Investigating Transport Through Sub-Nanometer Zeolites Pores
by
Thomas Humplik
B.S., Materials Science and Engineering, Lehigh University, Bethlehem, PA (2008)
S.M., Mechanical Engineering, Massachusetts Institute of Technology (2010)
SUBMITTED TO THE DEPARTMENT OF MECHANICAL ENGINEERING
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
Doctor of Philosophy in Mechanical Engineering
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2014
©2014 Massachusetts Institute of Technology. All rights reserved.
Signature of Author: ……………………………………………………………………………….
Department of Mechanical Engineering
May 9, 2014
Certified by: ………………………………………………………………………………………..
Evelyn N. Wang
Associate Professor of Mechanical Engineering
Thesis Supervisor
Accepted by: ……………………………………………………………………………………….
David E. Hardt
Chairman, Department Committee on Graduate Students
2
Dedicated to my parents.
In 1984, they risked everything to escape an oppressive, socialist-controlled Czechoslovakia so
that my siblings and I could grow up in a country where not only basic freedoms are guaranteed
but also where your motivation is the only barrier to success. It was their non-stop work ethic,
insurmountable persistence, and a refusal to accept any set back as permanent failure that
motivated me throughout the many obstacles I encountered during my PhD journey at MIT.
I hope that I have made you proud.
3
4
Investigating Transport Through Sub-Nanometer Zeolite Pores
by
Thomas Humplik
Submitted to the Department of Mechanical Engineering on May 27, 2014, in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy (Ph.D.) in Mechanical Engineering
Abstract
Membrane-based reverse osmosis (RO), which accounts for over 40% of the current
worldwide desalination capacity, is limited by the solution-diffusion mode of water transport
through a tortuous polymeric active layer. Alternatively, simulations suggest that introducing
rigid sub-nanometer pores into the active layer could substantially increase the water
permeability and achieve perfect salt rejection. In this thesis, we synthesized MFI (Mobil Five)
zeolites that have uniform, well-defined pores of ≈ 5.5 Å and experimentally investigated the
transport across these sub-nanometer pores. This porous structure serves as a model framework
to experimentally investigate water and salt transport and can be used to suggest the potential
performance of such microporous active layers in RO-based desalination.
We first developed an experimental methodology that combined vapor sorption analysis with
high-pressure infiltration to probe both the role of crystal size and internal surface chemistry on
the transport properties. For purely siliceous MFI zeolites, upwards of 100 MPa was required to
saturate the porous network to the framework capacity of 35 water molecules per unit cell.
However, by introducing hydrophilic (i.e, acidic) defects within the structure, this infiltration
pressure was reduced by five orders of magnitude (to ≈ 1 kPa). While increasing the defect
density increased the amount of water within the pores at typical RO pressures (≈ 5 – 6 MPa),
the diffusivity of this infiltrated water within the more defective zeolites was 1 – 2 orders of
magnitude lower than that of the water within the purely siliceous MFI zeolite. This decreased
diffusivity, which was attributed to the strong attraction of water to the hydrophilic defect sites,
resulted in ≈ 10x decrease in the estimated permeability.
We subsequently microfabricated sub-micron thick zeolite-based membranes to investigate
transport limited to the zeolite crystals. By quantifying the water flux generated by a
concentration gradient (i.e., forward osmosis), the flux across the more hydrophobic MFI zeolites
was ≈ 8 – 10x higher than that across the more hydrophilic MFI zeolites. Additionally, although
a small amount of meso/macro-sized defects existed in the membranes, no salt transport (within
5
experimental uncertainty) was detected across the zeolite pores, which demonstrated that the
pores of MFI zeolites were capable of selectively transporting water and rejecting hydrated salt
ions.
Collectively, this work presents an improved fundamental understanding of the transport of
water that is confined within the sub-nanometer pore structure of zeolites. The insights gained
demonstrate the potential of size-selective membranes in water desalination and offers
approaches toward improving the performance of future RO membranes.
Thesis Committee Chair:
Professor Evelyn N. Wang
Department of Mechanical Engineering, MIT
Thesis Committee Members:
Professor Rohit Karnik
Department of Mechanical Engineering, MIT
Professor Karen Gleason
Vice Provost, MIT
Alexander and I. Michael Kasser Professor of Chemical Engineering, MIT
Professor Michael Tsapatsis
Professor of Chemical Engineering and Materials Science, UMN
6
Acknowledgements
First and foremost, I must thank my advisor, Prof. Evelyn N. Wang, for taking me on as a
student and for supporting me throughout my Master’s and PhD work. Even when my research
was not progressing, she could somehow always encourage me to work through the difficulties
and, ultimately, achieve the results that are presented in this thesis. I believe that her enthusiasm
and dedication to her students are unrivalled at MIT and, in my opinion, she is the best advisor
that any graduate student could possibly have.
I also must thank my committee for all of the help and support during my thesis work. Prof.
Rohit Karnik’s knowledge of nanoscale transport phenomena and experimental procedures were
extremely helpful in deducing seemingly nonsensical data and Prof. Karen Gleason’s expertise in
membranes transport and fabrication helped me immensely in creating the microfabricated
zeolite-based membranes. I am also extremely grateful to Prof. Michael Tsapatsis for answering
an email of a random graduate student who was interested in zeolites. Being able to visit his lab
in Minnesota and interact with students shaved many years of frustration off of my research and
his insights into transport through zeolites was immeasurably helpful.
I am also greatly indebted to both Prof. Shalabh Maroo and Prof. Rishi Raj, the two postdocs
who assisted me with my research. Shalabh exposed me to the world of molecular interactions
where I really began to understand the complexities of the project. He also taught me how to read
between the lines of literature and the importance of paying attention to detail. Rishi was a joy to
work with. He always made time for my insane project ideas and somehow never got angry
about the lack of results (even though most of our ideas did not pan out to the countless Nature
and Science papers we envisioned). I learned a great deal from Rishi, and I hope that I was never
an encumbrance to him. Both Shalabh and Rishi were great role models for me and I am certain
they will be successful in their careers as Professors.
I owe a great debt of gratitude to Dr. Jongho Lee, Sean O’Hern, and Rong Yang. Each spent
countless hours with me in troubleshooting experiments or helping me fabricate the membranes.
There are not enough thanks that I could ever give to each of you for helping me over the years
of my PhD.
7
To the staff of the Center for Nanoscale Systems (Harvard), Center for Materials Science and
Engineering, and the Institute of Soldier Nanotechnologies, I thank you all for taking the time to
train me and help me with any tool problems. None of this research would have been possible
without their aid. I also have to thank Leslie Regan of the Mechanical Engineering graduate
office, Thea Szatkowitski and Theresa Werth (the two more ‘permanent’ admins our lab had). A
lot of their work is seemingly ‘behind the scenes’, but without their support, many graduate
students would not be able to do research. I am forever thankful for their aid.
Many many many thanks to all my friends in the DRL. I could not have envisioned a better
lab group to work with than the Device Research Lab. In the course of my six years within the
group, the size increased from three students to more than twenty-five. During the course of
those six years, I had the privilege in working with some of the best minds in science and
engineering. As I really had no concept of mechanical engineering when I started, the
interactions that I had with each of you really helped me with both classwork and research. I owe
a special thanks to Dr. Andrej Lenert, who was with me in the DRL for the entirety of my
Master’s and PhD at MIT. Some of my favorite times were spent brainstorming with Andrej
about crazy experiments for even crazier research projects. It is one of the things I will miss most
about MIT. Furthermore, to all of my friends that have shared an office with me (Heena Mutha,
Daniel Danks Hanks, Dr. Nenad Miljkovic, Jeremy Cho, Matt Thoms, Zhengmao Lu and even
Dr. Suman Bose), thank you! You made each and everyday enjoyable, and I may have learned
the most about science and engineering from our interactions. I hope that my endless supply of
animal videos brought some joy to the monotony of day-to-day research life.
Finally, thank you to my family. I would have never even dreamed about attending MIT
without your support and motivation. Thank you for always being there for me.
8
Table of Contents
1. Introduction ........................................................................................................................ 14
1.1
Motivation ................................................................................................................... 14
1.2
Zeolites as Molecular Sieves for Water Desalination ................................................. 18
1.3
Scope and Contents of the Thesis................................................................................ 22
2. The Framework Water Capacity and Infiltration Pressure of MFI Zeolites ...................... 24
2.1
Overview ..................................................................................................................... 24
2.2
Experimental Procedures ............................................................................................. 26
2.2.1 Zeolite Synthesis ................................................................................................... 26
2.2.2 Characterization .................................................................................................... 27
2.2.2.1 Imaging........................................................................................................... 27
2.2.2.2 X-ray diffraction (XRD)................................................................................. 28
2.2.2.3 Water Sorption and Infiltration ...................................................................... 28
2.2.2.4 Nitrogen Sorption ........................................................................................... 32
2.2.2.5 Nuclear Magnetic Resonance (NMR) ............................................................ 32
2.3
Results and Discussion ................................................................................................ 32
2.3.1 Framework Water Capacity .................................................................................. 34
2.3.1.1 Water Sorption and Infiltration Experiments ................................................. 34
2.3.1.2 Micropore Volume ......................................................................................... 37
2.3.1.3
29
Si MAS NMR spectroscopy ....................................................................... 40
2.3.2 Water Infiltration Pressure .................................................................................... 42
2.4
Conclusions ................................................................................................................. 44
3. Effect of Hydrophillic Defects on Water Transport in MFI Zeolites ................................. 46
3.1
Overview ..................................................................................................................... 46
3.2
Experimental Materials and Methods.......................................................................... 48
3.3
Results and Discussion ................................................................................................ 51
3.3.1 Sorption and Infiltration Behavior ........................................................................ 51
3.3.2 Water Diffusivity and Permeability ...................................................................... 56
3.4
Conclusions ................................................................................................................. 60
4. Osmotically-Driven Flow Across Microfabricated Zeolite Membranes............................ 62
4.1
Overview ..................................................................................................................... 62
9
4.2
Experimental Procedures ............................................................................................. 65
4.2.1 Zeolite Synthesis ................................................................................................... 65
4.2.2 Zeolite Orientation ................................................................................................ 66
4.2.3 Meso/Macropore Filling and Membrane Activation ............................................. 69
4.2.4 Transport Measurements ....................................................................................... 74
4.2.4.1 Water Transport Quantification ..................................................................... 75
4.2.4.2 Salt Transport Quantification ......................................................................... 77
4.2.4.3 Determining the Leakage Pathways ............................................................... 79
Results and Discussion .......................................................................................................... 80
4.2.5 Membrane Fabrication .......................................................................................... 80
4.2.6 Water Flux Quantification ..................................................................................... 82
4.2.7 Salt Diffusion Across Membranes ........................................................................ 86
4.2.8 Flux Comparison with RO Membranes ................................................................ 88
4.3
Conclusions ................................................................................................................. 90
5. Conclusions and Future Work ............................................................................................ 92
5.1
Conclusions and Contributions of this Thesis ............................................................. 92
5.2
Future Work ................................................................................................................ 95
5.2.1 Diffusivity Difference between Uptake and NMR ............................................... 96
5.2.2 Pressure Driven Transport in Sub-Nanometer Pores ............................................ 98
5.2.3 Effect of Pore Size and Tortuosity on Permeability and Salt Rejection ............... 98
6. References ........................................................................................................................ 101
10
List of Figures
Figure 1-1. Schematic diagrams of the current major desalination techniques. ........................... 15 Figure 1-2. Sub-nanometer length scales associated with creating a molecular sieve for water
desalination............................................................................................................................ 17 Figure 1-3. Pore structure and crystal shape of MFI-type zeolites. .............................................. 19 Figure 1-4. Past experimental results of zeolite-based membranes for desalination. ................... 20 Figure 1-5. Experimental results of interfacially embedded RO membranes with hydrophilic
zeolites. .................................................................................................................................. 21 Figure 2-1. High-pressure infiltration experimental setup. ........................................................... 29 Figure 2-2. Schematic of experimental setup used to determine if additional water entered into
the zeolites after the adsorption experiments but prior to the infiltration experiments. ........ 31 Figure 2-3. Electron microscopy (A) and x-ray diffraction (B) analysis of MFI zeolites. ........... 33 Figure 2-4. Combined water adsorption and infiltration isotherms for MFI-type zeolites (A)
Water adsorption isotherms at 25°C for varying crystal sizes of MFI zeolites..................... 35 Figure 2-5. (A) Nitrogen adsorption isotherms with the ordinate corresponding to the calculated
pore volume and the abscissa corresponding to the partial pressure of nitrogen at 77 K. .... 37 Figure 2-6. 29Si MAS NMR spectra for varying crystal sizes of MFI zeolites. ............................ 40 Figure 2-7. Defect density effects on the infiltration pressure and low-pressure water uptake. ... 42 Figure 3-1. SEM (A) and XRD (B) analysis of MFI zeolites. ...................................................... 49 Figure 3-2. Combined adsorption and infiltration isotherms for varying Si/Al ratio MFI zeolites.
............................................................................................................................................... 52 Figure 3-3. Comparison of the experimental data with Cailliez’s (A) weaker MFI zeolite defect
model and (B) stronger defect model. ................................................................................... 54 Figure 3-4. Magnified view of the comparison of the experimental data with Cailliez’s weak
defect model (A) and strong defect model (B). ..................................................................... 55 Figure 3-5. Example adsorption curve for various Si/Al ratio MFI zeolites taken at a relative
pressure of 0.4. ...................................................................................................................... 57 Figure 3-6. The complete data set of both the diffusivity (A) and solubility (B) as a function of
the relative pressure. .............................................................................................................. 59 Figure 3-7. The diffusivity, solubility and permeability of water within MFI zeolites as a function
of the defect density. ............................................................................................................. 59 Figure 4-1. Permeability and salt rejection of zeolite-based and current state-of-the-art
membranes............................................................................................................................. 63 Figure 4-2. Potential transport pathways through previous zeolite-based membranes. ................ 64 Figure 4-3. SEM images of synthesized zeolites. ......................................................................... 66 11
Figure 4-4. Manual direct assembly of MFI zeolites onto a support. ........................................... 67 Figure 4-5. SEM images of oriented (a) Na-MFI and (b) H-MFI zeolites on pieces of a silicon
wafer. ..................................................................................................................................... 67 Figure 4-6. Initial orientation results of smaller H-MFI zeolites on a porous AAO support. ....... 68 Figure 4-7. SEM images of the improved orientation technique, which utilized PVA as a smooth
adhesion layer on the AAO. .................................................................................................. 69 Figure 4-8. SEM images of H-MFI membranes after 93 nm of HfO2 via atomic layer deposition.
............................................................................................................................................... 71 Figure 4-9. SEM images of zeolite membranes that were over-etched with RIE using a CF4/O2
chemistry. .............................................................................................................................. 71 Figure 4-10. Fabrication schematic (shown with H-MFI zeolites) with corresponding SEM
images.................................................................................................................................... 73 Figure 4-11. Images of fabricated Na-MFI membranes. ............................................................... 74 Figure 4-12. Test cell to quantify water transport. ........................................................................ 75 Figure 4-13. Time-lapse images of the change in the water level in the graduated cylinder. ....... 76 Figure 4-14. Estimation of the active zeolite area using ImageJ. ................................................. 77 Figure 4-15. Schematic of (a) diffusion cell and (b) example for the reservoir conductivity as a
function of time. .................................................................................................................... 78 Figure 4-16. Setup used to quantify Allura Red Transport. .......................................................... 79 Figure 4-17. Adsorption isotherms for zeolites used for membranes. .......................................... 80 Figure 4-18. Transport properties of MFI zeolites. ....................................................................... 81 Figure 4-19. SEM images of the fabricated zeolite-based membranes. ........................................ 82 Figure 4-20. Flux results for membranes as a function of osmotic pressure. ............................... 83 Figure 4-21. Variation in active area for an H-MFI membrane (#8 in Table 4-1) ........................ 84 Figure 4-22. Cracks/defects present in membranes that lead to salt transport. ............................. 86 Figure 4-23. Diffusion normalized flux of AR and KCl across zeolite membranes. .................... 88 Figure 4-24. Thickness normalized flux as a function of the osmotic pressure. ........................... 89 Figure 5-1. Difference in the water flux predicted by Liu et al. (16) and the measured water flux
across the zeolites in this study. ............................................................................................ 96 12
Chapter 1 1. Introduction
1.1
Motivation
The ever-increasing water shortage around the world is well-documented: 88 developing
nations are currently affected by water scarcity, over 1.2 billion people lack access to potable
water, and tens of millions are affected each year by water-borne bacteria and viruses (1,2). Even
in developed nations, fresh water resources are becoming increasingly stressed due to expanding
agriculture, rising demand from industry, and growing human consumption (3). These water
stressed regions are only expected to grow as the human population is projected to increase by as
much as 50% in the coming decades. Since only ≈ 0.5% of the world’s total water supply is
accessible fresh water, there has been substantial effort in finding alternative means of producing
fresh water. Over the past few decades, desalination has become an essential tool in providing
clean water and, in addition to improvements in both water usage and reclamation, has helped in
beginning to alleviate the worldwide water crisis (1,3,4).
Desalination technologies and capabilities have drastically improved over the last 50 years
or so. The first major desalination plants appeared in 1938 in what is now Saudi Arabia and
utilized a simple distillation process to separate potable water from a seawater reservoir (5). Yet,
these early distillation systems used significantly higher energy than the theoretical minimum
required to separate water from salt due to the latent heat penalties associated with liquid to
vapor phase change (6). However, either with improved system design (e.g., utilizing waste heat
during desalination (1)) or by exploiting alternative mechanisms for transport (7-9), thermalbased desalination techniques have become much more energy efficient and, in some cases, are
only ≈ 10x higher than the theoretical minimum (≈ 1 kWhr m-3 for reasonable recovery rates).
Alternatively, by avoiding the liquid to vapor phase change, other desalination techniques have
demonstrated energy usage closer to the theoretical minimum (3). Among these methods,
membrane-based reverse osmosis (RO) has become the most economically viable solution and
currently accounts for ≈ 40% of the world’s desalination capability (Figure 1-1). Additionally,
more than 75% of the newly built or planned desalination plants are RO-based (4).
14
Figure 1-1. Schematic diagrams of the current major desalination techniques.
Both multi-stage flash and multi-effect distillation are thermal techniques that rely on phase
change to separate water from salt water. Reverse osmosis is a membrane-based method that
relies on pressurizing the seawater past the osmotic pressure to force water across a semipermeable membrane. Electrodialysis (and other techniques like capacitive deionization) utilizes
electric fields to separate salt from water streams. The percentages associated with each method
represent the respective portion of the desalination capacity as of 2011. Figure used from (3).
In RO, desalination of water is achieved by applying a high pressure to a salt water
solution to force fresh water across a semi-permeable membrane. This semi-permeable
membrane ideally allows only for the transport of water molecules while rejecting salts and other
contaminants found in seawater. To desalinate the water, the pressure must surpass the osmotic
pressure of the seawater, which is approximately 30 bar for typical seawater concentrations1.
Since most of the energy required for RO is in pressurizing liquid seawater, much of this energy
1
Seawater concentrations are substantially higher in regions in the Middle East. In these
regions, it can be more energy efficient to desalinate using thermal-based techniques (3).
15
can be recovered using turbines and pressure recovery devices (1). Thus, for large-scale RO
plants, the energy required to desalinate water is only ≈ 2 – 6 kWh m-3 (under normal recovery
rates), or only ≈ 2 – 4 times that of the theoretical minimum (4). Yet, many challenges with
reverse osmosis desalination still exist.
Current state-of-the-art RO membranes are comprised of a thin (100 – 200 nm) active
layer composed of an aromatic polyamide (AP) that is supported by a thicker (50 µm) porous
polysulfone mechanical support. Water (and other molecules) transport through the active layer
via the solution-diffusion mode of transport (10) (i.e., sorption to the surface followed by
diffusion through the layer). Although the AP doesn’t have well-defined pores, the movement of
the polymeric chains creates sub-nanometer gaps for water to transport through. These AP active
layers preferentially allow for significantly higher diffusion rates (1 – 2 orders of magnitude) for
water molecules in comparison to various salt ions, and thereby very high (> 98%) salt rejection
rates can be realized (4). For current AP membranes, water fluxes of 10 – 20 Lm-2hr-1 with
applied pressures in the range of 50 – 60 bar (3,4) are typically achieved. Yet, these fluxes have
not significantly increased since AP was introduced as the active layer in RO membranes
(replacing cellulose acetate in the early 1990s). While it is possible to reduce the thickness of this
active layers (which results in an increase in the flux), the salt rejection across the active layer
decreases (4). This permselectivity tradeoff is a well-known phenomena with most solutiondiffusion limited membranes, and therefore the potential increase in the permeability of APbased RO membranes, (which ultimately controls how much water is desalinated) is
limited (11,12). Furthermore, AP-based membranes also have poor boron rejection properties,
and subsequently RO desalinated water requires post-treatment processing such that it can be
used as drinking water (11,13).
Rather than focusing on creating new polymeric-based materials to improve both the
water permeability and salt/boron rejection, recent membrane research has focused on creating
an inorganic-based active layer that is comprised of well-defined, sub-nanometer pores to reject
salt but allow for water transport (i.e., a molecular sieve for water desalination, Figure 1-2) (1417). Recent molecular dynamics simulations have shown that if the diameter these subnanometer pores is less than the size of a hydrated salt ion (≈ 7 – 8 Å for the first hydration shell
of most monovalent ions) and larger than a diameter of a water molecule (≈ 3 Å), perfect
rejection can be achieved (15,16,18,19). Since the energy required to strip a monovalent salt ion
16
(e.g., sodium or potassium) of this hydration shell is significant (≈ 1700 kJ/mol), additional
enthalpic and entropic barriers (other than pure size-exclusion) exist for both the entry and
passage of ions through sub-nanometer sized gaps (3,20). To reject boric acid (the dissolved
form of boron found in seawater), the maximum pore diameter should be further restricted to
≈ 6 Å, which corresponds to the van der Waals diameter of the hydrated molecule. By using the
sub-nanometer pores to reject salt ions and boric acid molecules, the selectivity mechanism is
effectively decoupled from the mode of water transport. Thus, in contrast to the solutiondiffusion limited polymeric membranes, very thin (< 10 nm) size-exclusion based membranes
can potential provide both higher water fluxes (> 100 Lm-2hr-1) and perfect salt rejection (15).
Water Molecule
Boric Acid
~ 0.3 nm
~ 0.6 nm
Ideal Pore
0.4 nm < D < 0.6 nm
Solvated
Chlorine Ion
~ 0.8 nm
Figure 1-2. Sub-nanometer length scales associated with creating a molecular sieve for
water desalination.
