Physically Crosslinked Hydrogels: Impact of Interfaces and Stress on Structure and Properties

PHYSICALLY CROSSLINKED HYDROGELS: IMPACT OF

INTERFACES AND STRESS ON STRUCTURE AND PROPERTIES

A Dissertation

Presented to

The Graduate Faculty of The University of Akron

In Partial Fulfillment

of the Requirements for the

Degree Doctor of Philosophy

Clinton G. Wiener

May, 2017 PHYSICALLY CROSSLINKED HYDROGELS: IMPACT OF

INTERFACES AND STRESS ON STRUCTURE AND PROPERTIES

Clinton G. Wiener

Dissertation

Approved: Accepted:

______Advisor Department Chair Dr. Bryan D. Vogt Dr. Sadhan Jana

______Co-Advisor and Committee Member Dean of the College Dr. Robert Weiss Dr. Eric J. Amis

______Committee Member Dean of the Graduate School Dr. Nicole Zacharia Dr. Chand Midha

______Committee Chair Date Dr. Matthew Becker

______Committee Member Dr. Bi-min Zhang Newby

ii Adapted from Ref:

Wiener, C. G.; Weiss, R. A.; Vogt, B. D. Overcoming confinement limited swelling in hydrogel thin films using supramolecular interactions. Soft Matter 2014, 10 (35), 6705-

6712.

with permission from The Royal Society of Chemistry.

Wiener, C. G.; Tyagi, M.; Liu, Y.; Weiss, R. A.; Vogt, B. D. Supramolecular

Hydrophobic Aggregates in Hydrogels Partially Inhibit Ice Formation. Journal of

Physical Chemistry B 2016, 120 (24), 5543-5552.

This is an unofficial translation of an article that appeared in an ACS publication. ACS

has not endorsed the content of this translation or the context of its use.

Reprinted with permission from

Wiener, C. G.; Wang, C.; Liu, Y.; Weiss, R. A.; Vogt, B. D. Nanostructure Evolution during Relaxation from a Large Step Strain in a Supramolecular Copolymer-Based

Hydrogel: A SANS Investigation. Macromolecules 2017, 50 (4), 1672-1680

The methods to control hydrogels’ toughness, ultimate stress, and fracture energy have received attention by researchers in recent years. The toughness and fracture energy can be increased by incorporating energy dissipating mechanisms. Physical crosslinks that can break and reform during stressing provide a simple method to modulate hydrogel mechanics. Hydrophobic physical crosslinks are a well suited for increasing toughness and fracture energy. This work investigated a physically crosslinked hydrogel composed of a random copolymer containing hydrophilic segments [N,N-dimethylacrylamide or N- isopropylacrylamide] and hydrophobic segments [2-(N-ethylperfluorooctane sulfonamido)ethyl acrylate (FOSA)] that form a network by the hydrophobic aggregation of FOSA segments into nanodomains. This system provides a model system for tough physically crosslinked hydrogels. How physical crosslinks impact swelling in laterally confined thin films, how the physical hydrogel and nanodomains deform in stress relaxation, and how absorbed water is altered due to confinement between hydrophobic nanodomain crosslinks were studied.

Unlike the significantly reduced swelling of chemically crosslink thin film gels, the nature of the physical crosslinks’ allows rearrangements in laterally confined thin films.

These rearrangements allow the thin films to obtain an equilibrium swelling ratio similar to bulk. The osmotic pressure of the hydrophilic chain swelling induces these rearrangements. The result of this is the ability of physically crosslinked thin films ability to overcome thin film swelling constraints. An equilibrium water

iii content/distribution can be obtained by rearrangement of the network. The route by which physical crosslink domains breakup and reform was identified by performing 1D elongation during small angle neutron scattering with contrast matching (CM-SANS).

The stress induced relaxation time was found to span 5 orders of magnitude when fit with seven Maxwell elements. From the CM-SANS measurements, the relaxation times of the physical crosslinks and the interconnecting hydrophilic segments were found to correlate with distinct macroscopic stress relaxation times. The CM-SANS measurements also revealed that the domain breakup occurs primarily by the pullout of the FOSA segments only after the interconnecting swollen segments have become significantly strained.

Upon relaxation of the physical crosslinks, the physical domains display a spring-like rebound effect, increasing their spacing, and hydrophilic chain stretching, in the opposite direction in which the network is strained macroscopically. This strain-transverse domain spacing increase only relaxes after the physical crosslinks have reformed.

The effect of the hydrophobic moieties, which are impermeable to water, on the absorbed water behavior was studied. The fraction of supercooled water measured at

200K increases with increasing the copolymer FOSA fraction as found by differential scanning calorimetry. Hydrogels’ primarily supercool water by binding to hydrophilic segments, but here the hydrophilic fraction decreases with increasing FOSA content. The cause of this was found to be the result of water nanoconfinement. Probing the dynamics of the absorbed water with neutron scattering revealed that as the FOSA fraction is increased, water mobility below 240K is higher than in bulk supercooled water. This demonstrates that the nanoconfinements increase the supercooled water mobility. These physical crosslinks provide a route to supercool water in soft tissue-like hydrogels.

iv ACKNOWLEDGEMENTS

During my 5 years at the University of Akron many individuals have helped to make this dissertation possible. Special thanks to my tireless supervisor and friend, Dr.

Bryan Vogt. His mentoring gave me a strong start in the right direction when I began 5 years ago. His ability to let me find my own way, while also insuring I stayed on the right path, has been critical to what I have accomplished and allowing me to achieve success.

His knowledge of neutron scattering allowed me first chance at studying the microstructure and behavior of materials. This unknowingly sparked my interest in the powerful technique which has allowed me to discover many great things during my study and fostered my interest in the molecular interactions of materials. I would also like to thank my co-advisor Dr. Bob Weiss. His support in my studies and insightful questions of my work and results has continued to morph the way I think and understand my research.

I would like to especially thank him for being there not only as an advisor, but a friend when research was getting tough. He sought to not only see my work be successful, but for me to also be happy and seek enjoyment in everything I did.

I wish to thank Dr. Yun Liu, Dr. Madhusudan Tyagi, and Dr. Paul Butler. These scientists were critical in working through the experiments and results from my neutron scattering measurements. They offered insight that could not be found in papers or books.

Their input was a key part in my published work. I would also like to thank the beam line staff at NIST, especially Cedric Gagnon and Jeff Kryzon. Their help in experimental setup and troubleshooting was unmatched in assisting with the various neutron scattering

v experiments. Even when we brought setups crafted by hand, they worked to insure the best possible setup and suggested modifications to improve. A special thanks also goes out to all the friends I have made at my many trips to the NIST Center for Neutron

Research. The discussions have affected my dissertation and life in many subtle ways.

During my stay in UA, I made many great friends and colleagues. I would like to especially thank Chao Wang, for the many fruitful discussions on interpreting and understanding scattering results. Through these discussions, some getting quite heated, we were able to fine tune our perception of the behavior of the system and in the course he became a great colleague and friend. Special thanks also to Dr. Jeongwoo Lee and Dr.

Sarang Bhaway. Their guidance and insights as senior members of the research group were enormously helpful in my research and my success in my PhD. I would like to extend a thank you to all of my group members, come and gone, over my 5 years in UA.

Whether you helped me, or I helped you, both have shaped into the scientist I am today.

Lastly, I would like to thank my family, for their continued support in my ever increasing desire to become a scientist. To my brothers and sisters who endlessly (this could go on for some time I recall) had to answer my questions about why and what if. I want to finally and with special emphasis thank my parents, Alphonse and Anne, for nurturing of my scientific mind and continuing to support me in my increasing levels of study. If it were not for their persistence, this long journey of education I have taken could have never been achieved. Sometimes you may have not understood why I continued with my schooling, but you knew I wanted it, so you supported me in my choice to continue on to obtaining my PhD.

vi TABLE OF CONTENTS

Page

ABSTRACT ...... iii ACKNOWLEDGEMENTS ...... v LIST OF FIGURES ...... x LIST OF TABLES ...... xxix I. INTRODUCTION ...... 1 II. BACKGROUND ...... 12 2.1 HYDROGELS ...... 12 2.2 PHYSICAL CROSSLINKS ...... 20 2.3 THIN FILMS...... 28 2.4 MECHANICAL RESPONSE OF HYDROGELS ...... 32 2.5 SUPERCOOLING OF WATER ...... 38 2.6 SURVEY OF DISSERTATION ...... 43 III. OVERCOMING CONFINEMENT LIMITED SWELLING IN HYDROGEL THIN FILMS USING SUPRAMOLECULAR INTERACTIONS ...... 46 3.1 INTRODUCTION ...... 46 3.2 EXPERIMENTAL ...... 50 3.2.1 Materials...... 50 3.2.2 Sample preparation and measurement baseline...... 51 3.2.3 Characterization of swelling ...... 52 3.2.4 Data analysis ...... 53 3.3 RESULTS AND DISCUSSION ...... 63 3.4 CONCLUSIONS ...... 85 IV. NANOSTRUCTURE EVOLUTION DURING RELAXATION FROM A LARGE STEP STRAIN IN A SUPRAMOLECULAR COPOLYMER-BASED HYDROGEL: A SANS INVESTIGATION ...... 87 4.1 INTRODUCTION ...... 87

vii 4.2 EXPERIMENTAL SECTION ...... 89 4.2.1 Materials ...... 89 4.2.2 Stress relaxation measurements ...... 91 4.2.3 Time resolved SANS measurements ...... 92 4.3 RESULTS AND DISCUSSION ...... 96 4.4 CONCLUSIONS ...... 121 V. SUPRAMOLECULAR HYDROPHOBIC AGGREGATES IN HYDROGELS PARTIALLY INHIBIT ICE FORMATION ...... 123 5.1 INTRODUCTION ...... 123 5.2 EXPERIMENTAL SECTION ...... 125 5.2.1 Materials ...... 125 5.2.2 Differential Scanning Calorimetry (DSC) ...... 125 5.2.3 Small Angle Neutron Scattering (SANS) ...... 126 5.2.4 Disk Chopper Spectrometer (DCS) ...... 127 5.2.5 High Flux Backscattering Spectrometer (HFBS) ...... 128 5.2.6 Sample Preparation for SANS and QENS Measurements ...... 128 5.2.7 QENS Data Analysis ...... 130 5.3 RESULTS AND DISCUSSION ...... 131 5.4 CONCLUSIONS ...... 158 VI. VISCOELASTIC CHARACTERIZATION OF SOFT HYDROGEL THIN FILMS USING QUARTZ CRYSTAL MICROBALANCE (QCM-D): FROM THE SAUERBREY LIMIT TO BEYOND FILM RESONANCE ...... 159 6 .1 INTRODUCTION ...... 159 6.2 EXEPERIMENTAL ...... 162 6.2.1 Materials and sample preparation ...... 162 6.2.2 Characterization ...... 163 6.3 THEORY AND MODELING OF QCM OPERATION ...... 169 6.4. RESULTS AND DISCUSSION ...... 175 6.5. CONCLUSIONS ...... 191 VII. OVERALL SUMMARY AND FUTURE STUDIES ...... 193 7.1 CONCLUSIONS ...... 193 7.2 FUTURE WORK ...... 197

viii 7.2.1 In situ straining of physical gel ...... 197 7.2.2 Study of physical crosslink bond energy ...... 197 7.2.3 Ice inhibition versus hydration content ...... 198 IIX. REFERENCES ...... 199 APPENDIX 1. MATHEMATICA CODE FOR QCM MODEL ...... 222 APPENDIX 2. LIST OF PUBLICATIONS ...... 225 APPENDIX 3. REPRINT PERMISSIONS ...... 227 APPENDIX 4. FINANCIAL AND GRANT SUPPORT ...... 229

ix LIST OF FIGURES

Figure Page

Figure 2-1. Stress-strain curves of DMAA-NC-M1 gels with different clay content

(NC2.5 to NC7). All hydrogels had the same polymer/water ratio (1/10 (w/w)). 48

[Reprinted with permission from Ref. 48] ...... 14

Figure 2-2. Loading curve of PAMPS/PAAm DN gel under uniaxial elongation at an elongation strain rate of 0.13 s-1, and pictures demonstrating the necking process. The inserted letters represent the correspondence between the pictures and the arrowed data points .40 [Reproduced with permission from Ref. 40] ...... 15

Figure 2-3. Engineering tensile stress versus stretch ratio for DN (double), TN (tertiary),

QN (quaternary), DDN (true double), DTN (true tertiary), and DQN (true quaternary) hydrogels. Note that the QN hydrogel did not break during the tensile experiment.31

[Reprinted with permission from Ref.31] ...... 16

Figure 2-4. Engineering tensile stress versus stretch ratio for c-DN hydrogels made of

SAPS(1,4,2,9) as the first network. Black circles: virgin SAPS(1,4,2,9)/ AAm(2,0.1,0,97) c-DN hydrogel; blue circles: SAPS(1,4,2,9)/AAm(2,0.1,0,97) c-DN hydrogel predamaged with a compression load; red circles: virgin SAPS(1,4,2,9)/AAm(2,0.1,0,9) cDN hydrogel; green curve: subtraction of the blue data from the black data (see explanation in text).41 [Reprinted with permission from Ref.41] ...... 17

x demonstrating partial self-healing of the sample. f, Stress–strain curves of the virgin and self-healed sample P(NaSS-co-DMAEA-Q) 2.0–0.52. The self-healing in f is performed at 25 ◦C for 24 h in water, and others are indicated in the text.46 [Reprinted with permission from Ref.46] ...... 19

Figure 2-6. Fracture energy G of connected-DN (c-DN) gels and truly independent-DN

(t-DN) gels as a function of (a) the cross-linker density of the second network.. Open symbols indicate too weak samples to be measured by tearing test. Solid straight lines are the fitted curve to the experimental equation as described in the text.42 [Reprinted with permission from Ref.42] ...... 22

Figure 2-7. The stress–strain curves PVA reinforces by hemicelluloses and Chitin of Gel-

1, Gel-3, Gel-7, and Gel-9 (the number indicates the number of freeze-thaw cycles performed on the sample),36 [Reprinted with permission from Ref.36] ...... 23

Figure 2-8. Demonstration of the self-healing (a, c, d, e, f, g) and the stretching (b) properties of the DMAEMA–SCMHBMA hydrogel at pH 8. The gel in (a) and (b) was coloured with methyl blue for better imaging. Optical microscopy images (c–g) were obtained from a hydrogel film with an incision (c) after annealing at 50 1C (d, e) and subsequent cooling to 20 1C (f, g).44 [Reproduced with permission from Ref.44] ...... 24

Figure 2-9. Properties of hybrid hydrogels. (a) Stress-stretch curves of gels of various concentrations of alginate. Each test was performed by pulling an unnotched sample to rupture. (d) Stress-stretch curves of gels of various fractions of short-chain alginate to the total alginate.47 [Reprinted with permission from Ref.47] ...... 25

Figure 2-10. Rheology of bulk- and solution-synthesized L-polymer (crystalline) and R- polymer (amorphous) hydrogels. Mechanical properties of 20 wt % solution-synthesized

xi and 25 wt % bulk-synthesized L/R-polymer hydrogel. Molecular weight of end-block

71 total varies slightly (70 vs 72 DPPLA). [Reprinted with permission from Ref.71] ...... 27

Figure 2-11. The Flory-Rehner 1D swelling theory52 states that a constrained film will have a swelling ratio when compared to the bulk free swelling material by its power to the one-half...... 29

Figure 2-12. Comparison of the overall degree of swelling between the surface-attached

(b) and unconstrained bulk DMAAm networks (9).53 [Reprinted with permission from

Ref.53] ...... 30

Figure 2-13. Equilibrium swelling ratio (A) of NIPA gels in each fixed 0]1 as a function of temperature on heating. (B) Uniaxial strain dependence of the transition temperature of NIPA gels on heating. Solid circle denotes the transition temperature on heating without extension.98 [Reprinted with permission from Ref.98] ...... 32

Figure 2-14. Stress–strain diagram of the notched water-saturated hydrogel (ratio

ED2003:PGDGE = 1.5:2) reinforced with the 133x133 construct. The crack length (a) varied, with all other parameters being kept constant.34 [Reprinted with permission from

Ref.34] ...... 33

Figure 2-15. Recovery after the first loading. Each hybrid gel sample was first loaded to a stretch of λ = 7 , and then unloaded. The samples were then stored at a certain temperature for a period time, followed by a second loading at room temperature. Stress stretch curves are shown for samples stored at: a, 20 ℃; b, 60 ℃. The alginate-to- acrylamide ratio was 1:6. The covalent crosslinker, MBAA, was fixed at 0.0006 the weight of acrylamide. The ionic crosslinker, CaSO4, was fixed at 0.1328 the weight of alginate.43 [Reprinted with permission from Ref.43]...... 35

xii Figure 2-16. Comparison of measured relaxation times (in black) and calculated expectation based on diblock TR-SANS experiments (in red) at a reference temperature of 70 °C. (a) shows results for SEPS (17-53-17) and (b) shows results for SEPS (45-144-

45).101 [Reprinted with permission from Ref.101]...... 37

Figure 2-17. Crystallization points for water confined inside MCM-41-S samples having different pore sizes. The sharp negative-going peaks signal the freezing temperatures. It is noted that the samples with pore size 18 A do not show an obvious freezing peak. The open symbols represent the approximated glass transition point on the DSC curve for samples with no peak.112 ...... 40

Figure 2-18. Water in each state of hydration (mass water/mass dry gel). Peaks I, II, III, and IV indicate: free water and increasing states of being bound, from 2 to 4. With the strongly bound water indicated as UFW (unfrozen fraction of water). UFW denotes unfrozen water that remains unfrozen for sample cooled to 173K. 134 ...... 42

Figure 3-1. Dissipation curves for the 10 nm NF5 film at the start of a swelling run with temperatures for each step labeled. The three curves represent the 3rd overtone

(~15MHz) for: raw data ( ), with quadratic correction factor applied ( ), and with linear correction factor applied ( ). Note that the linear correction overcorrects the dissipation at high temperatures and results in a non-physical increase in dissipation upon collapse of the hydrogel when the temp is raised from 25 ˚C to 35 ˚C...... 55

Figure 3-2. Kinematic viscosity,  (▲) of water as a function of temperature (data obtained from NIST Chemistry WebBook)158. Solid black provides a quadratic fit of the data...... 56

xiii Figure 3-3. Overtone dependent temperature correction factors for ΔF and ΔD according to the best fit of a quadratic to a bare sensor in water (overtones labeled on the graphs).

These changes are subtracted from the raw data to correct for all temperature dependencies in the system. Error bars indicate the standard deviation for a measurement on three separate sensors...... 57

Figure 3-4. Influence of the correction on the temperature dependent thickness for a 32 nm (dry) NF5 film using the Sauerbrey approximation and viscoelastic modeling. Solid symbols represent the raw uncorrected data and open symbols are the temperature corrected data. Inset shows that after temperature correction, the Sauerbrey and viscoelastic modeled thicknesses converge in the high temperature limit. The overtones are determined by the Sauerbrey expression (3rd , 5th ) and viscoelastic model ( ).

For comparison, the thickness determined from SE ( ) is also included...... 59

Figure 3-5. The frequency (F) and dissipation (D) curves from QCM-D for 3rd, 5th, 7th, and 9th overtones associated with the cooling of the 32 nm (A and B) and 120 nm films

(C and D). The time is related to decreasing of the temperature from 35 ˚C to 5 ˚C. The

QCM-D at low temperatures behaves in an unusual manner with an increase in both frequency and dissipation, which is indicative of film resonance...... 61

Figure 3-6. Sigmoid fit of SE data for 75 nm (dry) NF5 film. Thickness components

(h/h0)max and (h/h0)collapsed are associated with the swelling (vertical) axis, while the

LCST and  are associated with the temperature (horizontal) axis...... 63

Figure 3-7. Volumetric swelling ratio (V/V0) for NF5 thin films equilibrated in water as the temperature is decreased from 35 ˚C to 5 ˚C, determined by SE (A) and QCM-D (B).

NF5 film thicknesses shown are: 10 nm (), 32 nm (▼), 52 nm (◂▸), 75 nm (), 100 nm

xiv (◆), and 120 nm (). The open symbol 10 nm film was modeled only in QCM-D due to its limited optical path length. Comparison to bulk (●) volumetric swelling measurement is provided as reference. (C) Thickness swelling ratios (left axis) from SE () and QCM-

D () at 5 °C and normalized volumetric swelling ratio for the films relative to the bulk hydrogel (right axis)...... 64

Figure 3-8. Comparison of hydrogel thickness from QCM-D and SE. For consistency, the thickness from SE and QCM-D should be identical (dashed line). A linear fit of these data before the frequency upturn at high swelling fractions is shown by the solid blue line. This suggests that the average thickness of the coupled water layer associated with

QCM-D is 26 ± 12 nm. NF5 film thicknesses shown are: 32 nm (▼), 52 nm (◂▸ ), 75 nm

(), 100 nm (◆), and 120 nm ()...... 67

Figure 3-9. Comparison of swelling curves for SE and coupled water layer corrected

QCM-D, where the average coupled water layer of 26 nm has been subtracted. The open symbols correspond to SE measurements and the solid symbols correspond to QCM-D measurements after applying the 26 nm coupled water layer correction. The measured thickness from QCM-D and SE at temperatures greater than 18 ˚C is very similar except for the 75 nm film...... 69

Figure 3-10. Analysis of the swelling data using equation (1) to determine (A) TLCST and

(B) LCST transition width (2σ) from the QCM-D () and SE () measurements as a function of film thickness. The dashed lines illustrate the average for each measurement along with the associated standard deviation for each data set. The bulk data (solid red line) is included as a reference...... 71

xv Figure 3-11. The viscoelastic properties of the hydrogels determined from QCM-D for the (A) viscosity and (B) shear elastic modulus as a function of temperature for the NF5

◂▸ ), 75 nm (), 100 nm (◆), and 120 nm (). At temperatures greater than 20-25 °C, the dissipation becomes sufficiently small that the viscoelastic properties cannot be accurately expressed by fitting of the QCM-D data as evidenced by the non-monotonic changes and large thickness variance especially for the shear elastic modulus...... 72

Figure 3-12. (top) Frequency and dissipation for F3 ( ), F5 ( ), D3 ( ), and D5 ( ) as a function of time (temperature) with fits using the viscoelastic model (solid black lines).

(middle) Corresponding thickness (h, green line) from the viscoelastic model fit and associated temperature (red line). (bottom) The shear elastic modulus (, green) and shear viscosity (, light blue) obtained from the fits using the viscoelastic model. Breaks in the data presented are transition regions where the film is not in equilibrium. The vertical dashed line illustrates where the fit parameters remain constant at each temperature step, which corresponds to a dissipation of approximately 10 × 10-6...... 75

Figure 3-13. The frequency dependence exponent for the shear elastic modulus using the

Q-Tools extended viscoelastic model for initial thickness of 10 nm ( ), 32 nm ( ), 52 nm ( ), 75 nm ( ), 100 nm ( ), and 120 nm ( ). A consistent exponent of approximately 0.4 is obtained for temperatures less than 20 °C, which indicates weak frequency dependence for the shear elastic modulus in the MHz regime. One outlier is the

10 nm film with no apparent dependence on frequency for the shear elastic modulus. ... 79

Figure 3-14. The frequency dependent exponent for the shear viscosity using Q-Tools extended viscoelastic model for initial film thicknesses of 10nm ( ), 32nm ( ), 52nm (

xvi ), 75nm ( ), 100nm ( ), and 120nm ( ). The exponent is consistently near zero for all of the films except the 10 nm film. This suggests no dependence on frequency for the shear viscosity for the thicker films...... 80

Figure 3-15. Swelling of 120 nm film as measured by QCM-D (thick green line) and SE

(black line) for 25 °C to 24 °C temperature steps. The inset shows the swelling for the entire step...... 82

Figure 3-16. Temperature cycling for 32 nm film. Blue filled symbols denote cooling, red open symbols denote heating. Circles are for first temperature cycling loop, squares are for second loop. The initial three cooling-heating-cooling (1-3) used 1-2 °C steps, while the final heating (4) used 5-10 °C steps. Significant hysteresis is consistent with film rearrangement, but the LCST remains almost unaffected...... 85

1 Figure 4-1. H NMR spectrum of DF10 (9.7 mol% FOSA) random copolymer in CDCl3.

...... 91

Figure 4-2. The repurposed wrinkling stage allowed the sample to measured first in the unstrained state. The sample was removed from the humidity chamber (Figure 4-3) and strained to the desired strain and screwed in place to hold the sample during the stress relaxation...... 93

Figure 4-3. The custom humidity chamber used to hold the sample stage during SANS measurements. Keyways on the bottom insured repeatable placement Aluminum foil was used as the wall on the sides of the container to allow transmission of the neutron beam.

The hole cut in the plastic wall was done to insure the maximum scattering angle possible on the beam line would not encounter the plastic shell and only the aluminum foil...... 93

xvii Figure 4-4. Scattering profile for unstretchedDF10 hydrogel swollen with 27/73

D2O/H2O (DMA match) over wide Q range. The interdomain spacing peak centered at approximately 0.09 Å-1is the dominant scattering feature...... 94

Figure 4-5. Scattering profile for unstretched DF10 hydrogel swollen with 50/50

D2O/H2O (FOSA match) over wide Q range. There is an upturn at low Q and a peak associated with the shell intracorrelation at high Q...... 95

Figure 4-6. Sector averages were performed in both parallel (blue sectors) and perpendicular (green sectors) directions in reference to the strain direction (arrows indicate direction of elongational strain on sample, with the sectors center about 0° and

90° with a width of ± 22° (=44°)...... 96

Figure 4-7. (A) Chemical structure of DMA-FOSA and (B) schematic representation of the DF10 hydrogel nanostructure that consists of FOSA aggregates (grey) surrounded by a water depleted DMA phase (dark blue) dispersed in a continuous hydrated DMA phase.

