaction of Na+ with the framework may not be facile in the presence

of H2O. Calculations of the hydrated state for zeolite N, including determination of

the hydration energy, show that zeolite N achieves equilibrium when the number of


water molecules approaches 8 H2O/cage. This calculated value is in excellent

agreement with the experimentally determined value for zeolite N19, 20. The modelling

approach described in this work offers important insight into the behavior of zeolite N

ion exchange and implies useful application to other aluminosilicate zeolites for a

range of multivalent cations.


4.6 REFRENCES

1. Yilmaz, B.; Müller, U., Catalytic Applications of Zeolites in Chemical Industry. Topics in Catalysis 2009, 52 (6), 888-895. 2. Gaare, K.; Akporiaye, D., Modified Zeolites as Catalysts in The Friedel-Crafts Acylation. Journal of Molecular Catalysis A: Chemical 1996, 109 (2), 177-187. 3. Hemelsoet, K.; Qian, Q.; De Meyer, T.; De Wispelaere, K.; De Sterck, B.; Weckhuysen, B. M.; Waroquier, M.; Van Speybroeck, V., Identification of Intermediates in Zeolite-Catalyzed Reactions by In Situ Uv/Vis Microspectroscopy and a Complementary Set of Molecular Simulations. Chemistry 2013, 19 (49), 16595-606. 4. Palomino, M.; Corma, A.; Jorda, J. L.; Rey, F.; Valencia, S., Zeolite Rho: a Highly Selective Adsorbent for CO2/CH4 Separation Induced by a Structural Phase Modification. Chemical Communications 2012, 48 (2), 215-217. 5. Palomino, M.; Corma, A.; Rey, F.; Valencia, S., New Insights on CO2−Methane Separation Using LTA Zeolites with Different Si/Al Ratios and a First Comparison with MOFs. Langmuir 2010, 26 (3), 1910-1917. 6. Zwingmann, N.; Mackinnon, I. D. R.; Gilkes, R. J., Use of a zeolite synthesised from alkali treated kaolin as a K fertiliser: Glasshouse experiments on leaching and uptake of K by wheat plants in sandy soil. Applied Clay Science 2011, 53 (4), 684-690. 7. Thornton, A.; Pearce, P.; Parsons, S. A., Ammonium removal from solution using ion exchange on to MesoLite, an equilibrium study. J. Hazard. Mater. 2007, 147 (3), 883-889. 8. Zwingmann, N.; Singh, B.; Mackinnon, I. D. R.; Gilkes, R. J., Zeolite from alkali modified kaolin increases NH4+ retention by sandy soil: Column experiments. Applied Clay Science 2009, 46 (1), 7-12. 9. Ghasemian, N.; Falamaki, C.; Kalbasi, M.; Khosravi, M., Enhancement of the catalytic performance of H-clinoptilolite in propane–SCR–NOx process through controlled dealumination. Chemical Engineering Journal 2014, 252, 112-119. 10. Cooney, E. L.; Booker, N. A.; Shallcross, D. C.; Stevens, G. W., Ammonia Removal from Wastewaters Using Natural Australian Zeolite. I. Characterization of the Zeolite. Separation Science and Technology 1999, 34, 2307–2327. 11. Deka, R.; Vetrivel, R., Developing the Molecular Modelling of Diffusion in Zeolites as a High Throughput Catalyst Screening Technique. Comb Chem High Throughput Screen 2003, 6 (1), 1-9. 12. Jia, W.; Murad, S., Separation of gas mixtures using a range of zeolite membranes: a molecular-dynamics study. The journal of chemical physics 2005, 122 (23), 234708. 13. Salmas, R. E.; Demir, B.; Yıldırım, E.; Sirkecioğlu, A.; Yurtsever, M.; Ahunbay, M. G., Silver–Sodium Ion Exchange Dynamics in LTA Zeolite Membranes. The Journal of Physical Chemistry C 2013, 117 (4), 1663-1671. 14. Nakamura, H.; Okumura, M.; Machida, M., First-Principles Calculation Study of Mechanism of Cation Adsorption Selectivity of Zeolites: A Guideline for Effective Removal of Radioactive Cesium. Journal of the Physical Society of Japan 2012, 82 (2), 023801. 15. Turgman-Cohen, S.; Araque, J. C.; Hoek, E. M. V.; Escobedo, F. A., Molecular Dynamics of Equilibrium and Pressure-Driven Transport Properties of Water through LTA-Type Zeolites. Langmuir 2013, 29 (40), 12389-12399.


16. Uzunova, E. L.; Mikosch, H., Cation site preference in zeolite clinoptilolite: A density functional study. Microporous and Mesoporous Materials 2013, 177, 113-119. 17. Valdivie´s-Cruz, K.; Lam, A.; Zicovich-Wilson, C. M., Chemical interaction of water molecules with framework Al in acid zeolites: a periodic ab initio study on H-clinoptilolite. Physical Chemistry Chemical Physics 2015, 17 (36), 23657-23666. 18. Valdiviés-Cruz, K.; Lam, A.; Zicovich-Wilson, C. M., Full Mechanism of Zeolite Dealumination in Aqueous Strong Acid Medium: Ab Initio Periodic Study on H-Clinoptilolite. The Journal of Physical Chemistry C 2017, 121 (5), 2652-2660. 19. Christensen, A. N.; Fjellvag, H., Crystal structure determination of zeolite N from synchrotron X-ray powder diffraction data. Acta Chemica Scandinavica 1997, 51, 969-973. 20. Christensen, A. N.; Fjellvag, H., Neutron Powder Differaction Study of the Dehydration of Zeolite N. Acta Chemica Scandinavica 1999, 53, 85-89. 21. Thornton, A.; Pearce, P.; Parsons, S. A., Ammonium removal from digested sludge liquors using ion exchange. Water Research 2007, 41 (2), 433-9. 22. Mackinnon, I. D. R.; Millar, G. J.; Stolz, W., Low temperature synthesis of zeolite N from kaolinites and montmorillonites. Applied Clay Science 2010, 48 (4), 622-630. 23. Mackinnon, I. D. R.; Millar, G. J.; Stolz, W., Hydrothermal syntheses of zeolite N from kaolin. Applied Clay Science 2012, 58, 1-7. 24. Sengyang, P.; Rangsriwatananon, K.; Chaisena, A., Preparation of zeolite N from metakaolinite by hydrothermal method. Journal of Ceramic Processing Research 2015, 16, 111-116. 25. Mackinnon, I. D. R.; Millar, G. J.; Stolz, W. Aluminosilicates of Zeolite N Ztructure. US 2006/0269472 A1, April 2, 2004, 2006. 26. Murad, S.; Jia, W.; Krishnamurthy, M., Ion-exchange of Monovalent and Bivalent Cations with NaA Zeolite Membranes : A Molecular Dynamics Study. Molecular Physics 2004, 102 (19-20), 2103-2112. 27. Hinkle, K. R.; Jameson, C. J.; Murad, S., Transport of Vanadium and Oxovanadium Ions Across Zeolite Membranes: A Molecular Dynamics Study. The Journal of Physical Chemistry C 2014, 118 (41), 23803-23810. 28. Nalaparaju, A.; Hu, Z. Q.; Zhao, X. S.; Jiang, J. W., Exchange of heavy metal ions in titanosilicate Na-ETS-10 membrane from molecular dynamics simulations. Journal of Membrane Science 2009, 335 (1-2), 89-95. 29. Refson, K.; Tulip, P. R.; Clark, S. J., Variational Density-Functional Perturbation Theory for Dielectrics and Lattice Dynamics. Phys. Rev. B 2006, 73 (155114), 1-12. 30. Clark, S. J.; Segall, M. D.; Pickard, C. J.; Hasnip, P. J.; Probert, M. I. J.; Refson, K.; Payne, M. C., First Principles Methods using CASTEP. Z. Kristallogr. 2005, 220, 567-570. 31. Sun, H., COMPASS: An ab Initio Force-Field Optimized for Condensed-Phase ApplicationsOverview with Details on Alkane and Benzene Compounds. The Journal of Physical Chemistry B 1998, 102 (38), 7338-7364. 32. Berendsen, H. J.; Postma, J. P.; van Gunsteren, W. F.; Hermans, J., Interaction models for water in relation to protein hydration. In Intermolecular Forces, Springer: 1981; pp 331-342. 33. Perdew, J. P.; Wang, Y., Pair-distribution function and its coupling-constant average for the spin-polarized electron gas. Physical Review B 1992, 46 (20), 12947. 34. Hockney, R. W.; Eastwood, J. W., Computer simulation using particles. crc Press: 1988.


35. Ewald, P. P., Ewald summation. Annalen den Physik 1921, 369, 253. 36. J. E. Dennis, J.; Moré, J. J., Quasi-Newton Methods, Motivation and Theory. SIAM Review 1977, 19 (1), 46-89. 37. Andersen, H. C. J. T. J. o. c. p., Molecular dynamics simulations at constant pressure and/or temperature. The journal of chemical physics 1980, 72 (4), 2384-2393. 38. Samoletov, A. A.; Dettmann, C. P.; Chaplain, M. A., Thermostats for “slow” configurational modes. Journal of Statistical Physics 2007, 128 (6), 1321-1336. 39. Leimkuhler, B.; Noorizadeh, E.; Penrose, O., Comparing the efficiencies of stochastic isothermal molecular dynamics methods. Journal of Statistical Physics 2011, 143 (5), 921-942. 40. Lin, J.; Murad, S., A computer simulation study of the separation of aqueous solutions using thin zeolite membranes. Molecular Physics 2001, 99 (14), 1175-1181. 41. Murad, S.; Jia, W.; Krishnamurthy, M., Molecular Simulations of Ion Exchange in NaA Zeolite Membranes. Chemical Physics Letters 2003, 369 (3–4), 402-408. 42. Pissis, P.; Daoukaki-Diamanti, D., Dielectric studies of molecular mobility in hydrated zeolites. Journal of Physics and Chemistry of Solids 1993, 54 (6), 701-709. 43. Maurin, G.; Bell, R. G.; Devautour, S.; Henn, F.; Giuntini, J. C., Modeling the Effect of Hydration in Zeolite Na+-Mordenite. Journal of Physical Chemistry B 2004, 108 (12), 3739-3745. 44. de S. Vilhena, F.; Serra, R. M.; Boix, A. V.; Ferreira, G. B.; de M. Carneiro, J. W., DFT study of Li+ and Na+ positions in mordenites and hydration stability. Computational and Theoretical Chemistry 2016, 1091, 115-121. 45. Gilberto Artioli, J. S., JJ Pluth, Å Kvick, K Ståhl, Neutron Diffraction Studies of the Hydrogen Bonding and Water Molecules in Zeolites. In Studies in Surface Science and Catalysis, 1985; Vol. 24, pp 249-254. 46. L. Leherte, J.-M. A., E. G. Derouane and D. P. Vercauteren,, Self-diffusion of Water into a Ferrierite-type Zeolite by Molecular Dynamics Simulations. J. Chem. Soc., Faraday Trans., 1991, 87 1991, 87, 1959. 47. Müller, K. J.; Hertz, H. G., A Parameter as an Indicator for Water−Water Association in Solutions of Strong Electrolytes. The Journal of Physical Chemistry 1996, 100 (4), 1256-1265. 48. Mackinnon, I.; Millar, G.; Stolz, W. Aluminosilicated of zeolite N structure. 2006.


Chapter 5: Evaluation of DFT Methods to Calculate Structure and Partial Atomic Charges for Zeolite N

In this chapter, the effects of density functional theory (DFT) models on zeolite

structural parameters and on partial atomic charges of framework atoms are assessed.

Section 5.1, briefly introduces DFT methods and how they are used here to optimise

the structure and properties of zeolite N. Section 5.2 describes the DFT functionals,

empirical dispersion corrections and basis sets used, together with all numerical

convergence criteria. In section 5.3 the optimized zeolite N structures are evaluated by

comparing the atom positions and framework T-O (T=Si or Al) bond lengths with

experimental data. Section 5.4 discusses the calculated results evaluating them by

comparison to previous experimental and computational data on zeolite N and other

zeolites. Finally, section 5.5 provides outcomes for this chapter and identifies a reliable

DFT model to optimize the structural parameters of zeolite N for further MD

simulations.

5.1 INTRODUCTION

Zeolites are charged alumino-silicate porous framework materials carrying

various ions and water molecules inside their pores and channels. The majority of

diffusion, adsorption, ion exchange and catalysis processes occur in these pores and

channels. In other words, the structure and physico-chemical behaviour of zeolites

depend on the electrostatic potential inside these pores and channels. This electrostatic

potential is Coulombic and directly dependent on the atomic charges and structural

positions of framework atoms and extra-framework ions inside a zeolite structure. For

instance, experiments show that structural positions and partial charges of zeolite

framework atoms critically influence water adsorption due to the dipole moment of

water molecules1. Therefore, the behaviour of these materials is strongly reliant on

charge distributions and specific structural configurations.

The structure and properties of a solid system, whether crystalline or amorphous,

are defined by the interactions between their electrons in three dimensions and their

electrostatic potentials that can be derived from experiments or theoretical quantum


mechanics calculations. Experimentally, the electrostatic properties of materials can

be estimated using electron density models derived from high-resolution X-ray

diffraction refinements2-5. However, it is difficult to define the atomic charges in

periodic systems, such as zeolites. Ghermani et al.6 used this method to calculate the

atomic charges and the electrostatic properties of natural natrolite. In theoretical

quantum mechanics, the distribution of the electrons, or their electron density, is

calculable using Density Functional Theory (DFT)7 and is a fundamental basis for

understanding functionality of structures and estimating atomic charge.

The advantage of DFT calculations is that it considers the three-dimensional

distribution of electrons which is much simpler to converge than the molecular orbitals

that need to be described in molecular orbital-based ab initio methods with a many

electron wavefunction used in Schrodinger’s equation8 9. In addition, the simple

formalism of DFT results in significantly reduced computational costs compared with

wave-function calculations9, 10 and increased calculation accuracy11, 12. Since the

precise exchange-correlation energy functional is unknown and possibly complex to

determine, the exchange-correlation potential can be computed using different types

of approximations or basis sets in DFT calculations13.

Zeolite N is a synthetic zeolite14-16 with chemical formula

K12Al10Si10O40Cl2.8H2O and is considered a fibrous zeolite. Christensen and

Fjellvag17, 18 determined the crystal structure of zeolite N as orthorhombic with space

group I22217, 19 using high resolution X-ray and neutron diffraction data. The high

capacity of zeolite N for selective ion exchange applications, compared with natural

zeolites, has been verified by experimental studies20-23. These experimental

investigations present valuable data on the ion-exchange capability of zeolite N. These

data serve as a verifiable template from which to understand, at an atomic or molecular

scale, the detailed behaviour of exchangeable ions in this structure. For example, in

chapter 4 the monovalent ion exchange behaviour of a zeolite N membrane modelled

using MD simulations to demonstrate that preference for exchange of K+ ions is in the

order NH4+> Rb+>Na+>Cs+. Further, this work provided a viable mechanistic model

for exchange of monovalent ions, in particular, the high affinity for exchange of

ammonium ions12.

DFT methods are used to establish key parameters of a structure prior to

invoking more extensive molecular dynamics (MD) models and, for complex


inorganic structures, this link between DFT and MD critically influences the outcome

of computational models. Several ab initio and DFT calculations have been performed

to calculate structural properties, atomic charges, electron density and electrostatic

potentials of high symmetry unit cells of zeolite and zeolitic materials24-33. For

example, Fischer et al.31 evaluated different DFT approaches to optimise the materials

constructed of SiO4 and/or (Al/P)O4 tetrahedra. They showed that the choice of DFT

methods influences the lattice parameters, T-O (T= Si or Al) bond lengths and T-O-T

angles (Si-O-Si , Si-O-Al or Al-O-Al) for a specific structure. Moreover, Wolffis et

al.32 employed various semi-empirical and ab initio DFT methods to calculate the

partial charges of BEA zeolite. They showed that the partial charges on zeolite

framework atoms depend on the zeolite geometry and available T-sites in the zeolite

framework. Moreover, they showed that the partial charges of zeolite framework

atoms obtained from DFT calculations affect the predicted gas adsorption for Faujasite

zeolite using Monte Carlo simulations

In order to define the atomic charges and structural parameters of zeolite N

before conducting molecular dynamic simulations described in Chapter 4, the same

DFT methods implemented by Salmas et al. 26 for zeolite LTA were applied. The Si:Al

ratio of zeolite N and LTA are similar and equal to one. However, LTA is a different

framework type to Zeolite N and contains Na+ as the extra framework cation compared

with K+ for zeolite N. Moreover, zeolite N has two different framework T-sites for Si

and Al atoms that invokes different tetrahedral arrangements compared to the LTA

structure which is cubic and so has a higher proportion of equivalent tetrahedral sites.

In this work, we provide detailed DFT calculations to precisely study the

structure of zeolite N, to determine the Mulliken partial charges of framework atoms

and then to compare with those used by Salmas et al.26 and calculations in Chapter 429.

In this study, we evaluate structural parameters and Mulliken partial atomic charges of

zeolite N derived from various DFT approximations, functionals, basis sets and

application of dispersion corrections under different SCF and convergence criteria.

These approximations are validated through the influence on the structure and partial

atomic charges for zeolite N framework atoms and comparison with experimental

data17.


5.2 COMPUTATIONAL AND THEORETICAL METHODS

In this study, all calculations and analyses were conducted using DMol3 34, 35

(DMol3; Accelrys Inc.: San Diego, CA, version 2016.) code in Accelrys Materials

Studio software package 2017. Using density functional theory (DFT), this code

allows prediction of structural, electronic, electrostatic, energetic and thermodynamic

properties of 3D organic and inorganic materials, periodic systems, solids and surfaces

with acceptable accuracy and reasonably low computational cost.

The initial unit cell of zeolite N was constructed using refined unit cell

parameters determined by Christensen and Fjelvag17 using high resolution X-ray

diffraction. The crystallographic lattice was built in the orthorhombic space group I222

with cell dimensions a=9.9041 Å, b=9.8860 Å and c=13.0900 Å and angles

α=β=γ=90º. All DFT calculations were carried out on one unit cell of zeolite N keeping

the symmetry, rigid lattice parameters and flexible Cartesian positions. The unit cell

of zeolite N contains ten silicon (Si), ten aluminum (Al) and forty oxygen (O)

framework atoms, twelve extra-framework potassium cations (K) and two chlorine

anions (Cl). Figure 5-1 illustrates the zeolite N structure highlighting potassium and

chlorine ions viewed from the four axial orientations of the 2x2x2 super cell.

The zeolite N unit cell was optimized with and without extra-framework cations

and water molecules to evaluate the partial charge and structural changes considering

the absence or presence of extra framework ions. We considered the full charge for K

and Cl ions in our calculations; hence, for unit cells without extra-framework ions, the

total charge of the unit cell was constrained to -10. Geometry optimizations were

performed under two different quality of convergence criteria (i) fine: 10-6 Ha. energy,

0.002 Ha./Å and 0.005 Å displacement and (ii) medium: 2x10-5 Ha. energy, 0.004

Ha./Å force and 0.005Å displacement. Also, in order to accelerate convergence in

some DFT calculations the thermal smearing of 0.005 Ha was applied to the orbital

occupation.

In this study, several different functional approximations were applied to

calculate structural parameters of zeolite N and estimate partial atomic charges of its

constructed atoms: (1) Local Density Approximation (LDA) with the PW9236

functional; this is the most straightforward DFT functional based on electron density

and assumes electrons in a homogeneous electron gas model. This method provides

results with lower computational cost, but with a lower level of accuracy37, 38. (2)


Generalised Gradient Approximation (GGA)39, which is based on electron spin density

and its gradients; this approximation accounts for heterogeneous electron density and

results in more accurate computational estimations. In this case, we considered two

functionals, PW9140 and PBE39, to be consistent with previous literature.

Figure 5-1 a 2x2x2 super cell of Zeolite N along different directions (a) (001), (b) (010) and (c) (100).

After selecting the DFT functional, self-consistent field (SCF) convergence

criteria were selected. We chose different levels of accuracy for SCF convergence

including 10-5 and 10-6 with 200 SCF cycles and hexadecapole expansion. The SCF

tolerance indicates the threshold for SCF density convergence and the number of SCF

cycles indicates the maximum number of SCF iterations allowed for an energy

calculation. In addition, all electrons in the system are treated for the SCF calculation.

The choice of basis set is critical for computing the SCF. In DMol3, numerical orbitals


are applied for basis functions related to atomic orbitals. In order to calculate the Kohn-

Sham orbital of the system, numerical basis sets are applied. In this study, the effect

of using a double-sized numerical basis set plus d-functional (DND) and polarized p-

functional (DNP) were assessed. We used both basic (3.5) and developed versions

(4.4) of the basis files for our calculations41. The 4.4 version is the most recent and is

an optimised and improved version of these basis sets.

In DFT methods, non-covalent forces are accounted using semi-empirical

dispersion correction schemes. In this study, three different schemes available in

Material Studio DMol3 version 2016, were employed to calculate the long range

interactions of extra-framework ions with framework atoms, including Grimme42, 43,

Orthmann, Bechstedt and Schmidt (OBS)44 and Tkatchenko and Scheffler (TS)45.

Table 5-1 lists all considered parameters for different DFT calculations in this study.

Each DFT method is labelled in order to simplify the following discussion for readers.

Table 5-1 DFT models employed in this study with different convergence quality, approximation, functional, basis sets, dispersion corrections and thermal smearing parameters

Code Structure Quality Functional/ Basis set DFT-D Smearing A UC Med LDA-PWC DNP-4.4 OBS

B UC Fine LDA-PWC DNP-4.4

C UC Fine LDA-PWC DNP-3.5

D UC Med GGA-PW91 DNP-4.4 OBS

E UC Fine GGA-PW91 DNP-4.4 OBS

F UC Fine GGA-PW91 DNP-3.5 OBS

G UC Fine GGA-PW91 DNP-3.5

H UC Fine GGA-PW92 DNP-4.4

I UC Med GGA-PBE DND-4.4 TS

J UC Med GGA-PBE DNP-4.4 Grimme

K UC Fine GGA-PBE DNP-4.4 TS

L UC Fine GGA-PBE DNP-3.5

M UC Fine GGA-PBE DNP-4.4

N UC Fine GGA-PBE DNP-3.5 0.005 O UC Fine GGA-PBE DNP-4.4 0.005 P UC-K-Cl Fine GGA-PW91 DNP-3.5

Q UC-K-Cl Fine GGA-PW91 DNP-4.4 OBS

R UC-K-Cl Fine GGA-PBE DNP-3.5 TS

S UC-K-Cl Fine GGA-PBE DNP-4.4 TS

T UC-K-Cl Fine GGA-PBE DNP-4.4

U UC-K-Cl Fine GGA-PW91 DNP-4.4 TS UC: unit cell (only zeolite N framework), UC-K-Cl: unit cell containing K and Cl ions


The structure of zeolite N was investigated by extracting and comparing the

atomic positions and T-O (T= Si or Al) bond lengths of optimised structures by DFT

methods with experimental synchrotron X-ray diffraction17 and neutron diffraction18

data. The atomic positions of optimised and experimental structures were compared

by calculating the atomic displacement and arithmetic mean of distances (dav.)46 using

COMPSTRU programme available at Bilbao Crystallographic Server47. Moreover, the

DFT results were evaluated by calculating the mean absolute deviation (MAD) of Si-

O and Al-O bond distances from experimental data17.