Although molecular-based simulations predict significant enhancements to both the water
permeability and salt rejection (15,16,19,21), experimental membranes have yet to conclusively
demonstrate this hypothesized improvement in performance. One challenge with these
experimental membranes is in controlling the pore diameter in the previously impermeable
materials (17,22). While advances in micro and nano-based fabrication techniques have allowed
researchers to manipulate matter to atomistic length scales (22-24), these techniques are usually
limited to very small areas (≈ 10-18 – 10-12 m2). These small active areas make transport analysis
(in particular water transport) particularly difficult to quantify. Thus, it is unclear how both water
and salt transport through these confined geometries and questions still exist whether the
transport models predicted by molecular simulations are accurate. Instead of improving the
17
fabrication procedures to better control both the pore diameter and the area over which these
pores are made, the work in this thesis seeks to experimentally probe the fundamental transport
mechanisms of both water and salt across materials in which these sub-nanometer pores
intrinsically exist. This improved understanding can first be used to determine if there is any
benefit (in terms of water permeability or salt rejection) in using porous sub-nanometer active
layers in reverse osmosis desalination over current AP-based membranes. Additionally, the
results from the experiments can be used to compare to the predicted transport results observed
with molecular simulations, and, ideally, be used to improve the molecular potentials to better
match the actual transport behavior. To probe the transport mechanisms as these length scales,
zeolites were used due to their well-known and uniform sub-nanometer pore structure. However,
as the subsequent sections and chapters in this thesis will show, even by using materials
containing these well-defined pores, considerable experimental challenges exist in elucidating
the fundamental transport properties. While the rest of this thesis will focus on studying water
transport in zeolites, I encourage readers to refer to our review on other nanostructured materials
used in desalination as well (see reference 3).
1.2
Zeolites as Molecular Sieves for Water Desalination
Zeolites are primarily aluminosilicate minerals containing a porous microstructure
composed of 3–8 Å pores (Figure 1-3). Zeolite crystals occur naturally or can be synthesized in a
laboratory environment (typically at high temperatures and high pressures) (25). Crystal sizes
can be controlled from a few nanometers to centimeters by varying synthesis temperature and
time (25,26). Properties such as adsorption characteristics, geometry, ion exchange capabilities,
and catalytic behavior differ amongst the zeolite crystal families and can be tailored to a specific
application by using the correct composition (25). Porosity varies among zeolites, typically
ranging between 30–40%. The combination of high porosity and high active surface area has led
to significant zeolite research for catalytic applications. In commercial applications, the most
common use for zeolites is as adsorbents during various chemical processes (9,27-29).
18
~"0.55"nm"
~"0.54"nm"
A.
B.
Figure 1-3. Pore structure and crystal shape of MFI-type zeolites.
(a) Schematic of three-dimension sub-nanometer porous microstructure of MFI-type zeolites.
(b) Scanning electron microscope image of MFI-type zeolites synthesized in this work. Scale bar
is 3 µm.
Since most zeolites have a tight pore distribution less than the diameter of a hydrated salt ion,
a membrane created from these crystals has the potential to completely reject salt ions while
permitting water molecules to permeate through. Molecular dynamics (MD) simulations have
provided mechanistic insights into these processes. Murad and Lin investigated water–ion
separation using a single ZK-4 zeolite with 4.4 Å pore diameters in a NaCl/water solution (18).
As expected, the solvated ions were too large to pass through the pores and only water molecules
could flow through the zeolite.
These MD simulations have since motivated researchers to fabricate zeolite-based
membranes for RO and experimentally investigate the possibility of achieving high flux with
excellent ion rejection. Li et al. (14) used hydrothermal synthesis to develop 0.5–3 µm thick
membranes consisting of hydrophobic MFI (Mobil Five) type zeolites with an average pore
diameter of 5.6 Å on a porous α-alumina support (Figure 1-4a). Under an applied pressure of
2.07 MPa (20.7 bar) and with 0.1 M NaCl feedwater, the membranes rejected 76% of Na+ ions
while permitting a water flux of 0.112 kg m−2 h−1 (≈0.11 L (m−2 h−1 )). This lower rejection
was attributed to ions transporting across nanometer-sized interstitial defects created during the
membrane synthesis process. In later work, Li et al. decreased the silicon/aluminum ratio of the
zeolite, which decreased the hydrophobicity and increased the flux to 10.21 kg m−2 h−1 with an
applied pressure of 3.5 MPa (35 bar) and 0.1 M NaCl feedwater (30). Ion rejection also
improved dramatically from ≈ 76% to 98.6% (Figure 1-4b).
19
Figure 1-4. Past experimental results of zeolite-based membranes for desalination.
(a) Cross-sectional view of a zeolite membrane grown directly on porous α-alumina support.
Adapted from (14) (b) Experimental results of ion rejection and water flux for direct growth
zeolite membranes as a function of ion concentration. Adapted from (31) (c) Schematic of
intercrystalline pore structure of direct growth zeolite membranes. Adapted from (14). (d)
Schematic of overlapping of the EDL between zeolite crystals at low salt concentration (solid
line) and high salt concentration (dashed line). Adapted from (14).
However, as the salt concentration of the solution increased, the salt rejection decreased
considerably to ≈ 90% for a 0.3 M NaCl solution (30). These results suggest that the salt
rejection depended on the formation and size of electric double layers (EDLs) at the surface of
the intercrystalline defects: in low salt concentrations, the EDLs can overlap and prohibit the
transport of salt ions, while, in high salt concentrations, EDLs become thinner and no longer
overlap, allowing ions to pass through the intercrystalline defects (Figure 1-4c,d). These
20
experimental results showed that nanometer-sized intercrystalline defects controlled the majority
of ion transport (and possibly the water transport) and presented a challenge in both the
fabrication of zeolite membranes and in understanding the mechanisms of transport across the
zeolite pores. In addition to synthesizing zeolites directly onto a ceramic support, zeolites have
also been incorporated into polymeric RO membranes. Jeong et al. synthesized thin polymercomposite (TFN) membranes interfacially embedded with ≈ 100 nm cubic LTA (Linde Type A)
zeolite crystals with 4.4 Å diameter pores (32), shown in Figure 1-5. These hydrophilic zeolites
were chosen to “create preferential flow paths for water to permeate through while also
simultaneously rejecting the transport of ions (32).”
Figure 1-5. Experimental results of interfacially embedded RO membranes with
hydrophilic zeolites.
(a) TEM micrograph of hand-cast nanocomposite membrane incorporating LTA zeolites in
polyamide with a polysulfone support layer. (b) Experimentally measured solute rejection and
permeability of the zeolite/polyamide composite membranes. The membranes exhibit increasing
permeability with increasing zeolite loading while the solute rejection remained approximately
constant. Open symbol indicates a zeolite pore-filled (non-calcined) TFN membrane. The TFC
arrow corresponds with 0% W/V zeolite loading. Other arrows indicate data corresponding to the
axes. Both images obtained from (32).
The membranes were characterized for permeability and salt rejection with increasing zeolite
content using a 2000 ppm NaCl/MgSO4 (≈ 0.03 M) solution and an applied pressure on the feed
side of 1.24 MPa (12.4 bar). Increased water fluxes were demonstrated with increasing zeolite
weight percentage. Additionally, membranes with embedded zeolite always showed higher
permeability when compared to experimental thin-film composite (TFC) membranes without
zeolites while maintaining high salt rejection (Figure 1-5). However, even with the optimal
21
zeolite loading of ≈ 40%, the water flux was found to be lower than some commercial state-ofthe-art RO membranes. Therefore, from the results of these experimental membranes, the role of
the zeolites in both the water and salt transport across the membranes is unclear.
Clearly, the primary challenge of studying transport with the zeolite membranes described
above is that water molecules and ions can permeate around the zeolite crystals. In this
configuration, it is difficult to determine the role that zeolites have in water transport and salt
rejection. While these aforementioned membrane fabrication methods are suitable for potential
scale-up, the yet unanswered question is whether the permeability/selectivity of zeolites is
warranting of such use.
1.3
Scope and Contents of the Thesis
The goal of this thesis was to develop novel experimental methodologies to probe the
mechanisms of water and salt transport within the sub-nanometer pores of zeolites. Using these
novel techniques, we investigated the zeolite properties that controlled the transport rate to find
routes to optimize the water permeability. This thesis expanded upon the work performed during
my Master’s research with the Device Research Laboratory in the Mechanical Engineering
Department at the Massachusetts Institute of Technology (33).
In Chapter 1, the motivation behind improving desalination capabilities is discussed, with
particular emphasis on improving the water permeability and salt rejection of polymeric reverse
osmosis membranes by using inorganic sub-nanometer pores as the separation medium in the
active layer. A thorough review of past zeolite-based membranes is presented.
In Chapter 2, the development and utilization of a combined water sorption and highpressure infiltration experimental technique is introduced. This technique is first used to identify
both the framework water capacity (i.e., the amount of water molecules that fill a unit cell) and
the infiltration pressure (i.e., the pressure required to saturate the unit cell with water) of MFI
zeolites, as significant variations in literature exist for these two properties. Furthermore,
experimental guidelines are set forth such that this combined sorption/high-pressure infiltration
technique can be applied so that the experimental analysis is limited to water within the zeolite
crystals.
22
In Chapter 3, the role of the internal surface chemistry on the water transport within MFI
zeolites was investigated using the sorption/high-pressure infiltration technique. Additionally,
the transient adsorption and desorption behavior was analyzed to estimate the water permeability
through the MFI zeolites as a function of the zeolite hydrophilicity.
In Chapter 4, osmotically-driven water transport and salt diffusion is studied across wellcharacterized microfabricated zeolite-based membranes. The insights from the previous chapter
are tested to determine the optimal internal surface chemistry that maximizes the water
permeability across the zeolite crystals.
In Chapter 5, the main contributions of this thesis are summarized. Recommendations for
future work are presented.
23
Chapter 2 2. The Framework Water Capacity and Infiltration Pressure of MFI Zeolites
2.1
Overview
The high specific surface area and the uniformity in sub-nanometer pores of zeolites, a
class of materials with well-known synthesis procedures (25), however, can overcome some of
these challenges and offer a route to study nanoscale transport phenomena using macroscopic
experimental techniques.
Specifically, water transport in MFI zeolites (a silica-based
microporous material with ≈5.5 Å diameter pores), where the pore structure can be utilized as a
molecular sieve for water desalination (14,16,34,35), has been studied extensively (35-46). In
spite of the advantages of uniform pore geometry, simplified synthesis, and over three decades of
widespread research (25), the water infiltration mechanisms and the subsequent transport through
the pores remain unclear. For example, the range of experimentally reported diffusivity values
for water within purely siliceous MFI zeolites spans seven orders of magnitude, from ≈10-7 m2/s
- 10-14 m2/s (36,47-50). Furthermore, even a basic quantity such as the total framework (internal)
capacity per unit cell of the MFI zeolite varies significantly (34-57 water molecules per unit cell)
amongst both experiments and simulations (Table 2-1).
Similarly, some studies show that the infiltration pressure is low (1– 30 kPa, adsorption
regime, (45,51,52)), while others report very high values for the infiltration pressure (>30 MPa,
high pressure infiltration regime (38,53,54)). A variation in the infiltration pressure could arise
from small changes in the zeolite composition or internal defect density, however, the
disagreement over the total internal water capacity, which ideally should remain consistent
amongst all MFI-zeolites (so long as the internal pore structure does not undergo significant
structural changes), highlights a lack of understanding of water transport within MFI-zeolites.
We attribute much of this discrepancy to the unknown contribution of the textural porosity (i.e.,
inter-crystalline pores) on the overall water uptake along with the unknown quantity of preadsorbed water into the pores of hydrophobic MFI zeolites.
24
Table 2-1. Infiltration pressure and water capacity into MFI zeolites pores compiled from
various studies.
(S) indicates simulation-based results and (E) indicates experimentally measured quantities. For
consistency, the framework capacity is taken at a value of 160 MPa to correspond to our
experimental conditions of this study. Note the large variation in the capacity and infiltration
pressure amongst the reported values. Olson et al. [37] did not investigate the infiltration
pressure.
Source
Framework
Capacity (N/UC)
Infiltration
Pressure (MPa)
Trzpit et al. [22]
34 - 37 (E), 41 (S)
75 - 125 (S, E)
Cailliez et al. [23]
35 (E), 40 (S)
80 (E), 120 (S)
Desbiens et al. [24]
39 - 45 (S)
0.01 - 100 (S)
Ahunbay et al. [25]
35 - 37 (S)
0.01 (S)
Lella et al. [26]
39 - 41 (S)
80 - 120 (S)
Ramachandran et al. [27]
57 (S)
0.001 - 0.003 (S)
Olson et al. [15]
53 (E)
n/a
In the work presented in this chapter, we performed controlled experiments and detailed
sample characterizations to investigate the total framework water capacity and infiltration
pressure, and identified the subtleties that may have led to the discrepancies in literature. We
synthesized and procured various sizes of purely siliceous MFI-type zeolites (Silicalite-1) and
confirmed the uniformity in structure and morphology using scanning electron microscopy
(SEM), transmission electron microscopy (TEM), x-ray diffraction (XRD), nitrogen sorption,
and nuclear magnetic resonance (NMR). Combined water adsorption and high-pressure
infiltration experiments on these MFI zeolites with characteristic crystal dimensions between
≈10 nm – 10 µm were performed. The framework water capacity for fully crystalline MFI
zeolites was determined to be 35 ± 2 water molecules per unit cell. Nano-sized zeolites, with
crystal diameters up to 100 nm, exhibited a reduction in both the measured micropore volume
(up to 25% compared to the larger zeolites) as well as the total internal water capacity (up to
50% compared to the larger zeolites). The decrease was attributed to the un-crystallized silica
25
regions infused within the crystal resulting from incomplete synthesis. Despite the differences in
synthesis procedure and crystal morphology for the various zeolites, the infiltration pressure was
approximately the same, ≈95 – 100 MPa. The experimentally determined framework water
capacity of 35 ± 2 water molecules per unit cell and infiltration pressure values of ≈95 – 100
MPa can now be utilized to validate and improve upon the existing water-zeolite interaction
potentials used in molecular simulations. The appropriate interaction potentials can then facilitate
design guidelines for practical applications such as porous membranes for water desalination.
2.2
2.2.1
Experimental Procedures
Zeolite Synthesis
The details of the zeolite synthesis (composition, temperature, and time) are provided in
Table 2-2. Tetraethyl orthosilicate (TEOS, Sigma Aldrich) and tetrapropylammonium hydroxide
(TPAOH, 40 wt % aqueous solution, Sigma-Aldrich) were used as the silica source and the
structure-directing agent (SDA), respectively. Class 2 deionized water (VWR) was used to
control the pH of the synthesis solutions. Measured amounts of TEOS, TPAOH, and H2O were
mixed at room temperature and stirred for at least 12 hours to obtain a clear solution. The
solutions were then transferred into their appropriate vessel for synthesis. For hydrothermal
treatment, a 45 mL PTFE-lined stainless steel autoclave (Parr, Inc) was used and was rotated at ≈
100 RPM inside of an oven (BlueM, Thermal Product Solutions) at the desired temperature. For
microwave synthesis, a Kevlar-lined PTFE vessel (HP-500, CEM) was used with a MARS 5
(CEM) microwave oven. For the smallest zeolites, PTFE tubes (VWR) were placed in a
temperature-controlled bath (PolyScience) and held at temperature without agitation. After
synthesis, the zeolites were recovered by centrifugation and washed with deionized water until
the pH of the solution was ≈ 9. This was followed by drying at 60 °C and calcination under air at
550 °C for three hours to remove the organic template. MFI-F was kindly provided by Dr.
Machteld Mertens (ExxonMobil, Machelen, Belgium). The zeolites are referred to by their
respective internal to external surface area ratio. For example, the ratio of internal framework
surface area to outer crystal surface area for the largest crystal investigated in this study
(MFI 12000 in Figure 2-3) is 12000. The internal surface area was estimated by calculating the
available surface area of the total number of unit cells in each crystal using a geometric
26
approximation (55) while the external surface area was estimated from the characteristic crystal
dimensions directly measured using the imaging techniques described in the Section 2.2.2.
2.2.2
Characterization
2.2.2.1
Imaging
The nano-sized zeolites (Figure 2-3, MFI 50, 120 and 250) were imaged by transmission
electron microscopy (2100, JEOL) (TEM) while the larger, micron-sized zeolites (Figure 2-3,
MFI 1150, 1500, 12000) were imaged using scanning electron microscopy (Ultra-55, Zeiss)
(SEM). For SEM analysis, the samples were coated with ≈ 10 nm of Pt/Pd to prevent charging
during imaging. The dimensions and exterior surface area of all zeolites were determined by
statistical sampling of >20 crystals. For MFI 50 and 120, the average crystal diameter estimated
from these images was subsequently confirmed using x-ray diffraction techniques. Due to the
variation in zeolite shapes from smaller spherical to larger prismatic crystals, the ratio of crystal
volume to external surface area (in units of nanometers) was used to define the characteristic
crystal dimension (Table 2-2) in this study.
Table 2-2. Synthesis conditions for the various MFI zeolites.
The composition refers to the molar ratio of TEOS:TPAOH:H2O used. The characteristic
dimension for each zeolite is the crystal volume to external surface area ratio and is given in
units of nanometers. MFI 12000 was procured from Exxon Mobil, Machelen, Belgium.
Zeolite
Composition
Conditions
Characteristic
Dimension (nm)
MFI 50
25 : 9 : 450
60 °C for 168 hours
7
MFI 120
25 : 9 : 450
80 °C for 196 hours
18
MFI 250
4 : 1 : 72
80 °C for 90 minutes,
180 °C for 30 minutes
40
MFI 1150
5 : 1 : 500
160 °C for 5 hours
175
MFI 1500
5 : 1 : 1000
160 °C for 5 hours
235
MFI 12000
n/a
n/a
1800
27
2.2.2.2
X-ray diffraction (XRD)
X-ray diffraction (PANalytical X’Pert Pro, Phillips) was performed using a Cu Kα target
and a nickel filter to collect the diffraction patterns (Figure 2-3B) and probe the crystallinity of
each zeolite sample. To confirm the sizes of MFI 50 and MFI 120, Scherrer’s equation (56) was
utilized and the full width at half maximum of the 7.94°, 14.77°, and 22.01° peaks were
obtained. The average diameter was then found by
𝑑=
𝐾𝜆
𝛽𝑐𝑜𝑠𝜃
(1)
where d is the crystal diameter, K is the crystal shape factor (i.e., K=1 in this case), λ is the
wavelength of the x-rays (for a Cu target, λ = 1.542 Å), β is the full width at half-maximum, and
θ is Bragg’s angle.
2.2.2.3
Water Sorption and Infiltration
To quantify the amount of water that adsorbs on the zeolites up to the saturation pressure
(3.14 kPa at 25 °C), physiosorption of water vapor was carried out at 25 °C using a gravimetric
vapor sorption analyzer (Q5000SA, TA Instruments). Before the tests, the samples were dried in
a furnace at 400 °C under air for at least 10 hours. The relative humidity was increased in steps
and the calculated adsorbed water was converted into the units of water molecules per unit cell of
zeolite.
A custom-built pressure vessel in a mechanical testing apparatus (Instron, 5582) was used
to examine additional water infiltration within the zeolite framework at significantly higher
pressures beyond the saturation pressure limit of the gravimetric vapor sorption analyzers (Figure
2-4B). The pressure vessel was made of type-304 stainless steel (McMaster-Carr) and used a
two-piston setup within a cylinder (see Figure 2-1). The vessel was sealed using polyurethane orings (90 Shore A, McMaster-Carr) and secured with glass-filled PTFE backup rings. 10 mL of
deionized water (class 2, VWR) and between 1 - 2.5 g of zeolite (the amount was varied to
ensure that the water capacity was independent of the zeolite amount and repeatable) were sealed
within the vessel. We machined the pressure vessel from 304 stainless steel-due to its high
28
strength, corrosion resistance and machinability. The inner diameter of the pressure was made to
be 0.502” and polished with burnishing tools. The pistons were also made from 304 stainless and
had grooves machined into them to hold two glass-filled PTFE backup rings and a urethane (90
Shore A hardness) o-ring. The other dimension of the o-ring was 0.504” so it had to compress
slightly to fit into the pressure vessel. This level of precision was required so that the vessel
would not leak under high (> 100 MPa) pressure. Polymer o-rings usually are not used for this
type of setup, however, we did not observe leaking during the experiments. O-rings and backup
rings were replaced after every experiment to significantly reduce the chance of o-ring failure.
The pressure vessel needed to be interfaced with an apparatus to apply a force and monitor
displacement. For our experiments, we used an Instron 5582 running the BlueHill 2 software.
The Instron is capable of applying loads of 100 kN (although we only needed 20 kN). To
determine the infiltration behavior into the pores, we used the ‘compliance correct’ feature on the
software. The compliance we corrected for was a control sample of 10 mL of water. Since every
experiment used 10 mL of water, we used to software to remove this displacement from the
experiments. Therefore, any additional displacement (due to water infiltrating into the zeolite
pores) would be recorded as the only displacement in the data. This removes time-consuming
post processing of the raw data and largely removes a source for any error in the data.
A"
B"
Figure 2-1. High-pressure infiltration experimental setup.
A) Schematic of the two-piston pressure vessel used for infiltration experiments. (B) Picture of
the Instron 5582 with pressure vessel loaded into the apparatus.
29
The vessel was compressed to a load of 20 kN, corresponding to a pressure of ≈160 MPa,
and the displacement data was corrected for water compressibility at these high pressures. A
volumetric compression rate of ≈254 mm3/min (corresponding to a displacement rate of 2
mm/min) was applied and the displacement and load were recorded. The compression rate was
varied to ensure that transient effects did not affect the behavior. After the completion of the
compression, the pressure was released and the vessel returned to the initial position confirming
the previous observations where water completely evacuated the pores. Consistency in the
experimental procedure and repeatability of the data was confirmed by running four to six
experiments each. The control displacement would be subtracted from the experimental, which
would therefore subtract the compliance of the mechanical test apparatus, the vessel, and the
zeolites. To convert to a number of water molecules, the net volume is multiplied by the density
of water at the corresponding pressure (found from NIST tables) and then divided by the mass of
a water molecule (3 x 10-23 g). The total capacity of the MFI zeolites was found by combining
the determined adsorption and infiltration capacities. To confirm this stipulation (i.e., if extra
water entered into the zeolite as it immersed into liquid water) an experimental setup depicted in
Figure 2-2 was utilized. Initially, a known mass of zeolites, enough to completely fill an
aluminum crucible with a known volume (Vtot, 160 µL) (DSC Al crucible, Mettler Toledo) was
dried at 400 °C under air and the total dry mass of the zeolites (Mzeolite) was recorded. Having
computed the dry mass of the zeolites, the intercrystalline volume (Vinter) was calculated by
subtracting the zeolite volume (Vzeolite,dry) from the total volume of the crucible (Vtot).