...... 97

Figure 4-8. Strain to break for DF10 sample swollen in H2O at 30 mm/min. The sample failed at 400% strain. The loading curve for the stress relaxation (400 mm/min) is shown for comparison where the strain rate is larger, which leads to an increased modulus...... 98

Figure 4-9. Schematic of the structure resolved by SANS using D2O/H2O mixture to contrast match (C) DMA (27/73 (v:v) D2O/H2O) and (D) the 2-D SANS pattern to determine the interdomain spacing, D...... 99

Figure 4-10. Schematic of hydrogel structure when (E) FOSA is contrast-matched (50/50

(v:v) D2O/H2O) and (F) the associated 2-D SANS pattern to determine the nanodomain

xviii size, ξ, from the form factor associated with the shell structure. The hydrogels were measured at ~23°C in the unstretched state...... 100

Figure 4-11. (A) Stress-strain behavior associated with the extension of the DF10 hydrogel to 150% strain at 400 mm/min. (B) Stress relaxation of DF10 at a constant applied 150% strain. The solid black line is the fit of a Generalized Maxwell Model with seven elements. The calculated relaxation times are listed above the fit curve at approximately the relaxation time. The residual (shown above the fit) is the difference between the measured and fit value for the normalize stress...... 101

Figure 4-12. Maxwell element fits using 1-6 exponentials with residual error of fit shown.

Here, the label “6 Exp” is the fit and residual associated with a 6 element Maxwell model for the stress relaxation data. Residual error is shown for all 6 model results, with the bottom error figure zoomed in to enlarge the see 4-6 element models...... 103

Figure 4-13. Time resolved 2D SANS profiles and azimuthal angle dependence of the average intensity of the peak at 4, ~25, and ~420 min. after the step strain to 150% to examine (A) the interdomain distance measured from FOSA scattering (DMA contrast match) where the intensity is averaged over Q = 0.092 ± 0.024 Å-1and (B) the shell size from DMA scattering (FOSA contrast match) where the intensity is averaged over Q =

0.14 ± 0.047 Å-. The = 0° azimuthal angle is shown on the first 2D pattern of each set of data for reference...... 105

Figure 4-14. Adjusted intensity scale for the 2D scattering data at 50/50 contrast at 418 min and the 27/73 contrast at 424 min. The slight anisotropy remaining for the 50/50 case

(FOSA matched, domain form factor scattering) can be distinguished in this case, while

xix the scattering is isotropic for the 27/73 case (DMA matched, interdomain structure factor scattering). Note in the 27/73 case the beam spot cannot be seen...... 106

Figure 4-15. The 3 Maxwell element fits of the Gaussian peak amplitude for the interdomain intensity (A) and shell intensity (B). The relaxation times for each process is noted on the figure. Residual error is shown above the fits. Red lines indicate 95% confidence interval of the fits...... 108

Figure 4-16. Time dependence of the stress and intensity anisotropy measured by SANS during stress relaxation following a step-strain. The green-circles data correspond to the normalized amplitude ratio for the azimuthal peak associated with ID (DMA was contrast matched) and the blue-squares data correspond to the normalized amplitude ratio for the azimuthal peak data for Iξ (FOSA was contrast matched). The solid lines provide a GMM fit of the structural relaxation. The solid black curve is the GMM fit of the stress relaxation data reproduced from Figure 4-11 with the stress data shown as the green squares. The inset shows the same data on a linear time scale...... 110

Figure 4-17. Broad peak model fits for (A) 27/73 contrast (DMA match, interdomain scattering peak) and (B) 50/50 (FOSA match, Shell diameter peak). Both cases are for sector average across the perpendicular direction (see Figure 4-6). Points indicate Q averaged intensity across the sector and solid line shows broad peak model fit to data in the overlapping range...... 114

Figure 4-18. (A) Evolution of ξ (nanodomain size) in the parallel (‖) and perpendicular

() directions to the deformation during relaxation. (B) The FWHM of the scattering peak associated with ξ that describes the change in the distribution of sizes...... 115

xx Figure 4-19. Temporal evolution in (A) D (interdomain spacing) upon stress relaxation and (B) FWHM of the scattering peak both parallel (║) and perpendicular (┴) to the deformation. The gray area represents isotropic interdomain distance before strain within one standard deviation. The FWHM provides a measure of the distribution of D distances. (C) I(Q) near the interdomain scattering peak for several relaxation times both parallel (║) and perpendicular (┴) to the deformation. These scattering profiles reveal the asymmetric rearrangement of D showing a significant loss on the low Q (larger spacing) side in parallel direction...... 120

Figure 5-1. (A) Schematic of the hierarchical structure of the DMA-FOSA hydrogels showing core-shell morphology with FOSA core and water-depleted DMA shell. The average center-to-center distance between the supramolecular crosslinks (d), measured by

SANS as a function of FOSA content of the copolymer:  this work;  ref(17). (B) DSC heating and cooling thermograms at 2 K/min for DF5 hydrogel (80 wt% H2O). (C) DSC heating and cooling thermograms at 2 K/min for DF22 hydrogel (45 wt% H2O)...... 132

Figure 5-2. DSC thermograms for DF22 on cooling at 5K/min (green, ‘5K’) and 2K/min

(blue, ‘2K’). The transition temperatures are offset due to the rate dependence, but are qualitatively similar in terms of the peak shapes...... 135

Figure 5-3. Temperature dependent SANS profiles for (A) DF22, (B) DF15, and (C) DF5 hydrogels equilibrated at 295K with D2O for maximum total scattering...... 136

Figure 5-4. DSC thermograms of DF15 on both cooling (blue) and heating (red) at 2

K/min. Inset shows zoomed in crystallization peak on cooling...... 137

xxi Figure 5-5. SANS profiles for DF22 hydrogel swollen with 27/73 (v/v) D2O/H2O to contrast match the DMA phase. These data provide the core size, interdomain spacing, and domain clustering of the hydrogel...... 138

Figure 5-6. Representative fit to the Broad Peak model (solid line) of the SANS profile for DF22 (circles) hydrogel swollen by 27/73 D2O/H2O at 260 K. The inset illustrates the fit of the correlation peak. The residual error at each Q is shown in the top panel...... 139

Figure 5-7. Contrast variation provides the interdomain spacing (using 27/73 v/v

D2O/H2O). Error bars represent one standard deviation, if not visible error bar are smaller than the size of the symbol...... 140

Figure 5-8. SANS profiles of DF22 swollen with 50/50 D2O/H2O to contrast match the

FOSA core on cooling from 295K to 210K. Fitting these data yields information on the shell thickness and DMA clustering...... 141

Figure 5-9. Representative fit to the core-shell model with a hard sphere structure factor

(solid line) of the SANS profile for DF22 (circles) hydrogel swollen by 50/50 D2O/H2O at 260 K. The inset illustrates the fit of the correlation peak. The residual error at each Q is shown in the top panel...... 141

Figure 5-10. Comparison of the goodness of the fit of SANS profiles of the DF22

(circles) hydrogel swollen by 50/50 D2O/H2O with (A) an additional Lorentzian term to account for DMA chain scattering in solution between nanodomains and (B) considering the DMA chain scattering as background. There is no statistical difference in the overall quality of the fit...... 142

xxii Figure 5-11. Contrast variation provides the effective volume fraction of the supramolecular aggregates separated by hydrated DMA (using 50/50 v/v D2O/H2O).

Error bars represent one standard deviation...... 143

Figure 5-12. QENS measurements of proton dynamics at Q=1.25 Å-1 for DCS at (A)

295K and (B) 260K for DF22 (red), DF15 (blue), and DF5 (green). All hydrogels were equilibrated at 295K with H2O prior to measurements. Error bars represent one standard deviation...... 144

Figure 5-13. Representative fits for the DF22 hydrogel swollen with H2O at (A) 295K and (B) 220K. The full data fit is shown as the solid line with contributions from the elastic peak associated with instrumental resolution (slower motions is shown with the dotted line), the background is the horizontal dashed line, and the dashed Lorentzian associated with the relaxation processes (data of interest). Error bars in the Figures represent one standard deviation...... 146

Figure 5-14. Analysis of DCS data provides insight into the water dynamics. FWHM at

295K (A), 270K (B), and 260K (C) for H2O swollen hydrogels of () DF5, () DF15 and () DF22 illustrates water motion is locally caged Fickian diffusion due to invariant

FWHM at low Q. (D) FWHM for DF22 hydrogels is only slightly impacted by temperature: (▲) 260K, (♦) 250K, (■) 240K, (+) 220K. The horizontal dashed line provides the slowest motions resolvable with DCS. The Fickian diffusion coefficient for

H2O is calculated from the linear fit of the FWHM (solid lines). The difference in the diffusivity of H2O within the hydrogels attributed to the variation in local hydration.

Error bars represent one standard deviation...... 148

xxiii

Figure 5-15. Calculated effective self-diffusion constant of H2O within the hydrogels from DCS, () DF5, () DF15, and () DF22 with comparison to bulk water ( ) as measured by NMR219 and QENS217 and predicted by simulations ( ).116 Below 260K, no motions in DF5 or DF15 hydrogel were resolvable, while water remained mobile down to

220K in the DF22 hydrogel. Error bars in all figures represent one standard deviation. 149

Figure 5-16. Representative fits of the EISF data from DCS measurements for the DF22 hydrogel at 295K and 270K...... 150

Figure 5-17. Comparison of cage size from EISF (empty symbols) and caged Fickian diffusion (filled symbols) for DF5 (green circle), DF15 (blue square), and DF22 (orange triangle). Error bars represent one standard deviation...... 151

Figure 5-18. The fit of the EISF from DCS provides the fraction of protons, n(t), in the sample that can be resolved by DCS. Sigmoid fits added to guide the eye. Error bars in all figures represent one standard deviation...... 152

Figure 5-19. Arrhenius fits (solid lines) to the average FWHM from HFBS data for D2O swollen DF22 (grey circles) and H2O swollen DF22 (blue squares) hydrogels. Error bars represent one standard deviation...... 153

Figure 5-20. The (A) cage radius and (B) fraction protons with resolvable motions by

HFBS are obtained from EISF fits of the HFBS data of DF22 swollen in H2O (blue squares) and D2O (grey circles). Error bars represent one standard deviation...... 153

Figure 5-21. Slower dynamics in the system were probed using HFBS with the Mean

Squared Displacement for DF22 and DF5 swollen in D2O, H2O, or dry (no water) during cooling at 2 K/min. Error bars represent one standard deviation...... 155

xxiv

Figure 5-22. The reversibility of both the dynamics and mechanical properties are shown

-1 for the DF22 gel swollen with H2O. A and B show the DCS signal at 1.1 Å as measured at 250K and 270-275K on cooling (orange) and heating (purple), respectively. The slightly different temperatures for 270-275K is explained in the text...... 156

Figure 5-23. The reversibility of mechanical properties is shown for the hydrogels swollen with H2O. A shows the frequency behavior of the dynamic (G’) and loss (G”) shear moduli measured at room temperature from a pristine sample of DF22 (orange) and the thawed state (purple) after cooling the sample in liquid nitrogen for 20 minutes. B shows an image of the DF5 and DF22 gels pre freezing and post freezing. Samples outlined in red for clarity. Scale bar (lower right) represents 1 cm...... 157

Figure 6-1. The raw Psi and Delta for the bare sensor with fits from the model in Table 6-

1 shown as black dashed lines...... 164

Figure 6-2. The raw Psi and Delta for the coated sensor with dry NF5 film with fits from the model in Table 6-2 shown as black dashed lines...... 165

Figure 6-3. The raw Psi and Delta for the coated sensor with swollen NF5 film in liquid cell at 65˚ with fits from the model in Table 6-3 shown as black dashed lines...... 166

Figure 6-4. Flow chart showing inputs and outputs for fitting of the QCM data with the viscoelastic (Voigt) model...... 168

Figure 6-5. Flow chart illustrating the inputs and outputs associated with fitting of the

QCM data when applying the rheological model from Shull and coworkers...... 168

xxv encountering an interface for reflection, which probes the “bulk” properties. For thinner films, the hydrogel-water interface is encountered by the shear wave, which can result in

(B) more commonly partial wave reflection from the interface to slightly decrease the frequency at the sensor surface or (C) coupling of the reflected wave with the propagating wave that effectively increases the frequency at the sensor (film resonance).

...... 170

S S Figure 6-7. Impact of D/λs on the real, RM , (green lines) and imaginary, XM , (blue lines) parts of impedance for the fundamental frequency (5 MHz). Four regimes of operation:

(1) Sauerbrey, (2) viscoelastic, (3) near film resonance, and (4) bulk-like (sensor is heavily dampened in most cases and the shear wave is predominately impacted by the film). A film that is highly elastic (dashed lines) (ϕ = 0.2, tan(/2) = 0.1) and fully viscous (solid lines) (ϕ = /2, tan(ϕ/2) = 1) are shown. The separation between (2) and (3) is dependent on the phase angle; the regimes shown are for the viscous case...... 172

Figure 6-8. Temperature corrected QCM-D frequency and dissipation behavior for 52 nm

(A and B) and 100 nm (C and D) thick dry films for the hydrogel immersed in water

(,3rd OT; , 5th OT; , 7th OT; and ◊, 9th OT). The fits using the V model (dashed colored lines) are shown for binary combinations of overtones. The fit with the R model is shown by the solid black lines...... 177

Figure 6-9. Comparisons in the temperature dependent film thickness of the hydrogels determined from SE and QCM-D using the Voigt (V) and rheological Kramers-Kronig

xxvi better illustrate the differences, the top panel shows a residual plot for the difference in the thickness (QCM-SE)...... 180

Figure 6-10. Temperature dependent shear modulus determined from fit of the QCM-D data with the V (dashed lines) and R models (open symbols) for hydrogels with dry copolymer film thicknesses of (A) 52 nm and (B) 100 nm...... 183

Figure 6-11. Shear viscosity of the hydrogel films predicted from the QCM-D data using the V (dashed lines) and R (open symbols) models for initial copolymer thickness of (A)

52 nm and (B) 100 nm...... 184

Figure 6-12. Temperature dependence of D/λs from the QCM-D measurements of the hydrogel film for (A) 52 nm and (B) 100 nm thick dry copolymer using film thickness from (●) SE or () QCM with V model viscoelastic results from the 3rd/5th overtone modeling. The hashed regions correspond to the predicted D/λs for failure of Sauerbrey model (5% deviation) with the rightward ascending hashed lines170 and the temperature range where this failure is expected (leftward ascending). The gold circle indicates where the viscosity and shear modulus are no longer dependent on the viscoelastic model. ... 186

Figure 6-13. Calculated phase angle from the fit results of the 52nm (Dry) NF5 film using the viscoelastic model...... 187

Figure 6-14. Calculated phase angle from the fit results of the 100nm (Dry) NF5 film using the viscoelastic model are shown with the solid color symbols. For comparison, phase angle calculated at higher temperatures for the 52 nm (dry) NF5 film is shown in the open symbols...... 187

Figure 6-15. Temperature dependence of D/λs from the QCM-D measurements of the hydrogel film for 52 nm thick dry copolymer using film thickness from (●) SE or (■)

xxvii

QCM with V model viscoelastic results from the 5th /7th overtone modeling. The hashed regions correspond to the predicted D/λs for failure of Sauerbrey model (5% deviation) with the rightward ascending hashed lines170 and the temperature range where this failure is expected (leftward ascending). The gold circle indicates where the viscosity and shear modulus predicted from the V and R models begin to agree – at higher temperatures, there is significant disagreement between the models...... 189

Figure 6-16. Frequency dependence of the storage moduli for the hydrogels. The modulus obtained from QCM-D data with the two viscoelastic models are nearly an order of magnitude greater than the apparent plateau in the modulus obtained from the bulk sample measured on a conventional rheometer...... 190

xxviii

LIST OF TABLES

Table Page

Table 3-1. The temperature correction parameters for the five overtones for both frequency and dissipation...... 57

Table 5-1. Fit parameters from broad peak model of DF22 hydrogel at 260K with the associated uncertainty (σ)...... 139

Table 5-2. Fit parameters from core-shell model of DF22 hydrogel at 260K with the associated uncertainty (σ)...... 142

Table 6-1. Model used for fitting of blank QCM sensor (QSX-335, silica coated sensor for ellipsometric characterization). Layer 2 is sputtered silica, with optical parameters obtained separately by manufacturer. Layer 1 is TiO2 intra-layer. Substrate is thick Ti layer. Fit protocol followed as described by Ramos et al.153 ...... 164

Table 6-2. Model used for fitting of coated QCM sensor. Layer 2, Layer 1, and Substrate are defined as fit from the previous bare sensor fit for the identical sensor. The polymer layer is modeled as a Cauchy layer (Layer 3). The layer only requires thickness and A and B fit parameters to obtain a good fit of the raw psi and delta measured...... 165

Table 6-3. Model used for fitting of coated QCM sensor in water in the liquid cell. Layer

2, Layer 1, and Substrate are defined as fit from the previous bare sensor fit for the identical sensor. The polymer layer is modeled as a Cauchy layer (Layer 3) that is now swollen in water. The layer only requires thickness and A and B fit parameters to obtain a good fit of the raw psi and delta measured. Additional Ambient index now being water

xxix must be accounted for. Also, window corrections are provided by Delta offset, Window

#1. These are found by previous measurements of known substrates...... 166

xxx

CHAPTER I.

INTRODUCTION

Hydrogels are defined as a polymer network (gel) in which its swelling agent is water by the International Union of Pure and Applied Chemistry (IUPAC).1 These networks can have greater than 99% water by mass. While they generally contain a majority of water, hydrogels behave solid like. The contained water does not drain from the network as it is held by osmotic swelling force. Additionally, due to their high water content and solid like behavior, hydrogels are used in medicine and biology, especially for applications of drug delivery and wound healing as the high water content aids in the loading of drugs and keeping the wound site properly hydrated.2-6 In this case the hydrogel can be imbibed with a drug or appropriate substance to slowly release by diffusion to the applied site.7-9 Some hydrogels are extremely transparent and thus offer the ability to see through the “bandage’ or application patch to the wound or treatment site.2 In recent years their application has been found in nearly any use where contact with water may occur: coatings,9-12 drug delivery,7-9 wound healing,2-4, 13 aqua-robotics,9,

14-19 tissue regrowth,20-23 stem cell differentiation,24, 25 sensing,14, 26, 27 etc….

One major limitation has been the brittle nature of the traditional hydrogel materials which are covalently crosslinked. These networks are much like rubbers, with low to intermediate crosslinking to allow moderate chain mobility. As hydrogels absorb water, the mobility of the chains allows swelling of the network by osmotic pressure. As a result the chains are in a strained state and with minimal force of elongation or

1 compression, hydrogels fail catastrophically. The material breaks into several jagged pieces, a clear indication of the sudden brittle type failure mechanism.28 To increase the modulus at break of the hydrogels, the network can be crosslinked to a much further degree at the lost of elasticity and they also imbibe less water and fail in a brittle catastrophic manner. With a low crosslinking density, they can undergo increased strain with a lower modulus at break as there is more freedom in the network. These hydrogels still fail in a brittle manner and at low stress levels as the propagation of a crack through the network meets minimal resistance.

Due to the many potential applications of hydrogel which require robustness, and the potential for new applications as a result of improved robustness, there has been significant work in the field on trying to improve the strength and toughness of the hydrogels to overcome their current limitations.16, 17, 29-39 Some of the pioneering advances on significantly improving the toughness and strength hydrogels are made by

J.P Gong and coworkers.28, 40 They used a double network design to make hydrogels with strain to break and toughness that far surpasses previous traditional hydrogels. Although this method yield materials which demonstrate improved properties, Nakajima et al. and

Es-Haghi and et al. showed that the toughness and strength only occur in the first strain cycle, as the first network is sacrificed on straining the sample.41, 42 This mechanism results in damage to the network and gives a mechanism by which the material can dissipate some of the strain energy without complete failure.

Inspired by this, several groups have followed the lead of incorporated dissipation mechanisms to yield improved toughness in hydrogels. One notable work was to use an ionic crosslinking mechanism in alginate gels as the secondary weak network in an

2 elastic low-crosslink covalent network.43 This gives a dissipation mechanism which is reversible, as the physical bond formed by the ionic coordination can be broken on straining, and reforms the same or new bonds thereafter. The material can be deformed repeatedly whilst still maintaining the same mechanical properties. One downside to this has been the recovery time of this bond can be as long as 24 hrs to obtain the original material properties. This is a result of the slow restructuring of this physical ionic bond after breaking.

Several groups have sought out other physical crosslink dissipation mechanisms and types to improve the mechanical properties of hydrogels. Many new types of systems have been developed, such as hydrogen bonded,36, 44, 45 polyampholyte,30, 46 covalent plus ionic dual networks,17, 43, 47 composite reinforced (both fiber and particle),3, 34, 48 and many more that far surpass the properties of the traditional hydrogels first developed.

Some of these systems have been characterized by their mechanical properties, and only now are researchers beginning to ask what is occurring at the microscopic level in these systems. From this understanding, the development of systems which take direct advantage of the known mechanisms can be produced, as well as matching a system’s mechanical properties to its intended use.

The key to the toughness of these new hydrogel types is due to the dissipation mechanisms, which includes but is not limited to: physical bond dissociation,33, 37, 49 chain sliding,42, 50, 51 weak network rupture,16, 17, 28, 40, 41 and fiber pullout34, 45.

Understanding the behavior of these dissipation mechanisms under load and in time is of great interest for their application and in new designs. This is especially important for the physically crosslinked systems where the crosslinks are transient. This does not pose a

3 significant issue in dual covalent/physical systems, but in some hydrogels, the network is formed only by physical associations. Although these can produce systems with extraordinary elongation to break and strength, the shape can deform and flow under deformation and in time.

In this work we focused on how physical crosslinks rearrange under difference applied stress and how physical crosslinks aspects of only physical crosslinked systems.

(1) To understand how the equilibrium swelling of the physical network is altered by lateral confinement in a thin film coating. Can the osmotic force of the hydrophilic segments which are not at their equilibrium water content induce rearrangement of the physical network to a new conformation allowing complete swelling? During the forces rearrangement of physical crosslinks by application of a force, how do the physical moieties breakup and reform and what is the mechanism by which they rearrange into their new conformations. (2) As the physical crosslinks form nanoaggregates that act to retard water motion as they are nearly impermeable to water, the water inside these gels experiences a tortuous and confined mean path of diffusion. We wanted to quantify this decreased diffusion and how the confinement of the water by the soft nanodomains may induce supercooling of the water as supercooled water in hydrogel is normally achieved by hydrogen bonding. The physical hydrogels may offer a potential route to increase the fraction of supercooled water, without addition of increased hydrogen bonding. Typical confinement induced supercooling occurs only in hard materials such as silica or carbon.

Soft tissue like materials for preservation at low temperatures may be possible with the ability to supercool water by confinement in physical hydrogels.

In the first part of this dissertation, the physically crosslinked hydrogel composed of a random copolymer of 95 mol% N-Isopropylacrylamide (NIPAAm) and 5 mol% 2-

N(ethylperfluorooctane sulfonamide)ethyl acrylate (FOSA), NF5, was used to test physical networks in substrate confined thin films. It is known that thin film coatings that undergo substantial swelling experience substrate confinement limited swelling as the network is allowed to only swell in one dimension.52 This is a result of the ability of the polymer network to undergo swelling/expansion in only a singular direction, rather than

3-dimensionally. This has been well demonstrated in hydrogels by Toomey et al.53 and it is found that the hydrogel follows the Flory Huggins 1D swelling theory52 of a network.

The theory states that any network when allowed to swell only in 1D dimension will obey a swelling rule where the swelling ratio (SR) is equal to the same network swollen in 3D

1/2 to the square root power, i.e. SR1D = (SR3D) . A physical network’s equilibrium SR is a balance of the crosslink (hydrophobe in this case) strength and the swollen hydrophilic segment (NIPAAm here). Our goal was to determine whether or not the substrate confinement effects which decrease the network swelling ratio could be overcome by the osmotic force of the hydrophilic chains which are not fully swollen. This osmotic force may have the potential to induce rearrangement of the physical crosslinks to a new more swollen state than expected from a confined thin film. We also sought to determine if there would be a film thickness effect critical for rearrangements to occur. The findings of the study confirmed that the thin films of the physical hydrogel did overcome the predicted SR (2.12) from the 1D swelling theory, obtaining a SR of 4 in the thickness films examined as compared to the bulk SR of 4.5 at 25˚C. The thinnest films demonstrated a unexpected SR of 5, surpassing the bulk SR. This may be the result of a

5 surface layer in the thin films which shows an enhanced swelling ratio. As the total film thickness decreases, this surface layer is an increased portion of the total film. This shows first that these thin films of physically crosslinked hydrogel are able to undergo rearrangement driven by osmotic pressure to obtain a new conformation of the network.

It was also found that the swelling ratio increases with decreasing film thickness which may be the result of a surface layer.

From the thin film rearrangement study, we also found the lower critical solution temperature (LCST) for the NF5 thermoresponsive copolymer was increased as compared to that measured in the bulk. This would suggest the rearrangement of physical crosslinks results in an altered state of the polymer chains thereafter. The increased LCST could be ascribed to a change in the crosslink density similar to the surface layer effect.