Additionally, zeolite N structures optimised using DFT methods were obtained

and compared with experimental structures based on X-ray powder diffraction data

using Reflex module in the Accelrys Materials Studio software package. X-ray powder

diffraction (XRD) patterns were collected with a copper source at a step size of 0.050°

2θ from 5° – 45° 2θ in Bragg-Brentano geometry and Rietveld correction. The

Mulliken partial charges on atoms were obtained from population analyses available

in the DMol3 module of Materials Studio.

5.3 RESULTS

In this study, the effects of DFT methods and constraints including DFT

functionals, basis sets and dispersion correction on partial atomic charges and zeolite

N structures were evaluated. For all structures of zeolite N optimised under different

DFT methods, we calculated and analysed the density of states, electron density and

electro-statistics. Moreover, to evaluate the partial charges of atoms, Mulliken charges

were calculated.

The atomic positions of optimised structures were compared with experimental

structure17 using COMPSTRU programme, Table 5-2 shows the calculated maximum

distance of displaced atoms (dmax) and the arithmetic mean of displacements (dav). The

results show that in structures without extra-framework atoms, O3 and O5 have the

maximum displacement distances in LDA and GGA approximations, respectively. In

structures contains extra-framework ions, O3 has the maximum displacement in

structures optimised by GGA-PW91 functional. However, in structures optimised by

GGA-PBE functional, the maximum displacement is related to K2 or O3.

The mean average deviation (MAD) of T-O bond distances of optimised

structures from experimental data17 were calculated. Table 5-3 shows the calculated


Si-O and Al-O bond distances of experimental and optimised zeolite N structures and

the MAD values. Initially, in order to evaluate the effect of different basis sets on

structure and atomic charges, DFT calculations were performed on the zeolite N unit

cell without extra-framework ions.

The first parameter investigated was the effect of quality of convergence criteria

on calculated results. The results indicate that structures calculated with higher

convergence criteria show results with lower dav (Table 5-2) and T-O bond distances

(Table 5-3) similar to experimental data17 (models B vs. A, E vs. D and K vs. I). Higher

quality of convergence criteria allows application of a double numerical polarised

basis set (DNP) that gives better results compared to a double numerical d-function

(DND). Moreover, we compared results with the two different basis set files, basic

(3.5) and developed (4.4)41, with double numerical plus polarisation function for all

DFT functionals, LDA-PWC, GGA-PW91 and GGA-PBE.

Table 5-2 Arithmetic mean of atomic displacements (d-av (Å)) and maximum atomic displacement (d-

max (Å)) obtained by COMPSTRU programme47. The Codes are explained in Table 5-1.

Code d-av (Å) d-max (Å) Atom*

A 0.0562 0.1038 O3 B 0.0551 0.1005 O3 C 0.0650 0.1223 O3 D 0.1133 0.2063 O5 E 0.1126 0.2058 O5 F 0.1264 0.2359 O5 G 0.1250 0.2327 O5 H 0.1112 0.2026 O5 I 0.1192 0.2254 O5 J 0.1229 0.2226 O5 K 0.1189 0.2251 O5 L 0.1313 0.2465 O5 M 0.1179 0.2175 O5 N 0.1313 0.2465 O5 O 0.1179 0.2175 O5 P 0.0997 0.1996 O3 Q 0.0986 0.1933 O3 R 0.1066 0.2307 K2 S 0.1055 0.2461 K2 T 0.1014 0.1990 O5 U 0.1007 0.2234 K2

* Atoms with maximum displacement


Table 5-3 Bond distances ( in Å

) between fram

ework Si/A

l atoms w

ith oxygen atoms derived from

DFT calculations and the calculated m

ean absolute

deviation (MA

D) of com

putational bond length from experim

ental data. The Codes are explained in Table 5-1.


The results show that the structures optimised with DNP-4.4 basis set exhibit

less atomic displacements compared to those optimised by DNP-3.5 basis set. Si-O

and Al-O distances obtained from 4.4 basis sets are closer to experimental17 (models

H vs. G, M vs. L and S vs. R). Comparing, all results from DFT functionals and basis

sets on zeolite N without extra-framework ions and water molecules, A, B and C

models (LDA-PWC models) resulted in lower dav and MAD values for bond distances.

Furthermore, it can be seen from the results that applying thermal smearing of 0.005

Ha has no effect on atomic positions and T-O bond lengths of structures without extra-

framework ions and water molecules (N and O models, respectively, vs. L and M

models in Table 5-2 and 5-3).

To account for the non-covalent van der Waals interaction between extra-

framework K+ and Cl- ions with framework atoms, we employed a semi-empirical

dispersion correction for DFT calculations on zeolite N unit cells with extra-

framework ions. The results indicate that using dispersion corrections OBS and TS

provide structures closer to experimental17, respectively, in PW91 and PBE functionals

(models S and U, Table 5-2 and 5-3).

The Mulliken partial charges of zeolite N framework atoms obtained from

various DFT models are presented in Table 5-4. The results show that using higher

quality of convergence criteria provides framework total charge closer to experimental

chemical formula of zeolite N17 (models E=-10 vs. D=-10.01). In LDA-PWC and

GGA-PW91 calculations, using DNP 4.4 basis set provides better results for

framework total charge compared to 3.5 basis set (models B=-10.01 vs. C=-9.988 and

H=-10 vs. G=-9.994). However, in the GGA-PBE calculation, the framework total

charges are the same for both basis sets (-10.002 in both models L and M). The total

framework charges in structures containing extra framework atoms ( P, Q, R, S and T

models) are lower than -10 (-7.00 to -7.86).

Even though the framework total charge of optimised structures is close to

experimental data17, the partial atomic charges of Si, Al and O atoms are different.

DFT models in this study provide two different Mulliken partial charges for framework

Si and Al atoms located at different T-sites as well as framework oxygen atoms.

Applying a dispersion correction has no effect on atomic partial charges of structures

optimised with GGA-PW91 functional (models E and H). However, using a dispersion

correction in GGA-PBE calculations slightly affects the partial atomic charges of Al1


and O3 with a 0.001e (models K and M). Use of thermal smearing on orbital

occupations did not influence the calculated Mulliken partial atomic charges (N and O

models, respectively, vs. L and M models).

Table 5-4 Calculated Mulliken atomic charges of zeolite N framework atoms derived from DFT calculations

Code Charge Framework

total charge Si1 Si2 Al1 Al2 O1 O2 O3 O4 O5

A 1.559 1.522 1.515 1.481 -1.007 -1.006 -1.008 -1.008 -0.994 -10.004 B 1.587 1.951 1.500 1.469 -1.011 -1.009 -1.012 -1.012 -0.999 -6.806 C 1.381 1.354 0.950 0.924 -0.827 -0.824 -0.828 -0.828 0.805 2.874 D 1.742 1.697 1.670 1.632 -1.090 -1.089 -1.093 -1.085 -1.087 -10.080 E 1.766 1.724 1.662 1.630 -1.094 -1.093 -1.090 -1.091 -1.095 -10.008 F 1.591 1.545 1.166 1.128 -0.927 -0.925 -0.921 -0.923 -0.924 -10.054 G 1.591 1.549 1.166 1.128 -0.927 -0.925 -0.921 -0.923 -0.924 -10.022 H 1.766 1.724 1.662 1.630 -1.094 -1.093 -1.089 -1.091 -1.095 -10.000 I 1.732 1.688 1.666 1.628 -1.087 -1.085 -1.081 -1.082 -1.082 -10.004 J 1.734 1.689 1.668 1.528 -1.087 -1.086 -1.081 -1.082 -1.083 -10.804 K 1.756 1.718 1.655 1.623 -1.090 -1.089 -1.088 -1.087 -1.090 -10.002 L 1.576 1.535 1.156 1.116 -0.921 -0.918 -0.915 -0.917 -0.914 -10.004 M 1.756 1.716 1.655 1.621 -1.090 -1.089 -1.086 -1.087 -1.089 -10.006 N 1.576 1.535 1.156 1.116 -0.921 -0.918 -0.915 -0.917 -0.914 -10.004 O 1.756 1.716 1.655 1.621 -1.090 -1.089 -1.086 -1.087 -1.089 -10.006 P 1.795 1.770 1.416 1.387 -0.963 -0.963 -0.965 -0.971 -0.985 -7.098 Q 1.948 1.928 1.788 1.768 -1.118 -1.121 -1.114 -1.120 -1.139 -7.856 R 1.797 1.769 1.402 1.378 -0.952 -0.957 -0.963 -0.971 -0.979 -7.002 S 1.935 1.905 1.796 1.786 -1.112 -1.118 -1.116 -1.128 -1.138 -7.666 T 1.951 1.931 1.805 1.783 -1.119 -1.122 -1.115 -1.122 -1.140 -7.720 U 1.938 1.928 1.782 1.769 -1.113 -1.119 -1.113 -1.124 -1.136 -7.824

5.4 DISCUSSION

The zeolite N framework consists of Si, Al and O atoms located in different

periodic positions. According to Christensen and Fjellvag17, 18 each unit cell of zeolite

N includes two T-sites for silicon (two Si1 and eight Si2), two T-sites for aluminium

(two Al1 and eight Al2) and five different sites for oxygen (eight of each type). Also,

each unit cell contains two different extra-framework sites for K cations (four K1 and

eight K2), one site for chlorine (two Cl ions) and two different sites for water

molecules (sixteen water molecules). The negative charge of the framework resulting

from ten Al atoms compensates for ten of the extra-framework K cations, and the

remainder compensate the chlorine negative charge. Each Si1 and Al1 tetrahedra are

surrounded by four Si2 and four Al2 tetrahedra, respectively, as shown in Figure 5-1.


Christensen and Fjellvag17, 18 using high resolution X-ray and neutron diffraction

indicated that K1 interacts (Coulomb interaction) with three framework oxygens, O3,

O4 and O5, and that K2 interacts with 4 framework oxygens, O1, O2, O3 and O4. In

addition, potassium cations interact with Cl anions and oxygen of water molecules. An

electron density profile along the (110) direction obtained by DFT (model S) is shown

in Figure 5-2. This figure illustrates the interactions between K1 and K2 cations with

Cl anions and with framework oxygens.

Figure 5-2 Electron density profile showing the interaction between potassium cations with chloride anions and framework oxygens in the (110) plane of zeolite N unit cell along z direction.

In this study, the effect of various approximations, functionals and basis sets of

DFT calculations on the zeolite N structure and partial atomic charges were assessed.

In order to reduce the computational costs, we kept the symmetry of zeolite N structure

during DFT calculations and released the atoms to find their structural positions. The

results show that the Si1 and Al1 stay in their positions during optimisations. The most

affected atoms in structures without extra-framework ions optimised by LDA and

GGA approximations, respectively, are O3 and O5 with the highest displacement

distances from their experimental positions (Table 5-2).


Our DFT calculations show that there are small differences in bond lengths

amongst all models. Furthermore, the calculated frameworks consist of more uniform

Si/Al tetrahedra compared with experimental determination of the zeolite N structure17

(Table 5-3). The tetrahedral bond lengths calculated experimentally are in the range of

1.584-1.638 Å for Si-O and 1.704-1.755 Å for Al-O. However, the Si-O and Al-O

bond lengths for structures without extra-framework ions are, respectively, in the range

of 1.612-1.623 Å and 1.730-1.746 Å in LDA-PWC models, 1.634-1.637 Å and 1.747-

1.766 Å in GGA-PW91 models and 1.631-1.64 Å and 1.749-1.769 Å in GGA-PBE

models. By adding extra-framework ions to the structures these distances increased up

to 1.653 Å for Si-O and 1.779 Å for Al-O after DFT calculations.

Even with these small differences in bond lengths for Si-O and Al-O in optimised

structures determined by different DFT models in this study, all were close to the range

obtained by Baur et al33 for Si-O (1.593-1.657 Å) and Al-O (1.728-1.776 Å) bond

lengths of EDI and other framework type zeolites. However, The average Si-O and Al-

O bond lengths obtained from GGA approximations in this study are slightly larger

than those obtained with previous theoretical studies on other zeolites and silicates6, 17,

18, 26, 31, 33 (Figure 5-3). This difference is due to the dissimilarity of the structural

topology and Si/Al ratio of the studied frameworks and considered DFT parameters.

Therefore, in order to choose the best DFT model, we considered other factors.

Figure 5-3 Comparison of average Si-O and Al-O bond lengths for zeolite N obtained from DFT models in this study with previous experimental and computational studies on zeolite N and other

zeolites


The first factor is the difference in atomic positions of framework and extra-

framework atoms between optimised structures and experimental data. This factor was

assessed by measuring the atom displacements from their experimental positions,

calculating the arithmetic mean of total displacements (dav) and the mean average

deviation (MAD) of Si/Al-O bond lengths of optimised structures from experimental

data. Comparing these calculated parameters (Table 5-2 and 5-3) for different DFT

models, models B, H and M were identified, respectively, as best models of LDA-

PWC, GGA-PW91 and GGA-PBE functionals with highest convergence and SCF

quality and DNP basis sets 4.4, for optimising structures without extra-framework

ions. Evaluating the impact of different dispersion correction schemes on DFT

calculations and optimized structures containing extra-framework ions, we conclude

that applying the TS45 dispersion correction schemes provides better results in

optimised structures with GGA-PW91 and GGA-PBE functionals (U and S models

respectively).

The second factor to evaluate is the value(s) for partial atomic charges. Most

theoretical calculations of zeolites report partial charges for framework atoms at about

half of their ionic charges 9, 10, 48, with a 0%-20% variation. It means the partial charges

can be between 1.6 to 2.4 for Si, 1.2 to 1.7 for Al and -1.4 to -1.2 for O. In this study,

The obtained Mulliken partial charges are in good agreement with this criterion.

In this study, we document noticeable differences in calculated Mulliken partial

atomic charges depending on specific parameters used in DFT models as well as the

position of atoms in zeolite N framework (Table 5-4). All DFT models in this study

provide different partial charges for Si and Al atoms located at two T-sites of the

zeolite N framework in good agreement with the recent study by Wolffis et al.32. Our

results clearly illustrate the dependency of Mulliken partial atomic charges on the

choice of basis set, as identified by previous researchers41. The charge difference for

Si and Al between the 3.5 and 4.4 versions of DNP basis sets, is on average, 0.2 and

0.5, respectively. Moreover, the application of dispersion correction has no effect on

the partial charges of zeolite N framework atoms in unit cells without extra-framework

ions. Among all DFT models tested, the GGA-PW91 and the GGA-PBE with DNP-

4.4 basis set (H and M respectively in Table 5-4) provide a value for framework total

charge closest to the zeolite N chemical formula.


Even though model B (LDA-PWC functional with DNP-4.4) gives the lowest

dav and MAD values, it provides partial charges that are less than a crystallo-

chemically reasonable range for Si atoms. The derived Mulliken partial charges

calculated from Models H and M are within reasonable ranges for framework Si, Al

and O atoms. However, the total discrepancy of framework atoms in model M is less

than model H (GGA-PBE vs. GGA-PW91). The Mulliken partial charges obtained by

these models are similar to charge distributions obtained by high resolution X-ray

diffraction refinement6 and by charges obtained by the REPEAT25 method on the

zeolite natrolite (Table 5-5). The REPEAT method is a simple error functional method,

introduced by Campana et al.25, to calculate the electrostatic potential (ESP) charge in

molecular systems and periodic nano-porous materials, such as zeolites and metal

organic framework materials. Campana et al.8 show that ESP charges of framework

atoms in natrolite derived from the REPEAT method are similar to those measured by

high resolution X-ray diffraction refinements6.

Table 5-5 Diversity of atomic charge of framework atoms of zeolites with Si/Al=1

Natrolite6 Natrolite25 LTA48 LTA26 Method XRD ESP Mulliken Mulliken Mulliken

Si 1.840 1.376 1.541 1.850 1.540 1.650 1.722 1.484

Al 1.510 1.616 1.165 1.270 1.110

O

-1.210 -0.898 -0.602 -1.030 -0.912 -1.030 -1.057 -0.629 -1.070 -1.107 -0.619 -0.870 -1.099 -0.615 -0.900 -0.767 -0.598

The third factor we considered in this study was computational accuracy versus

computational cost of the DFT method. A high quality of convergence criteria is

computationally expensive, but our results reveal that such criteria provide a calculated

framework structure that is close to models based on experimental data. In addition,

the use of thermal smearing up to 0.005 Ha hastens convergence, and the results show

that there is no effect on T-O bond lengths and the partial atomic charges of framework

atoms. Both the PW91 and PBE functionals provide acceptable structure and partial

atomic charges for zeolite N without extra-framework ions. However, the PBE

functional is a simpler functional while an improved form of PW91 involves both

electron correlation and electron exchange in an ideal system39. Moreover, a recent


study by Fischer et al.31 shows that the PBE functional with TS dispersion correction

is a reliable DFT method to optimise zeolitic materials.

The XRD patterns of our optimised structure obtained from model S (GGA-PBE

functional with DNP-4.4 and TS dispersion correction) are compared with

experimental data17 in Figure 5-4. This figure shows that even though the intensity of

some peaks differs from experiment due to movement of atoms (especially K2 and

O3) during optimisation, there is no difference in the peak positions for both structures.

The structural positions and interatomic distances of all atoms calculated for zeolite N

(Model S) are compared with values extracted from a high resolution X-ray diffraction

refinement17 in Table 5-6. This outcome, and the more appropriate values for T-O bond

lengths and framework atom charge(s), confirm the choice of this DFT method for

further molecular modelling and simulations.

Figure 5-4 XRD patterns for zeolite N unit cell containing K and Cl extra framework ions calculated using Reflex


Table 5-6 Refined positional parameters of zeolite N structure obtained from experiment17 and DFT calculation and calculated atomic displacement

Experimental data DFT model S Atomic Displacements WP Atom x y z x y z ux uy uz |u| 2d Si1 0.0000 0.5000 0.0000 0.0000 0.5000 0.0000 0 0 0 0 2b Al1 0.5000 0.0000 0.0000 0.5000 0.0000 0.0000 0 0 0 0 8k Si2 0.3862 0.1584 0.1920 0.3860 0.1623 0.1878 -0.0002 0.0039 -0.0042 0.0678 8k Al2 0.3430 -0.1151 0.3089 0.3482 -0.1115 0.3057 0.0052 0.0036 -0.0032 0.0746 8k O1 0.3967 0.0996 0.0753 0.3902 0.1058 0.0704 -0.0065 0.0062 -0.0049 0.1102 8k O2 0.4001 0.9003 0.4353 0.4035 0.8996 0.4322 0.0034 -0.0007 -0.0031 0.0530 8k O3 0.3106 0.0446 0.2569 0.3106 0.0551 0.2644 0 0.0105 0.0075 0.1427 8k O4 0.5406 0.1911 0.2291 0.5420 0.1981 0.2238 0.0014 0.0070 -0.0053 0.0984 8k O5 0.3044 0.2960 0.1894 0.3083 0.3077 0.1931 0.0039 0.0117 0.0037 0.1312 8k K1 0.0000 0.0000 0.2499 0.0000 0.0000 0.2578 0 0 0.0079 0.1029 4i K2 0.2115 0.2121 0.4363 0.2002 0.1906 0.4404 -0.0113 -0.0215 0.0040 0.2461 2c Cl 0.0000 0.0000 0.5000 0.0000 0.0000 0.5000 0 0 0 0

5.5 CONCLUSION

The results indicate that higher quality of convergence criteria in DFT

calculations delivers results close to experimental values and using smearing has no

effect on optimised structures and partial charges for zeolite N. We found that the

choice of numerical basis set file version considerably affects the calculated Mulliken

partial charges and the structures optimised using developed DNP basis sets (4.4) are

closer to the experimental structure17. Moreover, we found that Mulliken partial

charges of framework atoms are different from site to site. Applying the TS dispersion

correction scheme on structures containing extra-framework ions provides zeolite N

frameworks with Si-O and Al-O bond lengths in the range for EDI framework type

zeolites33 and close to experimental data17. In addition, we found that the framework

O5 and O3 atoms are the most affected atoms by DFT optimisations and show the

longest displacement distances.

The structure and Mulliken partial atomic charges of zeolite N framework

optimised using GGA-PBE functional with the DNP-4.4 basis set and TS dispersion

correction are more consistent with experimental data for zeolite N and available

theoretical and experimental studies on other zeolites.

Previous studies indicated that the choice of atomic charge for zeolite framework

atoms is arbitrary and could be reasonable if in the acceptable ranges48. However, in

this study we demonstrate that the calculated values for optimized partial charges


noticeably depends on the choice of DFT calculation method. This study, with the

support of recent work32, indicates that the partial charges for framework atoms are

strongly dependent on the DFT method and are different from one zeolite type to

another, due to changes in zeolite topology, the available T sites in the zeolite

framework and the Si/Al ratio of zeolite.

5.6 DATA AVAILABILITY

All raw and processed data to reproduce these findings are openly available in

QUT Research Data Finder and can be found at

https://researchdatafinder.qut.edu.au/display/n1225149.


5.7 REFERENCES

1. Castillo, J. M.; Dubbeldam, D.; Vlugt, T. J. H.; Smit, B.; Calero, S., Evaluation of various water models for simulation of adsorption in hydrophobic zeolites. Molecular Simulation 2009, 35 (12-13), 1067-1076. 2. Coppens, P.; Stevens, E. D., Accurate X-ray diffraction and quantum chemistry: The study of charge density distributions. Advances in Qantum Cemistry 1977, 10, 1-35. 3. Hansen, N. K.; Coppens, P., Testing aspherical atom refinements on small-molecule data sets. Acta Crystallographica Section A: Crystal Physics, Diffraction, Theoretical and General Crystallography 1978, 34 (6), 909-921. 4. Coppens, P.; Guru Row, T.; Leung, P.; Stevens, E.; Becker, P. t.; Yang, Y., Net atomic charges and molecular dipole moments from spherical-atom X-ray refinements, and the relation between atomic charge and shape. Acta Crystallographica Section A: Crystal Physics, Diffraction, Theoretical and General Crystallography 1979, 35 (1), 63-72. 5. Pearlman, D. A.; Kim, S. H., Determinations of atomic partial charges for nucleic acid constituents from x‐ray diffraction data. I. 2′‐Deoxycytidine‐5′‐monophosphate. Biopolymers 1985, 24 (2), 327-357. 6. Ghermani, N.; Lecomte, C.; Dusausoy, Y., Electrostatic properties in zeolite-type materials from high-resolution x-ray diffraction: The case of natrolite. Physical Review B 1996, 53 (9), 5231. 7. Levy, M., Universal variational functionals of electron densities, first-order density matrices, and natural spin-orbitals and solution of the v-representability problem. Proceedings of the National Academy of Sciences 1979, 76 (12), 6062-6065. 8. Kohn, W.; Sham, L. J., Self-consistent equations including exchange and correlation effects. Physical Review 1965, 140 (4A), A1133. 9. Kohn, W., Nobel Lecture: Electronic structure of matter—wave functions and density functionals. Reviews of Modern Physics 1999, 71 (5), 1253. 10. Cohen, A. J.; Mori-Sánchez, P.; Yang, W., Insights into current limitations of density functional theory. Science 2008, 321 (5890), 792-794. 11. Tsuneda, T., Density functional theory in quantum chemistry. Springer: 2014. 12. Lewars, E. G., Computational chemistry: introduction to the theory and applications of molecular and quantum mechanics. Springer: 2003. 13. Cohen, A. J.; Mori-Sánchez, P.; Yang, W., Challenges for density functional theory. Chemical Reviews 2011, 112 (1), 289-320. 14. Barrer, R.; Hinds, L.; White, E., 299. The hydrothermal chemistry of silicates. Part III. Reactions of analcite and leucite. Journal of the Chemical Society (Resumed) 1953, 1466-1475. 15. Mackinnon, I. D. R.; Millar, G. J.; Stolz, W., Low temperature synthesis of zeolite N from kaolinites and montmorillonites. Applied Clay Science 2010, 48 (4), 622-630. 16. Mackinnon, I. D. R.; Millar, G. J.; Stolz, W., Hydrothermal syntheses of zeolite N from kaolin. Applied Clay Science 2012, 58, 1-7. 17. Christensen, A. N.; Fjellvag, H., Crystal structure determination of zeolite N from synchrotron X-ray powder diffraction data. Acta Chemica Scandinavica 1997, 51, 969-973. 18. Christensen, A. N.; Fjellvag, H., nuetron powder diferaction study of the dehydration of zeolite N. Acta Chemica Scandinavica 1999, 53, 85-89.