𝑉!"! − 𝑉!"#$%&",!"# = 𝑉!"#$%
; 𝑤ℎ𝑒𝑟𝑒 𝑉!"#$%&",!"# =
!!"#$%&"
!!"#$%&"
(2)
Next, the samples were exposed to a 98% RH environment in Q5000SA vapor sorption
analyzer and the measured increase in mass (Mwater,ads) was used to calculate the total adsorbed
water within the zeolites. The calculated adsorbed water per unit mass was confirmed to be in
agreement with the numbers reported in Figure 2-4A.
30
A"
A"
Zeolite'
Vzeolite'
Zeolite&
Vinter'
Aluminum&
Crucible&
OHAUS&Discovery&Mass&Balance&
B"
B"
Zeolite/Water&
mixture&
Zeolite'
Water'
Aluminum&
Crucible&
OHAUS&Discovery&Mass&Balance&
Figure 2-2. Schematic of experimental setup used to determine if additional water entered
into the zeolites after the adsorption experiments but prior to the infiltration experiments.
(A) Schematic before water was added, highlighting the available volume of the zeolite and
intercrystalline space. (B) Schematic after water was added, showing that the full intercrystalline
space was filled with water.
Since the effect of adsorption on the external surface area and capillary condensation in
the intercrystalline pores was negligible for these larger zeolites (Figure 2-4A), we assumed all
of the adsorbed water to be within the zeolite pores, not changing the previously computed
intercrystalline volume (Vinter). Next, enough liquid water mass required to completely fill the
intercrystalline volume was introduced via a pipette into the aluminum crucible containing the
water-saturated zeolites and the extra mass was recorded using an mass balance (Discovery,
OHAUS) (with a resolution of 10 µg). This procedure mimicked the process of introducing the
water saturated zeolites (at 98% RH) into the water bath of the pressure vessel used in the
infiltration measurements. The volume occupied by the additional water added via the pipette
31
(assuming a bulk density of 1 g/cm3) was found to match the previously computed
intercrystalline volume (Vinter), confirming no additional water infiltrated into the zeolite and
thus justifying the use of the last data point in Figure 2-4A as the starting point for Figure 2-4B.
2.2.2.4
Nitrogen Sorption
The micropore volume of the zeolites was probed by carrying out physiosorption of
nitrogen at 77 K (ASAP 2020, Micromeretics). The samples were dried and degassed for 5 hours
at 400 °C and at a pressure of 10 µm Hg prior to the tests. The adsorption isotherms are shown in
Figure 2-5A and B and the t-plot micropore volume (estimated by extrapolating the y-intercept
from a linear fit in the partial pressure regions of 0.5 – 0.7) is shown in Table 2-3 (Vf).
2.2.2.5
Nuclear Magnetic Resonance (NMR)
Nuclear Magnetic Resonance (29Si MAS NMR) spectra were recorded with a TMS
chemical shift standard (DSX-500, Bruker) with the spectra shown in Figure 2-6.
29
Si MAS
NMR spectra (11.4 T, ωL = 8 kHz) were acquired at 300 MHz with a recycle delay of 100 – 120
seconds with a 4 mm rotor. The spectra, shown in Figure 2-6, were recorded in the frequency
ranges of -97 to -107 ppm and -108 to -119 ppm to probe the quantity and quality of defect
groups (i.e., Q3 groups) and to probe the localized order of the silica sites (i.e., Q4 groups),
respectively.
2.3
Results and Discussion
Based on the methods described above, we first confirmed the morphology and the
crystallinity of the zeolite samples. The TEM and SEM micrographs for the six zeolites
synthesized/procured are shown in Figure 2-3A. The two smaller zeolites, i.e., MFI 50 and MFI
120, which do not have well-defined morphologies, were roughly approximated as spheres.
Lattice planes were detected in >95% of the crystals examined in the TEM analysis, indicating
that the majority of the samples had crystallized. MFI zeolites 250, 1150, 1500, and 12000 had
either disc or prismatic shapes that are typically associated with the MFI zeolite framework
synthesized using tetrapropylammonium ions as the structure-directing agent (57).
32
Figure 2-3. Electron microscopy (A) and x-ray diffraction (B) analysis of MFI zeolites.
TEM was used for the nano-sized MFI 50, 120 and 250 zeolites while SEM utilized for the
micron-sized MFI 1150, 1500 and 12000 zeolites. The XRD patterns for each zeolite sample
matched well with the monoclinic phase of MFI zeolites. The boxes for MFI 50 and 120
highlight the diffraction peaks that were broadened due to diffraction of the nanoparticles
themselves. These peaks were used to quantify the size of MFI 50 and 120.
The XRD patterns of the calcined samples are shown in Figure 2-3B. All of the samples
have well-defined peaks that correspond to the monoclinic phase of MFI zeolites (58). The lack
of a broad peak in the 20° - 25° range (associated with amorphous x-ray diffraction) confirmed
that the samples were crystalline2. The spectra of the two smaller zeolites, MFI 50 and MFI 120,
had typical peak broadening associated with nanoparticles (Figure 2-3B, boxed regions) (56).
The average radius of MFI 50 and MFI 120, calculated using Equation (1), was found to be 20
nm and 50 nm, respectively (assuming a spherical geometry), and in good agreement with TEM
images in Figure 2-3A.
2
Note that a small amount of localized distortions may be undetectable by standard XRD
techniques, particularly for nano-sized crystals (59,60).
33
2.3.1
Framework Water Capacity
2.3.1.1
Water Sorption and Infiltration Experiments
In this section, we discuss a novel strategy where we combined water adsorption
experiments (< 3 kPa) with high-pressure infiltration experiments (>1 MPa) to isolate the preadsorbed quantity of water and elucidate the actual framework capacity of MFI-type zeolites.
The water uptake results for all the zeolites are shown in Figure 2-4A. The three larger zeolites,
MFI 1150, 1500, and 12000, show similar water adsorption behavior that is indicative of lowuptake Type I isotherms (61), with ≈ 4 water molecules adsorbed per unit cell (N/UC) of zeolite
at P/Po = 0.98. Low-uptake Type I isotherms suggest a weak interaction between the adsorbent
(zeolite) and adsorbate (water), which has been previously reported for these purely siliceous
MFI zeolites (35,41,45). While MFI 250 also exhibits a low-uptake Type I isotherm during the
initial stages (partial pressure < 0.6, Figure 2-4A), the uptake rises substantially afterwards. This
behavior is more indicative of a Type II isotherm (61) and a total of 17 N/UC was estimated to
be adsorbed at a partial pressure of 0.98. MFI 120 and 50 show more typical behavior of Type
II isotherms with a total adsorption of 22 and 57 N/UC, respectively.
The total adsorption at P/Po = 0.98 was then used as the starting point for the highpressure infiltration experiments (shown by the arrows in Figure 2-4A and 4B). Note that
separate experiments (refer to 2.2.2.3 Water Sorption and Infiltration in the Experimental
Procedures section) were performed to confirm that the total water inside the zeolites at the end
of the adsorption experiments was the same as that during the start of high-pressure infiltration
experiments. Little to no water entered into the zeolites after the adsorption experiments until the
applied pressure exceeded ≈ 60 MPa. At this point, water infiltrated into the porous framework
and subsequently saturated the zeolites once the pressure approached ≈130 MPa.
34
A
B
Figure 2-4. Combined water adsorption and infiltration isotherms for MFI-type zeolites (A)
Water adsorption isotherms at 25°C for varying crystal sizes of MFI zeolites.
The ordinate corresponds to the amount of water adsorbed per unit cell of the zeolite and the
abscissa corresponds to the vapor pressure of water (the partial pressure is also plotted for
reference). The measurement error was less than the symbol size. The increase in adsorption at
higher pressure for the nano-sized zeolites is indicative of capillary condensation in the pores of
the crystal agglomerates. (B) Experimental pressure infiltration curves of water into the six
zeolites. The ordinate is the calculated amount of water entering per unit cell of the zeolite where
the starting point was shifted by the capacity at the end of adsorption experiments in A. The
abscissa is the experimentally measured applied pressure. The error associated with displacement
was ± 2 N/UC. The arrows help connect the starting point for each plot, which corresponds to the
infiltrated water at ≈ 98% RH (P/Po = 0.98) from the adsorption experiments. The final measured
capacity for MFI 50, 120 and 250 overestimated the internal capacity due to the effects of the
textural porosity. This surface effect was negligible for MFI 1150, 1500 and 12000, and
therefore these measurements indicated the correct total internal water capacity associated with
A
B
MFI zeolites.
This high-pressure infiltration behavior of water into MFI-type zeolites has been
previously reported (38,53,54,62,63) and attributed to the decrease in both the number of
available water – water hydrogen bonds within the pore network (from ≈ 5 – 6 outside the zeolite
to ≈ 2 within the pores) and to a lack of zeolite – water hydrogen bonding sites within the zeolite
(38,64-67)3. The total capacity per unit cell (Ntot) at 160 MPa of applied pressure for all zeolites
is reported in Table 2-3. MFI 1150, 1500, and 12000 have similar capacities of ≈ 35 ± 2 N/UC,
3
It should also be noted that the water fully evacuated the pores as the pressure was released,
in agreement with previous studies (38,53,62,66).
35
which substantially increased as the crystal size decreased, with the maximum of ≈ 75 N/UC
observed for MFI 50. These results also suggest that a majority of the water infiltration for
smaller zeolites (MFI 50, 120 and 250) occurred in the low-pressure (≈ 2-3 kPa) adsorption
regime. The percentage infiltration during the high pressure increased with increasing crystal
dimension.
Table 2-3. Calculated micropore volume (Vf) and estimated internal water capacity (Nf) of
MFI zeolites.
Vtot and Ntot are the total adsorption/infiltration quantities determined at P/Po = 0.99 for nitrogen
sorption and P = 160 MPa for water infiltration, respectively. The ratio (R) is of the total
nitrogen adsorption (combined framework and textural) to the framework nitrogen adsorption,
which was then used to estimate the internal water capacity. Note the significant decrease in
internal water capacity for MFI 50 and 120, which is attributed to the incomplete crystallization
of incorporated primary units. MFI 250, 1150, 1500 and 12000 have comparable values for the
internal capacity, and are within the uncertainty of the infiltration experiments.
Nitrogen Sorption
Zeolite
Water Adsorption + Infiltration
Vf
(cm3/g)
Vtot
(cm3/g)
R
Ntot
(N/UC)
Nf = Ntot/R
(N/UC)
% decrease
in Nf
MFI 50
0.136
0.613
4.51
75
17
≈50%
MFI 120
0.137
0.316
2.31
43
19
≈45%
MFI 250
0.178
0.222
1.25
43
35
≈0%
MFI 1150
0.173
0.178
1.03
34
33
≈0%
MFI 1500
0.185
0.193
1.04
35
34
≈0%
MFI 12000
0.177
0.179
1.01
35
35
≈0%
The results from these combined adsorption and infiltration experiments show a wide
variation in both the water capacity and infiltration pressure of MFI-type zeolites, similar to what
has been reported in literature (Table 2-1). However, both the textural porosity and localized
distortions within the zeolite microstructure were found to significantly alter the framework
capacity and misrepresent the actual infiltration behavior (as will be explained in following
section).
36
AB
B
AA
2.3.1.2
B
Micropore Volume
The combined water adsorption and infiltration experiments demonstrated an increase in
water capacity as the crystal dimensions decreased, suggesting apparent crystal size dependence
A
of the water capacity of the zeolite network. This result was initially surprising since the
B
different zeolites crystals used in this study were, in principle, composed of repeating threedimensional unit cells, indicating that the framework water capacity (normalized with mass or
per unit cell) should remain constant.
A
B B
AA
B
A
Figure 2-5. (A) Nitrogen adsorption isotherms with the ordinate corresponding to the
calculated pore volume and the abscissa corresponding to the partial pressure of nitrogen
at 77 K.
The t-plot pore volume (Table 2-3) was estimated by extrapolating the y-intercept from a linear
fit in the partial pressure regions of 0.5 – 0.7 (B) Magnified view of the black dotted box from
(A) with the dashed lines showing the linear fit. There was a ≈25% decrease in the micropore
volume of MFI 50 and 120 compared to the larger MFI 250, 1150, 1500 and 12000 samples.
In order to explain this discrepancy, we performed nitrogen sorption experiments to
independently estimate the available micropore (framework) volume of the zeolites. It is wellknown that nitrogen vapor completely saturates the framework micropores of these zeolites at
low relative pressures (P/Po < 0.5) (61). As a result, any additional adsorption beyond P/Po of 0.5
is indicative of both adsorption at the external crystal surface and in the available volume
between crystals, both of which are associated with the textural porosity of the zeolite
agglomerates (68,69). The micropore volume, calculated using the t-plot method (see 2.2.2.4
Nitrogen Sorption in the experimental section for details) for the larger MFI samples (250, 1150,
37
B
1500 and 12000) was found to ≈0.18 cm3/g (Figure 2-5B and Table 2-3), which is in good
agreement with previously reported values in literature (61). However, a reduction of
approximately 25% (Figure 2-5B and Table 2-3, Vf) in the framework volume was observed for
the smaller sub-100 nm MFI zeolites (MFI 50 and 120). Furthermore, adsorption associated with
the textural porosity considerably increased the amount of adsorbed nitrogen (Vtot in Table 2-3)
for the nano-sized zeolite crystals (MFI 50, 120 and 250). These results suggest that the same
textural porosity may have artificially increased the total water capacity as shown in Figure
2-4A. Assuming that the ratio of the total capacity (combined framework and textural) to the
framework capacity was the same for both nitrogen and water4, the nitrogen sorption results were
used to estimate the actual value of framework water capacity (Table 2-3). The analysis
demonstrated that the role of the textural porosity on the water capacity was negligible and the
framework water capacity was 35 ± 2 N/UC for the three larger zeolites (MFI 1150, 1500 and
12000). This result is expected considering the fact that the internal framework surface area was
at least three orders of magnitudes larger than the estimated outer surface area. Interestingly,
zeolite MFI 250 which had a total apparent water capacity of 43 N/UC, was also found to have a
framework water capacity of 35 ± 2 N/UC. A framework capacity of 35 ± 2 N/UC for the four
larger zeolites is in good agreement with the previously reported experimental results of Trzpit et
al. (53) and Cailliez et al. (38). However, most of the simulation-based results still overestimate
the framework capacity by ≈20 – 40% (38,51,53,63), suggesting that the molecular interaction
parameters currently used require further tuning.
Conversely, the analysis (Table 2-3) also indicated that the MFI 50 and 120 zeolites,
despite having an apparent higher total capacity for both water and nitrogen compared to the
larger MFI counterparts, actually exhibited a decrease in the framework water capacity of ≈45%
to ≈50%. While this decrease in framework capacity of MFI zeolites with crystal diameters of
100 nm or less (which is more commonly investigated with nitrogen sorption) is in qualitative
agreement with those found in literature (Table 2-4) (53,61,70-75) and the observed decrease in
the framework volume estimated via the nitrogen sorption presented earlier, the underlying
mechanism for the reduction in framework volume/capacity is still unknown (61,72,76).
4
The total capacity was determined at a partial pressure of 0.99 for the nitrogen adsorption
experiments and at 160 MPa of applied pressure for the water adsorption/infiltration experiments
38
Table 2-4. Summary of past work highlighting reduction in measured micropore volume
for nano-sized MFI-type zeolites.
Note that the typical range for the micropore volume for crystalline MFI-type zeolites
determined with nitrogen sorption is between 0.18 and 0.2 cm3/g.
Source
Crystal Size
(nm)
Pore Volume
(cm3/g)
Kim et al. (71)
13
0.081
Kim et al. (71)
22
0.090
Hsu et al. (72)
40
0.110
Kim et al. (71)
42
0.138
Babeva et al. (73)
70
0.120
Majano et al. (74)
80
0.130
Kim et al. (71)
90
0.115
Aguado et al. (75)
10-100
0.140
Trzpit et al. (53)
20000
0.185
Zhang et al. (70)
70000
0.196
Kenny and Sing (61)
not given
0.190
We hypothesize this decrease in the normalized framework volume to be indicative of a
modification of the internal zeolite pore structure. While a long-range disordering of the
microstructure was not observed as confirmed by XRD results in Figure 2-3B, localized
distortion of the zeolite framework within the nanocrystals (known to be difficult to detect by
XRD analysis (60,77)) may have been present.
39
2.3.1.3
29
Si MAS NMR spectroscopy
B
Figure 2-6. 29Si MAS NMR spectra for varying crystal sizes of MFI zeolites.
Note that the curves have been arbitrarily shifted in the y direction for clarity. The decrease in
the number of peaks as well as the peak sharpness seen in the spectra of MFI 50 and 120
indicates an increase in localized disorder, which was a result of incomplete crystallization of
amorphous primary units during synthesis. The presence of the amorphous material explains the
decrease in the measured micropore volume.
To investigate the presence of localized disorders (which were undetected by XRD) in
the zeolite structure, we performed a more detailed investigation using
29
Si MAS NMR, a
technique that is better suited to investigate the structure of zeolites and probe the existence of
the non-crystalline material (59,60,77). The structure and crystallinity of the zeolites were
estimated by examining the spectra of the Q4 (Si-[(OSi)4]) groups that occurred in zeolites and
other silica-based materials. Typically, well-crystallized MFI zeolites exhibit between 9 and 16
sharp peaks in the range of -108 to -118 ppm TMS chemical shift (53). A decrease in number or
sharpness of the peaks indicates an increase in the defect density or a reduction in crystallinity.
40
For example, silica gels exhibit 1 – 3 broad peaks in the same frequency range, which is
indicative of a disordered structure (59). Figure 2-6 shows the NMR spectra collected for all
samples. MFI 250, 1150, 1500 and 12000 had between 9 and 16 peaks and appeared to be wellcrystallized. However, MFI 50 and 120 had fewer than 6 peaks and a significant lack of peak
sharpness, suggesting a loss in crystallinity. These results, when analyzed in conjunction with
nitrogen sorption (Table 2-3) and water adsorption/infiltration (Figure 2-4 and Table 2-3)
experiments, more clearly suggest that the decrease in the available framework capacity of the
sub 100 nanometer diameter MFI 50 and 120 zeolites was indeed due to the incomplete
crystallization of silica primary units during synthesis.
The incomplete crystallization of MFI primary units during synthesis can be explained by
the growth mechanisms proposed by de Moor et al. (78). They proposed that the primary units,
which are sub 5 nm diameter amorphous silica particles, were absorbed into growing MFI zeolite
crystals and subsequently crystallized into the zeolite structure during synthesis. However, due to
the low crystallization temperatures (≤ 80°C) and slow crystal growth rates (≈1 nm per hour) of
both MFI 50 and 120, it is plausible that some of these units did not completely crystallize during
the synthesis process(77). Furthermore, recent work has shown that if the nanoparticle synthesis
occurs at temperatures of at least 170 °C, the full micropore volume (≈ 0.18 cm3/g) can be
recovered (54,79).
41
2.3.2
Water Infiltration Pressure
A
B
A
B
A
B
Figure 2-7. Defect density effects on the infiltration pressure and low-pressure water
uptake.
(A) The change in infiltrated amount of water into the zeolites as a function of applied pressure.
The infiltration pressure corresponds to the maximum for each curve. Note that the infiltration
pressure for all zeolites studied here was between 95 – 100 MPa. (B) A magnified view of the
low partial pressure water uptake isotherms highlighting the difference in the low uptake data as
well as the approximate linear slopes (dashed lines) obtained to calculate the defect density.
The experimental results in Figure 2-4 showing significant water infiltration at low
pressures (< 3 kPa), as demonstrated in the previous section, was attributed to the textural
porosity and accordingly, did not correspond to the internal framework infiltration (Table 2-3).
These results imply that, at most, only ≈ 15% of the framework volume was occupied with water
at the end of the adsorption experiments and the majority of the available framework volume was
filled at high pressures (>60 MPa). The value of infiltration pressure was inferred from the
maxima of the plot of differential water capacity (ΔN/UC) and applied pressure. As seen in
Figure 2-7A, the infiltration pressure for all the zeolites was approximately the same and
occurred between 95 – 100 MPa, which is in agreement with some previous studies
(38,53,62,63).
42
These results suggest that the crystallized internal framework of all the zeolites was
approximately the same, with little to no difference in the internal surface chemistry and defect
density. To further probe the internal surface chemistry, the low-pressure water adsorption
isotherms were analyzed to quantify the defect density. The defects, which are in the form of
silanol groups (41), are proposed to act as sites for water adsorption, thereby increasing the
localized zeolite affinity for water. By analyzing the low (< 0.2) partial pressure water adsorption
behavior (Figure 2-7B), the defect density for each zeolite was approximated, as shown in Table
2-5 (45).
Table 2-5. Calculated defect densities of MFI zeolites obtained by extrapolating the
intercept from the water adsorption isotherms (see Figure 2-7B).
MFI 50 and 120 exhibited slightly higher defect densities, as only 0.5% of the total available
silicon sites with the unit cell were defective.
Zeolite
Defects/UC
MFI 50
0.448
MFI 120
0.450
MFI 250
0.063
MFI 1150
0.043
MFI 1500
0.048
MFI 12000
0.078
Following the procedure of Olson et al. (45), the defect density was approximated by
extrapolating a linear fit in the low partial pressure ranges, highlighted in Figure 2-7B, and
finding the y-axis intercept (which correlates to a number of water molecules per unit cell)5. The
defect density was calculated by assuming that each defect adsorbs ~ 4 water molecules at low
partial pressures (therefore, the defect density is ¼ of the number of adsorbed molecules at
5
the
For the micropore volume, a similar extrapolation procedure was used. As highlighted in
Figure 2-5B,
a linear fit in the partial pressure range of 0.5 – 0.7 in the nitrogen sorption data
was applied with the corresponding y-intercept as the available micropore volume.
43
hypothetic vacuum). The quantified defect density is in Table 2-5. While the defect density for
MFI 50 and 120 (which was estimated to be ≈ 0.45 defects per unit cell), was 6 – 10 times that of
the larger zeolites, this estimated defect density indicates that, at most, only ≈0.5% of the silicon
sites of each unit cell were defective for these nano-sized zeolites6. Previously, Trzpit et al.
reported a reduction in the water infiltration pressure when defects were introduced into a
previously perfect MFI zeolite, which contradicts the results presented here (53). However, the
defect densities studied in that work were higher (≈ 1 defect per unit cell), which could indicate
that a threshold defect density needs to be attained before a noticeable decrease in the infiltration
pressure can be observed. It should be noted that a rounding of the high-pressure infiltration
curve was observed for the nanosized MFI 50 and 120 zeolites, which is in qualitative agreement
with the results of Trzpit et al. in that an increasing defect density reduces the pronounced onset
of the infiltration.