From observation of the LCST versus film thickness we see no trend, thus as the hydrophobic to hydrophilic fraction of the chains remain the same, the only remaining explanation is a change in the chain conformation in the confined thin films. The second part of this dissertation motivated by the origin of the cause of the increased LCST, was to study process physical aggregates undergo upon rearrangement from a force. As the physical crosslinks are transient in nature, the mechanism by which they rearrange and reform will be very important to their application and respective environments. Using a similar random copolymer with 90 mol% N,N-dimethylacrylamide (DMA) and 10 mol%

FOSA, DF10, we probed the stress relaxation behavior of the physical aggregates in 1D extension. The DF10 was chosen as it has no LCST transition and thus makes investigation of the physical hydrogel less challenging as the temperature effects will have minimal effect. The 1D extension models a very similar strain as the network would

6 experience in the thin film confined swelling. The nanostructure of these hydrophobically crosslinked hydrogels can be described as having small nanodomains composed of a core of aggregated FOSA units. Due to the high density of clustering, and that each FOSA segment has a DMA monomer on each side of a FOSA monomer (assuming perfect random distribution) there is a water depleted shell of DMA (protonated chains) as determined from previous neutron scattering experiments.54 The individual domains are linked by hydrated chains of DMA. The MW of the linear chains was found to be about

40 KDa, so we believe each copolymer chain, of hydrophilic segments with hydrophobic crosslinking FOSA units, is connected to several domains. This structure is further described in the later chapters of the research investigations performed.54 By use of appropriate phase contrast matching with mixtures of D2O/H2O during SANS, a detailed picture of the structure change for each component as it changes during relaxation can be obtained. During relaxation, probing of the change in the domain aggregate shape and size, and the stretching of the interconnecting chain segments swollen in the water phase is possible. The network shows a distribution of relaxation times in the stress relaxation as fit by a summation of single Maxwell elements. From the contrast matched SANS analysis the domain and interconnecting chain segments were found to have distinct relaxation times that coincide with specific relaxation times found in the stress relaxation.

SANS also revealed that the aggregate becomes highly deformed following the step strain. The aggregate response is more like plastic flow, with a first a critical strain before yielding. Surprisingly there is rapid recovery to an almost uniform state (~90% recovery) with a long return of over several hours for the last few percent. This is most likely the result of the plastic deformation of domain moieties at room temperature and its glassy

7 nature preventing full rapid recovery. Upon breakup of the physical crosslinks after critical strain/time, the physical domains display a spring-like rebound effect, increasing their spacing (hydrophilic chain stretching) in the opposite direction in which the network is strained macroscopically. This transverse to strain domain spacing increase only begins to relax after the physical crosslinks have reformed (~90% recovery). These points would suggest there is a process by which breakup and relaxation of the network occurs, with each phase having a distinct contribution to overall macroscopic stress relaxation witnessed.

As the physical crosslinks are transient in nature, their deformation and time response behavior in hydrogels and other networks they incorporated into. One aspect that is not commonly addressed is the effect of these physical crosslinks on the absorbed water motions. They form a nanophase composite with a water swollen matrix and physical crosslink domains that are impermeable to water. As a result, the water motions are disrupted when encountering these nanodomains-physical crosslinks. Additionally, the water also experiences hydrogen bonding with the hydrophilic segments. In the third part of this dissertation, how these two factors can significantly influence the water self diffusion dynamics in the physical gels is shown. Using DF5-DF22, we investigated the effect of the nanodomain spacing on the ability of the hydrogels to induce supercooling of the water and reduce diffusion at room temperature. All three hydrogels studied showed over an order of magnitude reduction in the water self-diffusion constant at room temperature, but on cooling to 240K only the DF22 gel had a significant fraction of supercooled water with higher mobility compared to the bulk water at the same temperature. Using differential scanning calorimetry (DSC) it was found that the

8 unfrozen water fraction scales from 0.38 (unfrozen water/dry polymer mass) to 0.2, from

DF22 to DF5. As the increased FOSA gel, DF22, also has a lower fraction of hydrophilic

DMA, this demonstrates how physical aggregates can lead to confinement induced supercooling of the water.

In the fourth part of this dissertation, the physical gels were used as a medium to test the operational behavior of the quartz crystal microbalance (QCM) and compare methods of viscoelastic analysis. The physical gels allow precise control over thickness and the behavior is known for the thin film swelling. This allows us to perform a fundamental study on the in-operando behavior of the QCM. Using the NF5 copolymer, we are able to tune the swelling ratio by simple modulation of the temperature of the device and push the operation regimes of the QCM from the simple Sauerbrey behavior to viscoelastic regime and beyond into the film resonance and bulk like sensor response.

As the use of QCM has become a prevalent tool in many laboratories, the regard to its use and output results has not been well understood by many users. By use of the physical gel, we sought to provide to the QCM user community an overview of: 1) the operational regimes of the QCM, 2) comparison and contrast between two methods of analyzing the resultant data and cases for their best use, and 3) note of caution when performing viscoelastic analysis in regions of low dissipation sensitivity that may lead to incorrect analysis results and provide a simple factor to calculate the operation regime one may be in to access the plausibility of the full analysis. Using two dry film thicknesses of 52 and

100 nm of the previous NF5 copolymer, we are able to probe the 3 operational regimes of the QCM as the polymer shows a SR of up to 4.5 at the lowest temperatures and SR of

1.1 at higher temperatures due to its LCST responsive behavior. In both films, we show

9 the possibility for the occurrence of film resonance, where the frequency increases

(frequency change decreases) with increasing mass, opposite to what is normally seen in most QCM user conditions. Where the common convention is as the frequency change increases (fF-fI, negative result for all changes) the mass of the attached layer is increasing. In the film resonance condition, the frequency change decreases, while measurements from simultaneous spectroscopic ellipsometry (SE) show the film continues to swell, increasing its mass. Although this condition has been demonstrated before, it has required a very thick film (>1000nm)55 or very complicated system of stacked vesicles in solutions (with thickness approaching that of the film here).56 With this resonance condition established, we use this film which is changed from a glassy state at high temperature to highly swollen with significant viscous dissipation at low temperature to test the capability of the analysis software packages to follow the reversal in frequency, and yet predict an accurate film thickness. We confirmed that with recursive fits, the QCM Voigt model57 is able predict this frequency upturn and give thicknesses within 10% error of the measured thickness by SE. The model comparison between the QCM Voigt viscoelastic model and the Rheological solution by Shull and coworkers58, 59 showed good agreement in thickness, but differed slightly in the viscosity and shear modulus. The Shull model improves in the quality of solution as the film thickness increased while the Voigt model begins to produce erratic results for the thicker films as the QCM signal becomes highly dampening by the overly thick and dissipative film. These two models, although similar, are best suited for as follows: voigt model for thin films low dissipation and Shull model for thick films with large dissipation. For the

Voigt model, a phase angle associated with an elastic film (ϕ ≈ 0) is obtained in the low

10 dissipation limit, which corresponds to Sauerbrey behavior. In the low dissipation regime, the Shull model solution fails to the viscous limit (ϕ ≈ /2). Upon investigation to the region where the two models fail to different limits, a region of insensitivity to viscoelastic results can be non-dimensionalized by film thickness divided shear wavelength, and the resulting value corresponds well between the two measured here films, and with the previous work by White et al.60 suggesting the deviation of the

Sauerbrey model by ~5% error. At a value of film thickness divided by shear wavelength of 0.024 or less, viscoelastic modeling must be taken with extreme caution as the QCM operation is insensitive to predicting correct viscoelastic results in this regime.

CHAPTER II.

BACKGROUND

The background review in this chapter details the progression of hydrogels. The advancement of methods to improve the mechanical properties of hydrogels is discussed.

The reason for the need of improved mechanical properties in hydrogels and where these enhanced hydrogels may be used is presented. A review of the methods which have shown promise for improving the toughness of hydrogels and the relevant benefits and disadvantages of these methods are discussed. Physical crosslink types are outline with a brief explanation of each with benefits and shortcomings. Their conception and use in hydrogels are presented. Then discussing the use of hydrogels in thin film coatings and their shortcomings and why these occur. Following is a brief discussion on the transient nature of physical crosslinks as the bond is not permanent. How this transient nature is important and can be advantageous. A discussion on the need for why intimate understanding of the stress accelerated rearrangement is important is presented. Finally the possibility for soft materials for supercooling of water and the mechanisms by which this may be achieved are presented. With the current methods by which supercooling is achieved and why this occurs.

2.1 HYDROGELS

Hydrogels have many potential applications for use in vivo, medicine, and water borne applications.7, 8, 18, 21, 25, 39, 61 Hydrogels can be used as substrates for cell growth and stem cell differentiation.21, 24, 25, 62 Many of these applications are due to hydrogels’

12 high water content, sometimes > 99 wt% water, and their solid-like properties.

Traditional hydrogel systems tend to be very weak (peak stress of 0.2 MPa at break) and brittle.28, 45, 63 The traditional hydrogel fracture energy (10 Jm-2)64 is low when compared to natural tissues such as cartilage (1000 Jm-2) 65 and muscle tissue (2500 Jm-2 )66. As a result of traditional hydrogels poor performance, many researchers have sought methods to improve the mechanical properties. With improved mechanical properties, the potential hydrogel applications can be increased to uses such as cartilage replacement20 or even bone regrowth scaffolds23 where the mechanical properties must be comparable to the replaced tissue.

One obvious way of increasing hydrogel toughness is by composite reinforcement. By forming the gel inside/around a fiber network it has been shown that the obtained hydrogel is strong and tough.34, 38, 45 Use of particle reinforce systems also shows an increase in toughness and strength.48, 67 Particle systems show great promise for enhanced mechanical properties of the hydrogels, especially when using particles approaching the sub-micrometer size. In this case the particles appear to act as multifunctional crosslinkers.38, 67 This feature gives the network significantly increased toughness. Haraguchi et al. demonstrated that clay platelets can be used as crosslinking junctions to dramatically improve the toughness of DMA hydrogels with a strain to break of over 1000% at a stress of 250 KPa, Figure 2-1.48 The underlying mechanism for increased toughness is that for the network to propagate a crack/damage, it must break multiple chains across a single crosslink point. The multifunctional crosslinking is the origin of their improved toughness and crack blunting ability.38, 48, 67 The multifunctional

13 crosslink points are in contrast to covalent gels where a single chain breaking allows continued propagation of the crack.29

Figure 2-1. Stress-strain curves of DMAA-NC-M1 gels with different clay content

(NC2.5 to NC7). All hydrogels had the same polymer/water ratio (1/10 (w/w)). 48

[Reprinted with permission from Ref. 48]

One of the first notable methods on improving hydrogel toughness and strength without addition of foreign materials such as particles or fibers was shown by Gong et al. with their double network gels.28 These gels use a primary network which is then swollen in the same or different monomer which is then polymerized inside the primary network, creating a second network. The result is a brittle primary network and an elastic secondary network. When straining the gel, the primary (i.e. weak) network is broken.

The load is transferred to the remaining second elastic network and failure of the sample is prevented. This gel system was demonstrated to have stress at fracture of over 10 MPa and was resistant to slicing from a cutter.28, 40 The downside to this mechanism of toughening is that during cycling of stress in these networks, their strength and toughness is not preserved, Figure 2-2. The bond breakage of the weak primary network is the mechanism by which energy dissipation occurs. Due to the irreversible nature of the

14 covalent bond it can only act to dissipate energy once before the load is distributed to the second network as the primary network is broken.

Work by Es-haghi et al. showed that double network hydrogels forms a grafted second network and not a fully distinct separate second network.31 With removal of the double bonds in the first network, where grafting could occur, allowed synthesis of a true second network where the two networks only interacts by chain entanglements. This true- double network gel was shown to give slightly enhanced properties over the traditional single network.31 The increased toughness from the two distinct networks, not the graft networks, originates from the entanglement of the second network with the first and

15 results in a reversible dissipation mechanism, Figure 2-3. The second network being confined in the first network must flow and relax on stretching, and this dissipates strain energy in the process. A similar effect can result if the first network in the grafted double network is damaged by application of small strain, and the remains of this network must flow around the second elastic network. Es-haghi showed this by first destroying the primary network by loading the sample in compression, Figure 2-4. The compression damages the first network linkages and only the second network remains with regions of high crosslink density. The two networks remain highly entangled though. On testing, the resulting response will act more like a thermoplastic system, in that the residual chain segments are entangled and must relax and slide past each other. Even with the primary network significantly damaged from the compressive strain, a hydrogel with an increased toughness and elasticity can be obtained at a water content of >90%.

Figure 2-3. Engineering tensile stress versus stretch ratio for DN (double), TN (tertiary),

QN (quaternary), DDN (true double), DTN (true tertiary), and DQN (true quaternary) hydrogels. Note that the QN hydrogel did not break during the tensile experiment.31

[Reprinted with permission from Ref.31]

Figure 2-4. Engineering tensile stress versus stretch ratio for c-DN hydrogels made of

Using the concept of internal energy dissipation as a means to increase hydrogel mechanical properties, numerous ideas for ways to create tough, strong, and/or elastic gels have been developed. Using the idea of relaxation as a mechanism to dissipate energy, we see the advent of slide ring gels, which give the chains mobility and thus slight energy dissipation and load distribution, whilst also keeping the chains tied together in a network.50, 51, 68A further improvement was made by the incorporation of non-covalent bond (physical bonds) into the primary covalent network.16, 33, 37, 46, 47, 69, 70

Physical bonds can be defined as any bond which does not form covalent linkage between the polymer chains. These bonds have an energy per bond much lower than covalent bonds and tend to also be transient in nature with a finite relaxation time. In some cases the relaxation time of the bonds can be regarded as essentially infinite and

17 perform as strong as covalent bonds, such as in polymer crystals. The bonds of a crystal are a composite strength of many monomer to monomer crystal interactions that when summed give quite remarkable strength are very stable of long time. Most other forms of physical bonds are less permanent, such as ionic crosslinks formed by Ca-alginate bonds incorporated in the primary acrylamide network shown by Sun et al.43 This method of toughening hydrogels is similar in concept to traditional composite materials, in which two fairly poor performing materials on their own are integrated together to benefit from each others’ beneficial attributes while also counterbalancing each others’ shortcomings.

The resulting material has properties superior to both individual systems, with a total greater than the sum of the two individual components. These physical bonds allow breakage, giving energy dissipation, but they can then reform new bonds or return to their original state on removal of the load. This gives the system good repeatability and a mechanism by which energy dissipation can occur reversibly.

Moving further with the physical bonds, some systems use only physical bonds to form the network. These physical crosslink only networks rely on the distribution of crosslink aggregate size and thus a distribution of crosslink strengths.46 Some entirely physical hydrogels show the ability to self-heal when cut or damaged. Sun et al. demonstrated a hydrogel which can be cut in two and returned to another portion of the gel to heal, Figure 2-5. Allowing a healing time, then retesting shows the new healed sample is as tough as the virgin sample with nearly identical stress/strain response. This is because the ionic crosslinks formed by polyampholytes can rearrange and aggregate to neighboring physical crosslinks when cut or broken. The healing of this hydrogel can be done with no external impetus or application of substance, only that the material is

18 returned to intimate contact with itself.46 Entirely physical systems also offer ease of manufacture into a final parts, as they can molded, coated, or extruded much like traditional thermoplastics. They can also be easily recycled as the bonds can be dissociated with heat or solvent and reformed. These physically crosslink materials can be reused after one use or if severely damaged, simply recycled and cast as a new part.

Figure 2-5. Self-recovery, fatigue resistance, adhesion and self-healing behaviors of polyampholyte hydrogels. d, Self-healing and adhesion between either two freshly cut surfaces (red and blue), or a fresh and an aged surface (white) of samples. e, Images demonstrating partial self-healing of the sample. f, Stress–strain curves of the virgin and self-healed sample P(NaSS-co-DMAEA-Q) 2.0–0.52. The self-healing in f is performed at 25 ◦C for 24 h in water, and others are indicated in the text.46 [Reprinted with permission from Ref.46]

The one major issue of physical crosslinks lies in the nature of physical crosslinks: they are transient. Meaning the physical crosslinks have a lifetime in and out of the physical crosslink aggregate much like surfactants in micelle systems. This is especially important for non-glassy/non-crystalline type physical crosslinks, as the glassy/crystalline nature of a crosslink increases its crosslink lifetime significantly compared to a similar physical aggregate which may be rubbery or in a liquid state.71

This secondary feature of crystallinity/glassy adds an additional energy penalty which must be overcome to remove the crosslink from the aggregate/moiety.71

2.2 PHYSICAL CROSSLINKS

Physical crosslinks have been a mainstay in polymers for longer than we have fully understood their contributions.72, 73 With the first notable case being with semicrystalline polymers where the soft amorphous regions are reinforced by the hard crystalline domains that also crosslink the linear polymer chains together as a network.72,

74 This is especially important for polyethylene (PE). This is because the Tg of the amorphous chains is -150˚C while the Tm of the crystals is 130˚C. This difference between the melting of crystals and the solidification of the amorphous regions allow the non-crystalline regions to be in a melt state, giving mobility to the polymer chains and allowing flexibility and impact resistance while the crystalline domains link the single chains of the melt into a network. This is what gives PE its resilience and toughness. By understanding this effect researchers were able to incorporate similar mechanisms into other polymers to obtain robust novel systems.75-77

A second key improvement came with block copolymers, namely thermo-plastic- elastomers (TPE) made of triblock copolymers with a soft middle segment and hard end segments.75, 78-80 The hard segment forms glassy or crystalline domains while the soft segment remains rubbery or melt like at the operation temperature. A common TPE being styrene-butadiene-styrene (SBS), which uses the glassy nature of styrene and the rubbery nature of butadiene to produce an elastomeric material at room temperature which does

20 not need covalent crosslinking.75, 76 It was possible to injection mold and extrude final

TPE products that needed no additional processing.75 This was further developed and used in thermo-plastic-urethanes (TPU)77 which use blocky polymer chains with very high molecular weight (>300 kg/mol) to be produced with simpler methods as compared to that need for the block copolymers for TPEs. The hard segments form crystalline hard domains, while the soft segments permit mobility and flexibility in the material.75 From this many new products and applications were possible and many new products and ideas followed. The advent of these new easy to manufacture systems and their many new products was all made possible by the ability to physically crosslink polymers.

Recently, physical crosslinks have begun to be incorporated into hydrogels as mentioned previously.7, 8, 16, 18, 25, 30-32, 34-36, 38, 39, 45, 81 Physical crosslinks take on several forms, but the mains classes can be divided into: entanglement, hydrogen bonds, crystalline domains, ionic bonds, and hydrophobic aggregates. The physical crosslinks have been incorporated into covalent networks as well to produce standalone entirely physically crosslinked systems.

Entanglement enhanced hydrogels increase the toughness of hydrogel systems by allowing relaxation and flow of strands within the network; while the covalent crosslinks

(CX) provide the framework of the network.31, 42, 51 These chains do not contribute to the strength of the network, but instead soften and increase the toughness much like the mechanisms that make soft rubber tough and elastic. Es-haghi demonstrated how the double network system is transformed into an entanglement enhanced hydrogel after the first network is damaged by compression.31 Additional incorporation of more numbers of networks can be combined much like how the second networks was added to the first in

21 the distinct double network hydrogels. The increased number of networks allows a larger amount of dissipation to occur. This increased dissipation is due to the additional entanglements in the more frustrated system. To show the pronounced effect of an secondary network which is only entangled without the need to induce damage to the primary network, Nakajima et al. reacted all of the remaining double bonds in the primary network, where grafting form could occur, before proceeding with the addition of further network incorporation (Figure 2-6).42

Figure 2-6. Fracture energy G of connected-DN (c-DN) gels and truly independent-DN

The addition of entanglements as physical crosslinks can only slightly improve the toughness and strength of the hydrogels.31 Hydrogen bonds can be used in polymers for example in polyethylene by incorporation of carboxylic acid groups or by sulfonation of ethylene-propylene-diene rubber.82-84 Hydrogen bonds can be very strong, and they

22 typically occur in multiplets of many hydrogen bonds formed in succession along the polymer chain, further increasing the cohesion force between chains.36, 85 This effect is apparent in polyvinyl alcohol hydrogel which are formed by chains zipping together as they form hydrogen bonded crystallites.36, 86 Guan et al. demonstrated how with increasing number of freeze-thaw cycles a PVOH hydrogel can go to longer extensions before breaking, Figure 2-7.36 This increased toughness is from forming larger hydrogen bonded crystallites, while also increasing the hydrogel defects. This gives the network increased elasticity while also increasing the physical bond strength. These gels are not very tough though as they form inhomogeneous networks and frequently have brittle regions. Using these hydrogen bonds formed by 2-Ureido-4-pyrimidone (UPy) incorporated into selected monomer, Cui and Campo developed hydrogel films with self healing capabilities (Figure 2-8).44 The UPy can physically crosslink via quadruple H- bonding forming strong physically crosslinked hydrogels.

Figure 2-7. The stress–strain curves PVA reinforces by hemicelluloses and Chitin of Gel-

1, Gel-3, Gel-7, and Gel-9 (the number indicates the number of freeze-thaw cycles performed on the sample),36 [Reprinted with permission from Ref.36]

Although hydrogen bonds can add significant strength to hydrogels, they also act to increase the osmotic pressure due their propensity to form hydrogen bonds also with water. Hydrogen bond strength is also highly temperature dependent. On the contrary to hydrogen bonds, ionically crosslinked moieties can be incorporated into nearly any hydrogel system. Incorporation of an ionic second network in a covalent network by Sun et al. using alginate-Ca+ multivalent interactions produces hydrogels with elongation to break of over 1700%.16 Li et al. demonstrated that not only was the content of the ionic crosslinking agent important (Figure 2-9a), but that by tuning the relative molecular weight of these alginate chains that form the physical network inside the covalent network, they could tune the strain to break from 800% to 2500% (Figure 2-9d).47 Ionic

24 crosslinks offer the addition of a reversible energy dissipation mechanism and they can also be screened by addition of monovalent salts.87 The monovalent salt ions can screen the ionic interactions of the moieties on the polymer chain. When this occurs, the ionic moieties between chains are not as effective and this the cumulative number of physical bonds are decreased.

The final class of hydrogels to discuss is those with hydrophobic physical crosslinks. These systems are well adapted for hydrogels as the swelling of the gel by water increases the strength of the physical bond as the energy penalty for removal of a hydrophobic segment into the water matrix is increased beyond that of the polymer matrix.88 Typically these bonds are not exceptionally strong. To obtain stable gels most networks require incorporation of a large fraction of hydrophobic crosslinking segment, in some cases nearly 50% of the volume of the swollen gel is the hydrophobic moieties in

25 order to obtain hydrophobic physical aggregation strong enough to overcome the osmotic swelling force of the hydrophilic segments.89 One advantage of this method of physical crosslinking is the wide selection of possible chemical structures: hydrocarbons, fluorocarbons, lipid, and inorganic, and behaviors: glassy, rubbery, and crystallizable; that can be used in the physical crosslinking mechanism. All manner of materials can be used, as long as there is driving force to promote segregation from the water phase.29, 37

Hydrophobic aggregates tend to have low sensitivity to pH and ionic additions when compared to ionic crosslinks. Hydrophobic physical crosslinks tend to suffer to a lesser degree from temperature dependency as compared to hydrogen bonds, unless the group chose for the crosslinking is particularly temperature sensitive. This mechanism is very similar to that used in the TPE, TPU, and block copolymer systems, where the driving mechanism for crosslink strength can be related to the energy of the segregation between the hydrophobe segment and the matrix phase, in hydrogel case being mostly water.

In the work performed by Tew and coworkers, they showed that the hydrophobic crosslinker with crystallizable poly-Lactic Acid (PLA) gave improved properties over the amorphous PLA, serving as the hydrophobic crosslinker, Figure 2-10.71 They found an increase of over 2 times in the modulus simply by allowing the hydrophobic phase ability to crystallize.

Figure 2-10. Rheology of bulk- and solution-synthesized L-polymer (crystalline) and R- polymer (amorphous) hydrogels. Mechanical properties of 20 wt % solution-synthesized and 25 wt % bulk-synthesized L/R-polymer hydrogel. Molecular weight of end-block

71 total varies slightly (70 vs 72 DPPLA). [Reprinted with permission from Ref.71]

As hydrogels contain a large fraction of water, they differ significantly from the polymer only systems where much of our understanding of physical crosslinks originates.

The behavior and transient nature of the physical crosslinks in the water swollen matrix is very important to the properties of the material over long periods of time and under stress which induces rapid physical bond reformation. The swelling of interconnected chains provide an additional withdrawing force on the physical crosslink constituent. The lifetime of the physical crosslink must be balanced with the osmotic swelling force of the hydrophilic chain segments. Once a physical crosslink segment is removed, the solvated phase has much higher mobility and rearrangements to a new lower stress can occur very rapidly as compared to traditional polymer only systems. Where in the polymer only systems (TPE, TPU), the chains need to relax and slide past each other when a physical crosslink segment is removed. In these polymer-only systems once the physical crosslink

27 is formed the aggregate it is stable, lasting for a very long time. This is dependent on the degree of segregation and whether the physical domains are glassy/crystalline.75-78, 80, 90

2.3 THIN FILMS

One area of interest and potential application of physically crosslinked hydrogels is in thin films. Hydrogel coatings can provide an interface between a hard and soft material, such as on implants where it may be made of a metal and neighbors soft tissues of the body. The hydrogel can provide a smoother transition and better compliance between the implant and tissue on movement and during regrowth to reduce inflammation and irritation.91 These thin films can also be used in drug delivery as a medium interface between the reservoir and the deliver location, or as a drug containing gel.7, 21, 26, 92-94 They can also be used in coatings in aqueous environments to encapsulate, prevent, or increase adhesion of specific organism, cells, and/or compounds.

The area of film applications has some potential issues for covalently crosslinked hydrogels. The desired film thickness must be made by appropriately gapping the layer with an upper plate to the desired thickness before crosslinking.

Or alternatively, the hydrogel prepolymer is coated and allowed to dry on the surface before crosslinking is performed in the post-coated dry state. As a result the polymer chains of the network are crosslinked into a non-gaussian, pancake like state, this is especially prevalent for ultra thin films (<100nm dry).95, 96

Following the crosslinking the dry polymer chains into a network it is then swollen to form the gel. Upon inspection the swelling ratio of the network is found to be

28 significantly less than a bulk gel prepared with the same crosslinker ratio.53, 97 This reduced swelling is primarily the result of lateral confinement and is further decreased by the non-gaussian conformation seen in ultrathin films.