19. Baerlocher, C.; McCusker, L. B.; Olson, D. H., Atlas of zeolite framework types. Elsevier: 2007. 20. Mackinnon, I. D. R.; Barr, K.; Miller, E.; Hunter, S.; pinel, T., Nutrient Removal from waste water using high performance materials. Water Science and Technology 2003, 47, 101-107. 21. Zwingmann, N.; Singh, B.; Mackinnon, I. D. R.; Gilkes, R. J., Zeolite from alkali modified kaolin increases NH4+ retention by sandy soil: Column experiments. Applied Clay Science 2009, 46 (1), 7-12. 22. Zwingmann, N.; Mackinnon, I. D. R.; Gilkes, R. J., Use of a zeolite synthesised from alkali treated kaolin as a K fertiliser: Glasshouse experiments on leaching and uptake of K by wheat plants in sandy soil. Applied Clay Science 2011, 53 (4), 684-690. 23. Thornton, A.; Pearce, P.; Parsons, S. A., Ammonium removal from solution using ion exchange on to MesoLite, an equilibrium study. J Hazard Mater 2007, 147 (3), 883-9. 24. Di Lella, A.; Desbiens, N.; Boutin, A.; Demachy, I.; Ungerer, P.; Bellat, J.-P.; Fuchs, A. H., Molecular simulation studies of water physisorption in zeolites. Physical Chemistry Chemical Physics 2006, 8 (46), 5396-5406. 25. Campañá, C.; Mussard, B.; Woo, T. K., Electrostatic potential derived atomic charges for periodic systems using a modified error functional. Journal of Chemical Theory and Computation 2009, 5 (10), 2866-2878. 26. Salmas, R. E.; Demir, B.; Yıldırım, E.; Sirkecioğlu, A.; Yurtsever, M.; Ahunbay, M. G., Silver–Sodium Ion Exchange Dynamics in LTA Zeolite Membranes. The Journal of Physical Chemistry C 2013, 117, 1663. 27. Uzunova, E. L.; Mikosch, H., Cation Site Preference in Zeolite Clinoptilolite: A Density Functional Study. Microporous Mesoporous Mater. 2013, 177, 113. 28. Uzunova, E. L.; Mikosch, H., Adsorption and Activation of Ethene in Transition Metal Exchanged Zeolite Clinoptilolite: a Density Functional Study. ACS Catalysis 2013, 3, 2759−2767. 29. Murthy, V.; Khosravi, M.; Mackinnon, I. D. R., Molecular Modeling of Univalent Cation Exchange in Zeolite N. The Journal of Physical Chemistry C 2018, 122 (20), 10801-10810. 30. Awuah, J. B.; Dzade, N. Y.; Tia, R.; Adei, E.; Kwakye-Awuah, B.; Catlow, R. A.; De Leeuw, N. H., A density functional theory study of arsenic immobilization by the Al(iii)-modified zeolite clinoptilolite. Physical Chemistry Chemical Physics 2016, 18 (16), 11297-11305. 31. Fischer, M.; Kim, W. J.; Badawi, M.; Lebegue, S., Benchmarking the performance of approximate van der Waals methods for the structural and energetic properties of SiO2 and AlPO4 frameworks. The jJournal of Chemical Physics 2019, 150 (9), 094-102. 32. Wolffis, J. J.; Vanpoucke, D. E. P.; Sharma, A.; Lawler, K. V.; Forster, P. M., Predicting partial atomic charges in siliceous zeolites. Microporous and Mesoporous Materials 2019, 277, 184-196. 33. Baur, W. H.; Fischer, R. X., The floppiness of It all: bond lengths change with atomic displacement parameters and the flexibility of various coordination tetrahedra in zeolitic frameworks. An empirical structural study of bond lengths and angles. Chemistry of Materials 2019. 34. Delley, B., An all-electron numerical method for solving the local density functional for polyatomic molecules. The journal of chemical physics 1990, 92, 508.


35. Delley, B., From molecules to solids with the DMol3 approach. The journal of chemical physics 2000, 113, 7756. 36. Perdew, J. P.; Wang, Y., Accurate and simple analytic representation of the electron-gas correlation energy. Physical Review B 1992, 45 (23), 13244. 37. Vosko, S. H.; Wilk, L.; Nusair, M., Accurate spin-dependent electron liquid correlation energies for local spin density calculations: a critical analysis. Canadian Journal of physics 1980, 58 (8), 1200-1211. 38. Perdew, J. P.; Wang, Y., Pair-distribution function and its coupling-constant average for the spin-polarized electron gas. Physical Review B 1992, 46 (20), 12947. 39. Perdew, J. P.; Burke, K.; Ernzerhof, M., Generalized gradient approximation made simple. Physical review letters 1996, 77 (18), 3865-3868. 40. Perdew, J. P., Generalized gradient approximations for exchange and correlation: A look backward and forward. Physica B: Condensed Matter 1991, 172 (1-2), 1-6. 41. Delley, B., Ground-state enthalpies: evaluation of electronic structure approaches with emphasis on the density functional method. The Journal of Physical Chemistry A 2006, 110 (50), 13632-13639. 42. Grimme, S., Accurate description of van der Waals complexes by density functional theory including empirical corrections. Journal of computational chemistry 2004, 25 (12), 1463-1473. 43. Grimme, S., Semiempirical GGA‐type density functional constructed with a long‐range dispersion correction. Journal of computational chemistry 2006, 27 (15), 1787-1799. 44. Ortmann, F.; Bechstedt, F.; Schmidt, W., Semiempirical van der Waals correction to the density functional description of solids and molecular structures. Physical Review B 2006, 73 (20), 205101. 45. Tkatchenko, A.; Scheffler, M., Accurate Molecular Van Der Waals Interactions from Ground-State Electron Density and Free-Atom Reference Data. Physical review letters 2009, 102 (7), 073005. 46. Orobengoa, D.; Capillas, C.; Aroyo, M. I.; Perez-Mato, J. M., AMPLIMODES: symmetry-mode analysis on the Bilbao Crystallographic Server. Journal of Applied Crystallography 2009, 42 (5), 820-833. 47. de la Flor, G.; Orobengoa, D.; Tasci, E.; Perez-Mato, J. M.; Aroyo, M. I., Comparison of structures applying the tools available at the Bilbao Crystallographic Server. Journal of Applied Crystallography 2016, 49 (2), 653-664. 48. Demontis, P.; Gulín-González, J.; Jobic, H.; Suffritti, G. B., Diffusion of water in zeolites Na A and NaCa A: a molecular dynamics simulation study. The Journal of Physical Chemistry C 2010, 114 (43), 18612-18621. 49. Khosravi, M.; Murthy, V.; Mackinnon, I. D. R. Zeolite DFT calculations. https://researchdatafinder.qut.edu.au/display/n12251 (accessed 2019-04-12).


Chapter 6: Exchange Mechanism of Alkaline and Alkaline earth Elements in Zeolite N Membranes

This chapter includes a comprehensive study of exchange mechanism of alkaline

and alkaline earth cations in zeolite N membranes. Section 6.1 introduces previous

experimental and computational studies on zeolite N. Model constructions and

simulation settings are explained in section 6.2. Section 6.3 presents the results for

measurement of cation exchange, structural arrangements of cations and their mobility

inside zeolite N membranes obtained from MD simulations. The exchange

mechanisms of cations are comprehensively discussed in section 6.4 with outcomes

presented in section 6.5.

6.1 INTRODUCTION

The potassium-rich zeolite K-F(Cl), later renamed zeolite N with the general

formula K12Al10Si10O40Cl2.5H2O, was initially synthesised by Barrer et al.1 in 1953.

Christensen and Fjellvag determined the crystal structure of zeolite N using high

resolution X-ray and neutron diffraction data2, 3. The structure of zeolite N is

orthorhombic with space group I222 and lattice parameters a=9.9041, b=9.8860 and

c=13.0900. Zeolite N is in the EDI framework group and is considered a fibrous

zeolite. The Si/Al ratio of end member (un-exchanged) composition of zeolite N, is

equal to one that provides a high ion-exchange capacity. The framework of zeolite N

has low tortuosity and the predominant eight-membered channel along the c axis

provides an unimpeded path for ions to transfer or transport to exchangeable sites

inside the cages. Furthermore, the extra-framework cations, potassium, located at the

exchangeable sites inside the cages can be exchanged due to their accessible positions

and weak electrostatic bonds to water molecules and framework atoms. These

properties make zeolite N an interesting candidate for ion-exchange applications.

The high capacity of zeolite N for selective ion exchange applications, compared

with competitive natural zeolites, has been verified by experimental studies.


Mackinnon et al.4 and Thornton et al.5 indicate that zeolite N has a robust potential for

ammonia removal (up to 90%) from return side streams of wastewater treatment plants

with an inlet ammonium concentration ranging between 600 mg/L and 900 mg/L.

These investigations reported 45-55g NH4+-Nkg-1 ammonium loading capacity for

synthetic zeolite N, while the natural zeolite, clinoptilolite, which has been used

extensively for ammonium removal applications, shows a much lower loading

capacity for ammonium, in the range of 0.94-21.52g NH4+-Nkg-1 5, 6. In agronomy

applications of zeolite N, Zwingmann et al.7 demonstrated that adding small amounts

(0.4%) of zeolite N to sandy soils effectively increased NH4+ retention capability. In

controlled glasshouse trials, Zwingmann et al.7 showed that the performance of zeolite

N is 11 times higher than natural zeolite clinoptilolite under the same conditions.

Moreover, the exchange behaviour of zeolite N is investigated under different

experimental conditions. Mackinnon et al.8 and Thornton et al.6 show that the initial

solution concentration and pH impact the ammonium uptake by zeolite N. They

reported that increasing the ammonium concentration in solution results in an increase

in the rate and capacity of ammonium removal from solution. Thornton et al.6 found

that the pH 6-7 as the optimum pH for ammonium removal. Thornton et al.6 showed

that the capacity of zeolite N for ammonium uptake decreases by 30% in the presence

of competing cations, sodium, calcium and magnesium. Mackinnon et al.8 reported

that the presence of magnesium and calcium in a mixed solution (with different cation

concentration compared to the Thornton et al. study6) has no significant effect on

ammonium uptake. However, the presence of sodium slightly decreases the capacity

for removal of ammonium. These experimental studies indicate a preference by zeolite

N for univalent cation selectivity compared to divalent cations.

These experimental investigations present valuable data on the ion-exchange

capability and comportment of zeolite N. Furthermore, in our recent study9 we

simulated the exchange of univalent cations, NH4+, Na+, K+, Rb+ and Cs+, in a zeolite

N membrane using molecular dynamics calculations. We studied the structural and

dynamic behaviour of ions inside a zeolite N membrane. The results show that zeolite

N prefers K+ exchange with NH4+ rather than with Na+, Rb+ or Cs+. Moreover, the

behaviour of zeolite N at different hydration levels was investigated. The outcomes of

our molecular dynamics calculations are in good agreement with experimental data for

ammonium and sodium exchange with potassium and hydrated zeolite N. This


modelling approach and outcomes that, in general, conform with experimental data,

show that computational modelling can be used to understand detailed, atomic scale

mechanistic interactions for ion exchange of zeolite N.

In this study, we present further details of ion exchange mechanisms for zeolite

N based on exchange of monovalent cations inside a zeolite N membrane, as well as

the relative performance of Li+, Ca2+ and Mg2+ cations. We explore ion retention

within a zeolite N membrane along different crystallographic directions, [001] and

[110] as well as the site preference of exchanged cations in the zeolite N structure. The

outcomes in Chapter 5 show that the partial charges of Si and Al atoms can be

substantially different depending on the functional used in DFT calculations. In this

chapter, we also consider the effect of these different partial charges for framework Si

and Al atoms on the dynamic behaviour of zeolite N.

6.2 COMPUTATIONAL METHODS

The primary unit cell for zeolite N, used in this study, is based on the crystal

structure defined by Christensen and Fjellvag2 using synchrotron X-ray powder

diffraction. They identified two different sites for Si, two for Al and five for the O

atoms that construct the framework. Two different sites are identified for the potassium

extra-framework cations10. Site I is located in the middle of the eight-membered rings

along the (001) direction (K1) and Site II is located in the middle of the other eight-

membered rings along the [110] direction. Figure 6-1 illustrates the position of these

sites in the zeolite N unit cell. The Materials Studio (version 18.1) suite of programs

is used to construct the zeolite N models, DFT calculations and subsequent MD

simulations.

6.2.1 DFT Calculations

The partial charges of zeolite N framework atoms are calculated by periodic DFT

methods on a zeolite N unit cell without extra-framework atoms and water molecules.

The geometry optimization and population analysis is obtained using the GGA-PBE

functional11 with double numerical plus polarization basis sets 4.412. The convergence

tolerance criteria are 1x10-5 Ha, 0.002 Ha/Å and 0.005 Å for energy, force and

displacement convergence, respectively. The SCF convergence criterion is set to an

energy tolerance 1x10-6 Ha. The Mulliken partial charges are obtained from population

analysis of DMol3 13, 14 code (DMol3; Accelrys Inc.: San Diego, CA, 2016.) in Accelrys


Materials Studio. The geometry optimized unit cell is cleaved along two different

planes, (001) and (110), and then capped with –OH groups on both surfaces with a

vacuum slab of 5 Å. In order to obtain the partial charges of O and H atoms on the

surfaces, the cells are optimized by the DFT model described above. Table 1 in

Appendix B represents the calculated Mulliken partial charges for framework atoms,

O and H atoms at surfaces. The partial charges of extra-framework atoms are

considered equal to their ionic charge.

Figure 6-1 The illustrations of SI and SII for extra-framework K in zeolite N supercells along (a) [001] and (b) [110] crystallographic directions. The 8-membered ring pore openings in each channel

direction are highlighted with green colour. The black dashed lines indicate the interaction of potassium cations in site I and II with framework oxygen atoms. Atoms are coloured as

Silicon=yellow, Aluminium=pink, oxygen=red, potassium=purple and chloride=light green.

6.2.2 MD Simulations

In order to investigate the exchange capability of zeolite N along two different

channel directions, two membranes are built along the [001] and [110] directions. A

2x2x2 supercell is used to make the membrane along [001] and a 2x2x1 supercell is

used to make the membrane along [110]. Both supercells were cleaved and then caped

with –OH on the surfaces with vacuum slabs of 20 Å on either side. Thus, two different


membranes of zeolite N as shown in Figures 6-2 and 6-3 are generated and labelled as

ZM-001 and ZM-110. The size of ZM-001 is 19.8x19.8x28.2 Å3 and ZM-110 is

13.1x27.9x29.6 Å3. These two membranes are different sizes in order to maintain equal

numbers of framework and extra-framework atoms for both types of membranes.

Table 1 in Appendix B presents the number of framework, extra-framework and water

molecules in both membranes. Water molecules with a density of one g/cm3 are added

to either side of the vacuum slab of membrane models. The flexible SPC model of

water15 is used for all simulations.

Prior to all MD simulations, a geometry optimisation with periodic boundary

conditions is conducted on zeolite membranes in water. The minimization is carried

out by a quasi-Newton procedure with 500 iterations and the same convergence criteria

in DFT models are applied. The electrostatic interactions are calculated by Ewald

summation16 with accuracy 1x10-4 kcal/mol. However, the direct cut-off with 15.5 Å

distance is applied for determining the van der Waals interactions. We do not use the

Ewald summation for long range interactions in order to decrease the time for

computation. In addition, the 15.5 Å distance cut-off is sufficiently accurate compared

to the Ewald summation. Initially, the zeolite framework in both membranes is kept

rigid to allow extra-framework atoms and water molecules to displace with respect to

each other to reach minimum energy. These optimised models are used as starting

configurations for further MD simulations. In these MD simulations, all framework,

extra-framework atoms and water molecules are released in order to move freely

during the simulation. Our previous work9 shows that the structure of zeolite N

framework is quite stable under this condition and shows no significant change.

Both ZM-001 and ZM-110 membranes contain 80 Si and Al atoms, 96 K+ and

16 Cl- ions and 128 water molecules. The charge of the framework in each membrane

is -80e. In order to investigate the retention of ions inside membranes, a chemical

potential is created between the inside and outside of the membranes. The guest Mn

(n= +1 or +2) cations are placed randomly inside the membranes and 40 chloride ions

are distributed in the solvent on either side of the membranes. The number of guest

cations depends on their total charge compensating the -80e of frameworks. Therefore,

the number of added guest mono-and divalent cations are 80 and 40, respectively. The

guest ions include NH4+, Li+, Na+, K+, Rb+, Cs+, Mg2+ and Ca2+ cations. The models


created containing guest cations and extra chlorides, are designated as K+/Mn systems

in subsequent sections (represented in Fig.6-3a and 6-4a).

A 30 ps equilibrium MD simulation is performed in an NVT ensemble at 298 K

with NHL thermostat17 and with a time step of 1.0 fs followed by a production MD

simulation for 8.5 ns. This method allows evaluation of the dynamic, structural and

statistics properties as well as ions and water molecule retention inside zeolite N

membranes for the full MD simulation. To estimate ion localisation inside and outside

the membranes, radial distribution functions (RDF) of atom pairs and concentration

profiles of ions inside and outside the membranes as well as the electron density fields

of ions inside membranes are determined from the last 1ns of the MD simulation. The

dynamics of ions in the membrane and the solvent are studied by calculating the self-

diffusion coefficient (D). This parameter is computed using the mean square

displacement (MSD) of ions over time.

Moreover, the statistical errors in outcomes, resulted from MD simulation

sampling, are estimated with different approaches. In order to calculate the accuracy

of calculations for D values, five different 1ns MD simulations are conducted after

8.5ns simulations for each K+/Mn system and the average MSD of each ion is used to

calculate D. The five MD runs used different starting coordinates and random

velocities for each simulation. The framework was fixed in these simulations while

ions and water molecules were released to move inside and outside the membranes.

Moreover, Appendix C describes the example simulations for estimating the errors

resulted from the model construction and initial distribution of guest cations within the

zeolite N membrane.

6.3 RESULTS

The statistics, structural and dynamics results of ions and water molecules from

MD simulations are presented in this section. We compare results obtained for

different K+/Mn+ systems in the ZM-001 and ZM-110 membranes.

6.3.1 Ion Retention

The retention of ions and water molecules inside ZM-001 and ZM-110 over 8.5

ns simulation times are plotted in Figures 6-2a and b. These plots show the retention

ratio between guest cations and extra-framework K+ ions of zeolite N membranes for

ZM-001 and ZM-110. The retention ratio of guest to host ions in ZM-001 is higher


than that for all exchanging systems. However, ZM-110 shows different behaviour for

the retention of ions. NH4+ has the highest retention ratio in both membranes. The

ammonium to K+ retention ratio in ZM-001 fluctuates around 1.7 over time. In ZM-

110, the ammonium retention ratio increases up to 2.5 and after 4ns decreases but

remains higher than ZM-001. The Li+ and Na+ ions show roughly the same behaviour

in both membranes with slightly higher retention ratios over time in ZM-001. Both

types of membranes initially release Li+ and Na+ to the solution but over time,

especially after 6ns in ZM-001, the membrane exchanges more K+ with Li+ and Na+ in

the solution. The K+ retention equilibrates in ZM-001 faster than for ZM-110. The

retention ratios for Rb+ and Cs+ in ZM-001 are lower than ZM-110; however, the

retention ratios decrease over time in ZM-110. The retention ratios for Mg2+ and Ca2+

are higher in ZM-001 compared to ZM-110. The retention ratio of Mg2+ and Ca2+ in

ZM-110 are about one and below one, respectively and much lower than other ions

studied.

Experimentally, the zeolite N unit cell contains 12 monovalent cations

compensating the negative charge of framework (-10e) and two chloride ions. We find

that the average total compensating charge on cations per unit cell of ZM-001 and ZM-

110 (Figure 6-2c and d) in most exchanging systems is less than 12 except for the Li+

system in ZM-001 and Li+ and Na+ exchanging systems in ZM-110. The retention

results show that the membranes retain fewer cations compared to the original zeolite

N membranes without guest cations (K-ZM-001 and K-ZM-110). Here, we consider

the total compensating charge on the cations rather than the number of cations (in

Figures 6-2c and d). Investigating the average number of Cl- ions retained in each unit

cell of the membranes (Figure 6-2g and h) shows that the number of chloride ions

decrease over time especially in ZM-001. In the K+/Li+ system of ZM-110, the

membrane adsorbs more than 2 Cl- anions per unit cell at the initial stages of ion

exchange but subsequently releases excess Cl- ions into solution.

Zeolite N contains eight water molecules per cage. Counting the average number

of water molecules in each cage of ZM-110 (Figure 6-2i and j) demonstrates that the

membrane adsorbs water molecules in K+/Li+, K+/Na+, K+/K+, K+/Rb+, K+/Mg2+ and

K+/Ca2+ systems up to one extra water molecule per cage but releases water molecules

in K+/NH4+ and K+/Cs2+ systems (Figure 6-2j). However, in the ZM-001 membrane

only K+/Li+ and Mg2+/K+ systems adsorb water molecules (Figure 6-2i). Zeolitic water


molecules in K+/NH4+, K+/Na+, K+/K+, K+/Rb+, K+/Cs+ and K+/Ca2+ systems exit the

ZM-001 membrane within 1ns of MD simulations due to the stress created by addition

of excess cations inside the membrane. Over time, the membrane re-adsorbs water

molecules to reach an equilibrium condition. This feature is most obvious for the

K+/Cs+ system. Table 6-1 presents the number and percentage of retained extra-

framework K+ and Cl- ions, water molecules and guest ions inside ZM-001 and ZM-

110 after 8.5ns simulation time. As shown in Table 6-1 in the K+/K+ systems of ZM-

001 and ZM-110, ~50% of the potassium is retained in the membrane. The number of

potassium ions exchanged by both membranes in the K+/K+ systems after 8.5 ns

simulation is similar to that in the K-ZM-001 and K-ZM-110 membrane without guest

cations. Ca2+ and NH4+ ions have the highest retention in the ZM-001 membrane with

67.5% and 66.2% retention respectively. The Li+, Na+ and Mg2+ show more than 60%

retention in ZM-001 (with 63.7%, 63.2% and 62.5% retention, respectively). However,

Rb+ and Cs+, with 51.4% and 50.1% retention show the lowest capacity to remain in

the ZM-001 membrane. In ZM-110 membrane, NH4+ shows the highest retention of

70.8% while Na+, Li+, Cs+, Rb+ and Ca2+ show more than 50% retention of 64.3%

,59.4%, 58.9%, 56.7% and 54.9%, respectively. Mg2+ shows the lowest retention of

44.8%.