2.4
Conclusions
The effects of crystal size, morphology and synthesis procedure on the framework
capacity and the associated micropore volume of purely siliceous MFI zeolites was
systematically investigated using SEM, TEM, NMR, XRD, nitrogen sorption, water uptake and
infiltration experiments. We demonstrated that the total internal framework water capacity of
fully crystallized MFI zeolites is 35 ± 2 water molecules per unit cell. In principle, no size effect
on the total water capacity of the zeolites was inferred. However, an apparent decrease of up to
50% in internal framework water capacity and up to 25% in micropore volume for sub 100 nm
diameter MFI zeolite crystals was found. Analysis of
29
Si MAS NMR spectra of the zeolites
indicated this decrease in internal water capacity and micropore volume was due to an increase
in localized disorder from the incomplete crystallization of MFI primary units during synthesis.
The study also highlights the practical subtleties where the nano-sized MFI zeolites appear to
have a larger water capacity (≫ 35 water molecules per unit cell) due to water adsorption at the
external surface and in inter-crystalline pores of the zeolite structure. Thus, in our later
experiments, we utilized zeolites with an internal to external surface area ratio of at least 1000 to
6
No apparent peak associated with Q3 (or silanol groups) in the -100 ppm range existed in
the recorded NMR spectra (Figure 2-7), confirming that the quantity of the silanol defect density
was relatively small (< 0.5% of the total sites available).
44
more accurately probe the internal pore network of the zeolites. The experimental quantification
of the internal water capacity and the defect density as well as insight on the water infiltration
pressure provided in this work can be utilized to help validate and improve upon the existing
water and zeolite interaction models, as the current models still tend to overestimate the
framework water capacity of MFI zeolites. Such advancements will allow better understanding
of the transport mechanisms within the MFI zeolite pores, which can then be extended to other
nanoscale materials as well.
The combined results from these studies show that the purely siliceous MFI zeolites
exhibit a ‘hydrophobic’ behavior where over 95 MPa is required to saturate the internal
framework with water. These pressures are well in excess of the normal operating conditions
associated with typical membrane operating pressures (for example, sea water reverse osmosis
systems operate at pressures of 5.5 – 6.5 MPa (4)), which may explain the low water
permeability of current zeolite-based membranes compared to polymeric-based membranes (14).
In order to address this challenge, in the next chapter, we artificially increased the internal defect
density by varying the silicon to aluminum ratio to lower the water infiltration pressure (45,53)
and investigated the effects of the defect density on water transport.
45
Chapter 3 3. Effect of Hydrophillic Defects on Water Transport in MFI Zeolites
3.1
Overview
The physical and transport properties of water are usually altered due to the increased
effect of solid-liquid molecular interactions when water is confined at nanometer length scales
(3,16,80-83). Depending on the strength of interaction between a water molecule and the surface,
the local interface can either be defined as hydrophobic (i.e., the ratio of the water-water
interaction energy to water-solid interaction energy is high) or hydrophilic (i.e., the ratio of the
water-water interaction energy to the water-solid interaction energy is low)7. While a hydrophilic
surface is by definition ‘water loving’ and has long been expected to facilitate better water
transport (30,32,85,86), some recent studies have shown the opposite behavior where a
significant increase in the flow rate of water confined within a nanometer-sized hydrophobic
pore (e.g., a carbon nanotube) was observed (19,24,87). Such an interface could potentially be
exploited to improve the performance of membranes used in seawater reverse osmosis
desalination and other water-based separation applications. However, additional research is
needed to improve our understanding of the physical mechanism behind the water transport
within a generic hydrophobic pore.
Zeolites, a microporous (< 2 nm pores) crystalline material, offer a model structure to
experimentally investigate the properties of nanoconfined water. The ability to independently
vary the local internal defect density, and hence the surface-water molecular interactions, without
7
It should be noted that this hydrophobic definition does not imply that the surface repels
water as in the classical thermodynamic definition, but only that the water-water attraction is
stronger than the water-solid attraction. There is always an attractive force between two entities
due to van der Waals interactions, unless the separation is less than the hard sphere diameter (σ)
of the molecule (84).
46
affecting the pore size (25) provides an ideal platform to investigate the underlying physical
mechanism behind the transport at these length scales.
Numerous previous studies have shown that changes in the internal defect density (due to
variations in composition and synthesis procedures) have a significant effect on water transport
within zeolites (21,48,53,62). Purely siliceous or high silicon to aluminum ratio (Si/Al >1000)
MFI (Mobil Five) zeolites require pressures in excess of 100 MPa to be applied to completely
saturate the porous network with water (26,38,53,66,88,89).This large ‘infiltration pressure’ is
due to the lack of hydrogen-binding sites within the zeolite, thus creating both enthalpic and
entropic barriers for entry (64,88). However, when a large amount of acidic (i.e., hydrophilic)
defects are introduced within the zeolite, such as with FAU (Faujasite) or LTA (Linde Type A)
zeolites, water can completely saturate the porous network at pressures of 1 kPa or less (90,91).
The low-pressure pore filling is caused by the large attraction energy of water to these acidic
defects, the magnitude of which even surpasses the heat of sublimation for water, thereby
overcoming any entry barriers into the sub-nanometer pores (45,70,92,93). It is due to these
competing interactions that the mechanism of water transport within these nanoscale pores is not
well understood and the design of an optimal interface to enhance water transport within these
pores is complicated.
For transport processes that are limited by diffusion, the permeability of the fluid through
the material is based on the product of the diffusivity and the solubility (i.e., sorption
coefficient), indicating that both the rate of transport and the amount of infiltrated water are
important in controlling the permeability (94). While it is well known that increasing the defect
density increases the solubility of the zeolite (35,41,45), both molecular dynamics simulations
and nuclear magnetic resonance experiments demonstrated that the inclusion of defects decreases
the water diffusivity (47,48,95). These trends suggest that the effect of the defect density on the
interplay between the solubility and diffusivity, and ultimately the permeability, is complex. To
better understand the effect of the defect density on the overall water permeability and flux,
increased control over experimental techniques and analysis is needed to systematically
investigate the competing parametric effects of the diffusivity and the solubility within zeolite
crystals.
47
In this chapter, we experimentally investigated the role of the concentration of
hydrophilic defects in MFI (i.e., ZSM-5) zeolites on the permeability of water by independently
quantifying both the solubility and diffusivity. We synthesized MFI zeolites with a
compositional silicon/aluminum (Si/Al) ratio varying from 100 to infinite (i.e., Silicalite-1) and
confirmed the structure and morphology using x-ray diffraction (XRD) and scanning electron
microscopy (SEM). Through combined sorption and high-pressure infiltration experiments, we
investigated the solubility of water within the zeolite pores as a function of pressure and
chemical potential. The diffusivity, and subsequently the permeability, of water within the pores
was evaluated by analyzing the transient adsorption and desorption behavior. For a given
pressure, we found that the amount of water that infiltrated into the zeolite porous network
increased as the internal defect density increased. However, none of the zeolites studied were
found to be completely filled at the saturation pressure of water (3.14 kPa at 298 K), and each
zeolite sample required upwards of 40 MPa to reach the measured framework capacity of 35
water molecules per unit cell (N/UC) (26). The increasing defect density decreased the measured
diffusivity by up to two orders of magnitude and the permeability by upwards of an order of
magnitude compared to water within the near defect-free Silicalite-1 MFI zeolite. The results
from these experiments highlight the strong attraction of water to the hydrophilic defect sites
within the zeolite, which, although increased the solubility of the zeolites, had a pronounced
detrimental effect on the diffusivity and consequently on the permeability. The experimental
results from this study can be utilized to provide detailed physical insights into the transport
mechanisms, which can then help guide the design of high permeability membranes in various
water-based separation applications.
3.2
Experimental Materials and Methods
Five different MFI zeolites were synthesized for this study with a varying Si/Al ratio
(Figure 3-1A). For all but the purely siliceous MFI zeolite (which does not contain aluminum),
aluminum nitrate was used as the aluminum source (ACS Reagent >98%, Sigma Aldrich). First
(if needed), the appropriate amount of aluminum nitrate was dissolved in tetrapropylammonium
(TPA) hydroxide (TPAOH, 1M in H2O, Sigma Aldrich). Once the solution became clear after
agitation, the proper amounts of tetraethyl orthosilicate (TEOS, >98%, Sigma Aldrich) and
deionized (DI) water (Class 2, VWR) were added to the solution, after which the solution was
48
left to age overnight under agitation to fully hydrolyze the TEOS. The molar concentration of the
solutions was as follows, where X is the appropriate molar concentration of aluminum nitrate for
each respective zeolite.
35:10:X:1750 – TEOS:TPAOH:AlNit:H2O
The solution was then transferred into a PTFE-lined stainless steel autoclave (45 mL,
Parr, Inc.), placed into a preheated furnace at 175 °C (BlueM, Lindberg), and heated under
rotation for 5 hours. After synthesis, the resulting solution was centrifuged, decanted, and
washed with DI water at least three times to lower the pH of the solution to ≈ 9. The crystals
were then dried at 60 °C and calcined at 550 °C (with 3 °C/min ramp rates) to remove the
organic template.
MFI$INF$
MFI$1000$
MFI$300$
MFI$200$
MFI$100$
A
B
Figure 3-1. SEM (A) and XRD (B) analysis of MFI zeolites.
Although there is some variation in size amongst the different Si/Al ratio zeolites, the zeolites all
have an internal to external surface area ratio greater than 1000 (see Table 3-1) such that the
sorption/infiltration analysis was limited to the internal pore structure. The scale bar for all of the
SEM images is 5 µm. The XRD patterns for each zeolite sample matched well with that of
known MFI zeolites.
The zeolites were imaged using SEM (Ultra-55, Zeiss) (Figure 3-1A). For SEM analysis, the
samples were coated with ≈ 5 nm of Pt/Pd to prevent charging during imaging. The volume and
49
exterior surface area of all zeolites was determined by statistical sampling of >20 crystals and are
provided in Table 3-1. XRD (PANalytical X’Pert Pro, Phillips) was performed using a Cu Kα
target and a nickel filter to collect the diffraction patterns (Figure 3-1B) and was used to confirm
the MFI structure of the synthesized zeolites.
Table 3-1. Volume, surface area, and area ratio of MFI zeolites synthesized in this study.
Note that the surface area ratio for all samples is greater than 1000, thereby limiting the
experimental analysis to the internal pore volume.
Zeolite
INF
1000
300
200
100
Surface Area (µm) 2 Volume (µm) 3 Int/Ext Area Ratio
33.25
8.25
1618
22.25
5.25
1539
18
4.5
1631
17.2
4.2
1593
10
2
1305
To quantify the amount of water that adsorbed in the zeolites up to the saturation pressure
(3.14 kPa at 25 °C), physiosorption of water vapor was carried out at 25 °C using a gravimetric
vapor sorption analyzer (Q5000SA, TA Instruments). Before the tests, the samples were dried in
a furnace at 500 °C under air for at least 10 hours. The calculated adsorbed water was converted
into units of water molecules per unit cell of zeolite. Following the sorption experiments, the
zeolites were immersed within 10 mL of DI water and the solution was introduced into a custombuilt pressure vessel. Specific details on the pressure vessel can be found in the previous chapter
or in our report(26). A mechanical testing apparatus (5582, Instron) was used to examine
additional water infiltration within the zeolite framework by compressing the solution to a
significantly high pressure beyond the saturation pressure limit of the gravimetric vapor sorption
analyzer. A volumetric compression rate of ≈ 254 mm3/min (corresponding to a displacement
rate of 2 mm/min) was applied and the displacement and load were recorded. The compression
rate was varied to ensure that transient effects did not affect the behavior. The vessel was
compressed to a load of 20 kN corresponding to a pressure of ≈ 150 MPa and the displacement
data was corrected for water compressibility at these high pressure. The water capacity was
calculated by equating the displaced volume to an equivalent amount of water molecules
entering per unit cell of the zeolites.
50
3.3
Results and Discussion
We first confirmed the morphology and crystallinity of the zeolite samples using SEM
and XRD, respectively. The SEM images for the five synthesized zeolite samples are shown in
Figure 3-1A. The typical prismatic morphology associated with MFI-type zeolites synthesized
using TPA ions as the structure directing agent was observed for all zeolites (57). In addition, the
size of the crystal was controlled such that each zeolite crystal had an internal to external surface
area ratio of at least 1000 so that the role of the textural porosity (external surface) on the water
sorption analysis could be neglected (26). To make a fair comparison in the sorption behavior
between the different zeolites, significant effort was made to approximately maintain the same
crystal volume (± 2.5 𝜇𝑚! ) amongst all zeolite samples (45). The XRD patterns of the calcined
samples are shown in Figure 3-1B. All of the samples had well-defined peaks that correspond to
the known MFI zeolite diffraction pattern.
3.3.1
Sorption and Infiltration Behavior
Using the combined sorption and high-pressure infiltration experimental approach introduced
in our previous work (26), we investigated and quantified the effect of the defect density on the
pressure at which water entered into the zeolite pores (Figure 3-2). The uptake behavior
(recorded up to a relative pressure of 0.98 at 298 K) in Figure 3-2 showed an increase in total
water uptake as the Si/Al ratio decreased. As seen in Table 3-2, at near saturation conditions, the
purely siliceous MFI INF zeolite adsorbed only ≈ 4 water molecules per unit cell (N/UC), while
the most defective MFI 100 zeolite adsorbed ≈ 27 N/UC, indicating ≈ 7× increase in the sorption
capacity. It should be noted that since the overall capacity of the zeolites has been estimated to
be 35 N/UC (26,38,53), none of these zeolites were completely filled with water at the saturation
pressure. The low-pressure adsorption behavior provided information to better estimate the
internal defect density of the zeolites, since the compositional Si/Al ratio does not necessarily
reflect the actual defect density due to synthesis uncertainties (45). The defect density was
estimated via a low relative pressure linear extrapolation of the adsorption quantity (N/UC in
Figure 3-2) to zero relative pressure. This adsorption quantity at ‘zero’ pressure corresponds to
water that is specifically adsorbed to the defect sites, thereby giving a better approximation of the
defect density than the composition alone (26,45). Following the procedure of Olson et al., the
51
defect density was approximated by extrapolating a linear fit in the low relative pressure range in
Figure S5, and obtaining the intercept of the abscissa (which correlated to a number of water
molecules per unit cell). The defect density was calculated by assuming that each defect adsorbs
≈ 4 water molecules at zero relative pressure (i.e., the defect density is ¼ of the number of
adsorbed molecules at ‘zero’ pressure). The estimated defect density is provided in Table 3-2.
Figure 3-2. Combined adsorption and infiltration isotherms for varying Si/Al ratio MFI
zeolites.
Symbols indicate recorded data while the dashed lines are provided to guide the eye. The
pressure, P, is normalized by the saturation pressure at 298 K (3.14 kPa), Po. At lower relative
pressures, the water uptake increased as the Si/Al ratio decreased. None of the zeolites were
filled at the saturation pressure and all zeolites studied exhibited some amount of high-pressure
infiltration. The error associated with the water uptake is ≈ 0.5 N/UC, while the error associated
with the pressure infiltration is ≈ ± 2 N/UC.
Following the sorption experiments, the zeolites were immersed into water and placed within
the pressure vessel for high-pressure testing. The total adsorption amount at a relative pressure of
52
0.98 for each zeolite was used as the starting point for the high-pressure experiments8. For all
zeolites, water saturated the remaining pore network between the pressures of 40 and 120 MPa.
This high-pressure infiltration behavior of the hydrophobic MFI pores was first reported by
Eroshenko et al. (62) and was later confirmed by a number of experimental and simulation-based
studies including our recent work (26,38,53,63,64,66,88,96-98).
Table 3-2. Experimental results from the sorption/infiltration analysis.
The defect density was estimated by the low-pressure linear extrapolation method previously
used by Olson et al. (45)
Zeolite
MFI 100
MFI 200
MFI 300
MFI 1000
MFI INF
Adsorption @ P/PO = 0.98 Total Infiltration @ P = 150 MPa Estimated Defect Density
(N/UC)
(N/UC)
(N/UC)
26.9
18.5
14.9
6.9
3.8
36.5
36.6
35.4
34.3
33.5
1.33
1.14
0.97
0.42
0.1
The mechanism of water infiltration into the hydrophobic pores (and subsequent pore
evacuation upon pressure release) has been attributed to the decrease in both the number of
available water – water hydrogen bonds within the sub-nanometer pore network (from ≈ 5 – 6
outside the zeolite to ≈ 2 within the pores) and to a lack of zeolite – water hydrogen bonding
sites within the zeolite. As the defect density increased, the sharp onset that is associated with the
high-pressure water infiltration significantly decreased and, ultimately, the infiltration followed a
linear trend as a function of pressure. It should be noted that, for all zeolites studied in this work,
8
In our previous report (26), we show experimental results that confirm that the total water
inside the zeolites at the end of the adsorption experiments was the same as that during the start
of the high-pressure infiltration experiments.
53
the high-pressure infiltration was repeatable9 and a total number of ≈ 35 ± 2 N/UC entered the
zeolite framework up to a pressure of 150 MPa, which corroborates our previous work (26).
A
B
Figure 3-3. Comparison of the experimental data with Cailliez’s (A) weaker MFI zeolite
defect model and (B) stronger defect model.
The solid points correspond with the experimental data from this work while the solid lines
represent the modeling results. Note that the vapor to liquid phase transformation (saturation
pressure) occurs at a chemical potential of -0.1 kJ/mol and is indicated by the black dashed line
in each plot. Magnified plots highlighting the vapor to liquid transition region are shown in
Figure 3-4. The stronger defect model showed a better agreement with the experimental data
since increasing the defect density not only decreased the infiltration pressure but also
substantially increased the amount of adsorbed water prior to the saturation pressure. However,
the strength of the modeled defects was underestimated as the experimental defects contain an
aluminum cation, which provides an additional attractive force compared to the modeled silanol
groups.
To better understand the mechanistic effect of an increasing defect density on the
infiltration behavior, we compared the experimental data with the defective zeolite modeling
work of Cailliez et al.(38). Cailliez et al. probed the effect of the prescribed partial charge of
silanol defect groups within MFI zeolites on the water infiltration behavior. Although Cailliez et
9
As a further clarification, if the pressure was released, the piston-cylinder apparatus
returned to its initial displacement. As multiple runs were performed for each sample, the
recorded load-displacement data would follow along the same curve (within the measurement
error of the testing apparatus)
54
al. investigated a different type of acidic defect group within MFI zeolites (silanol groups rather
than the Al-substituted defects in this work), the qualitative comparison with the modeling is
useful in probing the effect of the quantity of defects on the infiltration pressure and
investigating the relative strength of the attraction of water to the Al-substituted defects.
Cailliez et al. reported that a higher assigned partial charge than is typically prescribed
(+0.65 atomic units (au) on the hydrogen atom versus +0.35 au on the hydrogen atom) better
matched their own experimental infiltration data, which indicated that the attraction of water to
the silanol defects was stronger than previously estimated values(52). Additionally, using this
larger prescribed partial charge, it was predicted that the full infiltration isotherm (i.e., an
isotherm from 0 Pa to hundreds of MPa) would transform from a Type V to a Type IV (61) as
the defect density increased (38). For an appropriate comparison with the experiments reported
in this work, the relative pressure (abscissa in Figure 3-2) was converted to a chemical potential
using the NIST properties database for water (99) (using the assumption that the temperature
remained at 298 K throughout the experiments (67)).
Figure 3-4. Magnified view of the comparison of the experimental data with Cailliez’s weak
defect model (A) and strong defect model (B).
Note that the vapor to liquid phase transformation occurs at a chemical potential of -0.1 kJ/mol.
Both the strong (+0.65 au) and weak (+0.35 au) defect models are compared with our
experimental results and are shown in Figure 3-3A and Figure 3-3B, respectively. The solid
points in both figures represent the experimental data while the solid lines correspond to the
55
respective defect models. Magnified plots of the vapor to liquid transition region are also
provided in Figure 3-4. For the simulated weak defect model (Figure 3-3A), nearly all of the pore
filling was estimated to occur over a small range of pressure (this behavior is more commonly
referred to as ‘pore condensation’ and is classified by a Type-V isotherm (38,61)), with a sharp
onset of infiltration with respect to the chemical potential. However, this type of pore filling was
not observed for the experimental data, as all of the zeolites studied exhibited a combination of
low-pressure adsorption and high-pressure infiltration (which is classified by a Type-IV
isotherm). In contrast, the simulated strong defect model (Figure 3-3B) showed this mix of
adsorption and high-pressure infiltration, particularly for the 1 D/UC (defect per unit cell) and
4 D/UC cases. A complete filling of the pores below the saturation pressure (as shown for the
modeled 12 strong defects per unit cell) was not experimentally observed, however, constraints
for the crystal morphology and size limited the defect density of this study10.
Nonetheless, even for the strong defect model, the simulated defect density
underestimated the strength of the Al-substituted defects investigated in this work, i.e., the
experimentally calculated defect density of 1 D/UC exhibited nearly the same behavior as the
simulated 4 D/UC. These results highlight the significant increase in the strength of attraction of
water to these defect groups compared to silanol groups, which is hypothesized to be a result of
the substituted aluminum cation providing an additional attractive force for the water to the
defect groups compared to the silanol defects (100). Furthermore, these experiments show that
even a small concentration of defects (≈ 1 D/UC) can significantly alter the sorption/infiltration
characteristics and increase the sorption capacity of MFI-type zeolites compared to a near defectfree Silicalite-1 zeolite at sub-saturation pressures. In the following section, we investigate the
role of these defects on the diffusivity of water within the zeolite pore network and investigate
the optimal internal surface chemistry to enhance water permeability.
3.3.2
Water Diffusivity and Permeability
As the results from the previous section demonstrated that small changes in the internal
defect density significantly affect the sorption/infiltration properties of MFI zeolites, it is
plausible that small changes in the internal defect density could also significantly affect the
10
It is well-known, however, that for low Si/Al ratio zeolites, such as zeolites X and Y (FAU
type), the pores become fully saturated at relative pressures of less than 0.5 (91).
56
transport behavior of water confined within zeolites. To probe the diffusivity of the water within
the pores, we utilized the transient adsorption and desorption behavior of water into and out of
each zeolite. Although this study is limited to water that infiltrates prior to the saturation
pressure, the high-pressure infiltration experiments (Figure 3-2) showed that no extra water
(within experimental uncertainty) entered into the zeolite pores between the saturation pressure
and typical pressures (≈ 1 – 10 MPa) used in membrane-based separations (4). Following the
work of Karger and Ruthven(101), the transient adsorption and desorption curves were
regression fit to the solution for the transient diffusion equation (Equation 1) for short times,
where 0.2 < Mt/M∞ < 0.6, to estimate the transport diffusivity (Dt).
!!
2! !! !
≈
!!
! !
!/!
!
(1)
Normalized Uptake (Mt/Minf)
0.6
0.4
0.2
MFI INF
MFI 1000
MFI 300
MFI 200
MFI 100
5
10
15
20
1/2
Square Root Time (sec)
Figure 3-5. Example adsorption curve for various Si/Al ratio MFI zeolites taken at a
relative pressure of 0.4.