A gel coated on a surface is only able to swelling in the direction normal to the substrate surface. This swelling only in the normal direction is known as the lateral confinement effect and is predicted by the Flory-1D swelling theory.52, 53

The adhesion to the substrate limits the deformation of the hydrogel such that the volumetric change is almost exclusively unidirectional through the thickness of the film. For one dimensional (1D), constrained swelling of a crosslinked polymer,53

Flory-Rehner theory52 predicts that ideal volumetric swelling ratio of the constrained network should be the square root of the isotropic bulk volumetric swelling ratio of the unconstrained network, Figure 2-11.

Figure 2-11. The Flory-Rehner 1D swelling theory52 states that a constrained film will have a swelling ratio when compared to the bulk free swelling material by its power to the one-half.

In 1D swelling, the volumetric swelling ratio is identical to the thickness increase in the swelling dimension as the other dimensions are held constant. This is especially true for thin films, where the lateral dimension is essentially infinite

29 when compared to the thickness. For chemically crosslinked poly(dimethylacrylamide) hydrogels covering a wide range of crosslink density,

Toomey, et al. found good agreement between the predicted constrained thin film swelling and experimental results as shown in Figure 2-12.53

Figure 2-12. Comparison of the overall degree of swelling between the surface-attached

(b) and unconstrained bulk DMAAm networks (9).53 [Reprinted with permission from

Ref.53]

The lateral confinement effects result in the thin film swelling ratio being lower as compared to their bulk counterpart.53 This is reduced swelling in the confined thin films is known to occur in covalently crosslinked systems as their swelling constraint is limited by the chain extension. In order for the chains to swell to their desired water to polymer ratio the osmotic pressure must overcome the covalent bond strength. As this is not possible the network is limited to the 1D swelling theory (Figure 2-11).

It is known that physical crosslinks are transient and can rearrange. In

Chapter 3, investigation on the effect of thin film confinement of a physical hydrogels will be presented. Physical gels are limited in their swelling ratio by an equilibrium balance of osmotic pressure and crosslink strength and lifetime. We

30 expect the nature of physical gel thin films to be able to overcome 1D confined swelling limitation if the residual osmotic pressure to swelling is sufficient to cause rearrangements of the physical network to a large enough extent to increase the swelling ratio by double.

It has also been shown that film or bulk gels with volume phase transitions can experience a change in their transition temperature. This change in transition temperature is the result of reduced chain conformation as it is held in and extended state (e.g. 1D elongation) or prevent from its desired expansion in two dimensions (e.g. confined swelling). As the transition temperature is thermodynamic transition, usually a result of changes in hydrogen bonding of the polymer with water, the added strain on the chains changes this transition temperature (e.g. for lower critical solution temperature, this transition temperature is increased52, 98). Suzuki et al demonstrated this effect for N-isopropylacrylamide

(NIPAAm) cylinders under 1D elongation, Figure 2-13.98 Chapter 3 also discusses the occurrence of transition temperature increase in the confined thin films and what implications this may have for the physical crosslinks on rearrangement. This also motivated our interest in the Chapter 4, the study of the physical crosslink relaxation under stress.

2.4 MECHANICAL RESPONSE OF HYDROGELS

The stress response of traditional hydrogel materials is very similar to elastic rubbers. With the advancement of hydrogels and the ability to increase their toughness and strength through use of dissipation mechanisms or reinforcing agents, their stress response has become more complicated. Fiber reinforced systems show J-type stress strain response (Figure 2-14)34 due to the fiber pullout during elongation, while double network gels show yielding and withdrawing behavior 40, 41.

Figure 2-14. Stress–strain diagram of the notched water-saturated hydrogel (ratio

ED2003:PGDGE = 1.5:2) reinforced with the 133x133 construct. The crack length (a) varied, with all other parameters being kept constant.34 [Reprinted with permission from

Ref.34]

Hydrogels are similar to rubber-elastic solids, but they also have an additional response factor due to the highly swollen nature of the network, much like a viscous polymer. The characterization of hydrogels’ stress-strain response and mechanical properties has been investigated by many, but there is still much which is not understood especially for the physically crosslinked systems. Because of this it is important that we understand the behavior of the physical crosslinks in hydrogels to best fit them to their end-use and application. A prime example of the transient nature of physical crosslinks is that the rate at which the physically crosslinked hydrogels are deformed can affect the sample differently. A difference in behavior is expected for timescales shorter versus longer than the average relaxation time of the physical crosslink. A timescale shorter or strain rate faster than the mean relaxation time of the physical crosslink results in an elastic

33 response. The physically crosslinked material can be rapidly deformed and returned to its original shape with no obvious change to shape. While a timescale longer or strain rate slower than the crosslink relaxation time would exhibit a viscous flow response in which the physically crosslink material shows a low modulus before beginning to yield and remaining at that stress level until the load is released. Once released, creep and permanent deformation of the sample is apparent. In some cases, with time and slight heating, the material can show a return of above 90% to its original shape,43 but this is highly dependent on the length of time at which the sample was held in the strained state allowing the physical bonds to obtain a new conformation.

The addition of physical crosslinks makes the behavior of hydrogels highly rate and time dependant.43, 46 These crosslinks add an additional behavior that is not fully appreciated in that they can respond in an adhesive and cohesive flow.99

With a dual network, one of physical and one of chemical, the chemical network can act almost as a fiber reinforced system. This permanent networks distributes stress while providing a strong resilient structure to the network that acts as a guide to return the system to its original state and form. The recovery of the network to its original shape can take an extended amount of time (< 24 hrs).43 The recovery requires the physical bonds to again rearrange as the covalent network returns back to its original order. This recovery process can take significant time as shown by

Sun et al., Figure 2-15.43 Using a hydrogel of covalent acrylamide and alginate chains crosslinked by multi-valent interaction with calcium they demonstrated very tough samples. In these samples, the recovery back to the original state took

34 several days. The covalently crosslinked acrylamide is suppose to act as an elastic restoring force after deformation. Although this is effective, its rate of recovery is rather poor as the restoring force is not adequate to cause rapid rearrangements of physical bonds and restricting to its original state.

One method to understand the mechanisms of flow and response in a physical network is to study its stress relaxation. From stress relaxation insight into the motions of chains and restructuring of physical crosslinks can be obtained.

Contributions of each component during stress relaxation should be propagated to macroscopic stress relaxation as they relate to the nanostructure. Rapidly loading the sample to the desired strain, the relaxation of physical bonds will be in-

35 significant due to their non-zero relaxation time. Following this loading, the relaxation spectrum can tell us what portion of stress is immediately relaxed and over what time frame, and how long residual stress remains. Both of these give insight into the use and potential application set of a material as it tells us what type of stress/strain environment it may successfully operate and when failure by flow and creep would be expected. An additional insight may be gained of how long a sample can be held in extended state before it notices significant creep or failure to return to its original form.

The relaxation of nanodomains that compose a physical crosslink can be thought of much like a micelle. The relaxation time at rest is driven by the thermal exchange of micelle molecules. Much like surfactants, block copolymers can form micelles with precisely controlled interactions between the groups of the micelles and the majority solvent phase. Using block copolymer surfactants of diblock and triblock architecture, Lodge and coworkers studied the thermal exchange kinetics of the micelle forming the block copolymer systems in organic solvent.100, 101

There was a slight differences of exchange times between the diblock versus triblock architecture. Moving further, they used concentrated triblock systems

(ABA), where the A block formed physical aggregates that crosslink the solvent swollen chains into an organogel. With the organogel they induced a stress to measure the relaxation time of the physical crosslinks when activated by stress.

They found that the relaxation time in the organogel is 4 orders of magnitude faster compared to the thermal exchange time found in their earlier micelle work, Figure

2-16. This clearly shows that the stress induced relaxation time can be significantly

36 different than the relaxation time of the physical crosslinks measured at equilibrium conditions driven by thermal exchange.

Figure 2-16. Comparison of measured relaxation times (in black) and calculated expectation based on diblock TR-SANS experiments (in red) at a reference temperature of 70 °C. (a) shows results for SEPS (17-53-17) and (b) shows results for SEPS (45-144-

45).101 [Reprinted with permission from Ref.101]

This work by Lodge and coworkers clearly demonstrates the need for understanding of the physical aggregate relaxation under stressed driven exchange.

The mechanisms for physical crosslink relaxation are understood in thermal exchange, but not as well understood in stress states. The network of physically crosslinked hydrogels contain distinct phases of physical crosslink domains, interconnecting domain chains, and a water swollen matrix. These different phases may give distinct responses under stress as they each have differing degrees of mobility.

Chapter 4 of this dissertation will investigate the nanostructure rearrangements of the physical hydrogel formed by hydrophobic physical crosslinks. The keys points of this investigation will be to understand the

37 relationship between microscopic changes as it relates to the macroscopic stress response, as well as understand how the separate nanophases of the physical gel behave and how intimately tied their response to stress is. With the understanding of how the separate phases behave under stress the future design of physically crosslinked materials can be improved. Additionally how a simple stress relaxation can be related to the intimate microscopic behavior of the system.

2.5 SUPERCOOLING OF WATER

The study of water’s many anomalous behaviors has been a large area of research for many years.102-113 From the first scientific noting of its density maximum at 4˚C114, 115 to the latest research on the glass forming nature of water102, 116. As we learn more about water and its many peculiar properties we also gain insight into its mechanisms of operation in living organisms.102

One area of research that has received significant amount of attention has been water’s ability to be supercooled.113, 117, 118 Supercooling is when a liquid is brought below its standard freezing (crystallization) temperature without forming crystals, thus remaining in a liquid phase.117, 119 On further cooling the liquid may freeze, and on reheating the melting point of crystals occurs at the standard temperature. This hysteresis in freezing and melting point indicates the degree of supercooling achieved. If no freezing occurs, the supercooled liquid can reach a glassy state and exhibit no freezing point. Water is particularly interesting as it must go through a volume increase to crystallize, so its supercooled state has unique properties unlike the liquid or solid (ice) phase. In this area of research on

38 supercooled water, the means to study the supercooled water has been achieved through two main routes, depending on the degree of supercooling wanted. Using ultrapure water, one can obtain modest levels of supercooling (250K) by careful cooling in a very clean uncontaminated vessel. More extreme supercooling to

239K requires ingenuity in separating the water into microcapsules such as droplets or in capillary tubes120 to avoid nucleation events. This can be done using microdroplets suspended in a non-miscible oil phase or through use of very small diameter capillaries120. Lowering of the freezing point can also be obtained by addition of salts, but this forms a solution with the water and hence alters its chemical potential as found by the freezing and melting point being suppressed to the same degree, so no true supercooling is achieved.121 The lower limit for bulk water freezing has been determined to be 235K, at which point spontaneous freezing occurs in the water.112, 122

A more common method of extreme supercooling has been through two means: nanoconfinement105, 107, 123, 124 and surface hydration layers12, 125-128. Surface hydration commonly uses proteins of known high surface water contents to induce supercooling by their attractive potential of the water.125, 127-129 The protein is concentrated and slowly hydrated to a very specific amount to insure only surface hydration occurs and no bulk free water is present. The bulk free water can prove detrimental to the structure and supercooling ability of the system. If the free water crystallizes, the formed ice has an attractive potential to draw the water from the neighboring proteins. This can also damage the proteins. By careful testing this can be avoided and the contained water can be significantly supercooled.

In contrast to this more natural means of water supercooling using proteins, some researchers employ nanoconfinement to induce strong supercooling of the water. In this method, the water is confined to pores of hard substrates with diameters of just a few nanometers. The pore size effects in hexagonally packed pores formed in mesoporous silica (MCM-41) on the supercooling of water in

Figure 2-17 shows the critical scale in which different degrees of supercooling can be achieved until the contained water no longer freezes.112 Two common materials for this are porous silica and porous carbon.107-111, 122-124, 130, 131 These two materials offer a nice contrast of substrates as they are hydrophilic and hydrophobic, respectively. From the study of confined water in these systems, a vast amount of understanding of the transitions that occur in water in the supercooled state has been gained.

The route to making these nanoconfined systems involves many steps that must be carefully carried out in sequence, or the end material will not perform satisfactorily or even be produced. This is the reason the route of protein surface hydration has begun to take hold, as the system is made by organisms which can precisely control their production and yield near perfect resultant materials.

Proteins also have to the additional benefit as understanding of the water supercooling ability may allow their use in tissue preservation. Similar to proteins, hydrogels can also exhibit water supercooling by their water hydration layers along the hydrophilic polymer chains.118, 132, 133 The hydration layers closest to the polymer where supercooled water may exist is referred to as the strongly bound hydration layer. In most hydrogel systems this is a standard amount, scaling directly with the solids content which contain the hydrophilic chains. The amount of supercooled water can only be increased by the amount of hydrogen bonded water in the strongly bound layer. One way which has shown modest amounts of supercooled water has been through the use of polymerized sugars, such as dextran with a large fraction of hydrogen bonded water approaching nearly 1/3 in ratio to the dry polymer (supercooled water fraction).133, 134

In work by Ito et al.,134 (Figure 2-18) is was shown in a polymerized dextran gel the bound water ratio is about 0.28 to polymer weight. This describes the ratio of water which is so strongly bound it will not freeze on cooling. As the hydration level of the gel is decreased from a water ratio of 1.4, the fraction of water in the different bound states decreases in order. First with the free water being removed and only once this is all gone is next the next lowest bound state of

41 water removed. This can followed on decreasing the hydration down until the strongly bound water hydration level is reached. At which point the hydration level equals the amount of water which is missing from the melting peak in accord with the total hydration at any level of hydration before. With the free water being a fraction of over 0.5 to dry gel mass. The medium bound states of water are of lower fractions, less than 0.2. The strongly bound water being is 0.28, mass water per mass dry gel. This highly bound water fraction remains the same as it is directly related to the number of hydrogen bonding sites on the polymer chains.

The strongly bound water is the ratio of surface hydrated water used in protein supercooling studies.125, 127 These hydrogel systems shows signs of damage on water freezing, as the force of growing ice crystals can easily overcome the polymer’s covalent bond strength. Cycling of these materials shows an increase in damage.36, 132, 135

These two techniques, hydration and confinement induced supercooling of water, both obtain the same end result: unfrozen supercooled water. In accord with the advent of improved hydrogel elasticity, toughness, and strength by the composite of networks or crosslink types, Chapter 5 demonstrates how the physical gels can obtain the same combinatorial effect as strongly bound and confined systems for water supercooling. The domains of physical crosslink moieties are impermeable to the water phase. Therefore, the domains act like a wall or obstruction to the water’s diffusion and hydrogen bonding structure. Then the hydrophilic interconnecting chains can create strongly bound hydration layers of water. Together the physical gel can act as a soft tissue-like medium for water supercooling and the inhibition of ice formation. This is achieved by the domains act as nanoconfining moieties, while the hydrophilic chains act as to strongly bind the water, inducing supercooling. Together, higher degrees of supercooling may be achieved. It is also shown that if the contained water does freeze, the system demonstrates healing of physical bonds once the sample is reheated. Thus the physical hydrogel may offer a new route to supercool water and prevent ice formation.

2.6 SURVEY OF DISSERTATION

In this thesis the results obtained in several different studies are shown using a model physically crosslinked hydrogel formed by hydrophobic association.

Chapter 3 demonstrates how the physical hydrogels can overcoming confinement limited

43 swelling. Driven by the osmotic force of the hydrophilic chain segments to induce rearrangement of physical network to a new equilibrium state.

Chapter 4 shows how in the model physically crosslinked hydrogel with a microphase separate system, clear evidence of step like progression of relaxation processes occurs. The max strain of the mobile swollen chain phase first occurs, followed by the breakup of the physical crosslinks, and finally the relaxation back to an equilibrium conformation. These processes can be observed to propagate to the macro- stress relaxation where similar relaxation times as found for the microstructure are also witnessed in the stress relaxation. Indicating not only do the molecular processes act distinctly, but these processes can be found in the macroscopic stress relaxation.

Additionally, it was found that on domain breakup and relaxation, segment pullout was the dominant method of rearrangement.

Chapter 5. The supercooling and inhibition of crystallization of water has been an objective of research for many years. With the achievement of several methods, the ability of a model physically crosslinked hydrogel may bring two key methods together: bound water and confinement induced supercooling. The result is a robust physical hydrogel that may have applications in several industries and can be easily manufactured.

The physical hydrogel also demonstrates self healing properties in the case where water crystallization does occur.

Chapter 6. Using the model physically crosslinked hydrogel the operation regimes of Quartz crystal microbalance are investigated. Particularly, the occurrence of thin film resonance at thicknesses thinner than expected is shown. Demonstration of the successful

44 ability of the analysis software to correctly fit the data in the region where the QCM operates in a highly non-linear regime is presented. A comparison between two methods of analyzing the QCM measurements to obtain viscoelastic and thickness results is discussed. The type of films and environments where successful use of these two models may be employed is found as well as defined regime of where the insensitivity to viscoelastic modeling occurs.

Chapter 7 provides a summary of the cumulative overview of the work performed and the achieved results. Outlining each chapter’s main achievements. Chapter 7 also includes a future work section. This section discusses some preliminary work in areas which hold promise to new findings and ideas for related projects that may prove worthwhile to pursue.

CHAPTER III.

OVERCOMING CONFINEMENT LIMITED SWELLING IN HYDROGEL THIN

FILMS USING SUPRAMOLECULAR INTERACTIONS

3.1 INTRODUCTION

The diverse chemistries available for the fabrication of hydrogel materials provides a pathway to tune their physical properties to enable utilization in a variety of applications, including drug delivery,136, 137 tissue scaffolds,138 sensors,61,

136 cell immobilization,25 control of flow in channels,9, 14 biomedical devices5 and soft machines.45 In many of these applications, the hydrogel is constrained to dimensions from the micrometer to nanometer scales, commonly as microparticles or thin film coatings. By inclusion of responsive groups in the chemistry of the hydrogel, such as polyacrylamide (PAAM)/poly(acrylic acid) (PAAC)139 or poly(N-isopropyl acrylamide) (PNIPAAm)140, 141, the particles or coatings can exhibit tunable properties that promote a response to environmental stimuli, e.g., the rise in local temperature in a tumor to trigger drug release142. Thermally responsive hydrogels, PNIPAAm in particular, have been extensively studied in the bulk,37, 69, 70 microparticles,143 and thin coatings144. The thermoresponsive behavior of PNIPAAm is resultant from its lower critical solution temperature

(LCST) and the associated differences in the swelling above and below this phase transition. However, the swelling behavior can be significantly different in thin

46 films as compared to the bulk hydrogel. Thus, the thin film properties of these hydrogels are important to ensure performance and operation requirements for applications are still met.

Two strategies are commonly used to generate thin PNIPAAm coatings: grafting a brush to the surface145-147 or in-situ chemical crosslinking using a thin coating.10, 136, 148 For brushes, the LCST can be dramatically shifted depending on the graft density.147 The thickness of these brush layers is controlled by the graft density and the molecular weight of the grafted chain. These properties also impact the relative difference in thickness between the collapsed (high temperature) and swollen (low temperature) states of the brush.145

Conversely, thin (ca. 150 nm dry) hydrogel films can be formed by crosslinking PNIPAAm on a substrate of interest. Such films generally exhibit an

LCST that is 2-4 ºC greater than the LCST of the analogous bulk hydrogel

(defined as a free standing film at least 0.5 mm thick with the same chemical composition and crosslink density).10, 136, 148 The change in the LCST for chemically crosslinked films is less than typically observed for PNIPAAm brushes.

However unlike the LCST, the swelling of thin chemically crosslinked PNIPAAm films is significantly reduced as compared to the bulk. For example, Harmon et al.149 observed a 100-fold increase in volume upon swelling of bulk covalently crosslinked PNIPAAm, but a thin constrained film exhibited only a 15 fold- increase in volume upon swelling. This significant decrease in the volumetric response of the hydrogel network can significantly alter their properties in thin films.

These differences between the thin film and bulk properties of hydrogels are generally attributed to the constraints in swelling by the substrate. For a thin film, if the hydrogel is not well adhered to the substrate, osmotic stresses associated with swelling will delaminate the film. The adhesion to the substrate limits the deformation of the hydrogel such that the volumetric change is almost exclusively unidirectional through the thickness of the film. For one dimensional (1D), constrained swelling of a crosslinked polymer,53 Flory-Rehner theory52 predicts that ideal volumetric swelling ratio of the constrained network should be the square root of the isotropic bulk volumetric swelling ratio of the unconstrained network. 1D constrained swelling infers that the thickness swelling ratio is indistinguishable from the volumetric swelling ratio. For chemically crosslinked poly(dimethylacrylamide) hydrogels covering a wide range of crosslink density,

Toomey, et al. found good agreement between the predicted constrained thin film swelling and experimental results.53

Additionally, component segregation to interfaces150 during synthesis of crosslinked PNIPAAM thin films10 can lead to non-uniform crosslink density through the film thickness. Moreover, the chain conformation may be altered due to confinement effects151 and those conformations may be locked-in by the crosslinking reaction carried out in the dry state of the thin films. For polystyrene thin films, Napolitano and Wubbenhorst reported that chains were compressed to produce a pancake-like conformation,151 and a similar anisotropy may be present in thin PNIPAAm films. That would impact the distribution of crosslinks and the

48 conformations available to the chain upon swelling, since the covalent crosslinks are chemically fixed and cannot rearrange.

A facile method to enable rearrangement of the network crosslinks is to use physical crosslinks. In that case, the crosslinks can dissociate as a mechano- response to the stresses generated from swelling, which allows the chains to rearrange if the osmotic stress is sufficiently large. Once the stresses are relaxed, however, the broken physical crosslinks reform to restore the network, with a similar crosslink density as the initial gel network.

One such thermally responsive, physically crosslinked hydrogel, consists of a random copolymer of NIPAAm and 2-(N-ethylperfluorooctanesulfonamido)ethyl acrylate (FOSA).14 The hydrophobic FOSA aggregates into nanodomains that are dispersed in a continuous poly(NIPAAm) phase, and the nanodomains behave as supramolecular crosslinks. The fluorocarbon chain in the FOSA produces stronger hydrophobic associations than typically provided by a hydrocarbon chain in other hydrophobically-modified hydrogels.152 NIPAAm-FOSA copolymers can form gels in water at FOSA concentrations in the copolymer as low as 2 mol%.69

Similar physical hydrogels based on N,N’-dimethylacrylamide (DMA) and FOSA copolymers exhibit strength and toughness15 similar to the ‘double networks’ hydrogels.27 The toughness of those physical hydrogels was attributed to the responsive character of the hydrophobic associations to stress,15 which allows the nanodomains to rearrange and reform in response to an applied load. In light of these observations, the motivation for the research reported herein was that

49 physically crosslinked hydrogels may provide a means to overcome the constraints from the substrate on the swelling of thin film hydrogels.

This paper describes the temperature-dependent swelling behavior of

NIPAAm-FOSA copolymers containing 5 mol% FOSA, denoted as NF5, for a wide range of film thicknesses using a combination of spectroscopic ellipsometry

(SE) and quartz crystal micro-balance with dissipation (QCM-D). These measurements provide complementary information as the SE measurements are challenged in the thin film limit by the coupling of refractive index and thickness,153 while for thicker hydrogels, the QCM-D response is strongly influenced by the viscoelastic nature of the film.53 This necessitates recursive modeling and significant assumptions regarding the frequency dependence of the viscoelastic properties.57 In contrast to chemically crosslinked PNIPAAm, where confinement effects produced over 50% reduction in thin film swelling,53 the volumetric swelling of the physically crosslinked hydrogel thin films was within

20 % of the bulk.

3.2 EXPERIMENTAL

3.2.1 Materials.

The synthesis of the NF5 copolymer by a free-radical copolymerization of

NIPAAm and FOSA has been previously described.14,16 The same copolymer

(NF5) sample that was used in Tian et al.70 was used in this work. The

4 4 characteristics of the copolymer were Mw = 6.4 x 10 Da, Mn = 3.4 x 10 Da and 5 mol % FOSA.70 Toluene (99.8 %, ACS Grade), isopropyl alcohol (99.5%, ACS

Grade), and 1,4-dioxane (99.8%, ACS Grade) were purchased from Sigma Aldrich and used as received for preparation of the thin films. Deionized water (~ 1-5 ppm inorganic) was used in all the swelling experiments.

3.2.2 Sample preparation and measurement baseline.

Quartz sensors (QSX-335, Q-Sense) with successive layers of titanium, titanium oxide, and silica oxide deposited on a standard gold quartz sensor were used as film substrates for all measurements. All sensors were cleaned prior to use by sonication in toluene, then isopropanol and finally deionized (DI) water. The sensor was then rinsed thoroughly with DI water and blown dry with nitrogen.

Prior to coating with the NF5, the optical properties of the bare sensors were determined using a spectroscopic ellipsometer (SE) (JA Woollam M-2000UI) and the baseline impedance of the quartz sensor was determined in air and in DI water at 25 ºC using the quartz crystal microbalance with dissipation (QCM-D) (Model

E1, Q-Sense). The sensors were again rinsed with DI water, blown dry with nitrogen and then further cleaned with ultraviolet ozone (UVO) (UVO

CLEANER®, Model 42, Jelight Company Inc.) for 90 s.

Immediately following the UVO exposure, the NF5 was spin-cast onto the sensor from a 1,4-dioxane solution at 2500 rpm for 30 seconds. After coating, the sensors were annealed at 150 ºC for at least 18 hours to aid in solvent removal, ensure that the film well adhered to the sensor surface, and promote the self- assembly of the FOSA nanodomains. Upon removal from the oven, the sensors were immediately placed in a desiccator under vacuum to ensure the samples remained free of water. The initial dry coating thickness was determined in air

51 using SE to obtain a reasonable optical model for the stack on the quartz sensor using the methodology described by Richter and co-workers.154

3.2.3 Characterization of swelling

Simultaneous measurements of SE and QCM-D were performed using a combined QCM-D/SE cell (Q-Sense, Ellipsometry Module). This system allows

QCM-D measurement of quartz sensors in fluid and simultaneous SE measurement through BK7 glass windows with the beam incident on the sensor surface at 65°.