Zeolite N contains two different exchange sites (SI and SII) for extra-framework

potassium, designated K1 and K2 (Figure 6-1). The K-ZM-001 and K-ZM-110

membranes without guest cations contain 32 K1 and 64 K2 cations. We estimate the

number of ions in K1 and K2 sites in order to validate their relative exchange

capability. The retention of K1 and K2 in ZM-001 and ZM-110 shows different

behaviour depending on the guest ions. In K+/NH4+, K+/Na+ and K+/Cs+ systems the

percentage of K1 retained in the membrane is higher compared to K2. However, in

K+/Li+ and K+/K+ systems of both membranes K2 retention percentage is higher than

K1. In the K+/Rb+ system, the K2 retention percentage is more than K1 in ZM-001 and

less than K1 in ZM-110. However, K1 and K2 retention in K+/Mg2+ and K+/Ca2+

systems of both membranes show the reverse behaviour to the K+/Rb+ system.


Figure 6-2 The retention ratio of guest to host ions in (a) ZM-001 and (b) ZM-110, the total charge on cations per unit cell of (c) ZM-001 and (d) ZM-110, the retention of total guest cations in (e) ZM-001 and (f) ZM-110, the number of chlorides in each unit cell of (g) ZM-001 and (h) ZM-110, and water

molecules per cage of (i) ZM-001 and (j) ZM-110, respectively.


6.3.2 Ion Distribution

The MD calculations allow determination of the relative distribution of ions and

bonding characteristics between different atom pairs within the membranes and in the

Table 6-1 Num

ber of initial and retained ions in K-ZM

without guest cations, ZM

-001 and ZM-110 m

embranes as

well as their com

parison with previous study. The potassium

retained in Site I and Site II after 8.5ns simulations

are presented as a percentage.


surrounding electrolytes. The proportion of host and guest ions inside and outside the

membranes are determined by concentration profiles. The location of ions inside the

channels and cages of membranes are estimated with ion density field maps. The

structural arrangements of host and guest ions around framework atoms is

characterised by calculating the radial distribution functions (RDFs).

The concentration profiles of host and selected guest cations along z direction

inside ZM-001 and ZM-110 and their surrounding solution after 8.5 ns MD simulation

are presented in Figures 6-3 and 6-4, respectively. Ion concentration profiles show the

occupation of middle parts of the membrane by NH4+, Li+, Na+, K+, Rb+, Cs+, Mg2+

and Ca2+. The distribution of these ions in the middle of each membrane is uniform for

NH4+, K+, Rb+ and Cs+ cations. However, Li+, Na+, Mg2+ and Ca2+ cations are unevenly

distributed in the middle of membranes. The opening of channels in both membranes

are occupied by guest ions in K+/NH4+, K+/Li+, K+/Na+, K+/Mg2+ and K+/Ca2+ systems.

However, the opening of channels are devoid of any cations in the K+/K+, K+/Rb+ and

K+/Cs+ systems of both membranes. Moreover, concentration profiles show the

adsorption of Li+ and Mg2+ guest ions on the surfaces of ZM-110. In all systems of

both membranes, K+ cations prefer to stay in the middle of membrane and the

concentration of K+ ions in the solution is higher near membrane surfaces.

The density field maps of K+ and Mn guest cations inside ZM-001 and ZM-110

are illustrated in Figures 6-5 and 6-6, respectively. The ion density field maps in both

ZM-001 and ZM-110 show that K+, NH4+, Rb+ and Cs+ cations occupy both the middle

of channels and close to cages surfaces. However, Li+, Na+, Mg2+ and Ca2+ cations

prefer to stay close to cage surfaces.

The RDFs of guest cations with O, Si and Al framework atoms, Cl- and oxygen

of water molecules inside membranes are calculated (see Figures 1 and 2 in Appendix

B). The first peak of g(r) shows the nearest distance of guest cations to the framework

atoms. These nearest distances are listed in Table 6-2. The peak intensities of g(r) show

the strength of the interaction between atom pairs and number of strong peaks show

regular arrangements of atom pairs.

The RDF results in Table 6-2 show that K+ cations are at the same distances to

the framework atoms, chloride ions and water molecules inside both membranes. The

average distances for O-K+, Si-K+, Al-K+, Cl--K+ and Ow-K+ are 2.43, 3.13, 3.19, 2.58,

and 2.94 Å, respectively. The O-Mn distances of guest cations in ZM-001 and ZM-110


are shorter than Si-Mn and Al-Mn distances. Also, the Mn guest cations are closer to

framework oxygen atoms compared to Cl- ions and oxygen of water molecules inside

membranes.

In addition to the first peaks, another one or two noticeable strong peaks, with

equal or higher g(r) intensities, are observed for Li+, Na+, Mg2+ and Ca2+ arrangements

around framework Si and Al atoms (Figures 1 and 2 in Appendix B).

Figure 6-3 (a) ZM-001 simulation box along z direction and (b-i) ion concentration profiles along z direction after 8.5 ns MD simulations. The two red dashed lines indicate the location of ZM-001 in

electrolyte solution.


Figure 6-4 (a) ZM-110 simulation box along z direction and (b-j) ion concentration profiles along z direction after 8.5 ns MD simulations. The two red dashed lines indicate the location of ZM-110 in

electrolyte solution


Figure 6-5 Density field maps of Mn guest cations in (a) K+/Li+, (b) K+/Na+, (c) K+/NH4+, (d) K+/Cs+,

(e) K+/ K+, (f) K+/Rb+, (g) K+/Mg2+ and (h) K+/Ca2+ systems retained inside ZM-001 after 8.5 ns MD simulations.


Figure 6-6 Density field maps of Mn guest cations in (a) K+/Li+, (b) K+/Na+, (c) K+/NH4+, (d) K+/Cs+,

(e) K+/ K+, (f) K+/Rb+, (g) K+/Mg2+ and (h) K+/Ca2+ systems retained inside ZM-110 after 8.5 ns MD simulations


Table 6-2 The nearest distances of Mn ions into fram

ework oxygen (O

-Mn), silicon (Si-M

n) and aluminium

(Al-M

n) atom

s, chloride ions (Cl-M

n) and oxygen of water m

olecules (Ow -M

n), inside mem

branes and their comparison w

ith previous study.


6.3.3 Ion Mobility

The MSD of ions inside and outside of membranes computed from five different

simulations and the average MSD for each ion was calculated. The self-diffusion

coefficient value (D) of each ion is estimated from the slope of the average MSD. The

average D values of ions inside and outside of the membranes are compared and

presented in Table 6-3, Figure 6-7 and 6-8.

The obtained D values shown in Figure 6-7 indicate that ions are more mobile

inside ZM-110 compared to ZM-001 except for Rb+ which moves faster in ZM-001.

Moreover, guest ions are less mobile inside membranes compared to extra-framework

K ions, except for the K+/Rb+ system in ZM-001 and K+/Ca2+ systems in ZM-110. The

self-diffusion of NH4+, Li+, Na+ and Cs+ cations are negative inside ZM-001 as well as

Li+ and Cs+ D values inside ZM-110. The mobility of ions in ZM-001 are in the order

Cs<Na<Li<NH4<Mg<Ca<K<Rb. The relative ion mobility in ZM-110 is in the order

Cs<Li<Mg<NH4<Na<K<Rb<Ca.

The ion D values shown in Figure 6-8 indicate that all ions are more mobile in

the electrolyte outside ZM-001 compared to ZM-110, as a result of the larger

simulation box for the ZM-001 system. The K mobility in the electrolyte is higher than

guest cations of each system outside ZM-001 and ZM-110, except for the K+/Rb+

system outside ZM-110.

Table 6-3 Self-diffusion coefficient of ions inside ZM-001 and ZM-110 membranes and outside in

the electrolyte solution.

ZM-001 ZM-110

Inside Outside Inside Outside

System Mn+ K+ Mn+ K+ Mn+ K+ Mn+ K+

K+/NH4+ -1.1E-17 1.9E-14 5.76E-09 7.85E-09 1.55E-15 3.35E-14 4.96E-09 6.18E-09

K+/Li+ -3.1E-16 3.37E-14 4.81E-09 9.46E-09 -2.3E-16 5.39E-15 3.57E-09 7.78E-09 K+/Na+ -3E-15 -2.1E-16 6.66E-09 8.78E-09 1.2E-14 1.23E-13 5.51E-09 7.37E-09 K+/K+ 1.18E-14 1.18E-14 7.11E-09 7.11E-09 1.41E-14 1.41E-14 5.95E-09 5.95E-09 K+/Rb+ 2.61E-14 1.16E-14 6.37E-09 7.16E-09 1.57E-14 2.4E-14 4.67E-09 4.37E-09 K+/Cs+ -3.9E-15 3.18E-15 3.62E-09 5.2E-09 -5.6E-16 3.84E-14 3.05E-09 3.78E-09

K+/Mg2+ 4.82E-16 5.22E-15 3.41E-09 7.47E-09 9.01E-16 2.28E-14 1.78E-09 7.14E-09 K+/Ca2+ 8.46E-16 1.07E-14 3.24E-09 6.28E-09 2.49E-14 2.06E-14 1.85E-09 4.1E-09


Figure 6-7 Self-diffusion coefficients for K+ and Mn guest cations of each exchanging system inside (a) ZM-001 and (b) ZM-110 membranes. The gray lines indicate uncertainties.


Figure 6-8 Self-diffusion coefficients for K+ and Mn guest cations of each exchanging system in the electrolyte outside (a) ZM-001 and (b) ZM-110 membranes. The gray lines indicate uncertainties.

6.4 DISCUSSION

Zeolite N is built from chains consisting of one-dimensional Periodic Building

Units (PBU). These tetrahedral PBUs consist of 5T units (T can be Si or Al) connected

together by bridging oxygen atoms along the a and b axes, translated along the c axis

to make connected channels. Zeolite N has channels of 3.6 Å along [001] direction

interconnected with channels along (110) direction with a different eight-membered


ring pore opening shape and size (2.5 Å) (Fig. 6-1). The intersection of these two

channels creates cages that surround the potassium and chloride extra-framework ions

and water molecules. The extra-framework K1 has interaction with three framework

oxygen atoms (shown as black dashed lines in Fig. 6-1a), one chloride and two oxygen

atoms from water molecules. The extra-framework K2 has interaction with four

framework oxygen atoms (shown as black dashed lines in Fig. 6-1b), one chloride and

two oxygen atoms from water molecules. The electrostatic and van der Waals

interactions can easily breakdown during exchange processes resulting in potassium

cations leaving their structural sites for locations of more favourable energy for zeolite

N.

Several experimental studies prove the high ion exchange capacity of zeolite N

and investigate the exchange isotherms for this zeolite. However, the exchange and

diffusion mechanism of cations inside the complex porous structure of zeolite N is still

unknown. In this study, we investigated the ion exchange characteristics of mono- and

divalent cations in zeolite N structure by molecular dynamics calculations. Here, we

discuss the retention, structural arrangement and mobility of monovalent NH4+, Li+,

Na+, K+, Rb+ and Cs+ as well as divalent Mg2+ and Ca2+ cations inside zeolite N

membranes. This allows exploration of the exchange and diffusion mechanism of

cations inside channels along [001] and [110] directions of zeolite N.

6.4.1 Ion Retention

The chemical formula of the zeolite N unit cell indicates 12 exchangeable cations

to compensate for the negatively charged -10e of the aluminosilicate framework and

two Cl anions. The results show that the total charge compensation on cations per unit

cell in most exchange systems is below 12, except for Li+ in ZM-001 and Na+ and Li+

in ZM-110. The potassium-rich ZM-001 and ZM-110 membranes without guest

cations, release 8% and 11%, respectively, of their K to solution and thus, contain less

than 12 cations per unit cell. Over the simulation time, each membrane loses up to one

K+ per unit cell. This calculated outcome is consistent with experimental observations

that show a neutral water solution will record a pH ~ 9 with addition of zeolite N at

room temperature8. Measurements of this solution with zeolite N also show the

presence of K+ ions8.

Our previous simulation outcomes9 show similar conditions regarding the total

number of cations retained in the membrane. These results suggest that the zeolite N


structure prefers to hold no more K or other cations than an approximately equilibrium

value. The one exception to this preference is for the small cation, Li+, for which

zeolite N is able to adsorb more ions than the charge compensating capacity.

Experimental studies provide valuable data on the tendency of zeolite N to

uptake ammonium ions from aqueous solutions5, 6, 8. Moreover, Zwingmann et al.7

showed that ammonium exchanged zeolite N is an ideal slow release fertiliser for

sandy soils due to the high retention capacity of zeolite N for NH4+ ions. Our

computational results from previous9 and this study for NH4+ retention are consistent

with these experimental results. This study shows that NH4+ has the highest retention

between all exchanging systems in zeolite N membranes along both crystallographic

directions. The NH4+ retention in ZM-110 is higher than ZM-001 (Fig. 6-2e and f).

However, the total number of cations that remain in the structure are similar for both

membranes. The small sized channel openings in ZM-110 does not allow NH4+ ions

to leave the membrane and thus ensures capture in 3D cages formed at the intersection

of the (001) and (110) planes. Release of K+ ions from the membrane provides

additional space for retained NH4+ ions. Our calculations suggest that only NH4

+ ions

close to the membrane surfaces can leave the ZM-110 structure. In contrast, NH4+ ions

can more readily leave the ZM-001 membrane due to the larger size of the channel

openings.

Li+ and Na+ are the next monovalent ions that show high and similar retention

behaviour in membranes. The K+/Li+ and K+/Na+ systems in this study hold more

cations compared to other systems due to the small size of Li+ and Na+ ions. In the

ZM-001 membrane Li+ and Na+ show similar retention rates (Fig. 6-2e and f). The

retention of Na+ in the ZM-001 membrane is the same as simulations shown previously

under similar conditions9. However, the K+/Li+ system retains more K+ in the

membrane compared to the K+/Na+ system during the simulation. In the ZM-110

membrane, Na+ retention is slightly higher than Li+. However, both K+/Li+ and K+/Na+

simulations show that a similar amount of K+ is retained in the ZM-110 membrane.

The K+/Rb+ and K+/Cs+ systems show the lowest retention and total number of

cations per unit cell among monovalent cations in both membranes due to the large

size of these ions. The retention of Rb+ and Cs+ is higher in ZM-110 than ZM-001

(Fig. 6-2e and f). Similar to the case for NH4+, the large size of these ions does not

allow passage through the small opening of the ZM-110 membrane channels to the


solute. The retention behaviour of Rb+ and Cs+ compared to other monovalent cations

in this study for ZM-001 is similar to our previous outcomes9. However, in this study,

we found lower retention ratios over time for Rb+ and Cs+.

Experimental data show that the potassic form of zeolite N can take up to three

times more Ca2+ than Mg2+ from a mixed solution of NH4+, Mg2+ and Ca2+ (with low

ammonium concentration, 30 mg/l) and this uptake of divalent ions is 10x lower than

the NH4+ uptake8. This data indicates a high preference of zeolite N for monovalent

cations over divalent ions. Computational results from this study for retention of Mg+2

and Ca+2 ions in zeolite N are in general agreement with experimental outcomes. For

example, the number of guest cations for the K+/Mg2+ and K+/Ca2+ systems show the

lowest value for total retained cations in the membranes (Fig. 6-2e and f). Accordingly,

in these systems the number of K+ ions retained in the membrane are higher than Mg2+

or Ca2+.

Based on the experimentally determined chemical formula for zeolite N, each

unit cell contains two Cl- anions. However, our simulations show that both membranes

release up to one Cl- per unit cell into solution during the exchange process (Figures

6-2g and h). A larger cation size results in greater reduction in the number of chlorides

inside each membrane. Consistent with this, systems with a lower amount of cations

per unit cell require a lower (or equivalent) amount of anions. These results are in also

agreement with experimental data8.

Experimental and computational studies show that hydrated zeolite N contains 8

water molecules per cage2, 9. The results from this study reveal that zeolite N

membranes along different directions show different hydration behaviour during the

exchange process (Figures 6-2i and j). For example, the number of water molecules

per cage changes during the ion-exchange process depending on the guest ions. A ZM-

001 membrane releases zeolitic water into solution at early stages of the exchange

process in all K+/Mn systems and subsequently continuously adsorbs water molecules

over time. This feature is noteworthy for the K+/Cs+ system of the ZM-001 membrane.

Concentration profiles confirm that this increase is due to adsorbed water molecules

in the opening of the pores to the membrane. However, the total number of water

molecules inside the membrane cages is constant (Figure 4 of Appendix B). All

exchanging systems in ZM-110 adsorb more than 8 water molecules per cage during

the simulation except for the K+/Rb+ and K+/Cs+ systems. Zeolite N channels are not


large enough to accommodate K+, Rb+ and Cs+ cations together with water molecules.

As with ZM-001, water molecules concentrate at the pore openings of the zeolite N

membrane.

6.4.2 Ion localization

Ion concentration profiles, electron density field maps and RDF results provide

different perspectives on the localization of guest and host ions within zeolite N

membranes.

The concentration profiles show that ions undergo similar localisation behaviour

within the ZM-001 and ZM-110 membranes (Figures 6-3 and 6-4). The ionic size

predominantly affects the distribution of ions inside membranes whereby larger ions,

such as NH4+, K+, Rb+ and Cs+, are distributed more uniformly compared with smaller

ions such as, Li+, Na+, Mg2+and Ca2+. On the other hand, these simulations show that

NH4+, Li+, Na+, Mg2+and Ca2+ ions concentrate in channel openings at the surface of

the membrane due to their strong interaction with framework oxygen atoms.

The electron density field maps indicate that for ion localisations inside both

types of membrane cages (Figures 6-4 and 6-5) K+ and NH4+ ions localise at SI, SII

and slightly shifted positions close to these two sites. However, ions larger than K+,

Rb+ and Cs+, exactly occupy both SI and SII sites. Ions smaller than K+, including Li+,

Na+, Mg2+and Ca2+, reside in disordered crystallographic positions closer to the

framework rather than at crystallographic sites. These ions localise differently in ZM-

001 and ZM-110. The Li+, Na+, Mg2+and Ca2+ ions occupy disordered crystallographic

positions in cages of ZM-001 that are closer to SII sites. However, their locations in

ZM-110 cages are closer to SI sites. The Na+ and Ca2+ ions are located at further

distances to the framework compared to Li+ and Mg2+ ions, due to their comparatively

larger ionic size.

The nearest distance values between atom pairs obtained from RDF in Table 6-

2 indicate that all guest cations have stronger interaction with framework oxygen

atoms than with oxygen atoms of the water molecules or with Cl- anions inside the

membranes. The RDF results show that K+ ions are located at the same distances to

the framework in all exchanging systems, though the membranes contain different

guest ions with different ionic sizes. The average calculated K distances to O, Si and


Al of the framework, Cl anions and oxygen of water molecules inside the membranes

are in good agreement with XRD data2.

The nearest distance of Mn cations to the framework atoms are identified by the

position of the first peak of the function g(r). The distances of Mn cations to the zeolite

N framework atoms of ZM-001 and ZM-110 membranes are similar except for Rb+

and Cs+ cations. In ZM-110, most Rb+ and Cs+ cations tend to locate in the middle of

cages at a further distance to the framework. However, in ZM-001these ions are

equally localised at both sites.

The nearest distances for NH4+ to the framework oxygen and to Si or Al atoms

of ZM-001 are larger and smaller, respectively, than previously obtained values9.

However, the nearest distances for Na+, K+, Rb+ and Cs+ are smaller than previous

results9. These differences in nearest distance of atom pairs between these models of

zeolite N exchange, are related to different partial charges on framework atoms used

in simulations.

Furthermore, the RDF graphs in this study show notable first peaks for Li+ as

well as for Na+, Mg2+ and Ca2+ around the framework Si and Al atoms (Figures 1 and

2 of Appendix B). These nearest distances for Li+, Na+, Mg2+ and Ca2+ to framework

Si and Al atoms are due to the small sizes of these cations as well as to the presence

of two different Si and Al atomic positions with different partial charges in the zeolite

framework. RDF plots and density field maps indicate that these ions are closer to the

Si/Al type 2 site(s) rather than the Si/Al type 1 site(s).

The strength of the interaction between atom pairs is estimated from peak

intensities of g(r). The peak heights for O-Mn for NH4+, Li+, Na+, Mg2+ and Ca2+ are

higher than Si/Al-Mn. Moreover, the O-Mn value for these ions is the nearest distance

to the framework atoms that show strong interaction with framework oxygen atoms

compared to Si and Al. The interaction increases from NH4+<Na+<Ca2+<Li+<Mg2+. In

contrast, the higher peaks for Si/Al-Mn where M=K+, Rb+ or Cs+ are not the nearest

distances and consequently have weaker interaction with framework oxygen atoms.

The interaction strength decreases from K+>Rb+>Cs+, in complete agreement with

previous outcomes9. RDF results using ion density profiles and ion density fields

confirm that ions are localised inside the framework relative to their ionic size. An

exception to this outcome is NH4+ for which hydrogen bonding provides stronger


interactions with framework oxygen atoms compared with other cations evaluated in

these simulations.

6.4.3 Ion Diffusion

We investigate the relative mobility of ions inside and outside zeolite N

membranes by calculating the self-diffusion coefficient (D) of ions from their mean

square displacement (MSD) over the simulation time. The diffusion behaviour of ions

inside the confined geometry of zeolites is clearly very different from their bulk

behaviour in solution. Simulations show that values of D for ions inside zeolite N

membranes are smaller than values obtained in the electrolyte solution by several

orders of magnitude (Figure 6-7, 6-8 and Fig. 3 of Appendix B).

The D values for cations inside both membranes are close to zero. Moreover, the

measured D values for NH4+, Li+, Na+ and Cs+ cations are negative inside ZM-001 as

well as for Li+ and Cs+ inside ZM-110. Close inspection of MSD curves reveal that

ions do not show diffuse behaviour while some curves show different behaviour

regimes over time with positive and negative slopes. This behaviour means that

movement of ions inside the membranes is significantly affected by a number of

mechanisms. The small or negative diffusion of ions in this study indicates that these

ions can not pass the free energy barriers inside zeolite N channels and jump from one

low energy site to another. Therefore, these ions localise in a specific position within

the structure and show an oscillatory behaviour.

These energy barriers are present in all directions and include dispersion-

repulsion and electrostatic energies between ions and the framework as well as the

activation energy that a particular ion requires to move between different

crystallographic positions18. The density field illustrations shown in Figures 6-4 and

6-5 are exemplars of this mechanistic interpretation for zeolite N. No systematic

dependence on ionic size is observed for D values of ions inside zeolite N membranes.

There are few experimental studies calculating the self-diffusion of cations

inside different zeolites (e.g. analcite19, chabazite20, mordenite21 and clinoptilolite22)

by measuring all activation, dispersion-repulsion and coloumbic energies. These

studies calculated the self-diffusion coefficient values for monovalent cations NH4+,

Na, K, Rb, and Cs, in different zeolites, in the range of 10-11-10-26 m2sec-1 and for

divalent cations in the range of 10-15-10-17 m2sec-1. The self-diffusion coefficient values


for cations obtained in this study are of similar magnitude to experimentally measured

values of self-diffusion for cations in other zeolites at 25 ᵒC.

6.4.4 Ion Exchange Mechanism

In general, the ion exchange property of zeolites and of cation diffusion in

zeolites depend on various parameters, including cation size, incipient charge on

cations, the smallest free diameter of channel pore-openings, the number of oxygens

involved in the pore-openings, the Si/Al ratio and finally, the water flux inside the

zeolite structure and its interaction with cations. Therefore, the difference between

ZM-001 and ZM-110 ion exchange properties is due to the difference between their

pore-opening diameter and the shape of the pore-opening that interfaces with the

ambient solution, since all other parameters are the same for both membrane types.

However, the interconnection of [001] and [110] channels in both membranes

complicates the intracrystaline exchange and diffusion of cations.