The approximate linear response indicates a Fickian diffusion process.
Figure 3-5 shows an example transient uptake curve (taken at a relative pressure of 0.4)
for the zeolites studied. All of the curves showed an approximately linear uptake behavior (as a
57
function of t1/2) at all relative pressures investigated. Accordingly, we modeled both the
adsorption and desorption behavior as a Fickian diffusion process.
The micrographs from the SEM analysis were used to estimate the external surface area
(A) and volume (V) of the crystal (Table 3-1). Since the Fickian transport diffusivity accounts
for gradients in concentration, it is more appropriate to utilize the thermodynamically corrected
diffusivity (which takes into account gradients in the chemical potential) for transport-related
processes. Following Zhang et al., the transport diffusivity was converted into the corrected
!"#$
diffusivity by utilizing Equation (2), where !"#$ is the gradient of the sorption isotherm in
logarithmic coordinates (50).
Dt = Do
∂ln(q)
∂ln( p)
(2)
The values for the corrected diffusivity taken at a relative pressure of 0.4 are shown as a
function of the estimated defect density in Figure 3-7A11 and are represented by the black
squares. The diffusivity of water within the near defect-free Silicalite-1 zeolite (MFI INF) is up
to two orders of magnitude higher than the most defective zeolite (MFI 100). Even for a slight
increase in the defect density (from the estimated value of 0.1 for the purely siliceous sample to
0.42 for the MFI 1000 sample), the diffusivity of water decreased by approximately an order of
magnitude. This trend occurs over the full range of pressures investigated in the sorption regime
(the corresponding data sets shown in Figure 3-6). Combined with the adsorption/infiltration
experiments and the insights gained from the Cailliez defect model, these results verify the
strong attraction between water and the hydrophilic defect sites, which cause a substantial
decrease in the local water mobility when compared to water within the hydrophobic defect-free
region of the MFI pore structure. However, since the permeability is dependent on both the
diffusivity and solubility, it is unclear which property is dominant in controlling the overall
transport behavior.
11
The estimated values of the diffusivity and solubility did not significantly change as the
relative pressure was varied. The complete data sets of the diffusivity and the solubility as a
function of the relative pressure are provided in the Figure 3-6.
58
A
B
Figure 3-6. The complete data set of both the diffusivity (A) and solubility (B) as a function
of the relative pressure.
A
B
Figure 3-7. The diffusivity, solubility and permeability of water within MFI zeolites as a
function of the defect density.
(A) Water diffusivity (black squares, left axis) and solubility (red triangles, right axis) plotted
against the estimated internal defect density taken at a relative pressure of 0.4. As the defect
density increased, the water diffusivity decreased by up to two orders of magnitude while the
solubility only increased ≈ 7x. (B) The estimated average water permeability of MFI zeolites as a
function of the defect density at 298 K. The decreasing trend as the defect density increased
highlights the detrimental effect of hydrophilic defects on water transport within the MFI
structure. The complete data sets of the diffusivity, solubility and permeability as a function of
the relative pressure are shown in Figure 3-6. Note that 1 Barrer is 3.06 x 10-16 (mol m)(m2 s Pa)1
at STP.
59
Using the definition of the solubility provided by Zhang et al., the solubility was
estimated and is shown in Figure 3-7A as red triangles 5. Although the solubility does increase
with more defective zeolites, the most defective zeolite (MFI 100) only adsorbed ≈ 7× more
water compared to MFI INF. By taking the product of the diffusion and solubility, it can be
clearly seen in Figure 3-7B that the water permeability was maximum for the least defective
zeolite and decreased by upwards of an order of magnitude as the defect density increased above
1 D/UC. Consequently, we can say that the diffusivity is the dominant property in controlling the
overall water permeability for MFI zeolites.
The combined results from these experiments demonstrate that less defective,
hydrophobic MFI zeolites should offer an advantageous internal surface chemistry to improve
the water transport properties for water-based membrane separations compared to the more
defective, hydrophilic zeolites. While these results are in qualitative agreement with past
transport studies of water within hydrophobic carbon materials (24,87,102,103) in that a surface
with a lower attraction energy to water can facilitate faster mass flow, they contradict past
experimental zeolite-based membranes studies, where hydrophilic membranes demonstrated an
increase in the measured water flux in comparison to hydrophobic membranes (30,32). This
inconsistency, however, highlights the difficulty in elucidating the transport mechanisms through
membrane experiments alone, where small changes in other properties, such as active layer
thickness, crystal size, localized variations in composition, contamination, and intercrystalline
mesoporosity are also known to affect the measured water flux (25,57,94). It should be
highlighted that, in this experimental study, the investigation was limited to water within the
zeolite crystals.
3.4
Conclusions
The effect of the defect density on the transport properties of water within sub-nanometer
MFI zeolite pores was systematically investigated through combined sorption analysis and highpressure infiltration experiments. We demonstrated that an increase in the internal defect density
increased the sorption capacity of zeolites from ≈ 4 N/UC for the near defect-free Silicalite-1
MFI zeolites to ≈ 27 N/UC for the most defective MFI 100 zeolites at the saturation pressure.
The remaining unfilled pore network of all zeolites required upwards of 40 MPa to become fully
saturated. While an increase in defect density increased the amount of water within the zeolite,
60
the diffusivity of this infiltrated water decreased by upwards of two orders of magnitude
compared to water within the hydrophobic Silicalite-1 zeolites. Subsequently, the permeability of
the infiltrated water within the more defective MFI zeolites was upwards of an order of
magnitude lower than that within the Silicalite-1 zeolites. These studies highlight the strong
attraction of water to the acidic Al-substituted defect sites, which, although increased the overall
affinity of water to the zeolite, effectively decreased the mass transport of water through the subnanometer pores. The insights gained from this study suggest that the intrinsic hydrophobic
internal surface of the MFI zeolite pores facilitates for faster water transport over more defective
hydrophilic MFI counterparts. The experimental results from this study can be utilized to
improve the performance of zeolite-based membranes used in water separation applications
while also providing additional understanding of the transport mechanisms of water at nanometer
length scales.
In order to address the transport challenges associated with current membranes and to
further study the effect of the defect density on water transport, better control of membrane
fabrication is required. In the next chapter in this thesis, we describe a novel technique to
fabricate thin zeolite membranes in which the transport is limited to the zeolite crystals. By
studying the water transport generated by a gradient in the salt concentration (i.e., forward
osmosis), we corroborated the flux predictions presented within this chapter.
61
Chapter 4 4. Osmotically-Driven Flow Across Microfabricated Zeolite Membranes
4.1
Overview
As the previous chapters highlighted, water transport within the sub-nanometer pores of
zeolites is complex, and small changes in the internal surface chemistry can significantly alter
both the diffusion and permeability of water. These previous results demonstrated that the local
interactions between water and the nano-confining surface controlled the transport process. Yet,
despite the insights gained from monitoring the effect of the concentration of hydrophilic defects
on the rate of water diffusion into and out of MFI-type zeolites, it is unclear if the permeability
calculations made using the sorption-diffusion model will accurately predict the rate of water
transport in membrane-based applications (as the calculations were based upon water vapor,
rather than liquid water, adsorbing and desorbing from a single zeolite crystal). Furthermore,
while insights into the effect of the surface chemistry can be gained from previous membrane
studies, the results between different studies (and materials) are inconsistent, which show the
challenges in studying transport with membranes composed of such microporous materials.
In studies investigating the use of carbon nanotubes as membranes, both molecular
dynamic simulations and experiments agree that the decreased attraction between the water and
the carbon surface (i.e., a hydrophobic interaction) explained the substantial increases in the flow
rate compared to predicted continuum hydrodynamic theory (i.e., super-lubricity) (19,24,87,103105). If the interior of the carbon nanotube was coated with an organic molecule with a higher
attraction energy to water (i.e., the interface was made more hydrophilic), the measured flow rate
of water through the nanotubes decreased by orders of magnitude (105,106). This decrease in
permeability with a more hydrophilic interface qualitatively agrees with the measurements and
predictions made with the sorption-diffusion modeling of water transport in MFI zeolites
described in Chapter 3 of this thesis. However, for experimental results investigating zeolites as
the active layer in a membrane, some literature showed the opposite result. In the work by Li et
al., it was found that, as the Si/Al ratio decreased (i.e., the membrane was made to be more
hydrophilic), an increase in the measured water flux was observed. Yet, it is unclear if these
62
results are conclusive, as there is a significant variation in the water flux reported with these
types of zeolite-based membranes. As shown in Figure 4-1, the reported permeability of these
membranes can span over two orders of magnitude. Furthermore, the permeability of these
zeolite-based membranes is substantially lower than that of the commercial aromatic polyamide
and also the permeability that is predicted by molecular simulations (16).
Figure 4-1. Permeability and salt rejection of zeolite-based and current state-of-the-art
membranes.
The variation in both the permeability and salt rejection across the zeolite-based membranes has
made it difficult to understand the transport across the zeolites. Additionally, even with the best
zeolite-based membranes, the permeability is about 5x lower than seawater AP membranes.
The difficulty in elucidating the water permeability across zeolite-based membranes is
that an unknown amount of transport can circumvent the zeolite crystals through intercrystalline
defects (Figure 4-2). The evidence for intercrystalline transport was demonstrated by the
imperfect salt rejection results shown in Figure 4-1. Thus, both improvements to membrane
fabrication and in the methodology of quantifying the water transport are needed.
63
Figure 4-2. Potential transport pathways through previous zeolite-based membranes.
(a) SEM image of the cross section of a zeolite-based membrane used for reverse osmosis. The
lighter region at the top of the image is the zeolites and the darker, porous region is an alumina
support. Taken from (14). (b) Schematic of the red boxed region in (a). The arrows (1 and 2)
show potential water and salt pathways through the zeolite membrane. Therefore, using this type
of membrane, it is unclear what the role of the zeolite is in the transport.
In this chapter, we present a novel membrane fabrication method utilizing various
microfabrication techniques (such as physical vapor deposition, atomic layer deposition, and
reactive ion etching) to produce an experimental platform that was designed to limit the transport
analysis to the zeolite crystals. We used a difference in salt concentration (i.e., an osmotic
pressure), rather than a hydraulic pressure gradient, to study and quantify the water flux. In
agreement with the sorption-diffusion model described in Chapter 3, the water flux across more
hydrophobic MFI zeolites was ≈ 10x higher than more hydrophilic (i.e., more defective) MFI
zeolites. Additionally, the linear increase in the recorded flux as a function of osmotic pressure
demonstrated that the zeolite pores were capable of selectively transporting water and rejecting
salt ions (i.e., the pores can sustain an osmotic flow). However, a small amount (< 5% of total
area) of meso/macroscale defects existed in the fabricated membranes, which created pathways
for salt transport across the membrane. By studying the diffusive transport of both potassium
chloride and Allura Red AC (a molecule with a diameter of ≈ 1 – 1.4 nm), it was found that
(within experimental uncertainty) this ‘back-diffusion’ occurred through the larger defects in the
membrane and, therefore, no observable salt transported through the zeolite pores. These
transport studies confirmed the molecular-based simulation predictions that the ≈ 5.5 Å pores of
MFI-type zeolites are capable of rejecting hydrated potassium and chlorine ions while still
allowing for water transport. The insights gained from this study demonstrate the increased
64
permeability of hydrophobic MFI zeolites and the size-selective molecular sieve properties of the
sub-nanometer pores. These results further suggest that the inclusion of sub-nanometer
hydrophobic pores into the active layer can potentially improve both the water permeability and
salt rejection of future RO membranes.
4.2
4.2.1
Experimental Procedures
Zeolite Synthesis
Two different purely siliceous MFI zeolites were synthesized for the study: a protonated
MFI zeolite (referred to in this chapter as H-MFI, and in the previous chapters as MFI INF,
Figure 4-3a), and a sodium-infused MFI (referred to as Na-MFI) zeolite12 (Figure 4-3b). Sodium
was introduced in the synthesis procedure such that zeolite was more hydrophilic than the H-MFI
form. Additionally, previous molecular simulations showed that the inclusion of extra framework
cations (such as sodium) significantly reduced the diffusivity of water compared to the more
hydrophobic H-MFI zeolite (52). The zeolites were synthesized similarly to the procedures
described in the previous chapters. Tetraethyl orthosilicate (TEOS, Sigma Aldrich) and
tetrapropylammonium hydroxide (TPAOH, 40 wt % aqueous solution, Sigma-Aldrich) were
used as the silica source and the structure-directing agent (SDA), respectively. Class 2 deionized
water (VWR) was used to control the pH of the synthesis solutions. For the sodium-infused
zeolite, sodium hydroxide (reagent grade beads 97%, Sigma Aldrich) was introduced with the
water. The molar concentrations were as follows:
5:1:1000 TPAOH:TEOS:H2O - (H-MFI)
5:1:1:1000 TPAOH:TEOS:NAOH:H2O - (Na-MFI)
Measured amounts of TEOS, TPAOH, H2O, and (if needed) sodium hydroxide were mixed at
room temperature and stirred for at least 12 hours to obtain a clear solution. The solutions were
then transferred into a 45mL PTFE lined stainless steel autoclave (Parr, Inc.) for synthesis. Both
12
Zeolites with a lower Si/Al ratio (300 or less) could not be oriented into a single monolayer on
the anodized aluminum oxide membranes (or a silicon wafer). We suspect the lack of orientation
arises from the rough morphology of the zeolite crystals, which results from the aluminum
incorporation.
65
zeolites were synthesized under rotation (≈ 100 RPM) for 5 hours. The H-MFI form was
synthesized at 160 °C and the Na-MFI form was synthesized at 185 °C (the higher temperature
was used to introduce more silicon vacancy defects). After synthesis, the zeolites were recovered
by centrifugation and washed with deionized water until the pH of the solution was ≈ 9. This step
was followed by drying at 60 °C. The zeolites were not calcinated until after the zeolite
orientation (described in the following sections).
B.
A.
Figure 4-3. SEM images of synthesized zeolites.
(a) Image of Na-MFI zeolites. Scale bar is 1 µm. (b) Image of H-MFI zeolites. Scale bar is 2 µm.
4.2.2
Zeolite Orientation
The orientation of the MFI zeolites implies that the straight-through pores were aligned
vertically (i.e., b-oriented) and in the direction of the pores of the underlying AAO. Previous
literature demonstrated that this orientation minimized the transport distance for a molecule to
diffuse through the porous network, and could increase the permeability (57). While many
techniques, such as a seeded growth procedure (57,107) and chemical functionalization (108),
have been established to achieve this type of orientation on a support, there are challenges in
obtaining highly-dense films. Recently, a novel technique known as manual direct assembly was
demonstrated to easily obtain high-density oriented monolayers on flat surfaces (109-111). The
technique exploits the anisotropic shape of various zeolites (such as the disc/coffin shape of MFI
zeolites) as the mechanism for orientation. For MFI zeolites, since the major flats of the crystal
66
are normal to the straight-through pore direction, the b-orientation of the zeolites can be easily
achieved.
Finger
d)
Zeolites
Substrate
Figure 4-4. Manual direct assembly of MFI zeolites onto a support.
(a) Zeolites are ‘sprinkled’ onto the support (as in physically deposited on the surface). (b) With
a finger enclosed in a latex glove, the zeolites are physically rubbed into the surface in a circular
manner, see (d). (c) A highly dense monolayer is formed after a few seconds of rubbing the
surface (see Figure 4-5 for SEM images). (a – c) are taken from (109).
Figure 4-5. SEM images of oriented (a) Na-MFI and (b) H-MFI zeolites on pieces of a
silicon wafer.
The zeolites were oriented on the silicon wafer using the manual direct assembly method.
Orientation is implied by the way the zeolite ‘straight-through’ pores are oriented, which in this
case is normal to the silicon surface. The scale bar for both images is 2 µm.
First, after zeolites are dried, a small amount (< 10 mg) was physically placed on the surface
such as glass or silicon (in the same method as shaking salt on food). Then, using a gloved index
finger (in my experience, powder free latex gloves provided the best results), pressure was
applied on the zeolites in a circular manner (Figure 4-4d). This pressure separates the zeolites
67
into individual crystals, and the zeolites subsequently oriented onto the support. Figure 4-5
shows SEM images of both the H-MFI and Na-MFI zeolites oriented on a small section of a
silicon wafer. However, utilizing the same methodology to orient zeolites on a porous support
(such as an anodized aluminum oxide membrane) did not produce a high-density oriented layer.
As seen in Figure 4-6, only a small amount of zeolites were attached to the underlying anodized
aluminum oxide (AAO) membrane using just the manual direct assembly method.
Figure 4-6. Initial orientation results of smaller H-MFI zeolites on a porous AAO support.
The zeolites were oriented (or an attempted was made to orient the zeolites) using the manual
direct assembly method on a clean AAO surface. The scale bar is 5 µm for (a) and 10 µm for (b).
To produce high-density oriented layers, the interface between the zeolite and surface
must be smooth enough such that the van der Waals attraction can withstand the physical
abrasion during the orientation process. Previously, both Lee et al. (111) and Liu et al. (112)
reported enhanced zeolite orientation on porous supports covered with a thin layer of spin-coated
polyvinyl-alcohol (PVA). Using these insights, the density of oriented zeolites on the surface
was substantially improved (see Figure 4-7).
First, 10 grams of PVA (Mowiol 8-88, Sigma Alridch) was dissolved in 50 mL distilled
water held at 80 °C. After cooling to room temperature, a small amount of the 20 wt% PVA
solution (< 1 mL) was deposited with a pipette onto an AAO membrane (150nm, 100 µm
thickness, Synkera) which was affixed to a glass slide with kapton tape. The membrane was then
spun at 4500 RPM for 40 seconds (WS-400A, Laurell Instruments) to produce a thin, uniform
layer. The membrane was then heated to 50 °C on a hotplate (Cimarec, ThermoScientific) to
68
evaporate the water. Typically, a second spin coating was performed to ensure that the AAO
surface was completely covered with PVA13.
Figure 4-7. SEM images of the improved orientation technique, which utilized PVA as a
smooth adhesion layer on the AAO.
(a) Orientation of H-MFI zeolite prior to calcination. (b) Orientation of H-MFI zeolites post
calcination. Note than the PVA decomposes to open the AAO pores for transport. The scale bar
for both images is 2 µm.
The zeolites were then oriented using the aforementioned manual direct assembly method. A
green-colored interference was usually observed once a monolayer was established over the
AAO surface (see inset of Figure 4-11A). To remove the PVA from the membranes and the
organic template (i.e., the structure directing agent) from the zeolites, the zeolite-coated
membranes were heat treated to 550 °C with 1 - 2 °C ramp rates. The membranes were covered
with an alumina block (Al-23, Alfa Aesar) to prevent the AAO membranes from warping during
the heat treatment process. SEM images of the results before and after the PVA was decomposed
are shown in Figure 4-7.
4.2.3
Meso/Macropore Filling and Membrane Activation
As seen in Figure 4-7A and B, a significant portion (in some cases, upwards of 30%) of
the surface was not covered by the zeolites and, thus, a large portion of the underlying AAO was
accessible for transport. To limit the transport to the zeolite crystals, these openings needed to be
blocked with a water-impermeable material. Limiting the transport to the zeolites in a membrane
13
The AAO surface becomes reflective to light once a uniform layer was produced.
69
has been attempted before with techniques such as secondary growth (14,111) or polymer/epoxy
deposition (113,114). With the secondary growth techniques, significant cracks/gaps existed
within the membrane, which made it difficult to establish the role of the zeolites in both the
transport and rejection. The results from the polymer/epoxy method, although ‘defect-free’,
reported lower than hypothsized flux results and raised the question of whether the polymer
molecules were infiltrating and blocking the zeolite pore network. Instead of these techniques,
we explored the application of atomic layer deposition (ALD) to fill the exposed AAO pores.
ALD is a self-limiting, layer-by-layer deposition technique known to achieve both high
conformality and high aspect ratio gap filling (115-117). Hafnium oxide (i.e., hafnia) was chosen
as the deposition material for its chemical stability and reported high conformality in comparison
to silica14. The hafnia was deposited (Savannah 100, Cambridge NanoTech) at 250 °C using
tetrakis-dimethylamidohafnium as the product and water as the reactant. The chamber pressure
was ≈ 300 mTorr under a 20 sccm N2 purge flow. 7 seconds was used as the relaxation time
between product and reactant purges. The hafnia was typically deposited such that the deposition
thickness was greater than the pore radius (i.e., 80 nm for a 150 nm pore diameter AAO).
14
Our deposition materials were limited to silica, alumina, or hafnia. Alumina had the ideal
coating properties, however, it deposits in an amorphous state and therefore easily dissolved in
water. Our experimental experience with silica did not produce uniform deposition thickness
across the membrane surface (> ± 10 nm for 100 nm deposited)
70
Figure 4-8. SEM images of H-MFI membranes after 93 nm of HfO2 via atomic layer
deposition.
Note that in the cross section image in (b), the pores beneath the zeolite were hafnia-free. The
scale bar for both images is 2 µm.
Figure 4-9. SEM images of zeolite membranes that were over-etched with RIE using a
CF4/O2 chemistry.
(a) 1 min over-etch of H-MFI membranes (the scale bar is 2 µm). The CF4/O2 chemistry is highly
selective for silica and can achieve etch rates on the order of 1 µm/min for MFI zeolites. The
residue within each zeolite ‘coffin’ was found to be primarily silica. (b) Outer edge of an NaMFI membrane. The zeolites were only exposed for a few seconds to the energized ions, which
created an interesting nanostructured surface. The scale bar is 1 µm.
As seen in Figure 4-8A and B, the hafnia coating closed the exposed pores of the
underlying AAO support. Furthermore, the cross section (Figure 4-8B) showed that the coating
did not fill the AAO pores beneath the zeolite crystals (which would have effectively blocked all
71
transport across the membranes). The final step in the membrane fabrication process was
removing the hafnia that covered the zeolite surface. Due to the chemical stability of hafnia, a
wet etching process was not used, as the chemicals used to etch hafnia (i.e., concentrated
hydrofluoric acid) would also dissolve the underlying AAO support and zeolite crystals. Instead,
directional, low-power reactive ion etching (RIE) was used to remove the hafnia blocking the
zeolite crystals. However, the zeolites were first masked as the chemistries used to RIE hafnia
(CF4/O2) had higher etch rates (at least 10x higher) for silica-based materials (zeolites) than for
hafnia. If the zeolites were not masked, the zeolites would be damaged (see Figure 4-9).
The schematic shown in Figure 4-10 was utilized to make the final membranes for the
transport studies. Various metals were investigated for the mask but, ultimately, gold was used as
the wet etchant for gold (dilute potassium iodide) does not damage the zeolite or AAO pore
structure.