For the SE measurements, wavelengths below 400 nm were not included due to a slight absorbance at these lower wavelengths by the BK7 windows.

The QCM-D was used to measure the mass adhered to the sensor surface

(which enables the calculation of film thickness, if the density is known, herein a density of 1 g/cm3 was assumed for the film) and its viscoelastic properties155. The latter determination is based on the oscillation frequency of the driven quartz, F, and the rate of decay of the oscillation amplitude, which is directly related to the dissipation, D156, when the potential is removed. Prior to filling the cell with DI water, the sensor was measured in air for 10 min to provide a stable baseline. The cell was then flushed with DI water for 10 min at100 μL/min, with the water pulled in using a 2 channel peristaltic pump. All ports were checked to insure removal of air. Subsequently, the flow of DI water was reduced to 50 μL/min for at least 40 min at 25 ºC until a frequency change of less than 6 Hz/hr was obtained.

The fundamental frequency and overtones (1-13, odd only) from the QCM-

D were again determined using a full frequency scan in order to correct for any perturbation that occurred due to the viscous loading of the QCM-D sensor by the

52 addition of water. The temperature-dependent swelling of the NF5 coating on the sensors was then determined by heating the cell to 35 ºC, restoring the water flow for 5 min to remove any air bubbles that may have formed during heating, and measuring the swelling under quiescent conditions to minimize the duty cycle demand. The temperature was decreased in 1-2 ºC steps every 60 min until 5 ºC was reached. Typical data collection time for the cooling cycle for each film was

24 h.

3.2.4 Data analysis

QCM-D and SE require recursive modeling to determine the film thickness and physical properties such as viscosity, , shear modulus, , and refractive index associated with the hydrogel. For QCM-D, the difference in F and D between the uncoated and

NF5-coated sensor in water at 25 ºC was used to calculate the swelling of the thin films.

For assessing the thickness, measurements over the last 20 min at each temperature were fit with the recursive model. The QCM-D data were averaged over 5 s intervals to reduce the effect of noise. System temperature effects, which are unrelated to the hydrogel film, can be significant especially for the thinnest films. These effects include thermal stresses applied to the crystal and the temperature dependence of the resonance frequencies. Prior to fitting the experimental data to the recursive model, the effect of temperature on the system was subtracted from the raw data. This correction accounts for both the changes in the kinematic viscosity of water and the intrinsic characteristics of the quartz sensor / electronics with temperature. In general, this correction is most critical for ultrathin films, where the effects of temperature contribute significantly to the signal. For example if we consider the 10 nm thick (dry) film, nearly 50% of the frequency shift measured with

QCM-D can be attributed to temperature dependent changes in physical properties of the water and the characteristics of the quartz sensor. As both components of the correction are critical, we utilize a single correction that can account for both effects in a facile manner.

To determine the correction factor, the frequency change (F) and dissipation change (D) of clean sensors were measured in DI water as a function of temperature from 35 ˚C to 5 ˚C using a similar temperature step protocol as used for the hydrogel swelling experiments. These baseline measurements on the clean sensors were performed in triplicate to insure the observed shifts are not artifacts, but a true measurement of the temperature dependencies and that the same effects are observed for different sensors.

For each temperature, the correction factor is determined at equilibrium (defined as F <

6 Hz/h) for each overtone. These correction factors for F and D in water are fit as a function of temperature. Although both linear and quadratic fits provide similar fit quality values, the quadratic fit was chosen as it provides an improved physical representation of the temperature dependence for to two reasons. First, with the linear fit correction of the

10 nm thick (dry) NF5 film, Figure 3-1 shows there is an increase in the dissipation at high temperature that would not be physically expected in a collapsed film as the difference in swelling between at 35 ˚C and 25 ˚C is quite small. Conversely, there is no substantial change in the dissipation when using the quadratic fit to correct the dissipation of the 10 nm thick (dry) NF5 film in the temperature range from 35 ˚C to 25 ˚C as would be expected by the behavior of the rigid film.

Figure 3-1. Dissipation curves for the 10 nm NF5 film at the start of a swelling run with temperatures for each step labeled. The three curves represent the 3rd overtone

A second justification for selection of the quadratic fit is the temperature dependence of the kinematic viscosity ( of water as shown in Figure 3-2, which shows a clear non-linear trend. As both F and D of viscous fluids can be related to ,157 this quadratic fit should provide an adequate approximate function form for the applied temperature correction if viscous effects are dominant. Using this quadratic expression, each overtone of the bare crystal in water is individually fit to develop a temperature correction factor to implement in the swelling experiments.

Figure 3-2. Kinematic viscosity,  (▲) of water as a function of temperature (data obtained from NIST Chemistry WebBook)158. Solid black provides a quadratic fit of the data.

Moving forward with the use of the quadratic function, we fit the 5 overtones to determine holistic effect to apply to all sensors based on a zero intercept at 25˚C. Figure

3-3 shows the determined correction factor for the five overtones (overtones 3-11) for both the frequency and dissipation; these correction factors include both the temperature dependence of the kinematic viscosity of water and the intrinsic sensitivity of the crystal.

In order to correct the data, this temperature dependent correction factor is subtracted from the raw data for each overtone. The reference temperature is 25 ˚C (standard temperature for AT-cut crystals), so the correction is zero at 25 ˚C. Table 3-1 provides the temperature dependent parameters obtained for each of the five overtones with a quadratic expression: ∆F(T) = F 2+ F + F, ∆ (T) = 2+ + . It should be noted that these parameters may also include some contributions specific to the electronics of the QCM-D, so these correction factors should ideally be determined for each QCM-D instrument before use to obtain the best possible correction.

Figure 3-3. Overtone dependent temperature correction factors for ΔF and ΔD according to the best fit of a quadratic to a bare sensor in water (overtones labeled on the graphs).

These changes are subtracted from the raw data to correct for all temperature dependencies in the system. Error bars indicate the standard deviation for a measurement on three separate sensors.

Table 3-1. The temperature correction parameters for the five overtones for both frequency and dissipation.

Frequency Dissipation

Overtone AF BF CF AD BD CD 3 -0.0836 10.8 -215 0.0288 -3.18 61.4 5 -0.0635 8.97 -182 0.0239 -2.54 48.1 7 -0.0565 8.17 -166 0.0189 -2.10 40.5 9 -0.0502 7.46 -152 0.0165 -1.85 35.9 11 -0.0465 7.09 -145 0.0150 -1.67 32.2

To illustrate the impact of application of these temperature corrections, Figure 3-4 shows the difference in the film thickness result using the two different models to fit the

QCM-D data with and without the temperature correction applied. At low temperatures in the highly swollen state, the Sauerbrey approximation for thickness is not strictly valid,159 but illustrates the over-prediction of the swelling without the temperature correction as compared to the thicknesses determined from SE in Figure 3-4. More strikingly, fitting both the frequency and dissipation with the viscoelastic model shows an even greater difference between the corrected and uncorrected data. Without the temperature correction, the fit suggests a nearly linear increase in thickness with temperature using this viscoelastic model to describe the QCM-D data. With temperature correction, the swelling dependence on temperature is found to be sigmoid-like, which is consistent with the SE and bulk measurements.70 For the 32 nm film (shown in Figure 3-4), the temperature correction has a drastic effect on the thickness with nearly 30% reduction in calculated thickness at low temperatures (highly swollen) and 25% increase in calculated thickness at high temperatures (collapsed). In the collapsed state, both the Sauerbrey and

Viscoelastic model used to analyze the QCM-D data predict the same thickness for the temperature corrected data (inset of Figure 3-4).

The temperature correction protocol results in improved agreement between the

QCM-D and SE results for swelling of the thin films, especially when applying the viscoelastic model to fit the QCM-D data. To our knowledge, there is no well-defined, standard protocol for temperature dependencies with QCM-D. These corrections are particularly important for the thinner films where the contribution from non-film effects can become a large portion of the total measured change. This method of correction applies a simple offset that includes temperature effects associated with the QCM-D sensor and equipment, which are not taken into account by simple changing model inputs

58 for the water based on its density and viscosity change with temperature. This correction also produces data that can easily be included in the standard Q-Tools program to fit the

QCM-D data range.

For comparison, the thickness determined from SE ( ) is also included.

Figure 3-4 illustrates the large error in the resulting thickness calculation if the device and bulk water temperature effects are not removed. From the recursive fit of the frequency (i.e., mass) and the dissipation (viscous loss), the thickness and viscoelastic properties of the hydrogel were determined using an average of the

59 last 20 minutes for each temperature step. The QCM-D data were analyzed using

Q-Tools software with a frequency-dependent Voigt model to calculate the viscoelastic properties of the sample from the measured dissipation and frequency change of the sensor.33,35 The zero baseline for all measurements and temperature corrections was chosen as 25 °C.

The temperature corrected QCM-D data for all films were fit with the extended viscoelastic model,154 which is a frequency-dependent Voigt based viscoelastic model.57 Due to anomalous behavior at low temperatures associated with a frequency upturn with swelling at higher overtones (Figure 3-5), only the 3rd and 5th overtones were utilized in the fit unless otherwise noted. In order to achieve a reasonable fit, the model parameters were bracketed as: h0 < h < 6h0,

5 8 where h0 was the dry film thickness, 10 < < 10 Pa), and 0.01 < < 1 Pa·s. Once

χ2 was minimized, the modeling was restarted to ensure that the modeling in Q-

Tools program could generate the same result globally. The final values were only used and reported herein when they could be reproduced.

An apparent difference between the QCM-D and SE thicknesses is that the thickness difference decreases at low temperatures, especially for the thickest films examined. Typically, an increase in frequency of a QCM is attributed to mass loss, but the SE data indicate that the film continued to swell as the temperature decreased as expected for PNIPAAm. Theoretical work based on modeling of the propagation of the shear wave through a lossy film by White and Schrag has predicted similar upturns in the frequency.60 This effect is also seen in models by Kanazawa and Johannsmann.160, 161

Chapter 6 explores the occurrence of this upturn and what we can learn from the model

60 results as the upturn occurs for this unusual behavior. Nonetheless, this behavior illustrates care must be taken when interpreting QCM data of lossy films without secondary measurements. An increase in thickness (as determined by SE) may be accompanied by an increase in frequency (from QCM-D), which is generally considered to be indicative of a loss of adhered mass to the sensor.

Figure 3-5. The frequency (F) and dissipation (D) curves from QCM-D for 3rd, 5th, 7th, and 9th overtones associated with the cooling of the 32 nm (A and B) and 120 nm films

(C and D). The time is related to decreasing of the temperature from 35 ˚C to 5 ˚C. The

QCM-D at low temperatures behaves in an unusual manner with an increase in both frequency and dissipation, which is indicative of film resonance.

The SE thin film data from wavelengths 400 nm to 1150 nm were modeled using a simple Cauchy layer162 to describe the optical properties of the hydrogel.

Temperature dependent optical properties of water were included in fitting of the

SE data to increase accuracy of the fits.

To quantify the lower critical solution temperature, LCST, and the breadth of the transition, the temperature dependent thickness as determined from both QCM-D and SE was fit to a sigmoid function, as shown in Equation 3-1, following the prior reports by Harmon, et al.149

 h   h          h   h  h  h   0 collapsed  0  max      h  h    T  T   0  0  max  1 exp LCST          (3-1)

where (h/h0)collapsed is the swelling ratio at high temperature (e.g., collapsed state),

(h/h0)max is the swelling ratio at low temperature (e.g., highly swollen state), TLCST is the inflection temperature that is defined as the LCST, and σ is the half-width of the sigmoid, which provides a measure of the width of the swelling transition. An example of the sigmoid fit can be seen in Figure 3-1 for the 75nm film for the thickness change with temperature as measured by SE. It should be noted that in this manuscript that we will use LCST to denote the volumetric swelling transition at atmospheric pressure.

Rigorously by definition, the LCST occurs if the isobar passes through the critical point and would require examination of both temperature and pressure as independent variable.163 However as the volumetric transition at 1 bar is commonly described as the

LCST in the hydrogel literature,164-166 we will utilize this more standard convention.

Figure 3-6. Sigmoid fit of SE data for 75 nm (dry) NF5 film. Thickness components

(h/h0)max and (h/h0)collapsed are associated with the swelling (vertical) axis, while the

LCST and  are associated with the temperature (horizontal) axis.

3.3 RESULTS AND DISCUSSION

Figure 3-7 illustrates how the apparent equilibrium volumetric swelling ratio (V/V0) of NF5 in DI water depends on the film thickness as the hydrogel is cooled through the LCST. The low temperature swelling is very similar to that for the bulk hydrogel. This similarity in swelling is striking given the significant differences reported between thin films and bulk for chemically crosslinked hydrogels.149 The Flory-Rehner theory for thin constrained films53 predicts a volumetric swelling ratio of 2.2 at the lowest temperature based on the bulk swelling, but the measured volumetric swelling ratio was greater than 3.9 for all the films examined. When examining the low temperature swelling data more

63 carefully, there was an increase in the volumetric swelling ratio as the film thickness decreases.

Figure 3-7. Volumetric swelling ratio (V/V0) for NF5 thin films equilibrated in water as the temperature is decreased from 35 ˚C to 5 ˚C, determined by SE (A) and QCM-D (B).

NF5 film thicknesses shown are: 10 nm (), 32 nm (▼), 52 nm (◂▸), 75 nm (), 100 nm

(◆), and 120 nm (). The open symbol 10 nm film was modeled only in QCM-D due to its limited optical path length. Comparison to bulk (●) volumetric swelling measurement is provided as reference. (C) Thickness swelling ratios (left axis) from SE () and QCM-

D () at 5 °C and normalized volumetric swelling ratio for the films relative to the bulk hydrogel (right axis).

Figure 3-7C illustrates this behavior more clearly with the swelling ratio at 5 °C increasing from 3.9 to 4.9 as the dry NF5 thickness was decreased from 120 nm to 10 nm

(for confined films, h/h0 is indistinguishable from V/V0). For the thinnest dry film thickness, 10 nm, the short optical path length of the hydrogel and the relatively small difference in refractive index between the hydrogel and water in the highly swollen state resulted in significant uncertainty in the SE fit due to the coupling of refractive index and thickness. As such, those data are not reported. Between 10 and 32 nm, there was no further increase in swelling as determined by the QCM-D, so it appears as if the swelling ratio reached a limit. In the thin film limit, the size of the hydrophobic domains (~6.4 nm, as measured in bulk54, 70) was of the same order as film thickness and thus rearrangement of these hydrophobic domains may be less constrained, which may explain the volumetric swelling ratio being very similar to the bulk.

The swelling of the two thinnest films at 5 °C was actually greater than the bulk hydrogel. That behavior is contradictory to expectations and prior reports for chemically crosslinked PNIPAAm.53, 149 Although water is known to accumulate at a silica-polymer interface,167 prior work concluded that a hydrophilic substrate had little effect on the swelling of thin PNIPAAm films.148 Consistent with the prior work, the swelling of NF5 on SiO2 and gold coated QCM-D sensors was identical.

Nonetheless, the low temperature volumetric swelling ratio was greater than 80% of the bulk for all films examined. Moreover, there was good agreement in the swelling ratio determined by QCM-D and SE measurements confirming that the swelling is not an artifact of the measurement.

At high temperatures (T > TLCST), the swelling for the films that were 30 nm

(dry) and thicker was significantly less than for the bulk hydrogel. Toomey, et al53 predicted that the swelling ratio of the film should be the square root of the bulk

65 swelling. Thus, the volumetric swelling ratio should be 1.4 in the collapsed state of the thin films, since the bulk hydrogel swelled to approximately twice its dry volume at the same conditions. That prediction agrees well with measurements for the hydrogel thin films with the volumetric swelling ratio for SE ranging from 1.2 to 1.4 and for that measured by the QCM-D ranging from 1.5 to 1.9 (excluding the

10 nm film).

To explain the consistently greater swelling reported from QCM-D (Figure

3-7B) than from SE (Figure 3-7A), the sensitivity of the respective instruments must be considered. The QCM-D reporting greater swelling than SE is consistent with prior reports associated with adsorption of proteins to the surface.12 This difference is attributed to the coupling of bound water to the surface of the adsorbed layer. Only QCM-D, not SE, is sensitive to coupled water168 at the hydrogel surface. Figure 3-8 clearly illustrates how the swelling ratio measured by

QCM-D is consistently greater than that measured by SE at high temperatures, irrespective of film thickness. The offset between QCM-D and SE was relatively consistent at lower swelling ratios, and that result should be due to the coupled water168 at the hydrogel surface, to which only the QCM-D is sensitive.

QCM-D is 26 ± 12 nm. NF5 film thicknesses shown are: 32 nm (▼), 52 nm (◂▸ ), 75 nm

(), 100 nm (◆), and 120 nm ().

One unusual observation was that the swelling for the thicker films at low temperature showed the QCM-D thickness approaching the thickness measured by

SE, which is unexpected based on prior comparisons of optical and acoustic measurements.12 That behavior is attributed to issues with the QCM-D measurement of these thick, lossy films. Although the films continued to swell at low temperature as determined by SE, the frequency of the higher overtones actually increased, which is typically associated with a loss of mass. This condition is known as film resonance, but is typically only seen in much thicker films.

Chapter 6 will discuss this in further detail as well as the model’s ability to fit the data and produce accurate results through this region. For this analysis only the 3rd and 5th overtones were used for modeling the QCM-D data to avoid any potential problems with fitting.

If one neglects the thickness measurements from QCM-D data where an upturn in any frequency was observed, the average thickness of the coupled water was 26 ± 12 nm for all of the films, as illustrated in Figure 3-8. At 30 °C, the film thickness calculated from QCM-D was on average 29 nm greater than that measured by SE, and that difference agrees well with the estimated coupled water thickness. However, application of that correction to the data at 5 °C results in significant mismatch between QCM-D and SE. Details of the comparison for each film with and without this coupled water layer are shown in Figure 3-9. For consistency, we have not included that correction to any data analysis, only showing the effect of applying this correction in Figure 3-9.

For a single film thickness as shown in Figure 3-9, the offset between QCM-D and SE appears to be almost invariant with temperature with greater swelling reported by the QCM-D at all temperatures. This behavior is actually observed for all film thicknesses measured with the swelling from QCM-D being greater than that from SE

(Figure 3-8). The average thickness difference between QCM-D and SE is 26 ± 12 nm.

Prior reports have also observed a similar phenomenon associated with the thickness of adsorbed layers between QCM-D and optical techniques.12 This difference is generally attributed to coupled water at the surface that adds to the mass measured by QCM-D, but does not contribute to the optical thickness measured by SE. To understand if such a

68 large coupled water layer (26 nm) might be present at the surface of the NF5 hydrogels, this thickness was subtracted from the thickness determined by VE fit of the QCM-D data with a direct comparison to the SE swelling as shown in Figure 3-9 for each film thickness from 32 to 120 nm (dry film).

Figure 3-9. Comparison of swelling curves for SE and coupled water layer corrected

In general, the agreement between SE and QCM-D after this subtraction is good for temperatures higher than 18 ˚C. The behavior of the 75 nm film is different from the other films as the thickness is still overestimated by QCM-D at high temperatures even

69 after the coupled water layer correction is applied. One explanation is that the film may have been rougher than the other films; this surface roughness could lead to additional coupling of water. However at low temperatures (<18 ˚C) for all films, the QCM-D thickness appears to be overcorrected with the swelling underestimated in comparison to the swelling measured by SE. However as this underestimation occurs in the highly swollen state, this difference could also be attributed to decreases in the surface roughness on swelling that would decrease the amount of coupled water.

With the good agreement in general between SE and QCM-D for the thickness dependent behavior of the NF5 hydrogel films (Figure 3-7), the utilization of only the 3rd and 5th overtones in the modeling of the QCM-D data does not appear to adversely impact the analysis. To further illustrate the general agreement between QCM-D and SE, Figure 3-10 illustrates the analysis result of the temperature dependent swelling ratio (Figure 3-7) using a sigmoid fit (equation

3-1) to determine TLCST and the width of the transition (2σ). As shown in Figure 3-

10A, an increased TLCST for all of the thin films of 3-5 ºC as compared to the bulk was determined in both SE and QCM-D measurements. There is a discrepancy of approximately 1 ºC in the TLCST between the SE and QCM-D measurements, but this difference is within the uncertainty of the measurement. There is a clear increase of TLCST in the thin films measured as compared to the bulk. A similar increase in TLCST for film thicknesses below 250 nm has also been reported for covalently crosslinked PNIPAAm films,10, 136, 148 so this behavior is not surprising.

Figure 3-10. Analysis of the swelling data using equation (1) to determine (A) TLCST and

Unlike TLCST, there appeared to be a difference in 2σ between QCM-D and

SE measurements (Figure 3-10B). Except for the 52 nm thick film, the width associated with the SE measurements was similar to the bulk. The more narrow

LCST associated with the QCM-D measurements may be an artifact of the film resonance frequency response of the QCM-D at high swelling ratios as shown in

Figure 3-5 or its clear offset at high temperatures (Figure 3-7). Harmon, et al.149 examined the width of the LCST in chemically crosslinked PNIPAAm and found the transition width to be about 0.5 ºC. The broad transition for NF5 is due to the physically crosslinked PNIPAAm and its distribution as described previously.54

In addition to thickness or mass, QCM-D can determine the viscoelastic properties of an adherent film that is sufficiently lossy. Figure 3-11 shows the temperature dependence of the shear viscosity () and shear elastic modulus () of

71 the hydrogel films calculated from the Voigt viscoelastic model57 with frequency dependence.169 The qualitative temperature dependence of  and  is similar for all film thicknesses, but there are quantitative differences. For the thicker films (> 50 nm dry), both the viscosity and shear modulus were nearly independent of thickness for temperatures below ~20 ºC, as shown in Figure 3-11.

Figure 3-11. The viscoelastic properties of the hydrogels determined from QCM-D for the (A) viscosity and (B) shear elastic modulus as a function of temperature for the NF5

At low temperatures  and were slightly lower for the 52 and 75 nm film than the thicker films, which was consistent with the greater extent of swelling

(increased hydration) for the thinner films (Figure 3-7C). Consistent with the large increase in swelling for the 32 nm film (Figure 3-7C), shear modulus and shear viscosity were significantly reduced at high swelling ratios. However, the data for

72 the 10 nm film did not fall within the construct of the arguments proposed for the variation in viscoelastic properties for the thicker films. This film had the second largest swelling ratio of the films examined (Figure 3-7C), but the viscosity was greater than all of the films examined in the low temperature limit. This is likely the result of the limiting ability of the QCM to sense the viscoelastic properties of very thin films, a point we discuss in detail with film resonance in Chapter 6.

At higher temperatures, there were significant variations in the viscoelastic properties for even the thicker films. The temperature dependence of  and  seemed to vary erratically. This behavior was likely associated with low dissipation, which provides limited losses associated with the film by which the viscoelastic properties can be extracted. The fit of the high temperature data was not well constrained by the simple 2 error criteria for determining the best fit in Q-Tools. At high temperatures, the low dissipation values resulted in a wide range of satisfactory fits of the viscoelastic (Voigt) model to the frequency and dissipation data. That produced a large variance in the calculated viscoelastic properties of the film in the low dissipation regime as shown in the

Figure 3-12 for a 100 nm thick film.

The swelling of the NF5 film transverses regimes for QCM-D operation from rigid (Sauerbrey regime), associated with low dissipation, to highly lossy, as evidenced by the large increase in dissipation. In the low dissipation regimes, the viscoelastic properties obtained from QCM-D tend to be rather noisy as illustrated in Figure 3-12. In this regime, the viscoelastic properties obtained from the model depend on the initial guess and thus these values associated with the viscoelastic properties are not unique for the fits. When the film is sufficiently swollen (at a swelling ratio approaching 2.4 from an

73 initial thickness of 100 nm), the viscosity and shear modulus obtained from the fit of the

QCM-D data becomes significantly more reproducible. To explain this behavior, the operation of the QCM must be carefully considered. White and Schrag illustrated mathematically that there exists a rigid to viscoelastic transition in the operation of QCM that is dependent on the viscoelastic character and thickness of the adhered mass,60 which has been confirmed experimentally using swelling of a glassy polyelectrolyte film by humid air; interestingly for a 96 nm thick film, the transition from Sauerbrey to viscoelastic regime occurs at a swelling ratio of approximately 2.3.170 This swelling agrees with the transition where consistent viscoelastic properties are obtained for the

NF5 hydrogel films from fitting the QCM-D data. This behavior suggests that the film must be sufficiently lossy to deviate from the Sauerbrey expression in order to effectively determine the viscoelastic properties of these hydrogel films.

For temperatures lower than 24 ˚C, nearly all of the films exhibit sufficient dissipation to enable effective elucidation of the viscoelastic properties of the films

(Figure 3-11). Figure 3-12 illustrates the differences in the consistency of the fit thickness and viscoelastic properties at high temperature for a thick film. As can be clearly observed, the fit of the data is excellent across all temperatures. The film thickness for each temperature is nearly invariant with time as the ‘equilibrium’ regime is only considered here. However when considering the viscosity and shear elastic modulus determined from the same viscoelastic model over the same temperature window, the data are quite noisy, especially at low swelling extents (high temperature). Additionally, the viscosity appears to increase as the film is initially swollen, which is counter to expectations for a glassy film absorbing solvent. Only for temperatures less than 24 ˚C is

74 a clear physically agreeable trend in viscosity and shear elastic modulus obtained from the viscoelastic model as denoted by the vertical dashed line in Figure 3-12.

Figure 3-12. (top) Frequency and dissipation for F3 ( ), F5 ( ), D3 ( ), and D5 ( ) as a function of time (temperature) with fits using the viscoelastic model (solid black lines).