Ammonium ions have the highest retention in these simulations and the closest

distances to framework atoms in both membranes, even though the ionic size for NH4+

is larger than K+. In our previous simulations, we indicated that hydrogen bonding

influences the selectivity of NH4+ in comparison to other monovalent cations9. The

hydrogen in NH4+ interacts with between one and three framework oxygen atoms and

between one and two water molecules or other NH4+ ions. Of these, the interaction

between the NH4+ hydrogen and framework oxygen atoms is the most stable.

These hydrogen bonds result in completely different diffusive behaviour for

NH4+ compared to other cations. The formation and elimination of hydrogen bonding

facilitates the movement of NH4+ ions within zeolite channels. However, K+ cations

leave the membranes more quickly than NH4+ ions due to weaker interactions with

framework and water molecules. As we have seen, the self-diffusion of K+ cations are

higher than NH4+ ions in K+/ NH4

+ systems.

In this study, we consider nitrogen as the centre of the NH4+ ion for calculation

of the ammonium self-diffusion. The D value for the total NH4+ ion is larger than the

D value of N by several orders of magnitude, as a result of changes in hydrogen

bonding. For example, the nitrogen and total NH4+ self-diffusions in ZM-001 are -

1.1x10-17 m2.sec-1 and 7.2x10-12 m2.sec-1, respectively. Although the NH4+ interaction

is same in both membranes, the ZM-110 membrane retains more NH4+. The small


diameter of channels along [110] direction increases the energy barrier for NH4+ ions

leaving the membrane with water molecules. Therefore, NH4+ ions localise within ZM-

110 membrane channels along a axis with limited or no access to the solution outside

the membrane.

In general, the exchange of monovalent cations and their location to the

framework inside zeolite N membranes depends on their ionic size and, respectively,

decrease and increase with increase in cation size. However, the exchanges of Li+ and

Cs+ do not follow this general principle for the ZM-110 membrane. Li+ and Na+ cations

have lower mobility compared to K extra-framework cations, due to their stronger

electrostatic interactions with the zeolite N framework. The Li+ cations show higher

levels of interaction than Na+ cations due to a higher charge density. Na+ cations show

between two and three electrostatic interactions with framework oxygen atoms, one

van der Waals interaction with water molecules and/or one interaction with chloride

anions. However, Li+ cations show interactions with two water molecules while they

have same number of electrostatic interactions with framework oxygen and chloride

atoms. As a result, Li+ can be more mobile within zeolite N channels compared to Na+.

Rb+ and Cs+ ions seem to follow similar diffusion mechanisms in zeolite N.

These cations prefer to localise at the SI and SII sites, which are the lowest energy

sites within the zeolite N structure. Rb+ and Cs+ ions interact with three to four

framework oxygen atoms and with two to three oxygen atoms of water molecules

and/or one chloride ion. As a result of these many interactions, and their large ionic

size, these ions oscillate at their site positions and show limited diffusivity. In

comparison to K+, Rb+ shows a higher self-diffusion value in the ZM-001 membrane.

The diffusion of Rb+ and Cs+ in the ZM-110 membrane is anisotropic. Rb and Cs in

the ZM-110 membrane can not transport through channels to the external solution due

to their ionic size in comparison to the small pore opening sizes of this membrane

direction. Thus, these larger ions prefer to move through channels along the a axis

inside the ZM-110 membrane. This attribute is evident by the rectangular shape of the

density fields for these cations shown in Figures 6-5d and f.

The behaviour of divalent Mg2+ and Ca2+ cations gives the impression that they

follow a similar exchange and diffusion mechanism as monovalent Li+ and Na+

cations, respectively. As shown in Figures 6-4 and 6-5, the localisation behaviour and

structural arrangements around framework oxygen atoms are similar, especially for


Mg2+ and Li+. However, the exchange and diffusion processes are completely

different. For example, the Mg2+ and Ca2+ mobilities inside zeolite N membranes are

higher than monovalent Li+ and Na+ cations even though they show the same number

of electrostatic interactions with framework oxygen atoms as Li+ and Na+. However,

Mg2+ and Ca2+ interact with more water molecules compared to Li+ and Na+. As a

result, these associated water molecules enhance the mobility of Mg2+ and Ca2+ inside

zeolite channels. Nevertheless, in comparison to extra-framework K+, the diffusion of

Mg2+ and Ca2+, is less facile due to the higher coulombic interaction with framework

oxygen atoms. Thus, for the ZM-001 membrane, these simulations suggest that

divalent Mg2+ and Ca2+ show analogous retention behaviour to monovalent Li+ and

Na+.

6.5 CONCLUSION

The outcomes of this chapter are consistent with the defined objectives of this

study; that is, to describe the exchange mechanism of zeolite N under different

conditions. In this study, we have investigated the ion exchange characteristics of

zeolite N membranes at atomic scale using molecular dynamics simulations. These

membrane models allow exploration of exchange and diffusion mechanisms for

univalent or divalent cations within the three-dimensional porous structure of zeolite

N.

We demonstrate that the exchange and diffusion of cations varies between

channels along the [001] and [110] directions depending on the nature of cation, ion

size and charge. The NH4+ diffusion mechanism is considerably different to the

mechanism(s) for other cations due to the influence of hydrogen bonding. The

diffusion behaviour of guest cations smaller than the extra-framework K+ ion is

isotropic and, in general, follows a similar mechanism in both directions. However,

the diffusion of cations larger than K+ is anisotropic in zeolite N due to different

diameter channels along the [001] and [110] directions.

Taking into account the retention behaviour and exchange mechanisms of the

cations evaluated in these simulations, we suggest that the cation selectivity series for

zeolite N as: NH4+ > Na+ > Li+ > K+ > Ca2+ > Rb+ > Cs+ >Mg2+. However, studying

the penetration of these cations from an electrolyte solution into the membrane can

provide a better estimation of zeolite N cation selectivity.


Furthermore, the results indicate that the structural arrangements of ions and

water molecules inside zeolite N membranes are influenced by the partial charges of

framework atoms. However, the general ion selectivity of zeolite N membranes is

independent of the partial charges of framework atoms.


6.6 REFRENCES

1. Barrer, R.; Hinds, L.; White, E., 299. The hydrothermal chemistry of silicates. Part III. Reactions of analcite and leucite. Journal of the Chemical Society (Resumed) 1953, 1466-1475. 2. Christensen, A. N.; Fjellvåg, H., Crystal structure determination of zeolite N from synchrotron X-ray powder diffraction data. Acta Chemica Scandinavica 1997, 51, 969. 3. Christensen, A. N.; Fjellvag, H., nuetron powder diferaction study of the dehydration of zeolite N. Acta Chemica Scandinavica 1999, 53, 85-89. 4. Mackinnon, I. D. R.; Barr, K.; Miller, E.; Hunter, S.; pinel, T., Nutrient Removal from waste water using high performance materials. Water Science and Technology 2003, 47, 101-107. 5. Thornton, A.; Pearce, P.; Parsons, S. A., Ammonium removal from digested sludge liquors using ion exchange. Water Research 2007, 41 (2), 433-9. 6. Thornton, A.; Pearce, P.; Parsons, S. A., Ammonium removal from solution using ion exchange on to MesoLite, an equilibrium study. J Hazard Mater 2007, 147 (3), 883-9. 7. Zwingmann, N.; Singh, B.; Mackinnon, I. D. R.; Gilkes, R. J., Zeolite from alkali modified kaolin increases NH4+ retention by sandy soil: Column experiments. Applied Clay Science 2009, 46 (1), 7-12. 8. Mackinnon, I.; Millar, G.; Stolz, W. Aluminosilicated of zeolite N structure. 2006. 9. Murthy, V.; Smith, H. D.; Zhang, H.; Smith, S. C., Molecular modeling of hydrotalcite structure intercalated with transition metal oxide anions: CrO4(2-) and VO4(3-). The Journal of Physical Chemistry A 2011, 115 (46), 13673-83. 10. Christensen, A. N.; Fjellvag, H., Crystal structure determination of zeolite N from synchrotron X-ray powder diffraction data. Acta Chemica Scandinavica 1997, 51, 969-973. 11. Perdew, J. P.; Burke, K.; Ernzerhof, M., Generalized gradient approximation made simple. Physical review letters 1996, 77 (18), 3865-3868. 12. Delley, B., Ground-state enthalpies: evaluation of electronic structure approaches with emphasis on the density functional method. The Journal of Physical Chemistry A 2006, 110 (50), 13632-13639. 13. Delley, B., An all-electron numerical method for solving the local density functional for polyatomic molecules. The journal of chemical physics 1990, 92, 508. 14. Delley, B., From molecules to solids with the DMol3 approach. The journal of chemical physics 2000, 113, 7756. 15. Berendsen, H. J.; Postma, J. P.; van Gunsteren, W. F.; Hermans, J., Interaction models for water in relation to protein hydration. In Intermolecular Forces, Springer: 1981; pp 331-342. 16. Ewald, P. P., Ewald summation. Annalen den Physik 1921, 369, 253. 17. Leimkuhler, B.; Noorizadeh, E.; Penrose, O., Comparing the efficiencies of stochastic isothermal molecular dynamics methods. Journal of Statistical Physics 2011, 143 (5), 921-942. 18. Smit, B.; Maesen, T. L. M., Molecular simulations of zeolites: adsorption, diffusion, and shape selectivity. Chemical Reviews 2008, 108, 4125–4184. 19. Barrer, R.; Rees, L. J. T. o. t. F. S., Self-diffusion of alkali metal ions in analcite. 1960, 56, 709-721.


20. Barrer, R.; Bartholomew, R.; Rees, L., Ion exchange in porous crystals part I. Self-and exchange-diffusion of ions in chabazites. Journal of Physics Chemistry Solids 1963, 24 (1), 51-62. 21. Rees, L.; Rao, A., Self-diffusion of various cations in natural mordenite. Transactions of the Faraday Society 1966, 62, 2103-2110. 22. Dyer, A.; White, K. J. J. T. a., Cation diffusion in the natural zeolite clinoptilolite. Journal of Thermochimica Acta 1999, 340, 341-348.


Chapter 7: Detailed Mineralogical Study on Natural Australian Zeolites

Over the last 20 years, Australian natural zeolites have been investigated for use

in various industrial applications. However, there are few, if any, mineral

characterisation studies on Australian natural zeolites since the early 1990s that use

modern techniques. This chapter includes a detailed mineralogical analysis conducted

on zeolite specimens from Avoca and Werris Creek deposits, located in Queensland

and New-South-Wales, respectively, in Australia. An introduction to their

characterisation and applications of these zeolite deposits are presented in section 7.1.

Section 7.2 describes the applied mineralogical characterisation methods, including

thermogravimetry, N2 adsorption/desorption, optical microscopy, XRF, XRD,

SEM/EDS and EPMA/WDS. The physical properties, thermal behaviour and porosity,

as well as mineral compositions for these samples are presented and discussed in

sections 7.3 and 7.4, respectively. Finally, implications from this study and the

potential applications of Australian zeolites are provided in section 7.5.

7.1 INTRODUCTION

Over the past five years, worldwide natural zeolite production is approximately

1,100 k tonne/year with China, Korea, USA, New Zealand, Turkey and Cuba as the

leading producers1. In Australia, early reports on the geology, exploration and

economics of zeolite deposits started in 19582 with mining commencing in 1987 at

Escott, New South Wales3. There are four documented zeolite provinces within

Australia located in: (1) the Tamworth Belt (New England province) of north-eastern

New South Wales4, (2) the Drummond Basin (Drummond zeolite province) in

Queensland5 (3) the Otway Basin in western Victoria6 and (4) the central Australia

lakes7. The first three of these zeolite occurrences are in altered carboniferous

volcanoclastic and pyroclastic rocks and are geologically related to an Andean margin

continental volcanic arc5. The Central Australia occurrence formed in the Tertiary-

Quaternary Period as saline-alkaline lake deposits7. Among these occurrences, the

New England and Drummond support active operating mines with approximately

10,000 tons and 4,000 tons annual production, respectively5.


Increased experience of both Australian producers and consumers about the

variety of applications for this mineral resulted in further geological, mineralogical,

experimental and economic studies. Three studies5, 7, 8 were undertaken in the 1990’s

focusing on eastern Australian deposits. Bulk mineralogical characterisations were

conducted using tools of the time which included X-ray diffraction and optical and

electron microscopy. Natural zeolites from the New South Wales mines at Werris

Creek and Castle Mountain – part of the New England Province – have been

characterized in some detail. These studies have focussed on their potential for a wide

range of applications including: gas adsorption or separation of hydrogen9, 10, helium11

and hydrocarbons12; to remove ammonium from wastewaters13-19, Na from coal seam

gas (CSG) production water20-22 and of heavy metals from acid mine drainage and

wastewater23-27. Other investigations include evaluation of these zeolites for

remediation of other environmental pollutants28-32 and for use in agriculture33-37.

However, there are few studies7, 38 on natural zeolites that occur in the Avoca or

Drummond deposits in Queensland. In this work, we provide a comprehensive

characterisation of natural zeolites from the Avoca deposit in Queensland (QLD) and

natural zeolites from Werris Creek in New South Wales (NSW).

The earlier microprobe analyses on both NSW and QLD zeolites conducted by

Flood and Taylor (1991) 4 and Cooper (1993)7 did not meet criteria for zeolite

compositions defined by Surdam and Sheppard (1978)39 and Birch (1989)40, and, as a

result, the authors considered their analyses as only a general indication of zeolite

compositions. As noted by other authors41 quantitative analysis of zeolite minerals

with electron microbeam methods can be a challenge. For example, sample heating

and electron mobility caused by interactions between the electron beam with the

sample result in dehydration and underestimation of light extra-framework cations and

problems with the determination of framework Si and Al. Subsequent inaccurate Si:Al

ratio calculations can be a significant factor in zeolite group identification42, 43.

However, more recent developments in microprobe technology, for example the

introduction of a field-emission gun (FEG) makes it possible to minimise beam

interaction issues particularly for in situ micron-sized grains44. As noted by Campbell

et al.41 recent improvements in microprobe technology provide an opportunity to

achieve more accurate and reliable analyses of zeolite mineral compositions.


We describe the mineralogy of the fine-grained Australian zeolite deposits based

on multiple electron microprobe analyses using more recent microprobe hardware and

software, including individual grains at spatial resolutions higher than obtained

previously by Flood and Taylor4 and Cooper7. These zeolite compositions are

informed by adaptations to the analytical protocol for quantitative analyses41 briefly

described below and considering nomenclatural changes approved by the International

Mineralogical Association (IMA)45.

7.2 EXPERIMENTAL

7.2.1 Source Materials and Geological Setting

Representative rock samples from two different locations of Australia: Avoca,

QLD and Werris Creek, NSW are selected for mineralogical analysis. The Australian

zeolite deposits are similar in geological age and process of zeolite formation and

related to an ancient early Carboniferous Period and formed during or following

deposition of volcanic glass5, 7. Polished and thin sections of these representative rocks

were prepared for optical and electron microscopy investigations. Representative

samples of each deposit were also comminuted, sieved and sorted. Particles with 1-2

mm diameter were collected for further physical and chemical characterization

experiments while sub-samples were micronized for X-ray diffraction and surface area

analyses. In addition, Avoca samples were crushed to particle sizes that allowed

separation of distinctive mineral assemblages based on the colour of layers: pink layers

(AV-P) and grey layers (AV-G).

7.2.2 Mineralogical characterisations

In order to provide a qualitative and quantitative mineralogical, physical and

chemical characterisation of zeolite samples, a combination of analytical methods

were implemented. Major element analyses of bulk rocks were obtained with a

PANalytical Axios 1 kW Wavelength Dispersive X-ray Fluoresence (XRF)

spectrometer using PANalytical’s WROXI software. For this purpose, 40 mm

diameter fused glass discs containing samples were prepared by the lithium borate flux

method. In this method, 1.15 g of powdered samples and 8.85 g of a commercial

Claisse lithium borate flux (50:50 mix of LiBO4:Li2B4O7 and 0.5% LiI as a non-

wetting agent) are mixed in a 95% Pt – 5% Au crucible and fused at 1070 °C using an

automated Claisse disc preparation furnace. In order to determine the gravimetric


Loss-On-Ignition (LOI), raw, micronized samples were weighed, heated at 1070 °C

for 2 hours, cooled in a desiccator and re-weighed. The LOI value is due to the mass

loss of H2O, CO2, organic and inorganic carbon and other volatile F, Cl, N compounds

from the ignition sample.

Cation exchange capacities (CEC) of zeolite samples were determined using an

NH4+/K+ ion exchange method. Samples were first agitated with 1 mol/L NH4Cl,

centrifuged, washed with ethanol and agitated with 1 mol/L KCl. The ammonium

exchange capacity and cation exchange capacity (CEC) were calculated by measuring

the NH4+ concentration in the supernatant using a steam distillation method by

Kjeldahl46.

A PANalytical X’Pert Pro diffractometer was employed to collect X-ray powder

diffraction (XRD) patterns in Bragg-Brentano geometry operating at 40 kV and 40 mA

with a cobalt source and an iron Kβ filter. Patterns were collected at a step size of

0.016° 2θ from 5° – 90° 2θ for 30 minutes. Corundum (Al2O3, supplied by Baikowski

International) was added as internal standard to micronized samples. Phase

identification was performed using PANalytical Highscore Plus (V4) and MDI Jade

(V4.1) and refined quantitative phase analysis was implemented in TOPAS (V5,

Bruker). The powder diffraction files of the International Centre for Diffraction Data:

(ICDDPDF-4 2013 database) were used for all phase identifications.

Optical microscopy on polished thin sections was performed using a Leica

DM6000 with polarised filter. Micrographs of each section were captured prior to

coating for electron microscopy analysis. Thin sections were coated with 20 nm carbon

for further electron microscopy and microprobe analysis. The morphology,

microstructure and chemical composition of samples were investigated by scanning

electron microscopy (SEM) and energy dispersive spectroscopy (EDS) using a JEOL

7001F electron microscope and an Oxford Instruments SDD XMax 50 mm2 detector.

All EDS analyses in the FE-SEM were obtained at 15 kV accelerating voltage, 10 mm

working distance and 60 s process time. These preliminary analyses using FE-SEM

and EDS allowed identification of specific areas for detailed elemental analysis using

EPMA.

Quantitative elemental analyses were performed using a JEOL JXA 8530F field

emission EPMA equipped with five wavelength-dispersive spectrometers and using

Probe for EPMA software. For these analyses, thin sections were re-polished to


remove electron beam damage from prior exposure, using a series of diamond pads

and cloths to a mirror finish suited to electron microprobe analysis. The EPMA

analysis routine includes the elements Si, Al, Fe, Mg, Ca, Sr, Ba, Na and K, and is

adapted from protocol recommendations by Campbell et al. (Campbell et al. 2016).

This protocol41 was developed for analysis under ambient conditions with a non-FEG

EPMA and includes consideration of instrument conditions, spectrometer

configuration, calibration standards, and order of element detection. Astimex

standards including albite (Na K ), barite (Ba L ), celestite (Sr L ), Cr2O3 (Cr K ),

hematite (Fe K ), and orthoclase (K K ) were used as calibration materials, along

with synthetic MgO (Mg K ), natural corundum (Al K ) and wollastonite (Si K , Ca

K ).

Following this protocol, initial count-rate monitoring experiments were

conducted at 15 kV with a 2 nA and 20 μm defocused beam on all calibration materials

for the assigned element X-ray lines, as well as for all routine elements on a selection

of zeolite unknowns, to provide a baseline for comparison with measurements taken

at 10 kV over a range of beam current (3–10 nA) and beam size (3–10 μm) conditions.

Particular concerns included detection of temporal instability in alkali element X-ray

count rates as a function of beam exposure time, measured concurrently with those of

Si and Al, and optimization of count rates and counting times based on count rate

profiling experiments. Peak counting times are as follows: 10 seconds for Na, Mg, Al,

K; 20 seconds for Ca, Sr, and Si; 30 seconds for Ba and 40 seconds for Fe, with Ca,

K, Al, Na and Fe measured first and concurrently in the analysis cycle.

PEG-EPMA instrument conditions of 10 kV accelerating voltage and 5 nA beam

current with a 5 μm defocused beam, are suitable for quantitative spot analyses in these

samples, along with use of the mean atomic number (MAN) background method and

correction procedure47. The Armstrong-Love/Scott z) matrix correction procedure

(modified from Brown and Bastin48) and mass absorption coefficients of Henke et al. 49 were used. Quality control and data reproducibility are monitored using secondary

standards as well as data reduction protocols specific to zeolites that consider non-

hydrous analytical totals, Si:Al using R values, and charge balance requirements as

outlined in Campbell et al.41.


7.2.3 Physical characterisations

Thermal stability analyses using Thermogravimetric (TG and DTG) and

Differential Scanning Calorimetry (DSC) were performed using a NETCHE STA

449F3 thermal analyser for the temperature range 25-800 °C with a heating rate 10 °C

min-1.

Porosimetry measurements were conducted using a Micromeritics ASAP 2020

instrument on 1 mm and micronized powders of each sample. The N2

adsorption/desorption isotherms were calculated at 77 K. The surface area, pore size

distribution and total pore volume are estimated using MICROMETRICS software

using the BET equation, t-plot and BJH models for adsorption and desorption branches

of the isotherms, respectively. Dehydrating and pre-degassing of the samples took

place under vacuum (10-2 Torr) at temperatures ranging from 25 °C –350 °C for six

days.

7.3 RESULTS

In this work, for Avoca and Werris Creek deposits, zeolite minerals comprise

the bulk of the rock formation which occurs as coherent, compact structures within the

sedimentary horizon.

7.3.1 Mineralogical characterisations

Initial mineralogical identifications were conducted with optical microscopy,

XRF, XRD patterns of bulk samples combined with SEM analysis of thin sections. In

addition, detailed mineralogical determination of zeolites were performed using

electron microscopy techniques, including EDS and WDS.

Optical microscopy

Representative samples from two zeolite deposits are shown in Figure 7-1.

Samples from the Avoca deposit consist of fine-grained, red to pink pyroclastic or

laminated silicic mudstone and siltstone layers interspersed with grey layers of

siliciclastic claystone. Zeolites are concentrated in these pink to red layers of the Avoca

deposit. In the Werris Creek sample, massive red layers of zeolitized vitric air-fall

tuffs, including thin green zeolitic speckles, are interlayered between altered

volcaniclastic sands. The optical petrography of these samples shows a uniform grain

size of minerals in the red and green layers.


Figure 7-1 Representative samples of (a) Avoca, QLD and (b) Werris Creek, NSW deposits.

Figures 7-2a and 7-3a show polarised light microscopy images of (a) Avoca and

(b) Werris Creek samples, respectively. In both samples, zeolite and feldspar minerals

are fine grained and distributed in a cement background containing clay, quartz and

silica-rich minerals. In these images, feldspars are recognisable by their bright white

colour and sharp edges. Zeolites in Avoca and Werris Creek samples are noticeable in

optical microscopy images by their red-brown colour resulting from the presence of

Fe.

Figure 7-2 (a) Polarised light microscope image showing the mineralogical distribution in Avoca thin section, (b) BSE image of pink layer of Avoca sample showing mineral diversity and (c) BSE image

of dark pink layer of Avoca sample showing the mineral diversity in this layer (red circles on the images represent the position of EPMA point analysis)


Figure 7-3 (a) Polarised light microscope image showing the mineralogical distribution in Werris

Creek thin section, (b) BSE image of general matrix of Werris Creek sample showing mineral diversity and (c) BSE image of dark red layer of Werris Creek sample showing the mineral diversity

in this layer (red circles on the images represent the position of EPMA point analysis)

Bulk rock chemical composition

The bulk chemical composition determined by XRF for the two samples from

this study are shown in Table 7-1 and compared with other bulk analyses of samples

from the same or similar deposits in Australia as identified in Table 7-1. The Avoca

and Werris Creek samples show similar calcium contents (3.66% and 3.64% CaO wt%

respectively), while the Werris Creek deposit shows higher sodium and potassium

contents (1.17% Na2O wt% and 1.17% K2O wt%) than Avoca (0.67 and 0.71 wt%).