First, 40 nm of gold was evaporated (TE-4, Sharon Vacuum) to mask the zeolites during
fabrication (Figure 4-10A). Following the evaporation, ≈ 80 nm of hafnia was deposited using
atomic layer deposition (Figure 4-10B). To expose the zeolites for transport, the hafnia was
reactive ion etched (Cirrus 150, Nexx Systems) at 1 min intervals using a CF4:O2 chemistry (20:1
sccm, 250 W microwave power, 125W RF power with a variable DC bias at 10 mTorr operating
pressure) until the underlying gold was exposed (Figure 4-10C)15. The membranes were dipped
into a dilute solution of potassium iodide (Gold Etchant, Sigma Aldrich) for ≈ 5 minutes to
remove the gold covering the surfaces of the zeolite crystals. This technique was used to create
membranes >20 mm2, which allowed for the measurement of water and salt transport across
>1013 sub-nanometer pores (Figure 4-11).
15
A spectroscopic ellipsometer was used to monitor the hafnia etch rates using hafnia deposited
on a silicon wafer as the control. Typically, the etch rates were ≈ 15 nm/min, although this value
could vary significantly. The best results in terms of maximizing the amount of zeolites ‘opened’
for transport occurred when the membranes were slightly over-etched (corresponding to an extra
5 nm etch time or ≈ 20 extra seconds under normal operating conditions).
72
Figure 4-10. Fabrication schematic (shown with H-MFI zeolites) with corresponding SEM
images.
(a) Following the zeolite orientation onto the AAO, 40 nm of gold was thermally evaporated
onto the membrane to mask the zeolites during processing. (b) The remaining exposed AAO
pore structure was filled by depositing hafnia (red in schematic) via atomic layer deposition. (c)
Any hafnia that covered the zeolite crystals was dry etched using a CF4/O2 chemistry. (d) The
gold was wet etched to expose the underlying sub-nanometer pore structure and remove any
residual hafnia covering the crystals. The scale bar for all images is 1 µm.
73
Figure 4-11. Images of fabricated Na-MFI membranes.
(a) Low magnification image demonstrating zeolite orientation over a large area (scale bar is 10
µm). Inset: Image of fabricated membrane on AAO (scale bar is 5 mm) (b) Higher magnification
image of red boxed region in (a) highlighting both the zeolite (darker cubes) and surrounding
ALD hafnia (white regions) (scale bar is 1 µm).
4.2.4
Transport Measurements
The water and salt transport was quantified using a custom diffusion cell (7.0 mL,
Permegear) shown in Figure 4-12 and was inspired by the work of Lee et al. (8). Due to both the
small size of the membrane and lack of mechanical robustness, only concentration-gradient
driven flows (i.e., osmotically driven transport) were investigated. Although this experiment
limited the study to water transport at approximately atmospheric pressures, the advantage of
forward osmosis experiments is that it is a tool that can determine if the zeolite pores are capable
of selectively transporting water and rejecting salt ions. If the zeolite pore structure was not
capable of allowing this type of semi-permeable transport, an osmotic water flow would not be
generated (118).
74
Figure 4-12. Test cell to quantify water transport.
Schematic (a) and image (b) of diffusion cell used to quantify both the water flux and salt/Allura
Red AC diffusion. Note that the scale bar for (b) is 3 cm.
4.2.4.1
Water Transport Quantification
After membrane fabrication, the membranes were soaked in DI water (Class 2, VWR) for
at least 12 hours. The diffusion cell was thoroughly washed with DI water, dried, and plasma
cleaned under air for 15 minutes (Plasma Cleaner, Harrick Plasma). To interface the membrane
with the diffusion cell (and to prevent any leaks around the membranes), two silicone gaskets
(JTR-2-2.0, Grace Bio-Labs) were attached to each side of the cell (orange areas in between the
cells shown in Figure 4-12). The membrane was subsequently clamped into place using the cell
holder provided by Permegear. After adding small magnetic stir bars into each vessel, ≈ 7 mL of
DI water was added to the right hand side of the cell, with extra care taken so that no air bubbles
were trapped16. Then, ≈ 9 mL of a high molar (between 0.5 M and 4 M) concentration of KCl
(≥ 99%, Sigma Aldrich) was introduced into the left side of the diffusion cell. The concentrated
salt side was plugged with a rubber stopper to ensure that any volume changing arising from the
osmotic flow changed the height of the fluid in the graduated cylinder (1725LT, Hamilton) that
was attached to the diffusion cell. Both sides of the diffusion cells were vigorously agitated with
stir bars to limit concentration polarization at the membrane interfaces. Water transport was
16
This was a particular problem at the interface of the membrane with the cell. To alleviate this
problem, the cell was tilted in such a way that the water would initially wet the membrane and
then fill the rest of the cell.
75
quantified by measuring the change in the corresponding height as a function of time (see Figure
4-13) using an SLR camera (Rebel t3i, Canon) that captured images at set intervals (typically
every 3 minutes).
Figure 4-13. Time-lapse images of the change in the water level in the graduated cylinder.
Each major notch corresponds to 5 µL of volume. This particular set of images corresponds to an
H-MFI membrane with an open area of ≈ 2.35x10-5 (± 3x10-6) m2 using a 0.5 M KCl draw
solution (≈ 24 bar). Scale bar is the same for all images and is 1 cm.
The open area of the membranes was quantified by image analysis of SEM images
(Ultra55, Zeiss) using the software ImageJ. As seen in Figure 4-14, there is a distinct contrast
difference between ‘opened’ (i.e., activated) zeolites and those zeolites that are still covered by
hafnia17. The contrast threshold was adjusted within ImageJ such that all ‘opened’ zeolites were
counted as ‘Particles’ with the ‘Analyze Particles’ tool. Various threshold values were used to
get an upper and lower bound on the ‘open’ area of each SEM image. Over 20 images for each
membrane were analyzed to obtain an average active area for each respective membrane.
17
The non-opened zeolites were also confirmed to still have hafnia covering the surface via
energy dispersive x-ray spectroscopy (EDS).
76
Figure 4-14. Estimation of the active zeolite area using ImageJ.
(a) SEM image of one H-MFI membrane with a small amount of zeolites activated for transport.
The darker zeolites were confirmed to be activated (i.e., the surface of the zeolite did not contain
any hafnia) using energy dispersive x-ray spectroscopy equipped on the SEM. The scale bar is
10 µm. (b) The estimated open surface that was obtained using a threshold analysis built into
ImageJ. The corresponding ‘open’ area percentage for this image was ≈ 0.8 ± 0.3 %.
The change in volume was converted into a change in mass assuming a water density of
0.998 g/cm3. The mass flow was normalized by the respective membrane active area to estimate
a water flux as a function of the draw solution (and of the osmotic pressure using the van’t Hoff
equation (119)).
4.2.4.2
Salt Transport Quantification
Salt transport/leakage was probed by measuring the conductivity of the DI water
reservoir during the experiments. Ideally, if the membrane is only permeable to water, the
change in reservoir conductivity during the duration of the experiment would be ≈ 0 mS/cm18.
However, if salt could either diffuse through the zeolite pores or through defects in the
membrane, the conductivity would increase over time.
18
A small increase in conductivity would occur since water molecules are diffusing into the high
salt concentration solution, however, this change in conductivity was negligible compared to the
salt leakage from the draw solution.
77
Conductivity Probe
A.
Reservoir Conductivity (mS/cm)
1 M KCl
1.2
DI Water
KCl Diffusion
B.
1.0
0.8
0.6
0.4
0.2
0
1000
2000
3000
4000
5000
Time (sec)
Figure 4-15. Schematic of (a) diffusion cell and (b) example for the reservoir conductivity
as a function of time.
This particular test corresponded to an H-MFI membrane and yielded a diffusive salt flux of
≈ 3.5x10-4 mol/m2s.
To quantify the transport, separate experiments were performed after the water transport
tests. Prior to each test, the conductivity probe (16-900 Conductivity Electrode, Microelectrodes,
Inc.), which was interfaced to a computer using an isoPod data acquisition system (eDAQ), was
calibrated with a 10 mM KCl salt solution (Conductivity Standard, 1.413 mS/cm, Sigma Aldrich)
to ensure conductivity accuracy at low salt concentrations. Then, exactly 7.0 mL of a 1 M KCl
solution (ISA - 1M KCl, Sigma Aldrich) and 7.0 mL of DI water (Class 2, VWR) were each
introduced into a compartment of the diffusion cell. The stir bars were spun at ≈ 800 RPM during
the experiments to maintain a uniform salt concentration in each bath. The conductivity of the DI
water compartment was measured every 10 seconds using a custom MatLab program. The
generated conductivity vs. time curves were fit to a linear slope using Origin (9.1, OriginLab).
The slope of the curve was converted to a salt concentration (mol/L) using the calibration values
(0.141 mS/cm equaled a change in concentration of 1 mM/L KCl19). Figure 4-15B shows an
example conductivity response for a membrane tested.
19
This change in conductivity as a function of the salt concentration was confirmed after each
test with a set of 1 mM, 2mM, 5mM and 10 mM KCl solutions to ensure that the ‘k-value’ of the
conductivity probe did not drift during experiments.
78
4.2.4.3
Determining the Leakage Pathways
When salt transport was observed to transport across the membranes, it was unclear
whether the salt was only diffused across the defects/cracks in the membrane (Figure 4-22) or
through the sub-nanometer zeolite pores. To probe the leakage pathways, we studied the
transport of a larger molecule, Allura Red AC (AR), which does not fit into the zeolite pores
since the Stokes-Einstein diameter is > 1 nm (17) (and the zeolite pores were ≈ 5.5 Å).
Spectrophotometer
-0.08
Absorptance (au)
-0.10
1 mM Allura
0.5 M KCl
A.
0 min
60 min
120 min
180 min
-0.12
-0.14
-0.16
0.5 M KCl
-0.18
Allura Diffusion
B.
450
500
550
600
650
700
750
Wavelength (nm)
Figure 4-16. Setup used to quantify Allura Red Transport.
(a) Schematic of setup for quantifying AR diffusional flux and (b) an example (Na-MFI, #3)
of the change in absorbance over time as AR diffused into the right side compartment in
(a).
As with the salt transport, AR transport was driven by a concentration gradient across the
membrane. A 1 mM AR (98% Allura Red AC, Sigma-Aldrich) solution was made by mixing 25
mg of AR in 50.35 mL of 0.5 M KCl (potassium chloride beads, Sigma-Aldrich) dissolved in DI
water (Class 2, VWR). Exactly 7 mL of the AR solution and 7 mL of 0.5 M KCl were each
introduced into a compartment of the diffusion cell (Figure 4-16A). The cells were vigorously
mixed using stir bars to ensure a uniform concentration in the cells at all times. To quantify the
AR transport, a UV-Vis spectrophotometer probe (Cary 60, Agilent) was placed in the 0.5 M
KCl side of cell, and an absorbance spectrum from 800 to 400 nm was obtained every minute for
3 hours. The difference in the absorbance peaks between 510 nm (AR absorbance) and 710 nm
79
(wavelength independent of AR concentration) was used to calculate the concentration of AR in
the solution as a function of time (Figure 4-16B).
Results and Discussion
4.2.5 Membrane Fabrication
Using the zeolite membrane fabrication process presented in previous sections, two types
of zeolite-hafnia membranes were fabricated. Due to the roughness of the lower (< 300) Si/Al
ratio MFI zeolites crystals that were synthesized, an oriented layer using the manual direct
assembly method could not be produced. However, a more hydrophilic form (Na-MFI) of the
purely siliceous zeolite (H-MFI) was synthesized by introducing sodium hydroxide into the
synthesis solution and by raising the synthesis temperature. It is well known that sodium cations
adsorb within the zeolite structure and act as ‘extra-framework cations’ (58). These cations act as
seeds for water adsorption and increase the adsorption capacity of the zeolites (Figure 4-17).
35
H-MFI
Na-MFI
Sorption Amount (N/UC)
30
25
20
15
10
5
0
-5
0.0
0.2
0.4
0.6
0.8
1.0
Relative Pressure (P/Po)
Figure 4-17. Adsorption isotherms for zeolites used for membranes.
The tests were run according to the procedures described in the previous chapter. Both zeolites
were dried at 500 °C under air in a box furnace for 12 hours prior to the test. The Na-MFI zeolite
(which has been previously referred to being hydrophilic) clearly had a higher sorption capacity
than the H-MFI zeolite. The hydrophilicity (which is defined the amount adsorbed at P/Po = 0.95
divided by the framework capacity of 35 N/UC) was found to be ≈ 0.4 for the Na-MFI zeolites
and ≈ 0.1 for the H-MFI zeolites.
Furthermore, it has been reported that these incorporated extra framework cations will also
impact the diffusion coefficient of water within the zeolite pores (52) due to the increased
80
attraction energy of water to the cation sites. We confirmed, using the sorption-diffusion analysis
presented in Chapter 3, that both the diffusivity and permeability of water was substantially
reduced (≈10 x) for the Na-MFI in comparison with the hydrophobic H-MFI form (Figure 4-18).
These results indicated that these extra framework cations act in a similar manner to the Alsubstituted defects in that the extra framework cations decrease the mass transport via the
increased attraction of water to the defect site.
1E-13
0.5
0.4
0.3
0.2
A.
H-MFI (MFI INF)
MFI 1000
MFI 300
MFI 200
MFI 100
Na-MFI
5
10
15
20
1/2
Square Root Time (sec)
Permeability (mol m)(m2 s)-1
Normalized Uptake (Mt/Minf)
0.6
1E-14
1E-15
1E-16
0.0
B.
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Hygroscopic
Capacity
Hydrophilicity
Figure 4-18. Transport properties of MFI zeolites.
The diffusion and the permeability plotted for the zeolites studied. The black curve and symbols
represent the H-MFI zeolites, and the orange curve and symbols represent the Na-MFI zeolites.
The permeability (estimated using the sorption-diffusion analysis) of the Na-MFI zeolites was
estimated to be about an order of magnitude lower than the H-MFI zeolites.
With these two types of MFI zeolites and the fabrication procedure described in the
experimental portion of this chapter, a number of membranes were fabricated (Figure 4-19). It
should be noted here that the fabrication process was not perfect and contained many steps that
could (and did) result in membrane failure. Small variations in the deposition or etch rates
(which is a well-known issue involving clean room equipment) created non-uniform active areas
in the membranes. Furthermore, only ≈ 10% of the membranes that ‘started’ the fabrication
process were useful for studying transport after the completion of the processing steps. This low
success rate was primarily attributed to over- or under-etching (using the RIE), to general defects
like large scale cracks propagating through the AAO, and to surface contamination that affected
the AAO pore filling during the ALD process. It would be a disservice to the reader to not
81
highlight these challenges. This process was not without its flaws and could be improved with
more work. Nevertheless, as will be discussed in the following sections, enough membranes
were fabricated to discern both the water flux and salt transport across the zeolite crystals.
A.
B.
Figure 4-19. SEM images of the fabricated zeolite-based membranes.
The images correspond to the (a) H-MFI (specifically #21, Table 4-1, scale bar is 3 µm) and to
the (b) Na-MFI (#3, Table 4-1, scale bar is 1 µm) membranes.
4.2.6
Water Flux Quantification
Using a concentration gradient to induce an osmotic pressure difference across the
membrane using the experimental setup shown in Figure 4-12, the water flux and permeability
across the fabricated membranes was quantified. This transport process, also known as forward
osmosis, is useful in studying the transport of water across microporous materials as a water flux
will only be generated if the ‘membrane material’ is perm-selective (i.e., allowing for water
transport while rejecting salt ions) (94). If larger scale defects/cracks (> 5 nm) exist in the
membrane, these areas will not generate an osmotic flow as both water and salt can transport
through the gap. If the defective area coverage is not high (> 5% area would allow for
measureable water backflow), these defects will not affect the water flux measurements, and
thus, any observed water flux can be assumed to be a result of flow across zeolite crystals.
Therefore, the transport across these membranes contrasts with that of hydraulically pressurized
flows, where a water flux will be produced both through the zeolites and the defective
regions (14).
82
100
H-MFI Flux Prediction
Na-MFI Flux Prediction
H-MFI Data
Na-MFI Data
2
Flux kg/(m hr)
10
1
0.1
0.01
1
10
Osmotic Pressure (MPa)
Figure 4-20. Flux results for membranes as a function of osmotic pressure.
The linear increase in water flux as a function of osmotic pressure indicates that both zeolites
selectively transport water and reject salt. Additionally, the H-MFI zeolites exhibited an ≈ 10x
increase in water flux compared to the Na-MFI zeolites. Additionally, both membranes match
the water flux predicted by the sorption-diffusion analysis.
The initial results from the water transport studies are graphically shown in Figure 4-20
(further details of each membrane are provided in Table 4-1). The linear increase in water flux as
a function of the osmotic pressure (i.e., concentration gradient) indicate that water was
selectively transported across the zeolite pores while salt ions were rejected. Although selectivity
(or salt rejection) cannot be determined using just these results, the results suggested that the size
of the zeolite pores (and possibly other effects such as local electrostatic interactions at the pore
interface) allowed for water transport and rejected salt ions (Section 4.2.7 has further details on
quantifying the salt transport across the membranes).
83
B.
A.
Figure 4-21. Variation in active area for an H-MFI membrane (#8 in Table 4-1)
Non-uniform reactive ion etching activated only portions of some membrane for transport. The
activated zeolites are the darker contrast zeolites and the non-activated zeolites are lighter in
color. Figure (a) is only ≈ 1 mm in distance from (b) on the membrane surface. The scale bar is
10 µm for both images.
The hydrophobic H-MFI zeolites generate an approximate 10x increase in water flux in
comparison with the more hydrophilic Na-MFI zeolites (Figure 4-20). The error bars with each
data point are associated with the uncertainty in the height/volume measurement (± 2 µL).
Additionally, error arose from variabilities in the etch rates across the membranes using RIE,
which resulted in some areas of a particular membrane exhibiting a higher ‘active’ area than
other sections (see Figure 4-21). This variance in the open areas across the membrane created a
significant uncertainty (in terms of the standard deviation of open area from one section to the
next) in the actual active area for some of the membranes (see Table 4-1). However, at least one
H-MFI and Na-MFI membrane exhibited uniform active areas over the full membrane surface
(refer to the results for H-MFI (#21) and Na-MFI (#3) in Table 4-1), which confirmed the large
difference in generated water flux between the two zeolites.
The permeability prediction from the sorption-diffusion model for the respective zeolites
is shown as the shaded regions in Figure 4-20. The trends predicted by the model quantitatively
match very well with the experimental flux measurements. This match between the data and the
model shows that the transport through the zeolite pores for these forward osmosis experiments
is limited by a sorption-diffusion mode of transport. Furthermore, it indicates that the difference
84
in the interfacial resistance (i.e., water entering and exiting the pores) does not appear to
significantly change (i.e., no difference is observed whether the water entering the pores was in a
liquid or vapor state).
Table 4-1. Compiled water flux results from membrane experiments.
The numbers (e.g., #3) are only for reference and correspond only to the numbering system in
which the membranes were distinguished. The two best membranes (in terms of highest active
area and therefore highest flux) produced were H-MFI (#21) and Na-MFI (#3). With these two
results, it can be seen that the H-MFI membrane has an ≈ 10x increase in generated water flux.
As a note, the maximum possible area for the membranes was 3.85 x 10-5 m2.
H-MFI (#8)
Area (10 -6 m 2 ) - 3.37
± Area (10 -6 m 2 ) - 2.80
H-MFI (#10)
Area (10 -6 m 2 ) - 2.67
± Area (10 -6 m 2 ) - 1.92
H-MFI (#17)
Area (10 -6 m 2 ) - 1.15
± Area (10 -6 m 2 ) - 0.77
H-MFI (#18)
Area (10 -6 m 2 ) - 5.77
± Area (10 -6 m 2 ) - 3.85
H-MFI (#21)
Area (10 -6 m 2 ) - 23.48
± Area (10 -6 m 2 ) - 3.08
Na-MFI (#1)
Area (10 -6 m 2 ) - 6.16
± Area (10 -6 m 2 ) - 1.96
Na-MFI (#3)
Area (10 -6 m 2 ) - 32.71
± Area (10 -6 m 2 ) - 1.54
Draw
Concentration
(mol/L KCl)
0.5
1
2
4
0.5
1
2
4
0.5
1
2
4
0.5
1
2
4
0.5
1
2
4
0.5
1
2
4
0.5
1
2
4
Pressure
(MPa)
Flux (kg m -2hr -1)
2.5
5.0
9.9
19.8
2.5
5.0
9.9
19.8
2.5
5.0
9.9
19.8
2.5
5.0
9.9
19.8
2.5
5.0
9.9
19.8
2.5
5.0
9.9
19.8
2.5
5.0
9.9
19.8
1.1
2.4
4.9
8.9
2.2
3.7
9.0
15.9
1.7
2.6
5.6
9.5
1.9
3.5
6.6
9.9
1.5
2.7
5.6
10.5
0.2
0.4
0.9
1.9
0.3
0.7
1.2
2.4
85
Upper Bound Flux Lower Bound Flux
(kg m -2hr-1)
(kg m -2hr -1)
7.9
15.4
30.2
53.6
9.7
14.8
33.8
58.5
6.2
8.6
17.7
29.3
6.9
11.4
20.7
30.4
2.1
3.4
6.8
12.4
0.3
0.6
1.3
2.9
0.4
0.8
1.3
2.6
0.5
1.2
2.5
4.7
1.0
2.0
5.0
9.0
0.8
1.4
3.2
5.6
0.9
1.9
3.8
5.8
1.1
2.1
4.7
9.1
0.1
0.3
0.7
1.5
0.2
0.6
1.1
2.3
4.2.7
Salt Diffusion Across Membranes
B.
A.
Figure 4-22. Cracks/defects present in membranes that lead to salt transport.
The arrow in (a) highlights ≈ sub-50 nm gaps that were observed at the interface between the
zeolites and the AAO. The lighter right around the zeolite is the hafnia and the darker cubes are
the Na-MFI zeolites. The scale bar for (a) is 1 µm. The black outline in (b) shows are area where
the portion of the zeolite layer delaminated from the AAO surface and therefore lead to a large
opening for salt transport. This type of defect was rare but it did occur in a few of the
membranes. The scale bar for (b) is 10 µm.
Although the zeolite membranes were capable of generating an osmotic water flux, salt
was also observed to diffuse from the high concentration cell into the DI water cell (Figure
4-15). This salt transport, however, did not correlate with the water flux (i.e., an increasing
amount of salt diffusion did not correspond with either an increasing or decreasing water flux),
which indicated that the amount of salt leakage across the membranes was low20 and also
decoupled from the mechanism of osmotic water transport. According to previous MD
simulations (16,18), salt should not be able to diffuse across the pores. Still, as with some
previous zeolite-based membranes, the challenge was in experimentally confirming that this salt
diffusion was not occurring through the zeolite pore network. After inspecting the zeolite
membranes post transport experiments, small cracks in the AAO/hafnia layer (Figure 4-20a) or
20
For the membranes that exhibited the highest salt leakage, the DI water concentration
never increased above 10 mM/L KCl after 1 – 3 hours of testing. This increase in concentration
for the DI water side corresponded to, at most, a 0.2% drop in osmotic pressure and therefore
was considered to be negligible.