This behavior is consistent with the data in Figure 3-4 where the Sauerbrey thickness begins to deviate from the thickness obtained from the viscoelastic model in approximately this same region. In general, the dissipation increases to ≈ 10 × 10-6 when consistent viscoelastic properties for the film are obtained from fits of the QCM-D data.

It is not clear why for some systems the viscoelastic properties was effectively modeled at low dissipation,171 while for others issues such as shown here were encountered. For example, Patra and Toomey97 reported high temperature  and  calculated from QCM-D for photo-crosslinked PNIPAAm films, but they also reported difficulties in fitting their low dissipation data to the

Voigt based viscoelastic model. Higher overtones were necessary and multiple thicknesses were fit simultaneously to the viscoelastic model in order to obtain the viscoelastic properties for the films. The same model was used in this paper to facilitate comparisons. A reason for this sensitivity in predicting the viscoelastic results and a suggestion to minimum requirements for effective modeling are discussed in Chapter 6.

The work by Patra and Toomey97 for photo-crosslinked PNIPAAm films

(36 - 144 nm) with a maximum volumetric swelling ratio of 2.6 provide an opportunity to compare the viscoelastic properties of chemically and physically crosslinked PNIPAAm. It should be noted that the swelling of the crosslinked thin film was less than for the NF5 films examined here (3.9 - 4.9), but the expected swelling of an analogous bulk photo-crosslinked PNIPAAm is 6.8 (greater than the bulk NF5, 4.7). For the highly swollen state (low temperature), the viscosities of the thicker physically crosslinked (~0.02 Pa·s) and chemically crosslinked films

(~0.04 Pa·s) were similar in comparison to the orders of magnitude change in modulus and viscosity that occurred as these hydrogels transversed across the

LCST. The lower viscosity for NF5 is consistent with its larger swelling ratio in the highly swollen state. However when the NF5 was less swollen, near the onset of the effective viscoelastic modeling (h/h0 ~2), the NF5 film viscosity was more than twice that of the fully collapsed covalently crosslinked PNIPAAm at high temperature. That behavior was unexpected since the water content in the NF5 was significantly greater than for the photo-crosslinked PNIPAAm films. Moreover, the shear modulus for the covalent network was consistently larger than that for

NF5 films at similar water content, which may be related to the differences in the frequencies utilized in the viscoelastic model for fitting the QCM-D data.

Nonetheless, in both measurements the viscoelastic properties of the hydrogels were within approximately an order of magnitude of each other despite the differences in the swelling ratios and nature of the crosslinks. That result justifies the use of the 3rd and 5th overtones in fitting the QCM-D data, despite the upturn in frequency (Figure 3-5) for higher overtones. Additionally, a factor for an overtone frequency dependence was obtained from the fits for both viscosity and shear elastic modulus as shown in Figure 3-13 and 3-14. At temperatures greater than 20 °C, the frequency dependence was highly variable, while for lower temperatures (< 20 °C) the frequency dependent exponents collapsed to similar values for all the films, except the 10 nm film.

Due to the highly swollen nature of the films in this work the QCM-D extended viscoelastic model169 is used to fit the ΔF and ΔD data; this model incorporates overtone

77 dependence in the shear elastic modulus Equation 3-2 (′ ) and the shear viscosity

7 Equation 3-3 (ηv) in the high frequency range (10 Hz) as a power-law dependence:

′ ′= ′0 (f/f0) (3-2)

″-1 ηv = ηv,0(f/f0) (3-3) where the frequency dependent exponents are: ′shear elastic modulus) and ″-1 (shear viscosity). As shown in Figure 3-13, the frequency dependence of ′is nearly independent of film thickness for temperatures less than 20 °C, except for the thinnest film examined (10 nm). In this temperature range, ′ is approximately 0.4, which illustrates that the shear elastic modulus for these hydrogels appears to be frequency dependent, even in the MHz frequency. The consistency between film thicknesses adds confidence to the physical significance of the physical properties obtained from the fits as the films all swell to similar extents and thus would be expected to have similar properties. The significant spread at higher temperatures is consistent with the prior discussion on the requirements for sufficient dissipation in order to consistently obtain viscoelastic properties for the films. One outlier is the 10 nm film where the frequency dependence becomes weaker (nearly zero). One plausible explanation is that a large fraction of chains may be bound to the surface that could also impact the elastic behavior of this thinnest film. This may also be due to the 10nm film being in the elastic limit of the QCM-D measurement ability as discussed in Chapter 6.

Figure 3-13. The frequency dependence exponent for the shear elastic modulus using the

10 nm film with no apparent dependence on frequency for the shear elastic modulus.

Similarly, Figure 3-14 illustrates the frequency dependence of ″-1, which is associated with the shear viscosity. Again, the behavior is similar for all films except for the thinnest (10 nm) film. Unlike the shear elastic modulus, the shear viscous modulus is frequency independent (Figure 3-13). This suggests that the hydrogel is in the terminal regime in the MHz frequency regime as would be expected. Interestingly, ″-1 is negative for the 10 nm film with an average dependence around -1.2. This suggests shear thickening for the hydrogels.

Figure 3-14. The frequency dependent exponent for the shear viscosity using Q-Tools extended viscoelastic model for initial film thicknesses of 10nm ( ), 32nm ( ), 52nm (

), 75nm ( ), 100nm ( ), and 120nm ( ). The exponent is consistently near zero for all of the films except the 10 nm film. This suggests no dependence on frequency for the shear viscosity for the thicker films.

The region where the model overtone dependence becomes similar correspond well with those where the primary viscoelastic properties were not consistently determined (Figure 3-11), which suggests that there is some dissipation limit at least for these NF5 hydrogels for effective fitting of the QCM-

D data to a generalized model. Moreover, this appears to be close to the Sauerbrey limit,60 where the thickness can be accurately determined solely from the frequency change – i.e., neglecting viscoelasticity. This point is addressed in detail in Chapter 6. One anomaly in the calculated viscoelastic properties is related to the behavior of the 10 nm film where the viscosity was larger than that of the other films. This is especially strange given that the swelling of that film was nearly the same as for the 32 nm film, which exhibited the lowest viscosity, determined by

QCM-D, as would be expected for the largest swelling ratio. One possible

80 explanation for the behavior of the 10 nm film is that the NF5 chains were strongly adsorbed to the substrate, which limited the mobility of the hydrogel. As the thickness of this film was initially commensurate with the size of the hydrophobic nanodomain crosslinks measured for a bulk hydrogel,70 all of the chains may be interacting with the substrate. Irreversible adsorption of polymer chains to substrates has been reported for other polymer systems,151 so this adsorption to the substrate may account for the increase in viscosity of the thinnest hydrogel examined.

It is proposed that an osmotic stress-induced rearrangement of the physical crosslinks28, 37 provides a mechanism that enables the hydrogel to swell to bulk- like levels in these thin films, while still maintaining a stable gel network. The volumetric swelling ratio increased with decreasing film thickness, but the swelling constraint was most prevalent nearest the substrate, so increased osmotic stress might be expected. With increasing stress, the hydrophobic FOSA nanodomains are more prone to re-arrangement. Additionally despite the long time allowed for the films to equilibrate (~1 h) at each temperature step, careful examination of the plateau region of low temperature swelling curves for a single temperature step showed a slow increase in thickness for the thicker films after 60 min (see Figure 3-15). Thus, the differences in swelling may also be attributed to kinetic effects. That is, the rearrangement of the FOSA domains may be less hindered in the thin films and thus reach equilibrium in shorter time. It is useful to note that the low temperature swelling ratio was nearly identical for the 10 nm and

32 nm films, where this additional slow increase in thickness after 1 h was

81 significantly reduced. Moreover, the swelling behavior is consistent with 1D

Flory-Rehner at low swelling extents (high temperatures) where the imposed stress was low, but then significant deviations from that theory were observed at high swelling (low temperatures) for these physically crosslinked hydrogels, where the imposed stresses were significant. That behavior is consistent with the hypothesis of a reversible crosslink mechanism of stress-induced re-arrangements.

Figure 3-15 illustrates the swelling behavior for a 120 nm film during a single temperature step from 25 °C to 24 °C. The single temperature step results in an increase in thickness of 20 nm in the 60 minutes the film is allowed to equilibrate. Although this is taken as the equilibrium swelling state, we see there is still a slight increasing in the thickness at the end of the step. Although in the last 10 minutes, an increase of less than

0.5 nm is seen. Suggesting the film change after 60 minutes is less than 0.2% increase in thickness per minute. This is therefore taken as the approximate equilibrium for all measurements.

Figure 3-15. Swelling of 120 nm film as measured by QCM-D (thick green line) and SE

(black line) for 25 °C to 24 °C temperature steps. The inset shows the swelling for the entire step.

The swelling behavior through multiple heating/cooling cycles was examined to further test the hypothesis of physical nanodomain rearrangement the in thin films. Stresses induced by swelling at low temperatures are hypothesized to induce relaxation of physical crosslinks under stress from the lateral confinement in the thin film. We would expect after cycling that the film should come to a stable equilibrium balance of the collapsed to swollen state.

Figure 3-16 illustrates the swelling behavior of a 32 nm film upon thermal cycling. Similar to the experiments discussed previously, the temperature was held at each step for 1 h. The film was initially swollen as the temperature was decreased from 35 °C to 5 °C (denoted by 1 in Figure 3-16). Upon heating

(denoted by 2 in Figure 3-16), the film did not contract to its initial swelling as expected, consistent with re-arrangement of the physical crosslinks. Moreover, the swelling on subsequent cooling increases substantially. The swelling is significantly greater than that for the bulk hydrogel with the same chemical composition. This suggests that osmotic stress-induced re-arrangement of the network acts to minimize the final stresses in the system to enable enhanced swelling as the physical crosslinks experience sufficient stress to undergo rapid relaxation.

One plausible explanation for the hysteresis in swelling illustrated in Figure

3-16 relies on the mechanism for the determining the extent of the swelling of these hydrogels: the balance between osmotic stress and chain stretching as well as the resistance of the physical crosslink to the imposed osmotic stress. This later

83 consideration is not present for covalently crosslinked hydrogels as the strength of the covalent bond is much greater than the forces associated with the osmotic swelling. For the confined swelling in thin films, the unidirectional stress through the thickness of the film will be greater than that imposed by the isotropic swelling of the bulk hydrogel. This increased stress could enable further re-arrangement of the FOSA domains to enable additional swelling on re-cooling (denoted by 3 in

Figure 3-16) that exceeds the swelling associated with the bulk hydrogel. However after the physical crosslinks are allowed to fully re-arrange, the stresses imposed on the network during swelling will be insufficient to disrupt the FOSA domains.

An additional heating step (denoted by 4 in Figure 3-16) appears to faithfully follow the thicknesses associated with the prior cooling step as would be expected for a constant network. It should be noted that the film used here was a 32nm dry thickness. This film in Figures 3-7 showed increased swelling compared to the other thin films. Due to the film thickness approaching the size scale of the nanodomains, this film may experience a further increased swelling. Further experiments were performed under 1D elongation during neutron scattering to determine the nanostructure changes during stretching and stress relaxation. This point is discussed in detail in Chapter 4. This work will examine the origins of this hysteresis and the associated nanostructure (FOSA domains) changes in these hydrogels.

With the ease in coating from solution without secondary crosslinking reactions and relative invariance in thermophysical properties with film thickness, these physically crosslinked and thermally responsive hydrogel coatings are promising candidates for thin responsive layers for biomedical and sensing applications.

3.4 CONCLUSIONS

The swelling behavior of a hydrophobically modified PNIPAAm hydrogel,

NF5, was examined in thin films using QCM-D and SE. The volumetric swelling of these NF5 thin films is similar to that of the bulk hydrogel. We attribute this greater than expected swelling for the thin films to re-arrangement of the physical crosslinks due to osmotic stress. This flexibility in the network overcomes the

85 conformational constraints associated with the substrate. This re-formation of the network leads to initial large hysteresis in the swelling on re-heating, but the second cooling-heating cycle exhibits completely reversible behavior as an equilibrium arrangement of the crosslinks is formed. These results demonstrate that the physics associated with physically crosslinked hydrogel films differ significantly from those of the more commonly examined covalently crosslinked systems.

CHAPTER VI.

NANOSTRUCTURE EVOLUTION DURING RELAXATION FROM A LARGE STEP

STRAIN IN A SUPRAMOLECULAR COPOLYMER-BASED HYDROGEL: A SANS

INVESTIGATION

4.1 INTRODUCTION

Rapid developments over the past decade have enabled the fabrication of hydrogels that are highly swollen, tough, and ductile through a variety of methods.28, 33, 34,

38, 43 Those methods tend to incorporate a network that provides a stress relief mechanism to dissipate energy on loading and prevent catastrophic failure.38 For double network hydrogels, the rupture of a sacrificial network172 provides toughness and these permanent changes to the network can be assessed using tensile hysteresis and equilibrium swelling.173 Small angle neutron scattering (SANS) was able to provide molecular insights into the structure and origins of toughness of these double network hydrogels.173,

174 Alternatively, the incorporation of transient, reversible physical bonds, such as hydrophobic associations,175 hydrogen bonds,176 or ionic bonds,30 provide a simple route to provide energy dissipation. Physical bonds can rearrange to relieve stress, but then reform to restore the network.175 However, there is little direct knowledge about how these physical bonds re-arrange to dissipate energy and provide high toughness to hydrogels. In order to indirectly examine the deformation of a network, nanoparticle

87 tracers have been added to hydrogels,177 and that work found that the networks deformed either in an affine or non-affine manner depending on the network connectivity and homogeneity. That technique, however, has issues with systems that can relieve stress by relaxation and network rearrangements, which are common attributes in tough hydrogels.

We have previously shown that the hydrophobic aggregates (nanodomains) that comprise the crosslinks in a hydrogel based on a random copolymer of 2-(N- ethylperfluorooctane sulfonamide)ethyl acrylate (FOSA) and N,N-dimethylacrylamide

(DMA) can be characterized by SAXS and SANS.54, 178 Contrast matching using

H2O/D2O in these DMA/FOSA hydrogels to isolate the scattering of the FOSA aggregates and the DMA chains has shown that the nanostructure consists of a relatively narrow distribution of spherical glassy FOSA nanodomains179 that are surrounded by a layer of water depleted DMA.54 The ability to isolate the effective crosslinks, i.e., the nanodomains in that system, provides a route to directly elucidate how those nanodomains change during deformation. The FOSA nanodomains are similar to the self- assembled cores of block copolymer micelles, where SANS using contrast matching techniques has been used to understand the chain exchange kinetics between micelles.100

In this case, chain exchange of the free micelles is driven by thermal energy fluctuations

(kT) that allow chain expulsion/insertion between micelles. More recently, Peters and

Lodge101 studied the rheological relaxation behavior of the same triblock block copolymer (BCP), but when swollen by solvent selective for the midblock of the BCP.

The characteristic relaxation time was nearly four orders of magnitude faster than that for the thermally induced chain exchange.101 Those results demonstrate the importance of

88 stress driven relaxation processes with regard to the properties and structure of supramolecular networks.

Herein, we describe the structural changes within a DMA/FOSA physical hydrogel containing 9.7 mol% FOSA (denoted as DF10) by in situ SANS measurements during stress relaxation in a 150 % step-strain experiment. The structural changes of the

FOSA nanodomains and the elastic DMA (network) chains that bridge the nanodomains were resolved by contrast matching SANS techniques. Immediately after stretching, the scattering pattern associated with the nanostructure became anisotropic, but relaxed towards the isotropic state during the stress relaxation when the sample was held at a constant macroscopic strain of 150 %. The relaxation times associated with the stress and structure relaxations were measured and compared to attain an understanding of how stress relief is related to changes in the hydrogel microstructure when the sample is held at a fixed strain.

4.2 EXPERIMENTAL SECTION

4.2.1 Materials

N,N-Dimethylacrylamide, (DMA, 99.9%, Sigma Aldrich), 2-(N- ethylperfluorooctanesulfonamido)ethyl acrylate (FOSA, 95%, Biolife Sciences), azobisisobutyronitrile (AIBN 99.9%, Sigma Aldrich), 1,4-dioxane (99.5%, Sigma

Aldrich),Diethyl Ether (99.5%, Sigma Aldrich), Methanol (99.5%, Sigma Aldrich) and

D2O (99.5%, Cambridge Isotope Laboratories) were purchased and used as received unless otherwise noted. The DMA was purified by distillation under vacuum and the

FOSA was purified by recrystallization in cold methanol prior to polymerization.69

MilliQ water (18.2 M) was used for H2O in all studies.

A random copolymer was synthesized by the free radical polymerization of DMA and FOSA using a monomer to initiator (AIBN) ratio of 1 to 1000. The reactivity ratios for DMA and FOSA have been reported to be near unity180 and thus the monomers incorporated randomly. The monomer mixture of DMA and FOSA was first sparged with

N2 for 1 h, and then the AIBN, dissolved in 5 mL of 1,4-Dioxane, was injected into the solution through the rubber septum. To initiate the polymerization, this mixture was heated to 60 °C and then continuously stirred for 24 h. The reaction was terminated by exposure to air and then the mixture was cooled to ambient temperature. The polymer solution was concentrated by rotary evaporation and subsequently precipitated in excess of diethyl ether. The composition of the copolymer was determined by 1H NMR (Varian

Mercury-300) to be 9.7 mol % FOSA (Figure 4-1). The molecular weight of the copolymer was 67 kDa with Đ = 1.97 as determined from GPC (Waters 1515 HPLC with

Waters 2414 refractive index detector) using THF as the eluent and calibrated against polystyrene standards.

1 Figure 4-1. H NMR spectrum of DF10 (9.7 mol% FOSA) random copolymer in CDCl3.

4.2.2 Stress relaxation measurements

In order to fabricate the hydrogels, the vacuum dried copolymer was compression molded under vacuum at 165 °C and hydrated in milliQ water for at least 7 days, which was found to be sufficient for the copolymer to reach its equilibrium swelling in water.

Dogbone shaped hydrogel samples (40 mm× 34 mm with 20 mm × 20 mm gauge section) were cut from the swollen sheet for mechanical testing. The stress relaxation was determined using a TA.XT Plus Texture Analyzer (Stable Micro Systems) at an extension rate of 400 mm/min to 150 % gauge strain, the sample was held at this strain, and the stress decay at constant strain was subsequently monitored for 2 h. The sample hydration was maintained during the stress relaxation measurement by spray application of water inside of a sealed flexible chamber. For strain to break measurements, the same dogbone samples were used with an extension rate of 30 mm/min until failure.

4.2.3 Time resolved SANS measurements

For SANS measurements, these same dogbone hydrogel samples were soaked in

D2O/H2O mixtures (accounting for the H2O already present in the hydrogel) for 48 h to allow D/H exchange to obtain the desired contrast in the system. In this case, the stretching stage was a simple extension stage composed of two flat plate clamps that has previously been reported for wrinkling experiments181 and enabled the strain to be applied and held during the duration of the scattering experiments. The sample was stretched to the desired strain (150 %) in approximately 5 s and then the stage was locked in place to maintain a constant strain for the experiment. The entire stretching stage was placed in a large sealed chamber with two large openings covered with aluminum foil to allow transmission of the neutron beam. Images of the stretching stage and the humidity chamber are provided Figures 4-2 and 4-3. Inside the container, the appropriate D2O/H2O mixture to match the sample was added and the sides of the container were lined with paper towels, which were moistened in the D2O/H2O mixture on the bottom. These paper towels increased the surface area for water evaporation to maintain a high humidity level in the chamber, which minimized evaporation from the sample. The SANS measurements were initiated approximately 3 min after stretching the DF10 sample to 150% due to the time required to load the sample into the chamber and align it on the beamline.

The hole cut in the plastic wall was done to insure the maximum scattering angle possible on the beam line would not encounter the plastic shell and only the aluminum foil.

SANS measurements were performed on the NGB 30 m beam line at the Center for Neutron Research at the National Institute of Standards and Technology

(Gaithersburg, MD).182 A wavelength, , of 5 Å with a wavelength spread, , of 14%

93 was used for all measurements with a rectangular beam (1 cm wide × 1.5 cm tall). This beam dimension was selected to maximize the neutron flux and insure that the beam did not scatter from edges of the sample. Initially, the hydrogels were measured in the unstretched state. For these static measurements, a broad scattering vector ( =4 sin

(where  is the scattered angle) range was examined with three sample-to-detector distances: 1.33m (8 beam guides), 4m (4 beam guides), and 10.5 m (1 beam guide). This larger Q range was examined to determine the best sample-to-detector where the critical features associated with the scattering from the DF10 hydrogel can be resolved. The wider Q range scattering profiles are shown in Figures 4-4 and 4-5 for both contrasts examined. For the time resolved measurements, the sample-to-detector distance was 1.33 m with 8 beam guides used. This distance was selected as the primary correlation peaks are well resolved in the associated Q-range. The scattering data were averaged over 180 s for the first 30 min of the relaxation and then over 300 s for longer times to improve the statistics when the structure was not changing as rapidly.

Figure 4-4. Scattering profile for unstretchedDF10 hydrogel swollen with 27/73

D2O/H2O (DMA match) over wide Q range. The interdomain spacing peak centered at approximately 0.09 Å-1is the dominant scattering feature.

Figure 4-5. Scattering profile for unstretched DF10 hydrogel swollen with 50/50

D2O/H2O (FOSA match) over wide Q range. There is an upturn at low Q and a peak associated with the shell intracorrelation at high Q.

The 2D scattering data were reduced to absolute intensity accounting for the beam intensity, detector sensitivity, sample thickness, and sample transmittance using the methods outlined by Kline and coworkers with Igor Pro softwave.183 The 2D scattering data were azimuthally averaged to obtain 1D profiles using sector averages with a width of ± 22° that were centered either parallel (90° azimuthal) or perpendicular (0° azimuthal) to direction of stretch (Figure 4-6) to account for anisotropy. The scattering peak in these

1D profiles were fit using the broad peak model184 for both contrasts. The anisotropy in intensity was quantified using an annular Q average about the scattering peak position to yield intensity as a function of azimuthal angle. The Q-range for this averaging was selected as the full width at half maximum (FWHM) of the scattering peak of the initial unstretched samples. For the hydrogel swollen with 27/73(v/v) D2O/H2O, the averaging was performed over the ΔQ = 0.092 ± 0.024 Å-1, and a ΔQ = 0.14 ± 0.047 Å-1 was used for the hydrogel with 50/50 (v/v) D2O/H2O contrast.

90° with a width of ± 22° (=44°).

4.3 RESULTS AND DISCUSSION

The chemical structure of the DF10 copolymer and a schematic of the microstructure of the DF10 hydrogel178 are shown in Figure 4-7A and 4-7B, respectively.

The network crosslink junctions were the core-shell nanodomains formed by aggregation of the hydrophobic FOSA. Despite the random nature of the copolymer, these aggregates of the FOSA are relatively uniform in size. This behavior is similar to a recent report by

Hirai et al. where uniform micelles were formed from amphiphilic random copolymers.185 In this case, the density of the FOSA aggregates is sufficient to form a network in water, which leads to a mechanically stable hydrogel.

AA DMA FOSA B

The schematic of the hydrogel in Figure 4-7B is color coded to the chemistry of the hydrogel where the FOSA nanodomains are represented in grey, DMA in dark blue, and the water in light blue. The water depleted shell is comprised of predominately DMA segments within approximately 1 nm of the FOSA core, so the FOSA domains are surrounded by the dark blue of DMA. The DF10 hydrogel exhibits a swelling ratio (mass hydrogel/mass dry polymer) of 3.3, which suggests approximately 10 vol % of the continuous phase is DMA chains with the remainder water. This DF10 hydrogel could be elongated over 400% before failure (Figure 4-8).

From an characterization perspective for the structure determination, the large difference in the scattering length density (SLD) between hydrogen and deuterium enables the SLD of the DMA or FOSA phase to be contrast matched with mixtures of

186 H2O/D2O. That contrast matching allows one to resolve the scattering from just the

FOSA phase using an aqueous mixture of 27/73(v/v) D2O /H2O, which matches the SLD of the DMA, Figure 4-9A. When the hydrogel is hydrated with 27/73(v/v) D2O /H2O, the

97 scattering is dominated by the core of the nanodomain composed of FOSA as the matrix of the hydrogel contains DMA and D2O/H2O with the same SLD. In this case, the shell around the FOSA nanodomains is not visible as the aqueous phase matches the DMA.

Figure 4-9B shows the 2-D isotropic scattering from the DMA contrast matched system prior to any deformation. This scattering profile is consistent with the structure illustrated schematically in Figure 4-9A. This scattering provides the structure factor associated with an average interdomain spacing between the FOSA aggregates, D = 6.95 nm, for an unstretched DF10 hydrogel at ~23°C. This spacing, D, is schematically shown by the arrows in Figure 1C. The intensity of the peak is related to the number density of domains with the same spacing and the width of the scattering peak is indicative of the breadth of the distribution of these interdomain spacings.

A B

Figure 4-9. Schematic of the structure resolved by SANS using D2O/H2O mixture to contrast match (C) DMA (27/73 (v:v) D2O/H2O) and (D) the 2-D SANS pattern to determine the interdomain spacing, D.

In a similar vein, the DMA can be independently resolved using a mixture of

50/50 (v/v) D2O /H2O to match the SLD of the FOSA. Figure 4-10A schematically illustrates the structure of the hydrogel gleaned from this contrast. The aqueous phase contains ~ 10 vol % DMA chains that comprises the continuous phase of the hydrogel, which are now visible to the neutrons as the water has the small SLD as FOSA and is thus grey in the schematic. The water depleted shell surrounding the FOSA core is also highlighted, as the SLD for the shell is approximately that of pure DMA and is thus shown as dark blue. As the shape and size of the nanodomains are well defined, the scattering is dominated by the form factor from the water depleted shell of the nanodomain, Figure 4-10B.

A B

Figure 4-10. Schematic of hydrogel structure when (E) FOSA is contrast-matched (50/50

(v:v) D2O/H2O) and (F) the associated 2-D SANS pattern to determine the nanodomain size, ξ, from the form factor associated with the shell structure. The hydrogels were measured at ~23°C in the unstretched state.