The Werris Creek sample contains the higher amount of iron (2.13 wt% Fe2O3)

compare to Avoca sample. The bulk Si:Al ratios of Avoca and Werris Creek are 4.9

and 4.7, respectively. The Si:Al ratio of pink layers of Avoca are similar to the bulk

Si:Al ratio of this sample whereas the grey layers show a higher Si/Al ratio ~5.4. This

higher Si:Al ratio confirms that high silica phases are present in Avoca grey layers


The ammonium exchange capacity and cation exchange capacity for both

samples in two different particle sizes are provided in Table 7-2. Due to the increase

of specific surface area in micronized particles, the CECs of these samples are higher

than that for the 1–2 mm size particles. Calculated CEC values for Avoca (120

meq/100g) sample is higher than the Werris Creek (107 meq/100g) sample.

Table 7-1 Cation Exchange Capacity of different particle size of zeolite samples

Sample Particle

size

NH4+

(mg/L)

CEC

(meq/100g)

Avoca, QLD Micronised 108.3 120.0

1-2 mm 102.6 112.9

Werris Creek, NSW Micronised 97.3 107.1

1-2 mm 88.2 98.6

Wang et al. (2012)50 <75 μm 119

Wang and Zhu (2006) )28,Vimonses et al. (2010)30 and Wang and Nguyen (2016)20

Various 120

XRD analysis

Typical XRD patterns of micronized samples from Avoca and Werris Creek are

displayed in Figure 7-4. The Avoca and Werris Creek samples exhibit similar patterns

and common phases such as quartz, feldspars and clays minerals. The principal zeolite

identified in both samples is clinoptilolite accompanied by heulandite and mordenite

in lower proportions. Alkali and plagioclase feldspars are present in both Australian

deposits (Figure 7-4).

Quantitative mineralogical analysis using powder XRD indicates that Avoca and

Werris Creek samples contain, respectively, 58.5% and 58.3% zeolites, 6.2% and 11%

feldspars, 10.3% and 11.6% clays and 25% and 19% Quartz, minor and amorphous

phases as shown in Figure 7-5. Also, the grey layers of the Avoca sample consist of

higher contents of amorphous phases and clays with lower zeolite content as shown in

Figure 7-5.

Micro surface morphology

SEM images on broken surfaces of samples are shown in Figure 7-6 for (a)

Avoca and (b) Werris Creek samples illustrating mixtures of microcrystalline minerals


and amorphous phases. The existence of platy/tabular-shaped heulandite crystals,

fibrous mordenite crystals, bulky feldspar laths and sheets of clay minerals all together

illustrates the heterogenous morphology of these natural zeolitic specimens.

Table 7-2 Bulk chem

ical composition of zeolite sam

ples according to the X

RF analysis, presented as w

t %oxides


Figure 7-4 XRD patterns of representative zeolite samples from Avoca and Werris Creek (Werris

Creek graph superimposed and offset upwards). Corundum was added as internal standard to micronized samples.

Figure 7-5 The proportion of mineral phases in Avoca and Werris Creek samples determined by XRD

quantitative analysis, compared with previous XRD studies by Flood et al.8

Figure 7-6 SEM image of (a) Avoca and (b) Werris Creek samples showing surface morphologies and different types of amorphous and microcrystalline phases, including platy/tabular-shaped heulandite

crystals, fibrous mordenite crystals, bulky feldspar laths and sheets of clay minerals.

(c)


Detailed Mineralogy by EM techniques

The detailed mineralogy of each sample is not evident nor unequivocally

determined from optical microscopy alone due to the fine grain size of the majority of

minerals. Nevertheless, indicative mineral assemblages can be differentiated as mm-

scale layers of alternating colours and, in some cases, gradations of colour tones that

show a well-defined spatial distribution as shown in Figures 7-2a and 7-3a. In this

section, the fine-grained mineralogy of key discernible layers in each sample using

Energy Dispersive X-ray Spectroscopy (EDS) and Wavelength Dispersive

Spectroscopy (WDS) analysis is outlined.

EDS investigations were performed on polished thin sections of Avoca and

Werris Creek samples. Approximate mineral identification was based on the major

elements detected and the Si:Al ratio calculated from EDS spectra. This qualitative

analysis by EDS revealed that the pink layers of Avoca sample included clinoptilolites

and heulandites containing Ca, Mg and Fe, alkali feldspars, plagioclases and quartz.

The grey layers in the Avoca sample are formed from micro-crystals including quartz

and high silica minerals containing K and Na. The back scattered electron (BSE)

images in Figures 7-2b and 7-2c and Figure 1b of Appendix D clearly illustrate the

diversity of minerals and crystal sizes between different layers. The darker pink layers

of the Avoca samples correlate with the presence of larger sized crystals.

EDS analyses of the Werris Creek sample show the presence of clinoptilolite

and heulandite zeolites containing Ca, Mg and Fe, alkali feldspars and plagioclases

(Figure 7-6b and c). The dark red layer in this sample shows higher Fe content. Figure

7-3c is a BSE image of the dark red layer of Werris Creek showing a high proportion

of Fe as light regions; many of which decorate the outside of, or interstices between,

other minerals. In addition, silica minerals containing Ti, Fe and Mn were identified

in this sample as well as minerals containing Ca, P, F and Cl in layers that are green

coloured in optical images (Figure 2 of Appendix D).

Detailed mineralogical and fully quantitative compositional analysis were

conducted using both mapping and point analysis using WDS with the EPMA. In

Figure 1a of Appendix D, different layers in the Avoca sample are visible from left to

right: large zeolite crystals, general matrix including zeolites, feldspars and smectite

minerals (pink layers), and a layer of clays (gray layers). The minerals illustrating

Si:Al ratio= 4-5 and a high amount of Ca and Mg are recognised as zeolitic phases


(Figure 7-7a, b and c). WDS compositional maps clearly show that Ca and Mg are

concentrated in the zeolitic phases of the Avoca sample. However, large zeolite

crystals at the left of Figure 7-7 show high concentrations of Ca while Fe is distributed

in both zeolitic phases and the background matrix of the Avoca sample (Figure 7-7e).

In this sample, higher amounts of Fe are accompanied by an increase in Si:Al ratio.

Feldspar minerals are discernible from other minerals in these images by their Si:Al

ratio ~3, high amounts of K and Na (Figure 7-7d and Figure 1e of appendix D) and

their Fe and Mg values (Figure 7-7c and e). These WDS maps of element wt% reveal

the presence of both plagioclase and alkali feldspars in the Avoca sample.

The WDS compositional map showing the Si:Al ratio (Figure 7-8a) for the

Werris Creek sample illustrates that a high silica cement agglomerate of fine-grained

(< 50 μm) zeolite and feldspar minerals comprise the general matrix in this sample.

Zeolites with average Si:Al ratio 4 are predominantly of the Ca variety containing Mg

(Fig.7-8b and Fig.3e of Appendix D) and with high iron content in larger crystals (see

Figure 7-8c). Similarly, with the Avoca sample, both plagioclases and alkali feldspars

are observed and are distinguished from zeolitic phases by their high Si:Al ratio, K

(Fig.7-8a and d) and Na content (see Figure 3f of appendix D). Figure 7-9a shows a

representative WDS map of Fe wt% distribution in a dark red layer of the Werris Creek

sample. In this layer the Fe wt% increase is due to zeolitised glass shards with

decreased Si concentration (Figure 7-9b).

In addition to compositional mapping analysis, a total of 195 points were

analysed using EPMA on polished thin sections of Avoca and Werris Creek to

quantitatively determine the chemical composition of zeolites and feldspars at specific

locations of the thin sections. EPMA data related to zeolites were refined with the

recommended reduction and quality control protocols by Campbell et al.41. Table 7-3

represents a summary of crystal size, Si:Al ratio, zeolite type, extra-framework cations

and feldspar types obtained from point analyses for different regions of each thin

section identified by the compositional mapping results.


Figure 7-7 WDS map showing stoichiometric proportion of (a) Si:Al and the element wt% for (b) Ca, (c) Mg, (d) K and (e) Fe in Avoca sample.

Figure 7-8 WDS map showing stoichiometric proportion of (a) Si:Al and the element wt% for (b) Ca, (c) Fe and (d) K in Werris Creek sample


Figure 7-9 WDS map image showing the relation of (a) Si element wt% and (b) Fe element wt% in the dark brown layer of Werris Creek sample.

Large crystals in Avoca samples were identified as sodium clinoptiloloite-Ca.

However, the common zeolite type in pink layers of this sample is magnesium

clinoptilolite-Ca. In addition, the presence of strontium in this sample is noticeable,

especially in the pink layers. The common zeolite types in Werris Creek samples were

determined as magnesium heulandite-Ca, but the iron containing dark red layer in this

sample shows both sodium and magnesium heulandite-Ca. The ternary graph in Figure

7-10a shows the diversity of the dominant extra-framework cations of clinoptilolites

in these samples. Moreover, the variety of feldspars in these samples is illustrated by

the ternary graph in Figure 7-10b. Sanidine, anorthoclase, oligoclase, andesine and

bytownite are the feldspars identified in Avoca and Werris Creek samples.


Figure 7-10 Ternary diagrams demonstrating (a) variation in major cation compositions for clinoptilolite and heulandite (green circles represents the Flood and Taylor (1991)4 and (b) feldspar

diversity. Compositions were obtained by EPMA.


Table 7-3 Summary of zeolite cationic compositions as determined using EPMA for different layers of

Avoca and Werris Creek samples

Sample Avoca Werris Creek General pink

layer Dark pink layer Large crystals General layers Dark red

layer Size of zeolite crystals (μm)

10-15 <100 >100 10-100 various

Zeolite type

Magnesium clinoptilolite-Ca

Sodium/ Magnesium clinoptilolite-Ca

Sodium clinoptilolite-Ca

Magnesium/ Sodium heulandite/ clinoptilolite-Ca

Sodium/ Magnesium heulandite-Ca

Si:Al ratio 4.20 4.28 4.25 3.93 3.96 Relative proportion of extra framework cations

Dominant EFW cations

Ca 71.97% 71.26% 82.35% 65.64% 70.92% Na 7.91% 10.17% 11.92% 10.56% 12.99% K 3.12% 2.74% 1.71% 4.82% 4.95%

Subdominant EFW cations

Mg 12.93% 11.98% 0.27% 17.99% 10.06% Ba 1.01% 0.99% 0.82% 0.08% 0% Sr 3.12% 2.85% 2.93% 0.91% 1.08%

Fe2O3 content <0.01% <0.0% <0.04% <0.01% >5%

Feldspars Sanidine (50-60% K), Oligoclase and Andesine

(10-40% Na)

Sanidine (40-50% K), Oligoclase and Andesine (20-40% Na)

7.3.2 Physical characterisation

The following sections describe the thermal behaviour and porous structure of

Avoca and Werris Creek zeolite samples analysed by thermogravimetic analysis and

N2 adsorption/desorption measurements.

Thermogravimetric analysis

In this study, TG/DTG curves were employed to estimate the amount of mass

lost during heating and DSC curves were applied to evaluate the thermodynamic

properties of zeolite. Typical TG/DTG and DSC curves of (a) Avoca and (b) Werris

Creek zeolite samples are provided in Figure 7-11.

Figure 7-11 TG, DTG and DSC curves of (a) Avoca and (b) Werris Creek zeolite samples.


Water molecules in zeolite pores and channels occur in two forms as hygroscopic

water molecules and as hydroxyl groups interacting with extra-framework ions and

framework atoms. Removal of these water molecules from typical zeolite structures

takes place at temperatures up to 500 °C51. During heating of zeolite samples up to 800

°C, physically adsorbed water of Avoca and Werris Creek samples, desorbed at

temperatures up to 69 °C, 84 °C and 82 °C, respectively. With an increase in

temperature, water molecules loosely bound to exchangeable cations are eliminated at

temperature ranges between 69–270 °C and 84–267 °C, respectively, for the Avoca

and Werris Creek samples. At higher temperatures, more strongly bonded and isolated

water molecules are gradually removed from the zeolite samples. The Avoca and

Werris Creek show no structural deformation by heating up to 800 °C.

Mass loss percentages at different temperature ranges are presented in Table 7-

4. Avoca and Werris Creek samples lose the majority of adsorbed water up to 400 °C

(11.32 weight % and10.83 weight % respectively).

Table 7-4 Mass loss of samples at different temperature ranges (all values are in %)

28-100 ºC 100-200 ºC

200-300 ºC

300-400 ºC

400-500 ºC

500-600 ºC

600-700 ºC

700-800 ºC

Total Loss

Avoca 4.51 3.37 1.94 1.50 0.96 0.67 0.33 0.06 13.33

Werris Creek

2.93 3.81 2.30 1.79 0.93 0.51 0.23 0.02 12.53

Porosity

Nitrogen adsorption/desorption isotherms of dehydrated samples are provided in

Figure 7-12. In order to minimise the influence of zeolitic water molecules on porosity

measurements, all samples were degassed for 6 days at 350 °C, on the basis of TG and

DSC analyses.

According to the IUPAC classification52, all isotherms are classified as type IV

with H3 type hysteresis loops known for slit-shaped pores and plate-like particles. The

initial part of the isotherms at lower relative pressures up to 0.01 are related to the

multi-layer adsorption of nitrogen in micropores of zeolites. However, capillary

condensation of nitrogen molecules in available mesopores due to the presence of

impurities such as feldspars, quartz and clays, and the resulting interstitial volume

between crystals is illustrated by the hysteresis loops at higher relative pressures52


(Lykiema et al. 1984). The Avoca isotherm (Figure 7-12a) illustrates a wide and open

hysteresis loop and shows a clear inflection at p/p0=0.46, that indicates the Tensile

Strength Effect (TSE)53.

Figure 7-12 Characteristic N2 adsorption/desorption isotherms of (a) Avoca with 1-2mm particle size and (b) Avoca micronized particles as well as differential pore size distribution of (c) Avoca samples

with 1-2mm particle size and (d) Avoca micronized particle size..

The surface area, pore volume and pore size distribution estimated from N2

adsorption/desorption isotherms54, 55 of zeolite samples are presented in Table 7-5. The

surface area is calculated from the N2 adsorption branch of each isotherm using the

BET equation56. Since the BET equation is more accurate for calculating the area of

meso- and macro-pores, t-plots are applied to estimate the area and volume of

micropores57, 58. Estimated values for surface area and volume of the micropores show

that the Semnan sample contains a larger microporous structure than Australian

samples (respectively 5.1 m2/g and 2.34 mm3/g for Semnan and 3.8 m2/g and 1.68

mm3/g for Avoca). As expected, the micronized sample of Avoca presents a larger

surface area and pore volume compared with the 1-2 mm sized particles.


The average pore width and volume are measured from the desorption branch of

isotherms using the BJH model59. Pore size distributions (PSD) attained from the

desorption branch using the BJH model show an average pore radius of 95.5 Å in the

Avoca sample. However, smaller and larger pores can be observed (Rp=49 and 140 Å)

as shown in Figure 7-12c.

Table 7-5 Parameters obtained from N2 adsorption/desorption isotherms for Avoca and NSW samples

Avoca Werris Creek

1mm micronised 0.5-1 mm19

<75 μm28

<70 μm29

BET Surface Area m²/g 11.8 25.8 13.69 16 8.31 t-Plot Micropore Area: m²/g 3.8 9.9

t-Plot Micropore volume: cm³/g 0.001682 0.004339

BJH Desorption cumulative volume of pores

cm³/g 0.038835 0.091539 0.032 0.039

BJH Desorption average pore width Å 95.5 194.2 174.5

7.4 DISCUSSION

In this study, detailed mineralogical analyses were conducted on zeolite

specimens formed by similar zeolitization processes from different deposits in

Australia. These comparisons provide an indication of the wide variation of zeolite

compositions possible within both clinoptilolite and heulandite structures in natural

environments. In addition, the degree of heterogeneity within each deposit varies

substantially, not only in specific zeolite compositional range(s) but also in accessory

mineral phase assemblages and physical properties. We discuss implications of these

mineral assemblages in zeolitic rocks and compare potential uses of such natural

deposits.

7.4.1 Chemical composition and detailed mineralogy

The major element oxides obtained by XRF show that Australian deposits

include calcic bulk compositions. In addition, the Werris Creek deposit contains a

significant amount of iron, unlike the Avoca deposit. The XRF data in this study are

consistent with previous studies as shown in Table 7-1. A general consideration of

XRF results suggests that Werris Creek samples would show higher CEC values than


Avoca samples. However, the CEC results in Table 7-2 demonstrate higher CEC

values for the Avoca sample in comparison to Werris Creek. The experimentally

calculated CEC results indicate that Werris Creek samples contain minerals that do not

participate in the cation exchange process or, alternatively, that exchangeable cations

are located in inaccessible micropores within the zeolite structure.

Moreover, micronized samples of bulk rock show higher CEC values than those

samples with larger 1-2 mm particle sizes. This result is due to an increase in specific

surface area and pore size distribution caused by an increase in accessibility to more

exchangeable cations in the micronized material. This accessibility of sites is likely

due to the comminution and separation of zeolite and matrix phases in Werris Creek

material caused by micronizing. The CEC of Werris Creek samples in this study are

different from reported in previous studies due to differences in chemical composition

and particle sizes20, 28, 29, 31, 50, 60 or experimental methods7, 50.

Qualitative analyses of XRD patterns from Werris Creek samples in this study

are similar to previous XRD studies on the NSW deposit3, 4, 17. These qualitative

analyses identified predominant phases as clinoptilolite accompanied by quartz,

mordenite, feldspars and clays. In addition, they considered clinoptilolite and

heulandite as a single mineralogical phase. However, quantitative analysis in this study

provides a more detailed phase identification of zeolite proportions in each sample

(Figure 7-5). For example, quantitative XRD analyses show that the zeolitic deposits

of Avoca and Werris Creek contain, respectively, 46.7% and 46.9% clinoptilolite,

8.1% and 8.5% heulandite, 3.8% and 3.4% mordenite as shown in Figure 7-5.

Furthermore, the results of SEM-EDS and EPMA investigations support the mineral

diversity and phase identifications by XRD.

Mineralogical investigations in the 1990s4, 8 showed that in the Werris Creek

deposit both glass shards and fine groundmass are replaced by zeolitization. Using an

electron microprobe, Flood and Taylor (1991)4 indicated that the interior of the large

shard pseudomorphs consists of clinoptilolite-Ca and Fe-rich rims of zeolitized shards

containing mordenite. Their work also showed quartz, albite, orthoclase (sanidine) and

biotite in the background matrix. In another study by Cooper (1993)7, microprobe data

of Queensland zeolite sites showed the presence of altered, zeolitized shards similar to

Werris Creek, but richer in calcium than the background. The composition of QLD

and NSW zeolites as reported in different studies compiled by Cooper 7 and plotted in


a ternary diagram (Ca+Mg-K-Na) show that zeolites from both sites are predominantly

clinoptilolite-Ca (green circles in Figure 7-10a). These mineralogical investigations

conducted by Flood and Taylor4 and Cooper7 did not meet defined criteria by Surdan

and Sheppard61 and Birch 40 for zeolite and clinoptilolite and were considered as only

a general suggestion of mineral compositions.

In this study, we have used a new generation field emission electron microprobe

along with the latest recommended data reduction protocols41 to describe the

mineralogy of the fine-grained zeolite deposits. A combination of point analyses and

compositional mapping across whole rock polished sections provides a reliable and

credible estimate of zeolite compositions at the scale of the individual grains. In

addition, this approach provides more explicit detail on the nature of iron in the Avoca

and Werris Creek deposits.

The chemical diversity of dominant extra-framework cations of zeolites of these

samples is clarified by (Ca+Mg-Na-K) ternary graphs as shown in Figure 7-10a.

Among 106 detected points, excluding feldspars and quartz, 86 points passed all

requirements for classification as a zeolite as defined by Campbell et al.41 that express

the reliability of our analysis. The calculated R values (i.e. Si:(Si+Al)) for heulandite

and clinoptilolite in this study (0.81 for Avoca and 0.80 for Werris Creek) are within

acceptable range for these type zeolites that is between 0.79 to 0.85 based on the

Passaglia and Sheppard study42. Also, the charge balance, E%, values (2.89% for

Avoca and 1.46% for Werris Creek) are between -10 to 10 that are acceptable range

for clinoptilolite and heulandite62. The average chemical composition and the

stoichiometry of Avoca and Werris Creek samples determined by refinement of EPMA

data are presented in Tables 7-6 and 7-7.

Zeolites in the Avoca deposit are predominantly sodium or magnesium

clinoptilolite-Ca, while Werris Creek zeolites are magnesium clinoptilolite-Ca or

magnesium heulandite-Ca. The average Si:Al ratio of Avoca zeolites (4.23) is higher

than Werris Creek (3.92). Werris Creek clinoptilolite and heulandite contain more Mg,

Na and K compared with Avoca zeolites, although the Sr and Ba content is higher in

Avoca samples.


Table 7-6 Average weight % oxide in zeolite composition obtained by EPMA point analysis following data reduction and quality control protocol of Campbell et al (2016) 41.

Avoca Werris Creek Number of points: 51 35

Average (1 SD) Average (1 SD) SiO2 66.5 1.4 65.0 1.7 Al2O3 13.38 0.56 14.10 0.71 Fe2O3 0.02 0.05 0.01 0.04 MgO 0.43 0.39 0.98 0.33 BaO 0.23 0.17 0.02 0.07 SrO 0.41 0.09 0.15 0.09 CaO 5.76 0.61 5.53 0.32 Na2O 0.44 0.15 0.52 0.11 K2O 0.16 0.05 0.35 0.08 Total 87.4 1.9 86.6 2.3 H2O* 12.6 1.9 13.4 2.3 * H2O calculated by difference

Table 7-7 Zeolite Formula (normalized to 72 oxygen atoms) obtained by EPMA point analysis following data reduction and quality control protocol of Campbell et al (2016) 41.

Avoca Werris Creek Number of points: 51 35

Average (1 SD) Average (1 SD) Si 4+ 29.14 0.23 28.69 0.26 Al 3+ 6.91 0.24 7.34 0.25 Fe 3+ 0.01 0.02 0.00 0.01 S T 36.05 0.07 36.02 0.06

Mg 2+ 0.28 0.25 0.64 0.21 Ba 2+ 0.04 0.03 0.00 0.01 Sr 2+ 0.11 0.02 0.04 0.02 Ca 2+ 2.70 0.26 2.62 0.18 Na + 0.37 0.12 0.45 0.11 K + 0.09 0.03 0.20 0.04 S EFW 3.59 0.14 3.94 0.17

R 0.81 0.01 0.80 0.01 E% 2.89 3.93 1.46 3.21 Si/Al 4.23 0.18 3.92 0.17

The combined bulk and individual mineral analyses suggest that Fe is distributed

in zeolitic phases as well as in the matrix of Avoca and Werris Creek samples, although

its form is not known. The average composition obtained from microprobe analyses

with data reduction using the method by Campbell et al.41 (i.e., rejecting zeolites with

Fe2O3 wt% >0.2 wt%) shows a negligible amount of Fe2O3 in the Avoca and Werris

creek zeolites (0.02 and 0.01 wt%, respectively; Table 7-6). According to Campbell et


al.41 Fe may be present in zeolites due to acceptance of both Fe2+ and/or Fe3+ in the

zeolite structure located in the framework, extra-framework, or both, sites or as a

separate iron oxide phase.