86
delaminated sections of the zeolite membrane (Figure 4-20b) were found. Therefore, we
hypothesized that the salt transport was around the zeolite crystals and through these larger scale
defects in the active layer. To test this hypothesis, the normalized diffusive flux of a molecule
that cannot fit in the 5.5 Å MFI pores was compared to the normalized diffusive flux of the
potassium and chlorine ions. If the fluxes were equal, then it could be assumed that little to no
salt transport occurred across the pores. The molecule chosen for this study was Allura Red AC
(AR), a food-coloring dye composed of ≈ 1 – 1.4 nm diameter molecules. This study was
modeled after the experiments of O’Hern et al. (17), which probed the transport of salt and AR
through microporous graphene membranes. The permeability, K (m s-1), was quantified both for
salt and AR (Equation 1). The molecular flux, J (mol m-2s), was measured either using
conductivity measurements (Section 4.2.4.2) for salt transport or with absorbance measurements
(Section 4.2.4.3) for AR transport. ΔC (mol m-3) was the respective driving concentration
gradient.
𝐽 = 𝐾∆𝐶
(1)
To compare the transport between salt and AR, the permeability was normalized by the
diffusivity D (m2 s-1) of the respective molecule21, which accounted for the differences in
molecular size and mobility. The preliminary results of these transport studies are shown in
Figure 4-23. The results demonstrate an approximate equal rate of transport for both the AR and
salt. Since the rates in transport between salt and AR are not significantly (i.e., an order of
magnitude) different, these preliminary results suggest that the salt diffusion is limited to pores
larger than 1 nm and therefore, little to no salt passed through the zeolite pores. To our
knowledge, this is the first experimental transport study that definitively demonstrated the
molecular sieving potential of these MFI zeolite pores for water desalination applications.
21
A diffusivity value of 1.9x10-9 m2 s-1 was used for both potassium and chlorine ions and a
diffusivity value of 3.6x10-10 m2 s-1 was used for AR (17).
87
Normalized Flux (K/D) (m-1)
1000
100
10
AR
KCl
H-MFI (18)
AR
KCl
H-MFI (21)
AR
KCl
H-MFI (8)
AR
KCl
Na-MFI (3)
Figure 4-23. Diffusion normalized flux of AR and KCl across zeolite membranes.
The red bars correspond to AR and the blue bars correspond to KCl. For all membranes tested,
the AR normalized flux was approximately the same as for KCl. While further work is needed to
prove that the salt is not diffusing through the zeolite pores, these results indicate that, at most, it
is not a significant amount. The numbers associated with each membrane (e.g., 18) correspond to
the specific membrane fabricated and show no other meaning.
4.2.8
Flux Comparison with RO Membranes
Finally, to compare the membranes with state-of-the-art RO membranes, the flux of each
membrane was normalized by the thickness of the respective active layer (i.e., the membrane
thickness) (Figure 4-3). This normalized flux (i.e., permeance) is useful in comparing transport
across different materials (94). There are two assumptions that we made use of for this analysis.
First, we assumed that the flux is inversely proportional to the thickness. While it is possible for
this assumption to break down with transport through nanoscale materials (e.g., for flow through
carbon nanotubes (24,103)), this assumption has been used for past zeolite-based membranes
88
(14,31,120)22. Second, as we were unable to probe a hydraulically pressure-driven water flow, it
is unclear if the water flux generated by the osmotic pressure difference is equal to that generated
by a hydraulic pressure difference (see Section 5.2.2). Yet, for flows at such small length scales,
the dominating thermodynamic potential is the chemical potential. Thus, we hypothesize that that
the generated flux of a pressure driven flow and an osmotically driven flow would be equal,
since both are produce equivalent gradients in the chemical potential (119).
Normalized Flux (mol m)/(m2s)
1E-4
H-MFI
Na-MFI
H-MFI Flux Prediction - (1 µm thick)
Na-MFI Flux Prediction - (0.7 µm thick)
SWRO Flux - (0.2 µm thick)
1E-5
1E-6
1E-7
1E-8
1E-9
1
10
100
Osmotic Pressure (MPa)
Figure 4-24. Thickness normalized flux as a function of the osmotic pressure.
The flux values were normalized by the thickness of the zeolite crystals (Figure 4-3) to compare
with the current state-of-the-art AP membranes (values taken from (3)). Under the assumption
that the water flux would not change under RO conditions, the water flux across the MFI
membranes would not surpass that of the RO membranes unless the active layer was decreased
to a thickness of 20 nm.
22
However, it is interesting to note that thicker zeolite-based membranes tend to exhibit a
higher thickness normalized flux in comparison to thinner zeolite membranes (14,30,121-123),
which may be evidence of a surface resistance dominating the transport process (see section
5.2.1).
89
As shown in Figure 4-24, the normalized flux for both of the zeolite membranes is lower
than that of the current state-of-the-art aromatic polyamide RO membranes. The decreased water
flux of the zeolite-based membranes in comparison to the AP-based membranes appears to
indicate that the transport through the rigid 5.5 Å MFI zeolite pores is not as fast through the
polymeric matrix in the active layers of RO membranes. Yet, in contrast with RO membranes,
decreasing the thickness of the zeolite active layer should not impact the salt rejection since the
rejection mechanism is decoupled from the water transport mechanism. Therefore, it may be
possible for zeolite-based membranes to surpass the permeability of polymeric membranes if the
active layer can be limited to a thickness of 20 nm. Previous research has shown that these purely
siliceous zeolite crystals can be made as thin as 2 nm (124). However, it may be (significantly)
challenging to incorporate these flakes into a continuous, defect-free membrane. Future research
in zeolite-based membranes should focus on developing scalable fabrication techniques in which
the active layer is, at most, 20 nm in thickness such that the water permeability an surpass that of
polymeric RO membranes.
4.3
Conclusions
The transport across microfabricated zeolite-based membranes was investigated by
quantifying the flux via forward osmosis. The water flux of the more hydrophobic H-MFI
zeolites was ≈ 10x higher than the more hydrophilic Na-MFI zeolites, which confirmed that an
interface with a decreased attraction to water facilitated faster waster transport. The forward
osmosis water flux also matched the predictions made using the sorption-diffusion modeling that
was described in Chapter 3 of this thesis. These results indicated that this forward osmosis
transport was a diffusion-limited transport process. Furthermore, by investigating both the
diffusion of salt (which potentially could transport through the pores) and Allura Red (which
should not be able to transport through the pores), it was determined that the salt transport across
the membranes was through cracks/defects within the active layer and not through the zeolite
pores. Future work is planned on improving the fabrication process to eliminate the presence of
these defects and cracks.
The generated water flux was compared to the flux measured across aromatic polyamide
reverse osmosis membranes used for seawater desalination. By normalizing the water flux with
respect to the membrane thickness, the flux of the hydrophobic H-MFI membranes 2 – 8x lower
90
than that of the AP-based membranes. However, since the salt rejection mechanism is decoupled
from the mode of water transport, it is possible that both the permeability and salt rejection of
zeolite membranes can surpass that of AP-based membranes if the active layer is made to be
20 nm in thickness or less. Future work with zeolite-based membranes should focus improving
the fabrication procedure to limit the thickness of the zeolite active layer to 20 nm or less.
91
Chapter 5 5. Conclusions and Future Work
5.1
Conclusions and Contributions of this Thesis
This thesis investigated the transport of water and salt across the sub-nanometer pores of
MFI zeolites with the goal of seeking to improve both the water permeability and salt rejection of
reverse osmosis membranes for water desalination.
In Chapter 2, a macroscale experimental method combining sorption analysis and highpressure infiltration was introduced as a means to investigate the water transport into and out of
zeolites. We performed controlled experiments and detailed sample characterizations to
investigate the total framework water capacity and infiltration pressure, and identified the
subtleties that may have led to the discrepancies in literature. We synthesized and procured
various sizes of purely siliceous MFI-type zeolites (Silicalite-1) and confirmed the uniformity in
structure and morphology using scanning electron microscopy (SEM), transmission electron
microscopy (TEM), x-ray diffraction (XRD), nitrogen sorption, and nuclear magnetic resonance
(NMR). Combined water adsorption and high-pressure infiltration experiments on these MFI
zeolites with characteristic crystal dimensions between ≈10 nm – 10 µm were performed. The
framework water capacity for fully crystalline MFI zeolites was determined to be 35 ± 2 water
molecules per unit cell. Nano-sized zeolites, with crystal diameters up to 100 nm, exhibited a
reduction in both the measured micropore volume (up to 25% compared to the larger zeolites) as
well as the total internal water capacity (up to 50% compared to the larger zeolites). The
decrease was attributed to the un-crystallized silica regions infused within the crystal resulting
from incomplete synthesis. Despite the differences in synthesis procedure and crystal
morphology for the various zeolites, the infiltration pressure was approximately the same, ≈95 –
100 MPa. The experimentally determined framework water capacity of 35 ± 2 water molecules
per unit cell and infiltration pressure values of ≈95 – 100 MPa can now be utilized to validate
and improve upon the existing water-zeolite interaction potentials used in molecular simulations.
Using these techniques, guidelines were set forth to study water that is confined within the subnanometer pores of zeolites.
92
In Chapter 3, this aforementioned technique was utilized to investigate the effect of the
internal surface on the diffusivity, solubility, and permeability of water within MFI zeolites. We
experimentally investigated the role of the concentration of hydrophilic defects in MFI (i.e.,
ZSM-5) zeolites on the permeability of water by independently quantifying both the solubility
and diffusivity. We synthesized MFI zeolites with a compositional silicon/aluminum (Si/Al) ratio
varying from 100 to infinite (i.e., Silicalite-1) and confirmed the structure and morphology using
x-ray diffraction (XRD) and scanning electron microscopy (SEM). Through combined sorption
and high-pressure infiltration experiments, we investigated the solubility of water within the
zeolite pores as a function of pressure and chemical potential. The diffusivity, and subsequently
the permeability, of water within the pores was evaluated by analyzing the transient adsorption
and desorption behavior. For a given pressure, we found that the amount of water that infiltrated
into the zeolite porous network increased as the internal defect density increased. However, none
of the zeolites studied were completely filled at the saturation pressure of water (3.14 kPa at
298 K), and each zeolite sample required upwards of 40 MPa to reach the measured framework
capacity of 35 water molecules per unit cell (N/UC). The increasing defect density decreased the
measured diffusivity by up to two orders of magnitude and the permeability by upwards of an
order of magnitude compared to water within the near defect-free Silicalite-1 MFI zeolite. The
results from these experiments highlight the strong attraction of water to the hydrophilic defect
sites within the zeolite, which, although increased the solubility of the zeolites, had a pronounced
detrimental effect on the diffusivity and consequently on the permeability. The experimental
results from this study can be utilized to provide detailed physical insights into the transport
mechanisms, which can then help guide the design of high permeability membranes in various
water-based separation applications.
In Chapter 4, we presented a novel membrane fabrication method utilizing various
microfabrication techniques (such as physical vapor deposition, atomic layer deposition, and
reactive ion etching) to produce an experimental platform that was designed to limit the transport
analysis to the zeolite crystals. We used a difference in salt concentration (i.e., an osmotic
pressure), rather than a hydraulic pressure gradient, to study and quantify the water flux. In
agreement with the sorption-diffusion model described in Chapter 3, the water flux across more
hydrophobic MFI zeolites was ≈ 10x higher than more hydrophilic (i.e., more defective) MFI
zeolites. Additionally, the linear increase in the recorded flux as a function of osmotic pressure
93
demonstrated that the zeolite pores were capable of selectively transporting water and rejecting
salt ions (i.e., the pores can sustain an osmotic flow). However, a small amount (< 5% of total
area) of meso/macroscale defects existed in the fabricated membranes, which created pathways
for salt transport across the membrane. By studying the diffusive transport of both potassium
chloride and Allura Red AC (a molecule with a diameter of ≈ 1 – 1.4 nm), it was found that
(within experimental uncertainty) this ‘back-diffusion’ occurred through the larger defects in the
membrane and, therefore, no observable salt transported through the zeolite pores. These
transport studies confirmed the molecular-based simulation predictions that the ≈ 5.5 Å pores of
MFI-type zeolites are capable of rejecting hydrated potassium and chlorine ions while still
allowing for water transport. The insights gained from this study demonstrate the increased
permeability of hydrophobic MFI zeolites and the size-selective molecular sieve properties of the
sub-nanometer pores. These results further suggest that the inclusion of sub-nanometer
hydrophobic pores into the active layer can potentially improve both the water permeability and
salt rejection of future RO membranes.
As is the case with most every PhD research project in science and engineering
(especially those with an emphasis on experimental work), there were significant challenges we
ran into in during my graduate work. Specifically with the work presented in this thesis, we ran
into considerable challenges in limiting the study to water within the zeolite pores network. As
stated within the Chapter 4 in this thesis, the membrane fabrication technique was not
particularly robust (i.e., only a few membranes of seemingly hundreds fabricated actually
worked). Furthermore, one of the biggest challenges we encountered was with filling in the
exposed AAO pores (and not filling in the zeolite-covered AAO pores or the zeolite pores). We
tried techniques such as directional physical vapor deposition, initiated chemical vapor
deposition and polymeric spin coating, but these methods did not every produce useable
membranes. Although ALD was not perfect, the kinks in the process were eventually fixed such
that the transport across the zeolite crystals could be studied. Even with the sorption/infiltration
experiments in Chapters 2 and 3 of this thesis, much of the actual experimental work was
developing experimental guidelines to ensure that the measurements were limited to water within
the zeolite framework. My hope is that the tools and techniques that were presented in this thesis
can be further developed such that a more improved understanding of nanoscale mass transport
within zeolites can be realized.
94
5.2
Future Work
Looking towards the future of membranes in desalination, there are still many challenges
that exist with utilizing sub-nanometer pores in the active layers of RO membranes. First, the
permeability must be shown to be at least as good as (if not higher than) current state-of-the-art
RO membranes for industry to consider switching active layer materials. Second, if these
microporous materials exhibit increases in water permeability and salt rejection, both the
biofouling and scaling behavior must be quantified to determine how these microporous
membranes will perform in long-term operation. While some of this work has already started in
various research groups (121,125,126), it is still unclear how the formation of these bio and
scaling films affect the permeability of zeolite-based (or other microporous) active layers over
time. Additionally, current AP-based RO membranes do not exhibit a high rejection of boron as
the kinetic diameter of boric acid (the dissolved form of boron found in seawater) is small
(≈ 6 Å) (4) and boric acid has similar transport properties (i.e., diffusivity) as water. Since the
zeolite pore size can potentially reject this form of boron, zeolite-based membranes may be used
as an alternative to the costly post-treatment processes that are currently used to remove boron
from desalinated water. We plan on investigating the boron rejection properties of the MFI
zeolite membranes fabricated in this study in the near future.
As is with the case of most PhD-related theses, I could not investigate every direction in
the limited timeframe of graduate school. However, I believe that future researchers interested in
nanoscale mass transport have many opportunities to impact the future of water desalination. If I
were to start a second PhD, I would investigate the topics presented in the following sections.
95
5.2.1
Diffusivity Difference between Uptake and NMR
Normalized Flux (mol m)/(m2s)
1E-3
H-MFI
Na-MFI
H-MFI Flux Prediction (1 µm)
Na-MFI Flux Prediction (0.6 µm)
MD (H-MFI) Flux Prediction (2 nm)
1E-4
1E-5
1E-6
1E-7
1E-8
1E-9
1E-10
1
10
Osmotic Pressure (MPa)
Figure 5-1. Difference in the water flux predicted by the simulations of Liu et al. (16) and
the measured water flux across the zeolites in this study.
Liu et al. studied the water flux across 2 nm thick MFI zeolite crystals for applications in reverse
osmosis using molecular based simulations (16). However, even if the measured water flux in
this study are normalized by the thickness of the zeolite crystals, the observed water flux is over
an order of magnitude lower than the molecular simulation predicted. We present possible
reasons for this decreased flux in the following sections.
In comparing the water flux across the microfabricated membranes (in addition to the
insights gained from the sorption-diffusion experiments) with the results from molecular
simulations (Figure 5-1), the experimental permeability of the MFI zeolite membranes is still
more than 40x lower than predictions made these simulations. One possibility is the existence of
additional mass transport resistances that arise from imperfections in the zeolite crystals.
The reported diffusion coefficients of water in purely siliceous MFI zeolites are shown in
Table 5-1. This substantial difference in diffusion coefficients (≈ 3 - 7 orders of magnitude)
96
between various measurement techniques have been widely reported for a number of
zeolite/sorbent systems (101,127). It has been accepted that the diffusivity obtained by NMR
(see Table 5-1) corresponds to the true diffusivity of the water within the microstructure of the
zeolite, as the NMR analysis is limited to the water already within the zeolite pore structure.
Therefore, for the uptake experiments (in which the water must first adsorb to the crystal surface
and then diffuse into the pore network), an additional transport resistance must exist to explain
the decreased diffusivity. For a perfect zeolite structure, the adsorption and the diffusion at the
surface should not induce an additional transport resistance (101,127), as the diffusion distances
along the surface to reach a pore are < 2 nm. Until recently, the origin of this decreased
diffusivity using the uptake experiments was not widely agreed upon (50). However, in the
recent review by Kärger (127) and in the work by Teixeria et al. (128), it appears that either a
collapse or blockage of the pore structure at the surface can create a surface barrier or surface
resistance. This blockage/collapse may explain the substantial difference in the diffusivity
measurements between the uptake and NMR experiments (127).
Table 5-1. Reported diffusion coefficients of water within purely siliceous MFI zeolites.
Diffusivity (cm2/s)
Measurement
Type
Source
10-14
10-10
10-9
10-7
Uptake
NMR
PFG-NMR, MD
PFG-NMR
(50,129)
(36,52)
(48,49,95,130)
(47)
If some type of structural distortion at the surface of the crystal is the source of the
decreased diffusivity measured with uptake experiments, it may be possible to remove this
disordered surface layer and achieve diffusivity values closer to that of the true diffusivity. As
the transport within the zeolites, at least during forward osmosis, is diffusion-limited, substantial
improvements in the permeability could be realized. Even an order of magnitude increase in the
diffusivity could increase the water permeability across purely siliceous zeolites to values above
those of current AP-based RO membranes. Advances in methods of nanoscale matter
manipulation (such as focused ion beam milling, low energy reactive ion etching, and chemical97
mechanical planarization) as well as improved techniques to study the mass transport within
zeolites (using interference microscopy (127) or quartz crystal microbalance techniques (131))
can help realize a higher water permeability for zeolite-based membranes.
5.2.2
Pressure Driven Transport in Sub-Nanometer Pores
Apart from the existence of the surface barriers that may be decreasing the flux, another
explanation for the lowered flux in comparison with molecular-based simulations (see Figure
5-1) could be due to the lack of a pressure-driven component of the water flux in the forward
osmosis experiments presented in this thesis. In contrast to the concentration-gradient driven
flow of forward osmosis, most molecular simulations generate a mass flow by imposing a
pressure gradient (i.e., a gradient in density) across the membrane (15,16). Additionally, most
membrane-based research (including the work with zeolite-based membranes) quantify the water
transport via a hydraulic pressure drop to drive the water flux (14,32). Yet, as the size of the
pores is less than a nanometer in diameter, it is unclear if the mass transport follows continuumbased hydrodynamic theory. In fact, this continuum description of the flow has been shown to
break down at length scales as large as 5 nm (24,81,82,106,132,133). Nevertheless, a pressuredriven flow has been observed in carbon nanotubes with diameters less than 2 nm. A
fundamental study investigating the difference between a mass flow generated by a hydraulic
pressure drop and one generated by a salinity gradient (i.e., an osmotic pressure) through pores at
these length scales would further assist in improving the understanding of transport in
nanoconfined geometries.
5.2.3
Effect of Pore Size and Tortuosity on Permeability and Salt
Rejection
Finally, it has been reported that the water flow through hydrophobic carbon nanotubes
(CNTs) is substantially higher that the flow rate predicted by the Hagan-Poiseuille equation (24).
This water flux (generated by a hydraulic pressure difference) through the ≈ 1.5 nm CNT pores
is approximately four orders of magnitude higher than the water flux generated through the
≈ 5.5 Å MFI zeolite pores. While the decreased water flux across the zeolites could be explained
by the surface resistance or by an additional water flow generated by a hydraulic pressure drop
(as described above), it is unclear if these reasons alone can account for the significant difference
98
in water flux between CNTs and zeolites. It would be a useful study to perform a parametric
investigation of the effect of the pore diameter on the generated water flux. This may be difficult
with zeolites, however, as most larger pore zeolites tend to have a high internal defect
density (25).
Furthermore, in contrast with CNTs, MFI zeolites have an interconnecting threedimensional pore structure. Although the ‘straight-through’ pores of the MFI zeolites were
oriented in the direction of the flow, the water transport was not limited to these straight
pathways. As other zeolites, such as MEL-type (i.e., ZSM-11) zeolites, contain only these
straight through pores, a study investigating the effects of pore tortuosity may elucidate more
insight into the transport through zeolites.
99
100
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
Miller, J. Review of Water Resources and Desalination Technologies. Sandia National
Laboratories Report, SAND2003-0800 (2003).
Shannon, M. A. et al. Science and technology for water purification in the coming
decades. Nature 452, 301–310 (2008).
Humplik, T. et al. Nanostructured materials for water desalination. Nanotechnology 22,
292001 (2011).
Fritzmann, C., Löwenberg, J., Wintgens, T. & Melin, T. State-of-the-art of reverse
osmosis desalination. Desalination 216, 1–76 (2007).
Cooley, H., Gleick, P. H. & Wolff, G. Desalination: with a Grain of Salt. (2006).
Research Council US Committee on Advancing Desalination Technology, N.
Desalination: a national perspective. 298 (2009).
Lee, J. & Karnik, R. Desalination of water by vapor-phase transport through
hydrophobic nanopores. J. Appl. Phys. 108, 044315–044315–11 (2010).
Lee, J., Laoui, T. & Karnik, R. Nanofluidic transport governed by the liquid/vapour
interface. Nature Nanotechnology 9, 317–323 (2014).
Pendergast, M. M. & Hoek, E. M. V. A review of water treatment membrane
nanotechnologies. Energy Environ. Sci. 4, 1946 (2011).
Wijmans, J. & Baker, R. The solution-diffusion model: a review. Journal of Membrane
Science (1995).
Lee, K. P., Arnot, T. C. & Mattia, D. A review of reverse osmosis membrane materials
for desalination—development to date and future potential. Journal of Membrane
Science 370, 1–22 (2011).
Bowen, T., Noble, R. & Falconer, J. Fundamentals and applications of pervaporation
through zeolite membranes. Journal of Membrane Science 245, 1–33 (2004).