Figure 4-10B shows the isotropic 2-D scattering pattern when the FOSA is contrast matched by 50/50 (v:v) D2O/H2O, as illustrated schematically in Figure 4-10E.

This scattering pattern includes contributions from the form factor of the DMA network chain, but the scattering is dominated by a peak that arises from the intra-shell correlation from the form factor associated with the core-shell contrast of the water depleted DMA layer surrounding the FOSA core. This peak in the scattering from the form factor occurs due to the shell having an SLD that is smaller than that for the core and matrix. Similar scattering behavior has been previously reported for selectively labeled dendrimers, which leads to a similar shell-core structure in solution.187. In Figure 4-10B, the peak location is the size of the nanodomains, ξ = 4.53 nm, which is the diameter of the FOSA core plus the thickness of the water-depleted DMA shell. ξ is hereafter used to describe the nanodomain size. The arrows in Figure 4-10A illustrate ξ schematically.

In addition to understanding the structure of the hydrogel in the undeformed state, the stress relaxation behavior of the DF10 hydrogel is another baseline requirement to

100 enable insight into the relationship between nanoscale structure and relaxations that lead to high toughness of these hydrogels. For the stress relaxation, the DF10 hydrogel was extended rapidly (400 mm/min) to 150% strain as shown in Figure 4-11A. The stress relaxation was then monitored for 2 h, while insuring the sample remained hydrated by using a humidity chamber. Figure 4-11B shows that about 60% of the stress relaxed in the first 2 min and ~80 % of the stress relaxed within the first 20 min. During the next

100 min, the stress relaxed to ~7 % of the initial load.

Stress relaxation data are commonly described by a stretched exponential function, which has been shown to provide a reasonable fit of stress relaxation data for the DMA-FOSA copolymers,37 but these relaxation measurements are generally performed at low strains.188 A stretched exponential, however, provides only an average

101 relaxation time and a measure of the breadth of the relaxation time distribution. That weakness is especially important with regards to multiphase systems, such as the DMA-

FOSA hydrogels, where there may be multiple structures that have their own distributions of relaxation times. An alternative approach for fitting stress relaxation data, used herein, is to employ a Generalized Maxwell Model (GMM),189 Equation (4-1), that provides multiple relaxation times, but it also requires a more complex mathematical fit of the experimental data (additional fit parameters),

 (t) N t / n (4-1)   Ane o n1

where (t) is the instantaneous stress, o is the initial stress (t = 0), n is the relaxation

th time of the n Maxwell element, An is an apportioning factor of how much each Maxwell element contributes to the total stress and N is the number of Maxwell elements needed to describe the stress relaxation data. A GMM with N = 7 fit the stress relaxation data for the DF10 hydrogel well, see black line in Figure 4-11B and residual error of fit in above the fit, with relaxation times of n = 0.0125, 0.0730, 0.432, 1.55, 7.94, 37.0 and 323, for n

= 1 – 7, respectively. Note that the relaxation times span five orders of magnitude. The choice of 7 relaxation times was arbitrary, but the fit was not significantly improved when N > 7. The GMM fits for N = 1 – 6 and the associated residual error of the fits are shown in Figure 4-12.

102

Figure 4-12. Maxwell element fits using 1-6 exponentials with residual error of fit shown.

Figure 4-13 shows both the SANS 2D scattering profiles and the 1D profiles for the average scattering intensity across the peak as a function of azimuthal angle, for step- strain experiments that match the stress relaxation experiment (Figure 2). The stretch direction for the sample corresponds to the arrows in the 2D scattering plots in Figure 4-

13. Figure 4-13A illustrates the scattering using a hydrogel with a 27/73 (v:v)

D2O/H2Omixture. In this case, the peak position in the 2D scattering is related to the interdomain spacing (D) between the FOSA aggregates. Shortly after stretching (4 min), the scattering profiles show a higher intensity on the ring along the equator (at  = 0º and

180º). The differences in the azimuthal scattering intensity are related to the concentration of the nanodomain correlations in the directions perpendicular versus

103 parallel to the stretching direction. This anisotropy can be better illustrated by the 1D patterns where the average intensity of the peak centered at Q = 0.092Å-1and averaged over ΔQ=0.024 Å-1, for the interdomain scattering is examined as a function of azimuthal angle. In Figure 4-13A, there is a clear peak in the scattered intensity at  = 0º and 180º at

4 min. At intermediate time (24 min), the azimuthal variance in intensity decreases, but a maximum in intensity remains at  = 0º and 180º. At long times (424 min), the average intensity across the peak is independent of azimuthal angle, which indicates that the anisotropy in the concentration of the nanodomain correlations imparted by the step- strain has relaxed over the time frame of the SANS experiment.

Figure 4-13B illustrates the scattering using a hydrogel with a 50/50 (v:v)

D2O/H2O mixture. Here, the peak position in the 2D scattering is related to the nanodomain size (ξ) that was determined from the intra-shell correlation peak (ξ = 2π/Q0) associated with the contrast for the core-shell structure of the nanodomain. The 1D scattering patterns in Figure 4-13B are averaged by integration over the width of the peak at Q = 0.14 ±0.047 Å-1. Similar to the scattering from the FOSA nanodomain correlations

(structure factor), the scattering patterns associated with the form factor of the nanodomain also became anisotropic on deformation as shown in Figure 4-13B. From the azimuthal dependent intensity, the peak (associated with the anisotropy in intensity) is much broader for the scattering associated with the nanodomain shape (Figure 4-13B) in comparison to the FOSA nanodomain correlations (Figure 4-13A) after 4 min of relaxation. The difference in the scattered intensity, associated with the directions parallel and perpendicular to the stretching, decreases as the sample relaxes. However, a discernible peak in the azimuthal dependence of the scattered intensity remains even after

104

418 min that is associated with the form factor (shape) of the nanodomains. To more clearly illustrate the anisotropy in the scattering, the intensity scale for the 2D scattering patterns has been re-scaled for both contrasts (Figure 4-14).

0.14 ± 0.047 Å-. The = 0° azimuthal angle is shown on the first 2D pattern of each set of data for reference.

105

Figure 4-14. Adjusted intensity scale for the 2D scattering data at 50/50 contrast at 418 min and the 27/73 contrast at 424 min. The slight anisotropy remaining for the 50/50 case

(FOSA matched, domain form factor scattering) can be distinguished in this case, while the scattering is isotropic for the 27/73 case (DMA matched, interdomain structure factor scattering). Note in the 27/73 case the beam spot cannot be seen.

In order to explain the anisotropy in the intensity, which is associated with the concentration of the water depleted DMA shells, the physics of the processes involved during a large step deformation must be considered. In the stretching direction, the DMA tie chains will become extended leading to deformation of the FOSA nanodomains and ultimately pull-out of FOSA moieties from the nanodomains. This pull out will disorder the DMA shell and likely lead to hydration of the DMA to reduce the scattering contrast associated with the shell. However, the scattering intensity associated with the shell does not fully recover over the time frame of these SANS experiments (7 h). As this scattering is associated with the water depleted DMA, it is instructive to examine the scattering more carefully to extract information associated with the DMA chain conformation. The

2D scattering profiles for the stretched hydrogel (Figure 4-13B) exhibited a butterfly pattern near the beam stop, which is characteristic of chain stretching67, 190-192 in the

106 direction parallel to the strain. This butterfly pattern is not present in the unstretched hydrogel (Figure 4-9F). When examining the temporal evolution of the 2D scattering patterns, the butterfly pattern persisted through the entire experiment, which indicates either a permanent deformation of the DMA chains (residual stress on the interconnecting

DMA chains) or that relaxation of the DMA chains was not complete after 7 h. As the hydrogel remains in a strained state (150%) at the completion of the SANS measurements, the limited shape recovery (~ 20 % decrease in the length of the hydrogel) on release of the hydrogel from the clamps provides some evidence for residual stresses remaining in the material. The deformed DMA chains that give rise to the butterfly pattern at the end of the relaxation experiment likely provides some of the retractive force to decrease the length of the hydrogel. The long recovery for these DMA chains could be associated with the glassy nature of the FOSA core, which significantly reduces the ability for FOSA monomers to rearrange their location.

To more quantitatively examine the anisotropy in the scattering intensity, the peaks in the azimuthal angle dependent intensity (Figure 4-13) were fit with a Gaussian function (solid lines). The amplitude of the Gaussian peak (difference from maximum intensity to baseline intensity) provides a measure of the relative anisotropy in the system. For direct comparison to the stress relaxation data, this Gaussian peak amplitude at any time, t, was normalized by the amplitude instantaneously upon stretching, t = 0, to provide quantification of scattering anisotropy in terms of an amplitude ratio, AR ≡

A(t)/Ao. However, since the first SANS measurement was obtained at t = 4 min after stretching, the actual value of Ao is not directly measured. In order to estimate Ao, the

107 amplitude determined from the Gaussian fit of the peak in I()was plotted as a function of time and fit Equation 4-2, the sum of three exponential functions (Figure 4-15):

3 t / n A(t)  ne (4-2) n1

where the n and n were constants. The sum of three exponentials produced an excellent fit of the data, and the value of Ao was then determined by using the regression equation to extrapolate the data to t = 0. That introduces some error in the absolute value of AR, but the relative changes in AR are the important factor to consider here, so this extrapolation is not seen as a critical deficiency of the analysis described below.

108

Figure 4-16 shows how the AR, which is a measure of the anisotropy in the scattering intensity of the peak, decays during stress relaxation following a step strain to

150%. It should be noted that here we are only describing the anisotropy in the averaged scattering intensity of the correlation peaks associated with either the interdomain spacing (ID) or the intradomain size(Iξ). AR does not include any information about the dimensions associated with the spacing or size (correlation peak position) as ID and Iξ are averaged across the width of the correlation peak. For both contrasts, there was first a rapid decay in the anisotropy of the scattering intensity. The AR decayed to 50 % of its initial value in ~4 min for Iξ and ~15 min for ID, which infers a relatively rapid structural recovery for a majority of the DF10 hydrogel. At these short relaxation times, the structural relaxation processes associated with the anisotropy in the intensity of the two phases of the hydrogels are suspected to be correlated due to the glassy nature of the

FOSA domains, which should limit the kinetics of structural recovery. For ID, the changes in AR are associated with recovery of nanodomains with a spacing of approximately D and the reformation of the water depleted DMA around the FOSA nanodomains to provide contrast for scattering. At longer times, the relaxation of the scattering associated with spacing of the FOSA nanostructure, ID, diverged from that associated with the anisotropy of the shell scattering, Iξ. The former decayed to zero, while the anisotropy associated with the water depleted DMA shell relaxed very slowly, and Iξ remained finite after 7 h.

109

The rate of the decay in the anisotropy in the scattered intensity for the two contrasts examined provides some insight into relative mobility of the two phases of the hydrogel and the differences in the structural relaxation processes. The interdomain spacing, D, is dependent on the conformation of the DMA-chains that connect the nanodomains. Stretching the hydrogel stretches and orients the network chains in the direction of stretching. That also produces a restoring stress in the DMA chains in accordance with rubber elasticity. The continuous, water-swollen DMA phase is

110 primarily water, so the network chains have considerable mobility, which generates a relatively fast stress relaxation process to restore a uniform density of nanodomains, which corresponds to a decay of A/Ao to zero, which matches the unstretched, isotropic hydrogel.

The intra-shell correlation from the form factor of the nanodomains, ξ, provides the approximate nanodomain size. This correlation is a result of the water-depleted DMA shell surrounding the glassy FOSA nanodomain as has been reported previously.54, 178

The intra-shell correlation is affected in two ways by stretching the hydrogel. First, the nanodomains can undergo a solid-like deformation, but this does not impact Iξ, which defines the AR, but rather impacts the relative position of the scattering peak. The retractive stress in the attached network chains can reorient the nanodomains to relax this stress. Second, since the FOSA groups are not covalently bonded, the stress from the network chains may rearrange the FOSA groups in the nanodomain core to pull FOSA groups out of the nanodomain, which produces stress relaxation. This pull out would likely hydrate the DMA segments near the FOSA and thus impact Iξ. Removal of a perfluorinated FOSA group from the nanodomain into the water-swollen polymer phase, however, produces a large enthalpic penalty, so there is a significant driving force for the re-aggregation of the FOSA into either the original nanodomain or an adjacent nanodomain. Since the nanodomains are glassy, the relaxation times of the aggregated

FOSA groups in the core and of the DMA segments in the shell are expected to be large.

Thus, once the stress in the continuous water-swollen DMA phase has mostly dissipated, such that the energy associated with the residual stress is less than the enthalpic penalty associated with removal of a FOSA group into the aqueous phase, one might expect little

111 or no further structural relaxation of the nanodomain anisotropy, which is consistent with the slow relaxation of the remaining intra-shell scattering, Iξ, shown in Figure 4-16 at the long times.

In order to better compare the structural relaxations with the stress relaxation, the decay of the anisotropy in ID and Iξ were each fit with the sum of three exponentials,

A(t) 3  A et / n (4-3) A  n o n1

where the An provide the relative contributions of each relaxation process and the n are relaxation times. The choice of three exponentials was arbitrary, and was solely based on the fact that three exponentials fit the data better than two and using four exponentials did not significantly improve the fit, but this follows the GMM used to fit the stress relaxation. The relaxation times generated from the structure relaxation fits, 6.28, 47.3 and 252 min and 2.63, 42.1 and 950 min for ID and Iξ, respectively, show the longer time relaxation behavior of the nanodomains. Note that the two slower relaxation times for both structure relaxations were similar, which is due to the connectivity of the two structures. That is the shorter time relaxation processes of the nanodomain anisotropy were coupled to the relaxation of the network chains. However, once the network chain relaxed ( = 252 min), the driving force for relaxation of the nanodomain structure was essentially removed, which explains the much longer relaxation time ( = 950 min) for the anisotropy associated with the water depleted DMA shell that persisted for over 7 h.

The continued relaxation of the nanodomain correlation anisotropy after the network

112 chains had relaxed may be a consequence of thermal processes, similar to what has been reported for the relaxation of structure in block copolymer micelles.100

In addition to the anisotropy of the scattering peak intensities, ID and Iξ, the location of the scattering peak, which is related to the spacing, D, or size of the nanodomains, ξ, changed on stretching in a manner dependent on the direction relative to the applied deformation. During the stress relaxation, the spacing and size also generally relaxed with time towards the dimensions for the unstretched, isotropic hydrogel. In order to analyze these dimensional changes, the 2D scattering profiles (see Figure 4-9 and 4-

10) for examples) were converted to directional1-D plots by using an azimuthal ( sector average to yield I(Q) for the scattering primarily parallel ( = 90° ± 22°) and primarily perpendicular ( = 0° ± 22°) to the stretching direction. An example of these azimuthal sector averages is shown in Figure 4-17. Those scattering profiles provided the scattered intensity as a function of Q and thus provide a route to obtain the average interdomain distance, D, or nanodomain diameter, ξ. I(Q)was fit using a broad peak model183, 184to determine the peak position (D and ξ) and full-width at half maximum (FWHM). The

FWHM provided a measure of the distribution of D and ξ. The temporal dependences of

D and ξ provide insight into how the nanodomains and network chains (interdomain distances) deform when the sample was stretched and during the stress relaxation process as shown in Figure 4-18.

113

Figure 4-18A shows that at 4 min after the step-strain (i.e., the earliest SANS data that were obtained) the FOSA nanodomains were stretched by nearly 5 Å (11.5%) parallel to the stretching direction and simultaneously compressed by 1.5 Å (3.4%) perpendicular to the stretching direction. The difference in the deformation of the nanodomain parallel and perpendicular to the stretching direction is consistent with the

Poisson’s ratio expected for a glass (≈1/3). This infers that the density of the nanodomains is decreased under tension (dilative deformation).In less than 100 min, the nanodomain size, ξ, in both directions recovers to its original value, with ξ appearing to recover more quickly.

114

Figure 4-18. (A) Evolution of ξ (nanodomain size) in the parallel (‖) and perpendicular

() directions to the deformation during relaxation. (B) The FWHM of the scattering peak associated with ξ that describes the change in the distribution of sizes.

It is important to note that the anisotropy in the sizes of the nanodomains decays to isotropic (Figure 4-18A) much faster than the long lived anisotropy in the scattered intensity (Iξ) that was determined as a function of azimuthal angle (Figure 4-13B and 4-

15B). This difference can originate with the origin of the scattering peak associated with

ξ, which is the form factor for the water depleted DMA shell surrounding the FOSA nanodomains. For the size ξ, only the peak location (Q) is important and small variation in the contrast will not impact the calculated size of the nanodomain. Conversely, the intensity of the peak (Iξ) is directly related to the contrast in the system and the number of nanodomains of similar size, so alteration of the distribution and hydration-level of connecting DMA chains near the FOSA nanodomains may have occurred that would impact the scattered intensity, but not the location of the correlation peak.

In addition to the size of the nanodomain, the breadth of the correlation peak provides information about the distribution of sizes as quantified by at the FWHM shown in Figure 4-18B. Upon stretching, the FWHM in the direction parallel to the deformation

115 is nearly double its isotropic value, with larger error bars. This would suggest the nanodomain is highly distorted in the direction of the stretching with a large distribution of sizes in the parallel direction. The large uncertainty in the fit arises from the decreased scattered intensity, which may be associated the nature of the water depleted shell. Again by 100 min, the distribution (FWHM) has nearly recovered similar to the recovery for ξ, but the FWHM in the parallel direction of strain remains larger than its initial isotropic value. This behavior aligns with Figure 4-13B and 4-15B, which indicates that the scattered intensity associated with the shell (Iξ) remains slightly perturbed and is not able to fully recover. The slightly broader distribution of sizes of the nanodomains is likely indicative of the distribution of stresses imposed to the nanodomains that result in plastic deformation as their glassy nature should inhibit significant rearrangements beyond elastic recovery. This deformation could be accompanied by an altered distribution and hydration-level of connecting DMA chains near the FOSA nanodomains in the direction parallel to the stretching. In contrast, the perpendicular direction FWHM for the nanodomain size remains largely unchanged throughout the relaxation, which suggests that the changes in ξ perpendicular to the applied deformation are likely limited to relatively uniform deformation of the FOSA nanodomains. The difference in the behavior of the FWHM in the directions parallel and perpendicular to the stretching may be a result of FOSA pull-out into the aqueous phase under tension (parallel) that alters the distribution of nanodomain sizes if the FOSA is subsequently inserted into a neighboring domain.

Figure 4-19A shows how the interdomain distance, D, evolves during the stress relaxation. As our first measurement is made at 4 min, we are unable to fully capture the

116 initial structural relaxation immediately on stretching. Upon stretching the sample, the

interdomain distance parallel to the stretching direction, D initially increased and the interdomain distance perpendicular to the stretching direction, D, initially decreased in the first few minutes. These dimensional changes in the spacing between the nanodomains are qualitatively consistent with expectations associated with the uniaxial stretching based on bulk dimension changes on deformation. However, the initial

increase of D and decrease of D were followed by a rapid change in the two spacings

that led to an increase of D by 1.4 Å from D0 (unstrained sample) and a decrease of D

by 1.4 Å. The decrease in D persisted for over 7 h. There was no statistical difference

between the spacing at 100 and 420 min. This lack of recovery of D to its initial dimensions is similar to the persistence of the butterfly pattern associated with stretched

DMA chains (Figure 4-13B). As the hydrogel remains macroscopically strained to 150 % during the stress relaxation, it is not unreasonable that the structure does not recover to its original state. In the perpendicular direction, D increased to a maximum of 71 Å, but then began to relax back towards the initial dimensions of the unstrained sample at ≅100 min. The spacing then slowly decayed over the next 300 min back D0 of the DF10 hydrogel. Note that the time at which D began to diminish coincided with the time at which ξ returned to its initial, unstretched value (c.f. Figures 4-18A and 4-19A).

The initial increase in D and initial decrease in D are consistent with expectations for the change in the spacing between crosslinks on a stretching deformation as is the relaxation towards the initial interdomain spacing, D0. However, it is not readily

apparent how and why D decreases to less than D0 (and D increases beyond D0) during the initial relaxation. This behavior appears to be almost an oscillation in the deformation

117 and may be related to the recoil of the chains on relaxation. This overshoot in the recovery may be related to the initial step-strain applied, through the amount of potential energy transmitted to the network chains. Determining the mechanisms associated with the overshoot in the structural recovery requires additional study.

Nonetheless, the distribution of D through the FWHM provides some preliminary insights into this behavior. As shown in Figure 4-19B, the FWHM, which corresponds to the distribution of D parallel to strain, decreases by 25%, so the interdomain spacing becomes better defined after the stretching. This decrease in FWHM is counter to the increase in the distribution associated with ξ (Figure 4-18B). The FWHM parallel to the stretching direction does not recover in the 7 h of the measurement, similar to how the spacing changes plateau at a lower D (Figure 4-19A) in this direction. In order to understand how the FWHM on recovery can be smaller than the initial FWHM of the unstretched hydrogel, I(Q) near the peak is examined for several times as shown in

Figure 4-19C. The scattering profiles are nearly invariant qualitatively when examining the direction perpendicular to the stretching, which is consistent with the analysis from the fit of these data in terms of D (Figure 4-19A) and FWHM (Figure 4-19B).

Conversely, there is a clear change in the shape of I(Q) in the parallel direction to the stretching with an asymmetric change in the shape of the peak as intensity decreases on the low Q side of the scattering peak, which corresponds with a decrease in the number of nanodomains with a larger than average D. This scattering profile does not appreciably change between 17 and 420 min, but there is a large change from 4 to 17 min with an increase in the total scattering intensity at longer times. To explain the loss of the low Q side of the scattering peak on stretching, one would expect that most extended network

118 chains, which should correspond to the larger D, are more efficient at transferring stress to the nanodomains to lead to preferential pull out of FOSA groups attached to these extended network chains during stretching. The larger spacings between the nanodomains do not appear to reform upon relaxation of the stress in the network. On relaxation, one could imagine that there is a probability distribution associated with DMA chain conformation that also controls the spacing between the nanodomains. As such, there is likely a greater probability to re-form crosslinks with a less extended state for the connecting chains as the extended chains were preferentially removed during the stretching process.

One additional possibility to explain these results is the change in volume of the hydrogel during the stretching and relaxation process. At low strains, hydrogels are generally treated as rubbers with a Poisson’s ratio of 1/2, which leads to isochoric deformation. As the initial deformation is large (150 % strain), the mechanical response is likely non-linear and Poisson’s ratio likely changes during the experiment as has been reported for other hydrogel systems.193 A Poisson’s ratio less than ½ leads to a reduced density for the material after deformation. During the stress relaxation, the FOSA nanodomains that comprise the crosslinks and DMA network chains re-arrange and this re-organization at a reduced density when initially stretched may also act to alter the equilibrium structure after relaxation. These SANS experiments exploring the structural evolution in a hydrophobically crosslinked hydrogel only begin to reveal how the reversible crosslinks relax in response to a tensile step-strain. Understanding of non- linear mechanics associated with reversibly crosslinked systems may be assisted by the

119 use of in-situ scattering experiments that provide insight into structural changes that occur on deformation.

Figure 4-19. Temporal evolution in (A) D (interdomain spacing) upon stress relaxation and (B) FWHM of the scattering peak both parallel (║) and perpendicular (┴) to the deformation. The gray area represents isotropic interdomain distance before strain within one standard deviation. The FWHM provides a measure of the distribution of D distances. (C) I(Q) near the interdomain scattering peak for several relaxation times both parallel (║) and perpendicular (┴) to the deformation. These scattering profiles reveal the asymmetric rearrangement of D showing a significant loss on the low Q (larger spacing) side in parallel direction.

120

4.4 CONCLUSIONS

The structural changes associated with stress dissipation were elucidated in a hydrogel based on an amphiphilic random copolymer of N,N-dimethylacrylamide (DMA) and 2-(N-ethylperfluorooctanesulfonamido)ethyl acrylate (FOSA). SANS and contrast variation enabled the evolution of the structural relaxations to be independently resolved for the two components of the copolymer that comprise the hydrogel. The stress relaxation data following a step-strain were fit to a Generalized Maxwell Model (GMM) with 7 elements whose relaxation times spanned five orders of magnitude. Structural relaxation was calculated from the anisotropy of the scattering intensity associated with correlation peaks associated with the size of the FOSA nanodomains that served as the physical crosslinks and the spacing between these FOSA nanodomains. The relaxation times associated with the decay in the anisotropy differed by an order of magnitude and agreed with the longer relaxation times determined from the stress relaxation fits to the

GMM.

In addition to the anisotropy in the scattered intensity, the dimensions associated with the size and spacing of the nanodomains also became anisotropic after stretching the hydrogel. On stretching, the interconnecting segments (network chains) initially elongated in the direction of the applied strain, but then rapidly decreased during stress relaxation. The spacing relaxed to dimensions less than that in the unstrained hydrogel.

We suspect that this decreased spacing is a manifestation of the preferred stress dissipation from FOSA moieties pulling out of the nanodomains for the most stretched

DMA tie chains during the stretching process. Even after 7 h of relaxation, the spacing between the FOSA domains remained less than that of the unstretched hydrogel in the

121 direction of the applied deformation. The FOSA nanodomains also deformed anisotropically, but their dimensions quickly recovered to close to those of the isotropic unstretched gel. The results suggested a physical origin for how these nanostructured supramolecular hydrogels dissipate stress through pull-out of FOSA from the nanodomains and their recovery to a structure similar to the initial hydrogel during stress relaxation. These measurements provide insight into the toughening of hydrogels with reversible interactions and their recovery.

122

CHAPTER V.