In the Avoca and Werris Creek samples, Na and K are mainly distributed in

feldspars and other minerals in the background matrix (see Figures 7-7, 7-8 and

Figures 1 and 3 of appendix D). Both plagioclase and alkali feldspars were identified

by EDS and WDS analysis, including sanidine, anorthoclase, oligoclase, andesine and

bytownite (Figure 7-10b).

7.4.2 Physical properties

Zeolites, including natural forms, may find application(s) at relatively high

temperature in a dehydrated or partially dehydrated state. This condition can occur

during catalysis reactions63, when used for a molecular sieve9 or as part of an energy

storage process64, 65. Thus, thermal behaviour is an important indicator of performance

for such processes. For example, An et al. (2011)9 showed the ideal temperature for

H2 and CO2 permeance into a zeolite membrane is 500°C. They used zeolite samples

from the Castle Mountain deposit which has similar chemical (see Table 7-1) and

mineralogical characteristics to the Werris Creek deposit and is part of the same

depositional setting. The work by An et al.9 confirmed the stability of this zeolite

deposit at high temperatures up to 500°C during molecular sieve experiments.

Thermal behaviour, dehydration and the framework collapse temperature of

zeolites are inherently dependent on chemical and structural factors. On the one hand,

the thermal stability of zeolites rises with increase in the framework Si:Al ratio66. On

the other hand, zeolites with high hydration energy exchangeable cations adsorb more

water molecules and retain them to higher temperatures67. The difference between total

mass loss 13.33% and 12.53%, respectively, for Avoca and Werris Creek samples

(Table 7-4) is evident in XRF and XRD results. For example, the Avoca and Werris

Creek Ca-rich zeolites have high water adsorption capacities and retain water to high

temperatures up to 400 °C. The TG/DTG curve obtained for the Werris Creek sample

is in good agreement with previous reported TG/DTG analyses7, 50. This work shows

a typical TG curve of clinoptilolite68 with a continuous weight loss from 50 °C and

some variations in the DTG curve at 170 °C and 320 °C due to the high Ca content of

clinoptilolite. Moreover, the percentage of water desorbed during heating of a zeolite

can provide an initial estimate of the adsorption capacity for other molecules that may


occupy the resulting vacant volume69, 70. Therefore, TG results suggest high adsorption

capacity for Ca-rich Australian zeolites.

Adsorption of relatively inert gases such as N2 at its boiling temperature of 77 K

over a wide range of relative pressures (p/p0) is a routine technique to characterise the

porous structure53, 58 of zeolites and estimate the adsorption capacity. In this study, the

zeolite samples, as with other natural zeolites, follow an adsorption behaviour typical

of multi-porous materials. The bulk materials in this study behave as microporous

materials, but due to occupation of their pores and channels with large hydrated extra-

framework ions and water molecules, N2 molecules are not able to access most of the

micropores71. Besides, the presence of accompanying minerals such as feldspars and

clays with corresponding voids between aggregated crystals, creates a meso- and

macro-porous system72. As is well known, various physical phenomena, such as

particle size, accompanying minerals and the tensile strength53 can influence

adsorption/desorption isotherms with consequent inaccuracies in estimating the

surface area, pore volume and pore size as is apparent with dissimilar values for Werris

Creek zeolite obtained by different studies (Table 7-5). Porosity estimates in this study

show that Australian Ca-rich zeolites have very condensed porosity structure.

7.5 CONCLUSION

This chapter addresses the final objective of this research to investigate the

mineralogical characteristics and chemical composition of Australian zeolites for

further MD simulations. We demonstrate that samples with particle sizes less than 20

μm can be consistently characterised with higher spatial resolution using an electron

probe with a field emission gun (FEG), a defocused 5 m beam and mean atomic

number background method. With this approach, we reliably estimate the composition

of fine-grained zeolites in the Avoca and Werris Creek deposits. In situ analyses of

zeolites and associated minerals in these altered and laminated volcaniclastic deposits

show subtle differences in assemblage and spatial distribution that reflect provenance

and subsequent diagenesis.

Taking into consideration the distinctive physical and chemical characteristics

of Australian zeolites, they are useful molecular sieve and ion exchange candidates for

various applications. Several factors are important in H2 and CO2 permeability and

hydrogen selectivity of membranes made of natural zeolites, including thermal


stability, porosity structure, the radii and charge of cations occupying the extra-

framework sites9, 10. The thermogravimetric and porosity analyses in this study

demonstrate that Australian natural clinoptilolites possess high thermal stability up to

400 °C and suitable condensed porosity structure that make them potential materials

for hydrogen separation processes.

Studies10 show that clinoptilolites containing Ca as a dominant extra-framework

cation have higher affinity for H2 unlike those containing Na/K, especially at high

temperatures over 300 °C. The EPMA results in this study show that the chemical

composition of zeolites in Avoca deposits are variable from sodium clinoptilolite-Ca

to magnesium clinoptilolite-Ca depending on the sedimented layer. However, the

Werris Creek zeolite deposit contains a more uniform chemical composition of

magnesium heulandite/clinoptilolite-Ca. Moreover, the presence of silica-rich

amorphous cement in the Avoca zeolite deposit is an undesirable feature for molecular

sieve applications. Hence, the presence of both Ca and Mg in the Werris creek deposit,

considering the cation radii and charge, may be a superior candidate for the hydrogen

separation industry.

Previous experimental studies18, 21 show that Australian zeolites, despite their

high ion-exchange capability, need pre-modifications in order to increase their

efficiency for ammonium or Na+ removal from waste waters or coal seam gas

production waters, respectively. This comprehensive mineralogical study provides

researchers better insight into selecting effective modification methods, considering

the application requirements and the diverse distribution of dominant and sub-

dominant extra-framework cations and accompanying minerals in Australian zeolites.

The outcomes of previous chapters on zeolite N ion-exchange behaviour indicate

that molecular modelling studies can enhance our understanding of material properties.

Similarly, the molecular sieve and ion-exchange properties of Australian zeolites can

be investigated by combining experiment and simulation. The key step in this process

involves determining the precise chemical composition of two natural Australian

zeolites by EPMA analysis. In the future, these compositions can be employed to

develop viable models to simulate the properties of these zeolites and to predict their

behaviour under real conditions for specific industrial applications.


7.6 REFRENCES

1. Nakhli, S. A. A.; Delkash, M.; Bakhshayesh, B. E.; Kazemian, H., Application of zeolites for sustainable agriculture: a review on water and nutrient retention. Water, Air, & Soil Pollution 2017, 228 (12). 2. Coombs, D. S., zeolitised tuffs from the Kuttung glacial beds near Seaham,NSW. Australian Journal of Science 1958, 21, 18-19. 3. Fredrickson, P. W. Charachterisation of natural zeolites-An information package for exploration and development in New South Wales; Geological Survey of New South Wales: 1987; pp 51-62. 4. Flood, P.; Taylor, J., Mineralogy and geochemistry of late carboniferous zeolites, near Werris Creek, New South Wales, Australia. Neues Jahrbuch fur Mineralogie-Monashefte 1991, (2), 49-62. 5. Flood, P. G. Sedimentary zolite deposits in Australia, International Zeolite Association. www.iza-online.org/natural/Catalog/Australia. 6. Duddy, I. R. The geology, petrology and geochemistry of the Otway formation volcanogenic sediments. University of Melbourne, Melbourne, 1983. 7. Cooper, S. A. Geology, development, & economics of zeolite mining in Australia. University of Tasmania, 1993. 8. Flood, P., Australian occurrences of Ca-type clinoptilolite. Natural Zeolites 1995, 93, 15-21. 9. An, W.; Swenson, P.; Wu, L.; Waller, T.; Ku, A.; Kuznicki, S. M., Selective separation of hydrogen from C1/C2 hydrocarbons and CO2 through dense natural zeolite membranes. Journal of Membrane Science 2011, 369 (1–2), 414-419. 10. An, W.; Swenson, P.; Gupta, A.; Wu, L.; Kuznicki, T. M.; Kuznicki, S. M., Improvement of H2/CO2 selectivity of the natural clinoptilolite membranes by cation exchange modification. Journal of Membrane Science 2013, 433, 25-31. 11. Arami-Niya, A.; Rufford, T. E.; Birkett, G.; Zhu, Z., Gravimetric adsorption measurements of helium on natural clinoptilolite and synthetic molecular sieves at pressures up to 3500 kPa. Microporous and Mesoporous Materials 2017, 244, 218-225. 12. Ma, J.; Stevens, G. W.; Mumford, K. A., The performance of diphenyl dichlorosilane coated ammonium exchange zeolite and its application in the combination of adsorption and biodegradation of hydrocarbon contaminated ground water. Chemical Engineering Journal 2018, 347, 415-423. 13. Booker, N. A.; Cooney, E. L.; Priestley, A. J., Ammonia removal from sewage using natural Australian zeolite. Water Science and Technology 1996, 34 (9), 17-24. 14. Komarowski, S.; Yu, Q., Ammonium Ion removal from wastewater using australian natural zeolite: batch equilibrium and kinetic studies. Environmental Technology 1997, 18 (11), 1085-1097. 15. Cooney, E. L.; Booker, N. A.; Shallcross, D. C.; Stevens, G. W., Ammonia removal from wastewaters using natural Australian zeolite. I. Characterization of the zeolite. Separation Science and Technology 1999, 34 (12), 2307-2327. 16. Cooney, E. L.; Booker, N. A.; Shallcross, D. C.; Stevens, G. W., Ammonia removal from wastewaters using natural Australian zeolite. II. pilot-scale study using continuous packed column process. Separation Science and Technology 1999, 34 (14), 2741-2760. 17. Widiastuti, N.; Wu, H.; Ang, H. M.; Zhang, D., Removal of ammonium from greywater using natural zeolite. Desalination 2011, 277 (1–3), 15-23.


18. Millar, G. J.; Winnett, A.; Thompson, T.; Couperthwaite, S. J., Equilibrium studies of ammonium exchange with Australian natural zeolites. Journal of Water Process Engineering 2016, 9, 47-57. 19. Wijesinghe, D. T.; Dassanayake, K. B.; Sommer, S. G.; Jayasinghe, G. Y.; P, J. S.; Chen, D., Ammonium removal from high-strength aqueous solutions by Australian zeolite. Journal of IEnvironmental Science and Health, Part A: Topxic/Hazardous Substances and Environmental Engineering 2016, 51 (8), 614-25. 20. Wang, X.; Nguyen, A. V., Characterisation of electrokinetic properties of clinoptilolite before and after activation by sulphuric acid for treating CSG water. Microporous and Mesoporous Materials 2016, 220, 175-182. 21. Millar, G. J.; Couperthwaite, S. J.; Alyuz, K., Behaviour of natural zeolites used for the treatment of simulated and actual coal seam gas water. Journal of Environmental Chemical Engineering 2016, 4 (2), 1918-1928. 22. Santiago, O.; Walsh, K.; Kele, B.; Gardner, E.; Chapman, J., Novel pre-treatment of zeolite materials for the removal of sodium ions: potential materials for coal seam gas co-produced wastewater. Springer Plus 2016, 5, 571. 23. Woinarski, A. Z.; Snape, I.; Stevens, G. W.; Stark, S. C., The effects of cold temperature on copper ion exchange by natural zeolite for use in a permeable reactive barrier in Antarctica. Cold Regions Science and Technology 2003, 37 (2), 159-168. 24. Woinarski, A. Z.; Stevens, G. W.; Snape, I., A Natural Zeolite Permeable Reactive Barrier to Treat Heavy-Metal Contaminated Waters in Antarctica. Process Safety and Environmental Protection 2006, 84 (2), 109-116. 25. Nguyen, T. C.; Loganathan, P.; Nguyen, T. V.; Vigneswaran, S.; Kandasamy, J.; Naidu, R., Simultaneous adsorption of Cd, Cr, Cu, Pb, and Zn by an iron-coated Australian zeolite in batch and fixed-bed column studies. Chemical Engineering Journal 2015, 270, 393-404. 26. Sounthararajah, D. P.; Loganathan, P.; Kandasamy, J.; Vigneswaran, S., Removing heavy metals using permeable pavement system with a titanate nano-fibrous adsorbent column as a post treatment. Chemosphere 2017, 168, 467-473. 27. Ryu, S.; Naidu, G.; Hasan Johir, M. A.; Choi, Y.; Jeong, S.; Vigneswaran, S., Acid mine drainage treatment by integrated submerged membrane distillation-sorption system. Chemosphere 2019, 218, 955-965. 28. Wang, S.; Zhu, Z. H., Characterisation and environmental application of an Australian natural zeolite for basic dye removal from aqueous solution. Journal of Hazardous Materials 2006, 136 (3), 946-952. 29. Vimonses, V.; Jin, B.; Chow, C. W.; Saint, C., Enhancing removal efficiency of anionic dye by combination and calcination of clay materials and calcium hydroxide. Journal of Hazardous Materials 2009, 171 (1-3), 941-7. 30. Vimonses, V.; Jin, B.; Chow, C. W.; Saint, C., An adsorption-photocatalysis hybrid process using multi-functional-nanoporous materials for wastewater reclamation. Water Research 2010, 44 (18), 5385-97. 31. Guan, H.; Bestland, E.; Zhu, C.; Zhu, H.; Albertsdottir, D.; Hutson, J.; Simmons, C. T.; Ginic-Markovic, M.; Tao, X.; Ellis, A. V., Variation in performance of surfactant loading and resulting nitrate removal among four selected natural zeolites. Journal of Hazardous Materials 2010, 183 (1-3), 616-21. 32. Li, Y. L.; McCarthy, D. T.; Deletic, A., Stable copper-zeolite filter media for bacteria removal in stormwater. Journal of Hazardous Materials 2014, 273, 222-30. 33. Stelting, S.; Burns, R. G.; Sunna, A.; Visnovsky, G.; Bunt, C. R., Immobilization of Pseudomonas sp. strain ADP: A stable inoculant for the bioremediation of atrazine. Applied Clay Science 2012, 64, 90-93.


34. Wijesinghe, D. T. N.; Dassanayake, K. B.; Scales, P.; Chen, D., Developing an anaerobic digester with external Zeolite filled column for enhancing methane production from swine manure - A feasibility study. Journal of Environmental Science and Health, Part B: Pesticides, Food Contaminants, and Agricultural Wastes 2018, 53 (11), 751-760. 35. Wijesinghe, D. T. N.; Dassanayake, K. B.; Scales, P.; Sommer, S. G.; Chen, D., Removal of excess nutrients by Australian zeolite during anaerobic digestion of swine manure. Journal of Environmental Science and Health, Part A: Toxic/Hazardous Substances and Environmental Engineering 2018, 53 (4), 362-372. 36. Wijesinghe, D. T. N.; Dassanayake, K. B.; Scales, P. J.; Sommer, S. G.; Chen, D., Effect of Australian zeolite on methane production and ammonium removal during anaerobic digestion of swine manure. Journal of Environmental Chemical Engineering 2018, 6 (1), 1233-1241. 37. Wijesinghe, D. T. N.; Dassanayake, K. B.; Sommer, S. G.; Scales, P.; Chen, D., Biogas Improvement by Adding Australian Zeolite During the Anaerobic Digestion of C:N Ratio Adjusted Swine Manure. Waste and Biomass Valorization 2018, 10 (7), 1883-1887. 38. Pickering, H. W.; Menzies, N. W.; Hunter, M. N., Zeolite/rock phosphate—a novel slow release phosphorus fertiliser for potted plant production. Scientia Horticulturae 2002, 94 (3), 333-343. 39. Surdam, R. C.; Sheppard, R. A., Zeolites in saline, alkaline-lake deposits. In Natural zeolites: Occurrence, properties, use, United States, 1978; pp 145-174. 40. Birch, W., Chemistry of vvctorian zeolites. Zeolites of Victoria 1989, 2, 91-102. 41. Campbell, L. S.; Charnock, J.; Dyer, A.; Hiller, S.; Chenery, S.; Stoppa, F.; Henderson, C. M. B.; Walcottt, R.; Rumsey, M., Determination of zeolite-group mineral compositions by electron probe microanalysis. Mineralogical Magazine 2016, 80(5), 781-807. 42. Passaglia, E.; Sheppard, R. A., The crystal chemistry of zeolites. Reviews in Mineralogy and Geochemistry 2001, 45 (1), 69-116. 43. Neuhoff, P. S.; Ruhl, L. S., Mechanisms and geochemical significance of Si–Al substitution in zeolite solid solutions. Chemical Geology 2006, 225 (3-4), 373-387. 44. Merlet, C.; Llovet, X. In Uncertainty and capability of quantitative EPMA at low voltage–A review, IOP Conference Series: Materials Science and Engineering, IOP Publishing: 2012; p 012016. 45. Coombs, D. S., Alberti, A., Armbruster, T., Artioli, G., Colella, C., Galli, E., Grice, J., Liebau, F., 600 Mandarino, J.A., Minato, H.,, Recommended numenclature for zeolite 601 minerals. In International Mineralogical Association, Melbourne, Australia, 2018; pp 1571-1606. 46. Kjeldahl, C., A new method for the determination of nitrogen in organic matter. Z Analitical Chemistry 1883, 22, 366. 47. Donovan, J. J.; Tingle, T. N. J. M.; Microanalysis, An improved mean atomic number background correction for quantitative microanalysis. Microscopy and Microanalysis 1996, 2 (1), 1-7. 48. Armstrong, J. J. M. a., Quantitative analysis of silicate and oxide materials: comparison of monte carlo, ZAF, and ψ (ρz) procedures. Microbeam Analysis 1988, 239-246. 49. Henke, B.; Lee, P.; Tanaka, T.; Shimabukuro, R.; Fujikawa, B., Low-energy X-ray interaction coefficients: Photoabsorption, scattering, and reflection: E= 100–2000 eV Z= 1–94. Atomic Data and Nuclear Data Tables 1982, 27 (1), 1-144.


50. Wang, X.; Ozdemir, O.; Hampton, M. A.; Nguyen, A. V.; Do, D. D., The effect of zeolite treatment by acids on sodium adsorption ratio of coal seam gas water. Water Research 2012, 46 (16), 5247-54. 51. Korkuna, O.; Leboda, R.; Skubiszewska-Zie¸ba, J.; Vrublevs’ka, T.; Gun’ko, V. M.; Ryczkowski, J., Structural and physicochemical properties of natural zeolites: clinoptilolite and mordenite. Microporous and Mesoporous Materials 2006, 87 (3), 243-254. 52. Likiema, J.; Sing, K. S. W.; Haber, J.; Kerker, M.; Wolfram, E.; Block, J. H.; Churaev, N. V.; Everett, D. H.; Hansen, R. S.; Haul, R. A. W., Prepared for publication by the Subcommittee on Reporting Gas Adsorption Data. Journal of Rouquerol 1984, 17. 53. Groen, J. C.; Peffer, L. A. A.; Pérez-Ramı́rez, J., Pore size determination in modified micro- and mesoporous materials. Pitfalls and limitations in gas adsorption data analysis. Microporous and Mesoporous Materials 2003, 60 (1-3), 1-17. 54. Lowell, S.; Shields, J. E., Powder surface area and porosity. Springer Science & Business Media: 2013; Vol. 2. 55. Sing, K. J. C.; Physicochemical, S. A.; Aspects, E., The use of nitrogen adsorption for the characterisation of porous materials. Physicochemical and Engioneering Aspects 2001, 187, 3-9. 56. Brunauer, S.; Emmett, P. H.; Teller, E., Adsorption of gases in multimolecular layers. Journal of the American chemical society 1938, 60 (2), 309-319. 57. Lippens, B.; Linsen, B.; De Boer, J., Studies on pore systems in catalysts I. The adsorption of nitrogen; apparatus and calculation. Journal of Catalysis 1964, 3 (1), 32-37. 58. Jacobs, P.; Flanigen, E. M.; Jansen, J.; van Bekkum, H., Introduction to zeolite science and practice. Elsevier: 2001; Vol. 137. 59. Barrett, E. P.; Joyner, L. G.; Halenda, P. P., The determination of pore volume and area distributions in porous substances. I. Computations from nitrogen isotherms. Journal of the American chemical society 1951, 73 (1), 373-380. 60. Vimonses, V.; Jin, B.; Chow, C. W. K.; Saint, C., Development of a pilot fluidised bed reactor system with a formulated clay–lime mixture for continuous removal of chemical pollutants from wastewater. Chemical Engineering Journal 2010, 158 (3), 535-541. 61. Surdam, R. C.; Sheppard, R. A., Zeolites in saline, alkaline-lake deposits. In Natural zeolites: Occurrence, properties, use, United States, 1978; Vol. 145, pp 145-174. 62. Passaglia, E., The crystal chemistry of chabazites. American Mineralogist: Journal of Earth and Planetray Materials 1970, 55 (7-8), 1278-1301. 63. Ghasemian, N.; Falamaki, C.; Kalbasi, M.; Khosravi, M., Enhancement of the catalytic performance of H-clinoptilolite in propane–SCR–NOx process through controlled dealumination. Chemical Engineering Journal 2014, 252, 112-119. 64. Fujii, S.; Horie, N.; Nakaibayashi, K.; Kanematsu, Y.; Kikuchi, Y.; Nakagaki, T., Design of zeolite boiler in thermochemical energy storage and transport system utilizing unused heat from sugar mill. Journal of Applied Energy 2019, 238, 561-571. 65. Kim, S. T.; Kurahashi, C.; Hoshino, H.; Takahashi, C.; Tamura, Y.; Takasu, H.; Saito, S.; Kurihara, M.; Kato, Y., Thermal driving demonstration of Li4SiO4/CO2/zeolite thermochemical energy storage system for efficient high-temperature heat utilizations. ISIJ International 2019, ISIJINT-2018-428.


66. Cruciani, G., Zeolites upon heating: Factors governing their thermal stability and structural changes. Journal of Physics and Chemistry of Solids 2006, 67 (9-10), 1973-1994. 67. Alver, B. E.; Sakizci, M., Influence of acid treatment on structure of clinoptilolite tuff and its adsorption of methane. Journal of Thermal Analysis and cCalorimetry 2015, 21 (5), 391-399. 68. Gottardi, G.; Galli, E., Natural Zeolites. 1985. 69. Zhao, D.; Cleare, K.; Oliver, C.; Ingram, C.; Cook, D.; Szostak, R.; Kevan, L., Characteristics of the synthetic heulandite-clinoptilolite family of zeolites. Microporous and Mesoporous Materials 1998, 21 (4-6), 371-379. 70. Ostrooumov, M.; Cappelletti, P.; de’Gennaro, R., Mineralogical study of zeolite from New Mexican deposits (Cuitzeo area, Michoacan, Mexico). Applied Clay Science 2012, 55, 27-35. 71. Tsitsishvili, G., Natural zeolites. Ellis Horwood: 1992. 72. Kowalczyk, P.; Sprynskyy, M.; Terzyk, A. P.; Lebedynets, M.; Namiesnik, J.; Buszewski, B., Porous structure of natural and modified clinoptilolites. Journal of Colloid Interface Science 2006, 297 (1), 77-85.