Hilder, T., Gordon, D. & Chung, S. Salt Rejection and Water Transport Through Boron
Nitride Nanotubes. Small (2009).
Li, L., Dong, J., Nenoff, T. & Lee, R. Desalination by reverse osmosis using MFI zeolite
membranes. Journal of Membrane Science 243, 401–404 (2004).
Cohen-Tanugi, D. & Grossman, J. C. Water Desalination across Nanoporous Graphene.
Nano Lett 12, 3602–3608 (2012).
Liu, Y. & Chen, X. High permeability and salt rejection reverse osmosis by a zeolite
nano-membrane. Phys Chem Chem Phys 15, 6817–6824 (2013).
O’Hern, S. C. et al. Selective Molecular Transport through Intrinsic Defects in a Single
Layer of CVD Graphene. ACS Nano 6, 10130–10138 (2012).
Murad, S., Oder, K. & Lin, J. Molecular simulation of osmosis, reverse osmosis, and
electro-osmosis in aqueous and methanolic electrolyte solutions. Molecular Physics 95,
401–408 (1998).
Corry, B. Designing carbon nanotube membranes for efficient water desalination. J Phys
Chem B 112, 1427–1434 (2008).
Song, C. & Corry, B. Intrinsic Ion Selectivity of Narrow Hydrophobic Pores. J Phys
Chem B 113, 7642–7649 (2009).
Hughes, Z. E., Carrington, L. A., Raiteri, P. & Gale, J. D. A Computational Investigation
into the Suitability of Purely Siliceous Zeolites as Reverse Osmosis Membranes. Nano
101
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
Lett 115, 4063–4075 (2011).
Boutilier, M. S. H. et al. Implications of Permeation through Intrinsic Defects in
Graphene on the Design of Defect-Tolerant Membranes for Gas Separation. ACS Nano
140103094953001 (2014).doi:10.1021/nn405537u
Ho, C. et al. Electrolytic transport through a synthetic nanometer-diameter pore.
Proceedings of the National Academy of Sciences 102, 10445–10450 (2005).
Holt, J. K. et al. Fast mass transport through sub-2-nanometer carbon nanotubes. Science
312, 1034–1037 (2006).
Čejka, J., van Bekkum, H., Corma, A. & Schueth, F. Introduction to Zeolite Molecular
Sieves. (Elsevier Science, 2007).
Humplik, T., Raj, R., Maroo, S. C., Laoui, T. & Wang, E. N. Framework Water Capacity
and Infiltration Pressure of MFI Zeolites. Microporous and Mesoporous Materials
(2014).
Ng, K. C., Thu, K., Kim, Y., Chakraborty, A. & Amy, G. Adsorption desalination: An
emerging low-cost thermal desalination method. Desalination 308, 161–179 (2013).
Bailey, S., Olin, T., Bricka, R. & Adrian, D. A review of potentially low-cost sorbents
for heavy metals. Water Res 33, 2469–2479 (1999).
Ayranci, E. & Duman, O. Adsorption behaviors of some phenolic compounds onto high
specific area activated carbon cloth. Journal of hazardous materials (2005).
Li, L., Liu, N., McPherson, B. & Lee, R. Enhanced water permeation of reverse osmosis
through MFI-type zeolite membranes with high aluminum contents. Ind. Eng. Chem. Res
46, 1584–1589 (2007).
Li, L., Liu, N., McPherson, B. & Lee, R. Influence of counter ions on the reverse
osmosis through MFI zeolite membranes: implications for produced water desalination.
Desalination 228, 217–225 (2008).
Jeong, B. et al. Interfacial polymerization of thin film nanocomposites: A new concept
for reverse osmosis membranes. Journal of Membrane Science 294, 1–7 (2007).
Humplik, T. Investigating the adsorption and transport of water in MFI zeolite pores for
water desalination. (2010).
Drobek, M. et al. Long term pervaporation desalination of tubular MFI zeolite
membranes. Journal of Membrane Science 415-416, 816–823 (2012).
Chen, N. Hydrophobic properties of zeolites. The Journal of Physical Chemistry 80, 60–
64 (1976).
Ahunbay, M. G. Monte Carlo Simulation of Water Adsorption in Hydrophobic MFI
Zeolites with Hydrophilic Sites. Langmuir 27, 4986–4993 (2011).
Bolis, V., Busco, C. & Ugliengo, P. Thermodynamic Study of Water Adsorption in
High-Silica Zeolites. J Phys Chem B 110, 14849–14859 (2006).
Cailliez, F., Stirnemann, G., Boutin, A., Demachy, I. & Fuchs, A. H. Does water
condense in hydrophobic cavities? A molecular simulation study of hydration in
heterogeneous nanopores. J Phys Chem C 112, 10435–10445 (2008).
Demontis, P., Gulín-González, J., Masia, M. & Suffritti, G. B. The behaviour of water
confined in zeolites: molecular dynamics simulations versus experiment. Journal of
Physics: Condensed Matter 22, 284106 (2010).
Han, A. & Qiao, Y. Infiltration of liquid water in an acid-leached zeolite. Journal of
Materials Research 22, 3538 (2007).
Jentys, A., Warecka, G., Derewinski, M. & Lercher, J. Adsorption of water on ZSM 5
102
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
zeolites. The Journal of Physical Chemistry 93, 4837–4843 (1989).
Masuda, T., Otani, S., Tsuji, T., Kitamura, M. & Mukai, S. R. Preparation of hydrophilic
and acid-proof silicalite-1 zeolite membrane and its application to selective separation of
water from water solutions of concentrated acetic acid by pervaporation. Separation and
Purification Technology 32, 181–189 (2003).
Valiullin, R., Kärger, J., Cho, K., Choi, M. & Ryoo, R. Dynamics of water diffusion in
mesoporous zeolites. Microporous and Mesoporous Materials 142, 236–244 (2011).
Zygmunt, S. A., Curtiss, L. A., Iton, L. E. & Erhardt, M. K. Computational Studies of
Water Adsorption in the Zeolite H-ZSM-5. The Journal of Physical Chemistry 100,
6663–6671 (1996).
Olson, D., Haag, W. & Borghard, W. Use of water as a probe of zeolitic properties:
interaction of water with HZSM-5. Microporous and Mesoporous Materials 35, 435–
446 (2000).
Thamm, H. Adsorption site heterogeneity in silicalite: a calorimetric study. Zeolites 7,
341–346 (1987).
Filippov, A. et al. Dynamic properties of water in silicalite-1 powder. Colloids and
Surfaces 30, 1022–1031 (2012).
Caro, J., Hǒcevar, S., Kärger, J. & Riekert, L. Intracrystalline self-diffusion of H2O and
CH4 in ZSM-5 zeolites. Colloids and Surfaces 6, 213–216 (1986).
Demontis, P., Stara, G. & Suffritti, G. Behavior of water in the hydrophobic zeolite
silicalite at different temperatures. A molecular dynamics study. J Phys Chem B 107,
4426–4436 (2003).
Zhang, K. et al. Diffusion of water and ethanol in silicalite crystals synthesized in
fluoride media. Microporous and Mesoporous Materials 170, 259–265 (2013).
Ramachandran, C. E., Chempath, S., Broadbelt, L. J. & Snurr, R. Q. Water adsorption in
hydrophobic nanopores: Monte Carlo simulations of water in silicalite. Microporous and
Mesoporous Materials 90, 293–298 (2006).
Özgür YazaydIn, A. & Thompson, R. Molecular simulation of water adsorption in
silicalite: Effect of silanol groups and different cations. Microporous and Mesoporous
Materials 123, 169–176 (2009).
Trzpit, M. et al. The effect of local defects on water adsorption in silicalite-1 zeolite: a
joint experimental and molecular simulation study. Langmuir 23, 10131 (2007).
Trzpit, M., Soulard, M. & Patarin, J. Water intrusion in mesoporous silicalite-1: An
increase of the stored energy. Microporous and Mesoporous Materials 117, 627–634
(2009).
maroo, S. C., Humplik, T., Laoui, T. & Wang, E. N. Water Infiltration in ZSM-5
Zeolites: Effect of Pore Volume and Water Structure. in ASME 2012 Third International
Conference on Micro/Nanoscale Heat and Mass Transfer 625–631 (ASME,
2012).doi:10.1115/MNHMT2012-75322
Patterson, A. The Scherrer Formula for X-Ray Particle Size Determination. Phys. Rev.
56, 978–982 (1939).
Lai, Z., Tsapatsis, M. & Nicolich, J. P. Siliceous ZSM 5 Membranes by Secondary
Growth of b Oriented Seed Layers. Advanced Functional Materials 14, 716–729
(2004).
Olson, D., KOKOTAILO, G., LAWTON, S. & Meier, W. CRYSTAL-STRUCTURE
AND STRUCTURE-RELATED PROPERTIES OF ZSM-5. J Phys Chem-Us 85, 2238–
103
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
2243 (1981).
Dupree, R., Holland, D., McMillan, P. W. & Pettifer, R. F. The structure of soda-silica
glasses: A mas NMR study. Journal of Non-Crystalline Solids 68, 399–410 (1984).
Zhang, W., Bao, X., Guo, X. & Wang, X. A high resolution solid state NMR study on
nano structured HZSM 5 zeolite - Springer. Catalysis Letters 60, 89–94 (1999).
Rouquerol, J., Rouquerol, F. & Sing, K. S. W. Absorption by Powders and Porous
Solids. (Academic Press, 1998).
Eroshenko, V., Regis, R., Soulard, M. & Patarin, J. Energetics: a new field of
applications for hydrophobic zeolites. J. Am. Chem. Soc 123, 8129–8130 (2001).
Di Lella, A. et al. Molecular simulation studies of water physisorption in zeolites. Phys
Chem Chem Phys 8, 5396–5406 (2006).
Desbiens, N., Boutin, A. & Demachy, I. Water condensation in hydrophobic silicalite-1
zeolite: A molecular simulation study. J Phys Chem B 109, 24071–24076 (2005).
Giovambattista, N., Rossky, P. J. & Debenedetti, P. G. Effect of pressure on the phase
behavior and structure of water confined between nanoscale hydrophobic and
hydrophilic plates. Physical Review E 73, 041604 (2006).
Karbowiak, T., Weber, G. & Bellat, J.-P. Confinement of water in hydrophobic
nanopores: effect of the geometry on the energy of intrusion. Langmuir (2013).
Zhao, J., Culligan, P. J., Germaine, J. T. & Chen, X. Experimental study on energy
dissipation of electrolytes in nanopores. Langmuir 25, 12687–12696 (2009).
Chen, H. & Wang, Y. Preparation of MCM-41 with high thermal stability and
complementary textural porosity. Ceramics International 28, 541–547 (2002).
Pauly, T. R., Liu, Y., Pinnavaia, T. J., Billinge, S. J. & Rieker, T. P. Textural
mesoporosity and the catalytic activity of mesoporous molecular sieves with wormhole
framework structures. Journal of the American Chemical Society 121, 8835–8842
(1999).
Zhang, K. et al. Adsorption of Water and Ethanol in MFI-Type Zeolites. Langmuir 28,
8664–8673 (2012).
Kim, S.-S., Shah, J. & Pinnavaia, T. J. Colloid-Imprinted Carbons as Templates for the
Nanocasting Synthesis of Mesoporous ZSM-5 Zeolite. Chem. Mater 15, 1664–1668
(2003).
Hsu, C.-Y., Chiang, A. S. T., Selvin, R. & Thompson, R. W. Rapid Synthesis of MFI
Zeolite Nanocrystals. J Phys Chem B 109, 18804–18814 (2005).
Babeva, T., Todorov, R., Mintova, S. & Yovcheva, T. Optical properties of silica MFI
doped acrylamide-based photopolymer. Journal of Optics A: Pure … (2009).
Majano, G., Darwiche, A., Mintova, S. & Valtchev, V. Seed-Induced Crystallization of
Nanosized Na-ZSM-5 Crystals. Ind. Eng. Chem. Res 48, 7084–7091 (2009).
Aguado, J., Serrano, D. P., Escola, J. M. & Rodriguez, J. M. Low temperature synthesis
and properties of ZSM-5 aggregates formed by ultra-small nanocrystals. Microporous
and Mesoporous Materials 75, 41–49 (2004).
Mallon, E. E., Jeon, M. Y., Navarro, M., Bhan, A. & Tsapatsis, M. Probing the
Relationship between Silicalite-1 Defects and Polyol Adsorption Properties. Langmuir
29, 6546–6555 (2013).
Scholle, K. F. M. G. J., Veeman, W. S., Frenken, P. & van der Velden, G. P. M.
Characterization of intermediate TPA-ZSM-5 type structures during crystallization.
Applied Catalysis 17, 233–259 (1985).
104
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
97.
98.
de Moor, P.-P. E., Beelen, T. P. & van Santen, R. A. In situ observation of nucleation
and crystal growth in zeolite synthesis. A small-angle X-ray scattering investigation on
Si-TPA-MFI. J Phys Chem B 103, 1639–1650 (1999).
Serrano, D. P., García, R. A., Linares, M. & Gil, B. Influence of the calcination
treatment on the catalytic properties of hierarchical ZSM-5. Catalysis Today (2012).
Savage, N. & Diallo, M. Nanomaterials and water purification: Opportunities and
challenges. Journal of Nanoparticle research 7, 331–342 (2005).
Mashl, R. J., Joseph, S., Aluru, N. R. & Jakobsson, E. Anomalously Immobilized
Water: A New Water Phase Induced by Confinement in Nanotubes. Nano Lett 3, 589–
592 (2003).
Duan, C. Anomalous ion transport in 2-nm hydrophilic nanochannels. Nature
Nanotechnology (2010).
Beckstein, O., Biggin, P. & Sansom, M. A hydrophobic gating mechanism for
nanopores. J Phys Chem B (2001).
Israelachvili, J. N. Intermolecular and Surface Forces. (Academic Press, 2011).
Semenova, S. I., Ohya, H. & Soontarapa, K. Hydrophilic membranes for pervaporation:
an analytical review. Desalination 110, 251–286 (1997).
Zhao, L., Chang, P. C.-Y. & Ho, W. S. High-flux reverse osmosis membranes
incorporated with hydrophilic additives for brackish water desalination. Desalination
(2012).
Hummer, G., Rasaiah, J. C. & Noworyta, J. P. Water conduction through the
hydrophobic channel of a carbon nanotube. Nature 414, 188–190 (2001).
Cailliez, F. et al. Thermodynamics of water intrusion in nanoporous hydrophobic solids.
Phys Chem Chem Phys 10, 4817–4826 (2008).
Karbowiak, T. et al. New insights in the formation of silanol defects in silicalite-1 by
water intrusion under high pressure. Phys Chem Chem Phys 12, 11454–11466 (2010).
Jaramillo, E. & Chandross, M. Adsorption of small molecules in LTA zeolites. 1. NH3,
CO2, and H2O in zeolite 4A. J Phys Chem B 108, 20155–20159 (2004).
Dzhigit, O. M., Kiselev, A. V., Mikos, K. N., Muttik, G. G. & Rahmanova, T. A. Heats
of adsorption of water vapour on X-zeolites containing Li+, Na+, K+, Rb+, and Cs+
cations. Trans. Faraday Soc. 67, 458–467 (1971).
Nagao, M. & Morimoto, T. Differential heat of adsorption and entropy of water
adsorbed on zinc oxide surface. The Journal of Physical Chemistry 73, 3809–3814
(1969).
Demir, H., Mobedi, M. & Ulku, S. A review on adsorption heat pump: Problems and
solutions. Renew Sust Energ Rev 12, 2381–2403 (2008).
Geankoplis, C. J. Transport Processes and Separation Process Principles. (Pearson
College Division, 2003).
Bussai, C. et al. On the diffusion of water in silicalite-1: MD simulations using ab initio
fitted potential and PFG NMR measurements. Applied Catalysis A: General 232, 59–66
(2002).
Coudert, F., Cailliez, F., Vuilleumier, R., Fuchs, A. & Boutin, A. Water nanodroplets
confined in zeolite pores. Faraday Discuss 141, 377–398 (2009).
Pellenq, R. J. M., Roussel, T. & Puibasset, J. Molecular simulations of water in
hydrophobic microporous solids. Adsorption 14, 733–742 (2008).
Framework Water Capacity and Infiltration Pressure of MFI Zeolites.
105
99.
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
116.
117.
Linstrom, P. J. & Mallard, W. G. NIST Chemistry WebBook; NIST Standard Reference
Database No. 69. (2001) (2001).
Redondo, A. & Hay, P. J. Quantum chemical studies of acid sites in zeolite ZSM-5. The
Journal of Physical Chemistry 97, 11754–11761 (1993).
Kärger, J. & Ruthven, D. M. Diffusion in zeolites and other microporous solids. (WileyInterscience, 1992).
Pascal, T. A., Goddard, W. A. & Jung, Y. Entropy and the driving force for the filling of
carbon nanotubes with water. Proceedings of the National Academy of Sciences 108,
11794 (2011).
Falk, K., Sedlmeier, F., Joly, L., Netz, R. R. & Bocquet, L. Molecular Origin of Fast
Water Transport in Carbon Nanotube Membranes: Superlubricity versus Curvature
Dependent Friction. Nano Lett 10, 4067–4073 (2010).
Fornasiero, F. et al. Ion exclusion by sub-2-nm carbon nanotube pores. Proceedings of
the National Academy of Sciences 105, 17250 (2008).
Majumder, M., Chopra, N. & Hinds, B. J. Mass Transport through Carbon Nanotube
Membranes in Three Different Regimes: Ionic Diffusion, Gas, and Liquid Flow. ACS
Nano (2011).
Majumder, M. & Corry, B. Anomalous decline of water transport in covalently modified
carbon nanotube membranes. Chemical Communications 47, 7683–7685 (2011).
Lai, Z. et al. Microstructural optimization of a zeolite membrane for organic vapor
separation. Science 300, 456–460 (2003).
Ha, K., Lee, Y. J., Lee, H. J. & Yoon, K. B. Facile Assembly of Zeolite Monolayers on
Glass, Silica, Alumina, and Other Zeolites Using 3-Halopropylsilyl Reagents as
Covalent Linkers. Advanced Materials 12, 1114–1117 (2000).
Lee, J. S., Kim, J. H., Lee, Y. J., Jeong, N. C. & Yoon, K. B. Manual assembly of
microcrystal monolayers on substrates. Angew Chem Int Edit 46, 3087–3090 (2007).
Lee, I., Buday, J. & Jeong, H. µ-Tiles and mortar approach: A simple technique for the
facile fabrication of continuous b-oriented MFI silicalite-1 thin films. Microporous and
Mesoporous Materials (2009).
Lee, I. & Jeong, H.-K. Synthesis and gas permeation properties of highly b-oriented MFI
silicalite-1 thin membranes with controlled microstructure. Colloids and Surfaces 141,
175–183 (2011).
Liu, Y., Li, Y. & Yang, W. Fabrication of highly b-oriented MFI monolayers on various
substrates. Chemical Communications (2009).
Hayhurst, D. T. & Paravar, A. R. Diffusion of C1 to C5 normal paraffins in silicalite.
Zeolites 8, 27–29 (1988).
Shah, D. B., Chokchai-Acha, S. & Hayhurst, D. T. Measurements of transport rates of
C4 hydrocarbons across a single-crystal silicalite membrane. J. Chem. Soc., Faraday
Trans. 89, 3161–3167 (1993).
Gordon, R. G., Hausmann, D., Kim, E. & Shepard, J. A kinetic model for step coverage
by atomic layer deposition in narrow holes or trenches. Chemical Vapor Deposition 9,
73–78 (2003).
Farmer, D. B. & Gordon, R. G. Atomic Layer Deposition on Suspended Single-Walled
Carbon Nanotubes via Gas-Phase Noncovalent Functionalization. Nano Lett 6, 699–703
(2006).
Dendooven, J., Deduytsche, D., Musschoot, J., Vanmeirhaeghe, R. L. & Detavernier, C.
106
118.
119.
120.
121.
122.
123.
124.
125.
126.
127.
128.
129.
130.
131.
132.
133.
Modeling the Conformality of Atomic Layer Deposition: The Effect of Sticking
Probability. J. Electrochem. Soc. 156, P63 (2009).
Probstein, R. F. Physicochemical hydrodynamics : an introduction. (New York : Wiley,
1994).
van't Hoff, J. H. The role of osmotic pressure in the analogy between solutions and
gases. Journal of Membrane Science 100, 39–44 (1995).
Pendergast, M. M. & Hoek, E. M. A review of water treatment membrane
nanotechnologies. Energy Environ. Sci. 4, 1946–1971 (2011).
Zhu, B. et al. Desalination of seawater ion complexes by MFI-type zeolite membranes:
Temperature and long term stability. Journal of Membrane Science 453, 126–135
(2014).
Liu, N., McPherson, B., Li, L. & Lee, R. Factors Determining the Reverse-Osmosis
Performance of Zeolite Membranes on Produced-Water Purification. (2007).
Lind, M. L. et al. Influence of Zeolite Crystal Size on Zeolite-Polyamide Thin Film
Nanocomposite Membranes. Langmuir 25, 10139–10145 (2009).
Varoon, K. et al. Dispersible exfoliated zeolite nanosheets and their application as a
selective membrane. Science 334, 72–75 (2011).
Li, D. & Wang, H. Recent developments in reverse osmosis desalination membranes.
Journal of Materials Chemistry 20, 4551–4566 (2010).
Lu, J., Liu, N., Li, L. & Lee, R. Organic fouling and regeneration of zeolite membrane in
wastewater treatment. Separation and Purification Technology 72, 203–207 (2010).
Kärger, J. In-depth study of surface resistances in nanoporous materials by microscopic
diffusion measurement. Microporous and Mesoporous Materials (2013).
Teixeira, A. R. et al. Dominance of Surface Barriers in Molecular Transport through
Silicalite-1. J Phys Chem C 117, 25545–25555 (2013).
Humplik, T., Raj, R., Maroo, S. C., Laoui, T. & Wang, E. N. Effect of Hydrophilic
Defects on Water Transport in MFI Zeolites. Langmuir
Arı, M. U., Ahunbay, M. G., Yurtsever, M. & Erdem-Şenatalar, A. Molecular dynamics
simulation of water diffusion in MFI-type zeolites. J Phys Chem B 113, 8073–8079
(2009).
Sasaki, I., Tsuchiya, H., Nishioka, M., Sadakata, M. & Okubo, T. Gas sensing with
zeolite-coated quartz crystal microbalances—principal component analysis approach.
Sensors and Actuators B: Chemical 86, 26–33 (2002).
Boukhvalov, D. W., Katsnelson, M. I. & Son, Y.-W. Origin of anomalous water
permeation through graphene oxide membrane. Nano letters 13, 3930–3935 (2013).
Won, C. Y. & Aluru, N. R. Structure and dynamics of water confined in a boron nitride
nanotube. J Phys Chem C 112, 1812–1818 (2008).
107