Chapter 8: Conclusions

Advances in computational chemistry techniques in the field of zeolites makes

them effective tools to model the properties of zeolites. These valuable techniques

allow explanation of chemical and physical mechanisms at molecular and atomic

scales that many experiments may not be able to describe. Experiments reveal the high

ion-exchange capability of zeolite N by studying the exchange isotherms at macro

scale. In this study, a combination of computational calculations are applied to study

the structure and ion-exchange property of zeolite N at atomic scale. The outcomes of

this research are summarised as below:

8.1 SUMMARY

The outcomes of DFT calculations on the zeolite N structure demonstrate that

the calculated values for optimized partial charges of zeolite framework atoms

noticeably depend on the choice of DFT calculation method. The calculated Mulliken

partial charges are influenced by quality of convergence and the choice of numerical

basis set file version. The Si-O and Al-O bond lengths of zeolite N optimised with the

developed DNP basis sets (4.4) and TS dispersion correction scheme are closer to the

experimental structure for zeolite N. In addition, the location of framework O5 and O3

atoms (as shown in Fig. 5-2) is particularly sensitive to the nature of DFT

optimisations. Moreover, the Mulliken partial charges of framework atoms are affected

by the available crystallographic T-sites in a zeolite framework. The outcomes reveal

that the optimised structure and corresponding partial atomic charges on the

framework atoms in zeolite N obtained from the GGA-PBE functional with the DNP-

4.4 basis set are more consistent with our knowledge from experimental data and other

theoretical studies on zeolites.

MD simulations on potassic zeolite N demonstrates that the models and

methodologies developed and applied in this study provide results on hydration and

ion-exchange behaviour of zeolite N in agreement with available experiments. The

simulation outcomes of the hydrated state and the hydration energy for zeolite N

indicate that zeolite N adsorbs and retains eight H2O/cage at equilibrium and this value

agrees with the experimentally obtained values. Moreover, MD calculations simulate,


in general, experimental outcomes for NH4+, Na+, Ca2+ and Mg2+ exchange with K+.

Simulations also show potential high capacity for K+ to exchange with Li+ and partially

with Rb+ and Cs+.

Moreover, the ion-exchange MD simulations of zeolite N membranes with

different partial charge on framework atoms indicate that the general ion selectivity of

zeolite N is independent of the partial charges of framework atoms. However, the

structural arrangements of ions and water molecules inside zeolite N membranes are

influenced by the partial charges of framework atoms.

From simulation outcomes, it can be concluded that different parameters affect

the ion exchange behaviour of zeolite N, including the nature, size and ionic charge of

exchanging cations, as well as the direction of channels governing the flux of water

molecules and exchange process to the outside aqueous solution.

Consistent with experimental data, simulations show that zeolite N prefers K+

exchange with NH4+ ions with a high exchange ratio (~70%) as compared to other

mono- and divalent cations. Compared with other cations, hydrogen bonding causes a

considerably different diffusion mechanism for NH4+ that facilitates its exchange

process inside zeolite N channels in the presence of H2O. In general, for other

monovalent cations (Li+, Na+, K+, Rb+ and Cs+), the exchange and interaction with

framework zeolite N decrease with increase in ionic size. However, the localisation

behaviour of divalent cations Mg2+ and Ca2+ are similar to monovalent Li+ and Na+

cations, respectively. These ions follow completely different exchange and diffusion

mechanisms due to a higher charge density of these cations.

Furthermore, the exchange and diffusion mechanisms inside zeolite N channels

along the [001] and [110] directions depend on the ionic size. Cations smaller than the

extra-framework K+ ion show an isotropic diffusion. However, the diffusion of cations

larger than K+ is anisotropic in zeolite N channels with different diameter and

direction.

Taking into consideration the retention behaviour, diffusion and exchange

mechanisms of the cations evaluated by these simulations, the cation selectivity for

zeolite N is predicted as below series:

NH4+ > Na+ > Li+ > K+ > Ca2+ > Rb+ > Cs+ >Mg2+.


Mineralogical characterisation of Australian zeolites demonstrates that

improved spatial resolution for quantitative analysis of zeolite particles <<20 μm can

be achieved using field-emission electron probe microanalysis (FE-EPMA) under

special conditions (10kV accelerating voltage, 5nA beam current, 5 μm defocused

beam) and using the mean atomic number background method. This approach allows

estimation of the composition of fine-grained zeolites in the Avoca and Werris Creek

deposits with accuracy and spatial precision. Australian zeolites are identified as

magnesium clinoptilolite/heulandite-Ca. The distinctive physical and chemical

characteristics of Australian zeolites make them candidates for various applications

including hydrogen separation processing and water treatment (under special

conditions). This study provides a precise understanding of two Australian natural

zeolites’ chemical composition, Si/Al ratio as well as type and amount of extra-

framework cations. These data will be of value in the development of a reliable model

for simulating their gas separation and ion-exchange properties.

8.2 CONCLUSION

This research investigated the ion-exchange properties of synthetic zeolite N.

For the first time, the exchange behaviour of zeolite N has been explored by

computational modelling. Different types of modelling approaches were employed to

qualify and quantify the exchange mechanism of this material. This study provides a

reasonable and self-consistent atomistic description of chemical and physical

interactions at a solid-liquid interface and, in principle, may be extended to phases

other than zeolite and water.

The new methodology applied in this study can be a guide for other researchers

studying the exchange behaviour of all materials with ion-exchange properties. These

simulations, in combination with precise chemical and physical data obtained from

experiment, can be effective predictors of behaviour that allows the design of new

materials with desired properties for a specific application.

8.3 LIMITATIONS

Although this research study delivers a comprehensive view of the ion exchange

mechanism in zeolite N at atomic scale, some limitations have been recognized.


Simulation conditions: In ion exchange simulations reported here, the various

cations and chloride anions were located inside zeolite N membranes and outside

in the aqueous solution. This approach makes a strong chemical potential that

reduces computation cost. However, this chemical condition is unrealistic and is

not possible in experiments. Moreover, in this study, the diffusion and exchange

of cations have been studied during their retention inside zeolite N membranes.

Therefore, they are not the real diffusion rates for exchange of ions. The effective

pore size of zeolite N channels and the largest cage diameter are 3.6 Å and 6.3

Å, respectively. Therefore, some ions are captured in the internal cages within

zeolite N membrane and can not diffuse through the zeolite channels to the

outside solution on the time scale of the simulation. Hence, the simulation results

have a degree of uncertainty especially for large cations Rb+ and Cs+ and divalent

cations Mg2+ and Ca2+.

Simulation scale: Since in this study, simulations are conducted on nano-time

and length scales, the ion exchange and self-diffusion coefficients may not be

representative of macro scale experiments and need to be validated by alternative

methods, especially, for exchange of slow-diffused Mg2+ and Ca2+ cations.

However, nano-scale simulations allow a qualitative comparison for zeolite N

ion selectivity at macro scale.

Experimental Deficiency: The simulation outcomes in this study have been

evaluated with experimental data where available. There are few experimental

studies on ion-exchange behaviour of zeolite N mostly focused on exchange of

NH4+, Na+ and K+ and rarely of Mg2+ and Ca2+. However, there are to date no

published experimental data on exchange of Li+, Rb+ and Cs+ that allows

validation for these cations. Therefore, the ion selectivity series for zeolite N

should be considered a prediction of behaviour.

8.4 FUTURE RECOMMENDATIONS

In this study, a comprehensive understanding of zeolite N structure and ion-

exchange behaviour has been achieved by conducting a combination of DFT and MD

calculations as well as comparison with experiments. However, in order to overcome

the above limitations, additional research is suggested including:


(a) Investigate the ion-exchange behaviour of zeolite N by developing

simulation models constructed from zeolite N membranes in an electrolyte

solution environment containing both cations and anions. This approach

simulates realistic experimental conditions.

(b) Study the diffusion and exchange of ions during their penetration from an

electrolyte solution into the zeolite N membrane. This methodology can

provide a better estimation of zeolite N cation selectivity. However, longer

simulation times are required to accomplish a complete ion-exchange

process. Advances in computation facilities makes it possible to conduct

multi-scale simulations and to study the atomistic behaviour of large-scale

systems for longer times.

(c) Simulate the effect of concentration of cations, contact time, temperature, pH

and competitive ions on ion-exchange behaviour of zeolite N.

(d) Perform comprehensive and corresponding ion-exchange experiments for

zeolite N in order to evaluate the predicted cation selectivity series obtained

by simulations.

(e) Simulate the ion-exchange behaviour of Australian clinoptilolite using the

applied simulation methodology and defined chemical formula obtained in

this study.

(f) Utilise the modelling approach described in this study to gain important

insight into the behaviour of ion exchange for other exchangeable materials

and zeolites, such as zeolite A and zeolite Y.

Appendices 161

Appendices

Appendix A

Modelling Hydration behaviour of zeolite N

To study the zeolite structural behaviour and mobility of water molecules,

potassium and chlorine ions during hydration, MD simulations are used to calculate

the radial distribution function (RDF), concentration profiles, mean square

displacement (MSD) and self-diffusion coefficients of each component under similar

conditions for the zeolite membrane in water.

Results and Discussion

In order to characterise the structural arrangements of water molecules in the

membrane system, the RDF of oxygen atoms of water molecules (Ow) around the Si

and Al atoms of zeolite N framework and the RDF of hydrogen atoms of water

molecules (Hw) around O atoms of zeolite N framework were analysed. A summary

of these results is shown in Figure 1. The diagrams in Fig.1 reveal that, generally, Hw

atoms are at a closer distance to the framework than Ow atoms. For a lower amount of

added water molecules (Fig.1a and b), water molecules have stronger interactions with

the zeolite N framework (both Ow and Hw) because water molecules are mostly located

inside the membrane. At the equilibrium point (Fig.1c and d), the RDF of Ow around

the Si and Al atoms illustrate peaks at approximately equal distances (Fig. 1c and d),

which suggests that water molecules are well arranged in the membrane. However,

with higher amounts of water added to the system, interactions between water

molecules and the framework decrease due to overloading of the membrane with water

molecules (Fig. 1e and f).

162 Appendices

Figure1: Radial distribution function of Ow (O atoms of H2O molecules) around Al and Si atoms of zeolite N framework and Hw (H atoms of H2O molecules) around O atoms of zeolite N framework

(W: Water molecules)

The density variations of water molecules and other ions in the membrane

system give an understanding of the hydration process. Figure 2 represents the

concentration profiles of H2O, K+ and Cl- in the membrane for different hydration

levels at equilibrium. To estimate the relative concentrations, the slab was divided into

small boxes with 1Å width along the c axis and the density per Å3 was calculated.

In Figure 2a, Cl- ions show four peaks at 13.58, 19.62, 26.66 and 32.7 Å that

relate to their structural locations at the center of the zeolite N cages. In addition, K+

ions show eight peaks around these Cl- ions related to their structural positions inside

the zeolite N channel and cages that are consistent with X-ray studies1. Figure 2 also

shows that water molecules are located within the zeolite N cages. By increasing the

0

1

2

3

4

0 2 4 6 8 10

g(r)

r (Å)

a- 64 W Hw-OOw-AlOw-Si

0

1

2

3

4

0 2 4 6 8 10

g(r)

r (Å)

b- 128 W Hw-OOw-AlOw-Si

0

1

2

3

4

0 2 4 6 8 10

g(r)

r (Å)

c- 160 W Hw-OOw-AlOw-Si

0

1

2

3

4

0 2 4 6 8 10

g(r)

r (Å)

d- 192 W Hw-OOw-AlOw-Si

0

2

4

0 2 4 6 8 10

g(r)

r (Å)

e- 256 W Hw-OOw-AlOw-Si

0

2

4

0 2 4 6 8 10

g(r)

r (Å)

f- 318 W Hw-OOw-AlOw-Si

Appendices 163

amount of water, the intensity of K+ and Cl- ion peaks decreased. This response shows

that these ions left their structural locations and move around to different sites. With

increasing amounts of water, the zeolite membrane needs to absorb more water

molecules. In order to find extra space in the structure, the water molecules appear to

push out the K+ and especially Cl- ions out of their structural positions in the membrane

into outside of the membrane. This observation is clear by inspection of the water

loaded profiles (at 318, 352 and 384 water molecules, respectively) in Figures 2f, g

and h.

The self-diffusion coefficient (D) of water molecules, K+ and Cl- ions inside and

outside the membrane were calculated by computing the mean squared displacement

(MSD) of each component (Eq. 1) for the hydrated zeolite N membrane systems during

2ns MD simulation (Table 1).

Equation 1

Table1: Self diffusion coefficient values of water molecules, K+ and Cl- ions

Total number of added

water molecules

Self-diffusion coefficient (Å2/sec) Inside framework Outside framework

K+ Cl- water K+ Cl- water 0 6.67E-07 1.67E-07 64 5.00E-06 5.00E-07 1.67E-04

128 3.33E-05 1.67E-06 1.33E-04 1.91E-02 160 8.33E-05 1.67E-05 5.00E-05 3.17E-04 1.19E-02 192 3.33E-05 1.50E-05 1.00E-04 5.12E-02 9.61E-02 256 1.00E-04 1.00E-04 3.33E-04 2.03E-03 3.26E-02 1.85E-01 318 4.17E-04 2.00E-04 4.67E-04 4.47E-02 1.12E-01 2.52E-01 352 6.67E-05 6.67E-05 1.67E-04 5.66E-02 1.07E+03 2.41E-01 384 2.00E-04 2.17E-04 4.33E-04 4.41E-02 6.67E-02 1.65E-01

164 Appendices

1) Figure2: Concentration profiles of hydrated zeolite N membrane for different amounts of water

loading (W: Water molecules)

0

5

10

15

0 5 10 15 20 25 30 35 40

Conc

entr

atio

n (io

ns/Å

3 )

Z distance (Å)

a) 64 W WatKCl

0

5

10

15

0 5 10 15 20 25 30 35 40

Conc

entr

atio

n (io

ns/Å

3 )

Z distance (Å)

e) 256 W WatKCl

0

5

10

15

0 5 10 15 20 25 30 35 40

Conc

entr

atio

n (io

ns/Å

3 )

Z distance (Å)

b) 128 W WatKCl

0

5

10

15

0 5 10 15 20 25 30 35 40

Conc

entr

atio

n (io

ns/Å

3 )

Z distance (Å)

f) 318 W WatKCl

0

5

10

15

0 5 10 15 20 25 30 35 40

Conc

entr

atio

n (io

ns/Å

3 )

Z distance (Å)

c) 160 W WatKCl

0

5

10

15

0 5 10 15 20 25 30 35 40

Conc

entr

atio

n (io

ns/Å

3 )

Z distance (Å)

g) 352 W WatKCl

0

5

10

15

0 5 10 15 20 25 30 35 40

Conc

entr

atio

n (io

ns/Å

3 )

Z distance (Å)

d) 192 W WatKCl

0

5

10

15

0 5 10 15 20 25 30 35 40

Conc

entr

atio

n (io

ns/Å

3 )

Z distance (Å)

h) 384 W WatKCl

Appendices 165

Before reaching the equilibrium point, as water molecules are loaded into the

structural sites of zeolite N (in hydrated membranes with 64, 128 water molecules) the

diffusivity of water molecules inside the membrane decreases and at the equilibrium

point (160 water molecules) reaches its lowest value (Fig. 3a), However, subsequent

added water molecules occupy sites with weaker binding energies (192, 256 and 318

water molecules) and push out extra-framework ions from the membrane structure.

These effects sharply increase water diffusion. When all available adsorption sites are

occupied (at hydrated zeolite by 384 water molecules), any remaining water molecules

can only locate in interstitial sites that could not occupy states with lower energies.

This results in a decrease in water diffusivity inside the membrane. Similar behaviour

was recognized for zeolites LTA and 4A 2, 3.

Moreover, by adding water molecules more than the favourable amount of

zeolite N structure (8 water molecules/cage), the diffusivity of water molecules outside

the membrane (Fig. 3b) increases until the density of water outside the membrane

reaches the density of water (1cm3/gr). Then, due to the increase of pressure on the

system, the mobility of these molecules decreases.

Figure 3: Self diffusivity of H2O molecules, K+ and Cl- ions inside (a) and outside (b) of the membrane

Overall, the diffusivity of Cl- and K+ ions inside and outside of the membrane

increased, because by increasing the amount of water molecules, the interaction of

these ions with the framework decreased and consequently their mobility increased.

0

0.0001

0.0002

0.0003

0.0004

0.0005

0 50 100 150 200 250 300 350 40

Self

diffu

sion

coef

ficie

nt (Å

2 /se

c)

Total number of added water molecules

a- Inside the membrane

WaterKCl

0

0.04

0.08

0.12

0.16

0 50 100 150 200 250 300 350 40

Self

diffu

sion

coef

ficie

nt (Å

2 /se

c)

Total number of added water molecules

b- outside the membrane

Water

K

Cl

166 Appendices

Conclusion

MD simulation studies of zeolite N hydration as well as structural and dynamic

properties of the system have been determined by computational modelling. These

studies on the zeolite N structure present macro-scale parameters that are compatible

with experimental data and validates the model and applied methodologies. Moreover,

the simulation results show that extra-framework ions are progressively extracted from

their initial sites in the dehydrated structure as the number of water molecules per cage

is increased in hydrated states. This represents a decrease in interaction of extra-

framework ions with the zeolite framework and increases their mobility with increased

water adsorbed. Hence, in order to increase the mobility of K+ cations and to enhance

the ion exchange process the amount of water in the zeolite structure should be far

from the equilibrium number (eight water molecules per cage), as it is in the

environment of ion exchange processes.

References

1. Christensen, A. N.; Fjellvag, H., Crystal structure determination of zeolite N from synchrotron X-ray powder diffraction data. Acta Chemica Scandinavica 1997, 51, 969-973. 2. Faux, D. A., Molecular Dynamics Studies of Hydrated Zeolite 4A. The Journal of Physical Chemistry B 1999, 103 (37), 7803-7808. 3. Turgman-Cohen, S.; Araque, J. C.; Hoek, E. M.; Escobedo, F. A., Molecular dynamics of equilibrium and pressure-driven transport properties of water through LTA-type zeolites. Langmuir 2013, 29 (40), 12389-99.

Appendices 167

Appendix B

Table 1 The partial charges, force field assigned types and number of framework, extra-framework and water atoms used in this study.

atom Atomic charges

Force fields assigned

Number of atoms (001) (110)

Si 1 1.756 si4z 16 16 Si 2 1.718 si4z 64 64 Al 1 1.654 al4z 16 16 Al 2 1.623 al4z 64 64

O -1.0886 o2z 304 312 O-OH -0.8453 o2z 32 16 H-OH 0.299 h1o 32 16 K1 +1 k+ 32 32 K2 +1 k+ 64 64 Cl -1 cl+ 16 16 OW -0.82 o2* 128 128 HW 0.41 h1o 256 256

168 Appendices

Figure 1 (a-h) RDFs, g(r) for guest cations to framework atoms, chlorides and water molecules inside ZM-001 membrane

Appendices 169

Figure 2 (a-h) RDFs, g(r) for guest cations to framework atoms, chlorides and water molecules inside ZM-110 membrane

170 Appendices

Figure 3 Self-diffusion coefficient of ions (D) vs. ionic radius. The black labelled points are D values of guest cation in each system. The D values of K cations in each system are identified with different

colours.

Appendices 171

Figure 4 The density profile of water molecules along z direction in different time of MD simulations for K+/Cs+ system of ZM-001. The two red dashed lines indicate the location of ZM-110 in electrolyte

solution

172 Appendices

Appendix C

The uncertainties in MD simulation samplings

The results of MD simulations may have some degree of uncertainty. In this study, the

statistical uncertainties resulting from model construction were investigated. The

initial distribution of guest cations within a zeolite N membrane can influence the final

ion retention values and ion distribution within the zeolite N membrane. To address

this concern, a conventional method in MD simulations is to conduct several MD

simulations with different initial distribution for targeted atoms and measure the error

in results affected by atoms distribution.

Method

In order to investigate the error of model construction, three different MD simulations

were conducted with the same computational settings described in section 6.2.2. The

only difference between these three simulations was the initial distribution of guest

Na+ cations within ZM-001. The Na+ cations were distributed within zeolite N

membrane with a 2 Å location difference in each system. The original K+/Na+ system

mentioned in Ch.6 was named as system Na-1 and the two repeated MD simulations

were named as Na-2 and Na-3. The retention of Na+ cations over 8.5 ns was compared

for three systems. The localisation of Na+ cations within zeolite N membrane were

compared for three systems by analysing the concentration profiles and RDFs after 8.5

ns MD simulations.

Results

Table 1 and Figure 1 represents the retention of Na+ cations over time for three

different systems. The results show, on average, a 2% difference between obtained

values for retention of Na+ cations inside ZM-001.

Table 1 Retained Na cations inside ZM-001 over simulation time for three different systems. Time (ns) 0 1 2 3 4 5 6 7 8 Last ps Na-1 80.0 45.8 46.9 47.7 47.4 47.3 48.1 49.6 51.3 50.0 Na-2 80.0 46.9 48.0 48.4 48.6 48.9 49.5 49.7 51.4 50.0 Na-3 80.0 47.6 48.6 48.4 50.4 51.8 52.2 51.9 52.5 53.0 Average 46.8 47.9 48.2 48.8 49.3 50.0 50.4 51.8 51.0 STD 0.7 0.7 0.3 1.2 1.8 1.7 1.1 0.6 1.4

Appendices 173

Figure 1 Retained Na cations inside ZM-001 over simulation time for three different systems.

Figure 2 represents the distribution of Na+ cations in three systems after 8.5 ns MD

simulations. The Na+ cations show a similar distribution behaviour inside the ZM-001

membrane, though they show different concentration intensities in some cages of

zeolite N.

Figure 2 Na concentration profiles along z direction after 8.5 ns for three different MD simulations.

The two red dashed lines indicate the location of ZM-001 in electrolyte solution.

The distribution of Na+ cations inside the cages of ZM-001 for three systems were

analysed by calculating the RDF graphs for framework atoms. The position of first

peaks in RDF graphs (Figure 3) indicate the nearest distances between Na+ cations and

framework O, Si and Al atoms. These estimated nearest distances are presented in

Table 2. The obtained values show, on average, less than 1% difference between the

results of three simulations.

174 Appendices

Figure 3 RDFs, g(r) for Na cations to framework (a) oxygen, (b) silicon and (c) aluminium atoms

inside ZM-001 for three different MD simulations.

Table 2 Comparison of the nearest distances of Na cations into framework oxygen (O-Na), silicon (Si-Na) and aluminium (Al-Na) atoms inside membranes of three different systems

O-Na Si-Na Al-Na

Na-1 1.965 2.725 2.705 Na-2 1.985 2.815 2.685 Na-3 1.975 2.805 2.685 Average 1.975 2.782 2.692 STD 0.008 0.040 0.009

Conclusion

The uncertainty of MD simulation sampling with designed methodology in this thesis

were investigated for K+/Na+ system of ZM-001. The results show very small errors

arise from the initial distribution of Na+ cations within zeolite N. Since the same

method was used for initial distributions of guest cations in other K+/M+n, the same

relative uncertainty can be expected for the results of other systems.

Appendices 175

Appendix D

Figure 1 (a) polarised microscope image showing mineral and crystal size diversity in different layers of Avoca (b) BSE image of the areas on Avoca thin section analysed by EPMA, and WDS maps

showing the element wt% for (c) Si, (d) Al, (e) Na and (f) analytical total.

176 Appendices

Figure 2 BSE images and EDS spectra from Werris Creek sample demonstrating the presence of (a)

spinal group and (b) apatite minerals.

Appendices 177

Figure 3 (a) BSE image of the area of Werris Creek thin section selected for WDS analysis and WDS maps showing the element wt% for (b) Si, (c) Al, (d) Na, (e) Fe and (f) analytical total

Monireh Khosravi nasab Master of Science nasab_Thesis.… · Monireh Khosravi nasab Master of...

Documents

Transcript of Monireh Khosravi nasab Master of Science nasab_Thesis.… · Monireh Khosravi nasab Master of...