SUPRAMOLECULAR HYDROPHOBIC AGGREGATES IN HYDROGELS

PARTIALLY INHIBIT ICE FORMATION

5.1 INTRODUCTION

Control of ice formation is a critical industrial challenge for numerous applications from renewable wind power194 and preservation of biologics,126 where inhibition of freezing is desired, to food products,195 where precise control of ice crystallization is desired. Decades of work involving fundamental investigations of water have identified numerous crystalline and amorphous ice structures.196-198 One key finding from these fundamental studies is that the confinement of water to dimensions less than

10 nm tends to inhibit water crystallization;128, 199 typically this inhibition is only a kinetic limitation. These confinement studies tend to be limited to hard materials where water concentrations are generally low and water is confined within the pores of the material.200-202

Conversely, Nature has evolved unique methods for mitigating ice formation with soft materials in aqueous environments.203 Typically these inhibition strategies involve adsorption of water-soluble macromolecules to the surface of nuclei that prevent the growth of ice crystals. Synthetic mimics of these proteins can inhibit ice growth in aqueous solutions,126 but these strategies are limited to relatively limited degrees of undercooling (<10 K). An alternative evolutionary strategy for proteins involves water

123 confinement between hydrophobic residues to inhibit freezing.128 However, soft synthetic mimics that inhibit ice by hydrophobic confinement are, in general, lacking and represent an unexplored route to enable the supercooling of water.

Hydrogels represent one of the most widely examined forms of water-rich soft matter, as they can mimic the mechanical properties and aqueous environment of biological tissue.204 However, the freezing of water is generally only modestly impacted by inclusion in a hydrogel, while the polymer network of the hydrogel is irreversibly damaged structurally by ice crystallization.126,135, 205 Antifreeze proteins with hydrophobic residues can inhibit ice formation128 and thus provide inspiration towards the development synthetic analogs. Within this context, it is hypothesized that nanoscale confinement of water between hydrophobic moieties within hydrogels should also inhibit ice formation.

Herein, we describe the ability to tune water freezing within supramolecular hydrogels. The water is confined by hydrophobic nanodomain crosslinks that are formed through the physical association of the hydrophobic constituent (FOSA) of a DMA-

FOSA random copolymer hydrogel. Prior small angle neutron scattering (SANS) measurements have demonstrated <10 nm hydrophobic regions between the FOSA nanodomains,17 here we confirm these SANS results and demonstrate the persistence of the hydrogel nanostructure, at sufficiently high FOSA content, deep into the supercooled state for water, while the scattering from the nanostructure was lost at low FOSA content during crystallization of the water in the hydrogel. The dynamics of water within the hydrogel examined by quasielastic neutron scattering (QENS) were significantly altered from that of bulk water and demonstrated that extremely mobile water with liquid like

124 diffusivity can persist to 205K within these hydrogels. These characteristics appear to prevent damage to the hydrogel on thawing, similar to strategies in nature for survival in cold climates. This demonstration of unexplored strategy for ice inhibition will enable the design of new materials for applications, such as the preservation of biologicals and anti- icing surfaces.

5.2 EXPERIMENTAL SECTION

5.2.1 Materials

A statistical random copolymer of N,N-dimethylacrylamide and 2-(N- ethylperfluorooctane sulfonamido)ethyl acrylate was synthesized by free radical polymerization as described previously in the literature.69, 70 The comonomer feed composition was used to effectively tune the composition of the resulting copolymer. For

H2O, MilliQ water (18.2 MΩ resistance) was used in all studies. For select neutron experiments, D2O (99.9%, Cambridge Isotope Laboratories, Inc.) was used in place of

H2O. Hydrogels were formed simply by immersing compression molded sheets of the copolymer in an excess of MilliQ water.

5.2.2 Differential Scanning Calorimetry (DSC)

Copolymer pieces (<20 mg, or <1mm square) were soaked in H2O for a minimum of 3 days to form equilibrated hydrogels. Small pieces (3-6 mg) of the hydrogel were cut from the sample and blotted with lint free wipes to remove excess surface water. We found no dependence of the thermal transitions on mass loading. The hydrogel was placed into an aluminum DSC pan (TA) and hermetically sealed. Calorimetric

125 measurements (TA DSC 8500) were performed between 293K and 205K at cooling rates of 0.5, 2, and 5 K/min. The sample was isothermally held at 205K for 10 min prior to reheating to 293K at the same rate. This cool-heat cycle was repeated to check for reproducibility and any history dependence on the freezing/melting behavior.

Additionally, the DSC pans were weighed before and after measurement to insure that no water was lost during the measurement that could impact the analysis. No statistically significant change in mass was observed during the course of the measurement.

5.2.3 Small Angle Neutron Scattering (SANS)

All SANS measurements were performed using NGB30 meter SANS (NCNR

NIST, Gaithersburg, MD) with a beam wavelength of 6 Å, a spread of 14% and a beam size of 1.5 cm. Three sample to detector distances were used, 133 cm, 350 cm, and 1250 cm, to provide Q ranges of 0.020-0.31, 0.0085-0.083, and 0.0036-0.026 Å-1, respectively. For 133 and 350 cm distances, 4 neutron guides were used before the sample. For 1250, one neutron guide was used. The scattering at each distance and temperature was collected for 5 min. The sample was cooled with a closed-cycle refrigerator (CCR) at 2K/min to 2K above the desired temperature and then slowly stepped to the set point to avoid overshoot in cooling. A similar procedure was followed on re-heating to check the reversibility of structural changes induced on cooling. For each hydrogel (DF5, DF15, DF22), 3 contrasts were measured: pure D2O (maximum contrast and minimal background), 50/50 v/v D2O/H2O (contrast match FOSA, allows measurement of interdomain spacing), and 27/73 v/v D2O/H2O (DMA contrast match, allows measurement of domain size from shell scattering). The scattering data was

126 corrected to absolute intensity using the sensitivity of the detector, sample transmission, sample thickness, and an empty cell background. SANS results were reduced and analyzed using the NIST SANS tool package.183

5.2.4 Disk Chopper Spectrometer (DCS)

All DCS measurements performed on NG4 DCS (NCNR NIST, Gaithersburg,

MD) using a wavelength of 6 Å. The hydrogel sample was 1 mm thick total cross section in an aluminum pouch and placed in an aluminum cell for the measurement purged with

He and sealed with indium (explained in more detail below). The neutron beam on the sample was masked to 7 cm × 1.5 cm by cadmium. The sample was cooled with a low temperature Closed Cycle Refrigerator (CCR) with a temperature range of 325 K to 4 K based on the Gifford-McMahon Refrigeration scheme with helium as the coolant. An effective cooling rate of ≤2 K/ min was used for all measurements. The sample was equilibrated at the measurement temperatures (295 K, 270 K, 260 K, 250 K, 240 K, and

220 K) for 15 min prior to beginning the measurement. The scattering at each temperature was measured for 5 h to obtain good statistics of the quasi-elastic scattered neutrons. The reversibility was examined by re-heating the samples from 220K to 250 K and 275 K at 2 K/min ramp. The background was determined using empty can and dark count measurements for 1h each at room temperature. The resolution was calibrated using a vanadium filled cell with statistics based on 4 h of scattering.

127

5.2.5 High Flux Backscattering Spectrometer (HFBS)

All HFBS measurements performed on NG2 HFBS (NCNR NIST, Gaithersburg,

MD). The neutron beam wavelength used was 6.27 Å. A standard top loading CCR with a temperature range of 4K – 700K was used for measurements with temperature accuracy of 0.1K. The Mean Squared Displacement (MSD) was calculated using a standard

Gaussian approximation based on the elastic intensity. With the intensity I0 calculated at the lowest temperature measured of 4K for the samples. A cooling rate of 0.8K/min was used from 345 K to 4 K for these measurements. The MSD was averaged and recorded every minute. For these scans, hydrogels were fabricated with both D2O and H2O to examine the dynamics primarily associated with the copolymer and water, respectively.

Additionally as a control, the scattering of a dry copolymer was also measured with

HFBS. To elucidate the energies associated with the local motions, the hydrogels were measured using an energy window of ± 16 μeV. For these measurements, the full energy window configuration was limited to the DF22 sample as this was the most intriguing hydrogel. The scattering of the DF22 hydrogel was measured at 295 K, 270 K, 260 K,

250 K, 240 K, and 220 K. Each temperature was measured for 6 h and a cooling rate of

1K/min was used to minimize overshoot between set temperatures. The sample was equilibrated for 10 min prior to each measurement.

5.2.6 Sample Preparation for SANS and QENS Measurements

For SANS measurements, the sample cell was 50 mm by 50 mm with a 4 mm gap between front and back plates. The hydrogel samples were cut to fit inside and wrapped in an aluminum foil pouch before loading into the cell. The sample pouch was loaded into

128 the sample cell and the cell was sealed with lead wire and mounting screws every 5 mm around the perimeter. The CCR heat shields were designed to reduce scattering from the shields by removing the shield wall in the direction of the beam and replacing with ultrapure aluminum foil. For the outer shell (of ¾ inch thick aluminum), large silicon windows are used in the beam path to increase transmission of the direct beam and still allow a vacuum to be pulled inside the CCR chamber.

The hydrogel samples for the DCS (QENS) measurements were prepared by first pressing the DFx copolymer at 155 °C within an aluminum foil pouch. The mass of the

DFx copolymer within the pouch was based on the calculated mass to generate approximately 0.5 mm thick swollen hydrogel within the pouch (55 mm by 90 mm). The sample was loaded in a compression molder heated to 160 °C and then pressed to 5000 psi for 15 minutes. After pressing the copolymer, the sample was cooled to room temperature and the aluminum pouch was opened with the copolymer adhered to one side of foil. The copolymer was equilibrated in water (Millipore, 18.2 M) for 24 h, then the swollen hydrogel was blotted dry with a lint free task wiper to remove excess surface water and finally re-encapsulated in a aluminum pouch with crimped edges. The pouch acts to both retain H2O and assist in loading the sample into the measurement cell. The packaged sample was rolled into a 17 mm diameter cylinder that is 90 mm long and placed into the sample cell (20 mm diameter). The samples were loaded to insure the foil seam was perpendicular to the neutron beam. The cell was purged with Helium gas briefly and sealed with indium wire. Cadmium plates were used to reduce the beam to a size of 70 mm by 15 mm, centered on the sample. A cadmium strip imbedded into the outer aluminum shell of the CCR outer shell was used as the beam stop and boron paste

129 was added to the top and bottom of the cell to minimize scattering from the ends of the sample.

The same sample preparation methods were used for HFBS samples except the rolled sample was 28 mm diameter and 50 mm tall to match the beam shape on HFBS

(25 mm by 50 mm). The low temperature heat stick was used for mounting the sample and to hear and cool. This stick allows placement of the sample in the HFBS standard top loading CCR setup. The low temperature stick has a temperature range of 325K to 4K.

5.2.7 QENS Data Analysis

Data reduction and analysis were performed using DAVE software provided by NIST.206 The analysis followed customary use of delta function, which represents the elastic peak as measured by a vanadium sample during measurement. For the 295K and 270K on DCS only, the sample resolution was modeled as a Lorentzian.

There was a process nearly the same as the resolution width of the instrument. To reduce fitting parameters, this process was incorporated as part of the delta resolution. And as this process is within the standard resolution of the instrument, it cannot be identified with confidence. A flat background was used to account for motions much faster than the instrument capability and for general background noise. Finally a Lorentzian curve was used to fit the broadened quasielastic signal that represents the dynamic motions measured for the sample in both HFBS and DCS data. This method provided fits with a reduced Chi squared of below 2 on average across all data curves. The standard deviation of all fits has been included in the final represented results in the text.

130

5.3 RESULTS AND DISCUSSION

Figure 5-1A shows the hierarchical structure of the supramolecular hydrogels.

These hydrogels are physically crosslinked by FOSA nanodomains. Small angle neutron scattering (SANS) showed that a water depleted DMA shell is formed around the FOSA nanodomains.17 By adjusting the FOSA content of the copolymer from 5 mol% (DF5) to

22 mol% (DF22), the center-to-center spacing between the nanodomains was varied between 7.1 nm and 6.4 nm (inset in Figure 5-1A). From prior SANS measurements, the radius of the FOSA nanodomains is approximately 2 nm with a 1 nm water depleted shell of DMA54 and confirmed by the SANS measurements here with the shell thickness independent of FOSA content in the hydrogel. From geometry, the average hydrated distance between the FOSA nanodomains is 1-2 nm and is dependent on the FOSA content in the hydrogel. These length scales for confinement of water are similar to those where large changes in the freezing of water occur in carbon nanotubes,207 so it appears that the nanodomains act to confine the water in a manner similar to that of hard porous matter. However, the efficacy of these hydrogels in inhibiting freezing of water appears to be intimately tied to the composition of the copolymer. For the DF5 hydrogel (low crosslink density), water freezing was suppressed to 256K for a hydrogel that was swollen to an equilibrium water concentration of ~80 wt% at 295K (DSC thermogram in

Figure 5-1B). This suppression in the freezing point is reminiscent of the influence of some electrolytes commonly observed in pharmaceuticals.121, 208 However, all of the water within the hydrogel (Figure 5-1B) melts at approximately the normal melting point of water. This is in contrast to the influence of electrolytes where both the freezing and melting points are suppressed to a similar degree.127 The lack of reversibility within the

131 hydrogel is suggestive of a change in the local environment for water on freezing, which is not unexpected as freezing of water within hydrogels is known to degrade mechanical properties (directly correlated with the network structure of the hydrogel).126 The normal freezing point provides evidence that the water phase separates from the copolymer on freezing.

132

The exotherm and endotherm in these studies provide a direct route to estimate the fraction of water that has crystallized. Recently, a cubic form of ice with nearly the same normal melting point has been reported for water confined within alumina nanochannels,21 but the crystallization is not confined to nanoscopic dimensions in the hydrogel as will be discussed later. Thus, we assume that the ice within the hydrogel is Ih with the specific enthalpy of fusion of 334 J/g. With this assumption, we calculated that

~95 wt% of the water in the DF5 hydrogel was crystallized on cooling from the enthalpy of the melting endotherm of the hydrogels. This near complete freezing of water is expected for hydrogels.

A more complex freezing behavior is observed when the FOSA content in the hydrogel is increased to 22 mol% (DF22), which decreases the water content to ~ 45 wt

%. The thermogram on cooling exhibited three exotherm peaks at 254K, 244K and 227K

(Figure 5-1C). These transitions suggest three distinct local environments for water with the DF22 hydrogel; these different environments are not clearly observed when freezing water within the DF5 hydrogel despite the chemical similarities. Recent work by Floudas and coworkers demonstrated how slight modulation in the confinement could lead to significant variation in the freezing point of the supercooled water.209-211 However, both hydrogels (DF5 and DF22) exhibit a single peak in the heating endotherm near the normal melting point of ice. These endotherms suggest that the ice in both hydrogels is

21 likely Ih, but cubic ice also melts at a similar temperature. For the DF22 hydrogel and again assuming the enthalpy of Ih, only 55 wt% of the water is frozen, based on the heating endotherms, in comparison to the near complete freezing (~95%) in DF5. One potential explanation is the difference in hydration between the samples as bound water

133 to hydrophilic polymers can inhibit crystallization. However, normalizing mass of amorphous water by that of the dry copolymer in the hydrogel still has more amorphous water for DF22 (0.36 w:w) than for DF5 (0.2). Note that the DF22 copolymer contains less of the hydrophilic DMA, but the fraction of amorphous water increased in comparison to DF5. Thus, the change in the freezing behavior does not appear to be associated with favorable interactions between the water and the hydrophilic component of the copolymer. While the baseline value for the DF5 may indicate the degree of bound water which is prevented from freezing.

Moreover, the partial suppression of ice formation in the DF22 hydrogel does not appear to be a consequence of kinetic factors; using cooling rates that differed by more than 2 times and annealing the sample at low temperature do not significantly alter the exotherm, nor change the fractional freezing as determined from the heating endotherm

(Figure 5-2). We hypothesize that this large fraction of amorphous water in the DF22 hydrogel is due to confinement of water between the FOSA nanodomains. However the supramolecular structure of the hydrogel can rearrange when stressed,212 e.g., from stresses associated with water expansion when it crystallizes, so understanding the nanostructure of these hydrogels at low temperature will provide insight to the mechanical vs. thermodynamic origins of confinement that appears to partially supress ice formation.

134

Figure 5-2. DSC thermograms for DF22 on cooling at 5K/min (green, ‘5K’) and 2K/min

(blue, ‘2K’). The transition temperatures are offset due to the rate dependence, but are qualitatively similar in terms of the peak shapes.

In situ freezing studies using SANS experiments illustrate differences in the nanostructure of the hydrogels and how these nanostructures evolve through the freezing transition(s). Figure 5-3 demonstrates the significant influence of the FOSA content in the hydrogel on the temperature dependence of the scattering profiles. As shown in

Figure 5-3A, the absolute scattering for the DF22 hydrogel swollen with D2O shows that the peak position associated with the correlation of the FOSA domains (analogous to d- spacing) was nearly invariant, thus the nanostructure persists to 205K although the scattering intensity decreased on cooling. However, there is an increase in low q scattering at temperatures below 260K; this is associated with the generation of macroscopic heterogenities in the hydrogel (likely small D2O ice crystals within the hydrogel). For the DF15 and DF5 hydrogels (Figures 5-3B and 5-3C), the correlation peak associated with the FOSA domains was no longer well resolved at temperatures below 260K. This behavior is attributed to the exclusion of the copolymer when the water crystalized. This effectively decreases the concentration of D2O within the DMA phase

135 upon freezing to decrease the spacing between FOSA nanodomains and the scattering contrast associated with the nanodomains and the matrix of DMA/D2O. For these hydrogels, the scattering was essentially unchanged at lower temperatures (<260 K), which is consistent with the DSC data (Figure 5-1B and 5-4), indicating that >90 % of the water crystalllized when the DF5 and DF15 hydrogels were cooled to 260K.

Figure 5-3. Temperature dependent SANS profiles for (A) DF22, (B) DF15, and (C) DF5 hydrogels equilibrated at 295K with D2O for maximum total scattering.

136

Figure 5-4. DSC thermograms of DF15 on both cooling (blue) and heating (red) at 2

K/min. Inset shows zoomed in crystallization peak on cooling.

To explain this evolution in the nanostructure of these hydrogels, the impact of ice formation must be considered. For more conventional covalently crosslinked hydrogels, stresses that develop due to the expansion of water upon ice crystallization break covalent bonds.85 However for the supramolecular hydrogels examined herein, the physical crosslinks can re-arrange to accommodate the water expansion that accompanies ice formation. This re-arrangement will distort the nanodomain structure and decrease the correlation peak in the SANS profile. As the water crystallizes, it phase separates to exclude the copolymer from the ice. This will impact the scattering through a significant decrease in contrast by the loss of D2O between the FOSA nanodomains.

To better understand the structural changes in these hydrogels during cooling, contrast variation SANS was performed using mixtures of H2O and D2O to match the contrast of the DMA or FOSA phase of the hydrogel. SANS experiments using 27/73

(v/v) D2O/H2O mixtures to contrast match the neutron scattering length density (NSLD) of the DMA phase quantify the changes in the FOSA nanostructure (Figure 5-5) induced

137 by freezing water. These scattering profiles were fit with Equation 5-1, the Broad Peak

Model183, 184 (representative fit shown in Figure 5-6):

(5-1) where A and C are scaling factors, B is the background,  is the Lorentzian screening length, n is the Porod exponent and m is the Lorentzian exponent. These fits determine the temperature dependence of the interdomain FOSA spacing, d = 2/Q0, and a Porod term that describes the fractal dimensions of the aggregates responsible for the low Q upturn in scattering. Table 5-1 shows the results of example fit in Figure 5-6.

Figure 5-5. SANS profiles for DF22 hydrogel swollen with 27/73 (v/v) D2O/H2O to contrast match the DMA phase. These data provide the core size, interdomain spacing, and domain clustering of the hydrogel.

138

Table 5-1. Fit parameters from broad peak model of DF22 hydrogel at 260K with the associated uncertainty (σ).

Parameter Fit Result σ Porod Scale (x107) 6.05 0.41 Porod Exponent 3.08 0.01 Lorentzian Scale 2.78 0.01 Lor Screening Length [Å] 70.0 0.4 -1 Q0 [Å ] 0.101 0.0004 Lorentzian Exponent 1.98 0.01 Bgd [cm-1] 0.68 0.001

Figure 5-7 summarizes the evolution of the interdomain spacing (d), on cooling for the DFx hydrogels. For DF5 and DF15, d decreased abruptly between 270-260K and remained almost constant on cooling to 210K. The large decrease in d was attributed to the water phase separating from the copolymer during ice crystallization. In contrast, the gradual decrease in the d-spacing upon cooling the DF22 hydrogel (nearly exponential between 270K and 242K) was a consequence of the suppression of ice crystallization in

139 that system. This evolution in the structure on cooling for DF22 is consistent with the large amorphous water fraction calculated from DSC.

Figure 5-7. Contrast variation provides the interdomain spacing (using 27/73 v/v

D2O/H2O). Error bars represent one standard deviation, if not visible error bar are smaller than the size of the symbol.

Similarly, matching the NSLD of the FOSA with 50/50 (v/v) D2O/H2O mixture enabled direct interogation of the DMA structure (Figure 5-8). These scattering profiles

(representative fit shown in Figure 5-9 with fit parameter result shown in Table 5-2) were fit with a core-shell model with a Schulz distribution of radii and hard sphere interaction.213 This model, Equation 5-2, provides the thickness of the dehydrated shell surrounding the FOSA core and a Porod term that describes the composition fluctuations within the hydrogel,

(5-2) where A is the scale factor for the Porod fit at low Q with exponent n, B is the background scattering, F(q) is the form factor from a core shell structure with a Schulz probability density distribution and S(q) is the structure factor for a core-shell system214 using the Percus-Yevick215 closure for a polydisperse hard sphere structure. The small

140 scattering contribution of the DMA chains between the nanodomains were lumped into the background as this scattering was minimal (Figure 5-10).

Figure 5-8. SANS profiles of DF22 swollen with 50/50 D2O/H2O to contrast match the

FOSA core on cooling from 295K to 210K. Fitting these data yields information on the shell thickness and DMA clustering.

Figure 5-9. Representative fit to the core-shell model with a hard sphere structure factor

141

Table 5-2. Fit parameters from core-shell model of DF22 hydrogel at 260K with the associated uncertainty (σ).

Parameter Fit Result σ Porod scale (x108) 31.4 4.5 Porod exponent 3.07 0.03 Bgd [cm-1] 0.496 0.0008 Volume fraction 0.443 0.0013 Avg core rad (Å) 22.3 0.04 Core polydisp (0,1) 0.213 0.0009 Shell thickness (Å) 10 - SLD core (Å-2 x 10-6) 3.15 - SLD shell (Å-2 x 10-6) 0.899 - SLD solvent (Å-2 x 10-6) 2.28 0.008

Figure 5-10. Comparison of the goodness of the fit of SANS profiles of the DF22

From these fits, the effective volume fraction of FOSA nanodomains (separated by DMA/water) was determined for these DFx hydrogels (Figure 5-11). For DF15, a sharp increase in the effective volume fraction was found on cooling from 270K to 260K.

This behavior was attributed to the deswelling of the DMA phase when ice was formed.

142

In contrast, the nanodomain volume fraction only modestly increased for the DF22 hydrogel between 275K and 240K, and the volume fraction was invariant on cooling further below 240K. That result is consistent with the suppression of ice formation in

DF22 (limited deswelling of DMA phase).

Figure 5-11. Contrast variation provides the effective volume fraction of the supramolecular aggregates separated by hydrated DMA (using 50/50 v/v D2O/H2O).

Error bars represent one standard deviation.

These structural characteristics from SANS provide indirect evidence to crystallization/supercooling of water within these hydrogels that depends on the FOSA content. In order to more directly examine the potential supercooling, quasielastic neutron scattering (QENS) was used to provide a direct measure of the water dynamics in the hydrogels. Figure 5-12 illustrates the QENS spectra at 295K and 260K for the three

DFx hydrogels obtained using the Disk Chopper Spectrometer (DCS) at the NIST Center for Neutron Research (Gaithersburg, MD, USA).216 The DCS has an accessible measurable range of relaxation times, , from 2.1 ps to 75 ps with slower protons contributing to the elastic peak in the spectra. The mobility of the protons is directly related to the width of the peak. Within the measureable range, shorter relaxation time for

143 the proton will lead to more energy being transferred and a broader peak in the spectrum.

In these hydrogels, the protons are primarily from H2O. At 295K, the spectrum for DF5 is broader than that for DF22 (Figure 5-12A), which suggests a decrease in the water mobility in the DF22 hydrogel. This decreased mobility is likely a consequence of the

FOSA-content dependent water fraction in the hydrogels (Figure 5-11E) and the length scale of confinement between FOSA domains (Figure 5-7).

On decreasing the temperature to 260K, the QENS spectra narrowed significantly

(i.e., decreased dynamics of the protons) as was expected as shown in Figure 5-12B. The compositional dependence of the hydrogel (FOSA content) on the resolvable proton dynamics by DCS is inverted at 260K relative to that at 295K. At 260K, the narrowest spectrum is for the DF5 hydrogel (low mobility) and broadest spectrum is for the DF22 hydrogel (higher mobility). This narrowing is likely partially due to an increase in the intensity of the elastic peak as the dynamics of ice are below the resolution of the instrument as has been reported previously.217

Figure 5-12. QENS measurements of proton dynamics at Q=1.25 Å-1 for DCS at (A)

295K and (B) 260K for DF22 (red), DF15 (blue), and DF5 (green). All hydrogels were equilibrated at 295K with H2O prior to measurements. Error bars represent one standard deviation.

144

In order to quantify the diffusive motions within these hydrogels, the QENS data were fit by a single Lorentzian that is convoluted with the energy-independent background and an elastic delta function that is smeared by the instrumental resolution as shown in Figure 5-13. Figure 5-13A shows the fit of the DCS spectrum for the DF22 hydrogel at 295K. The Lorentzian describes the broadening of the spectrum and provides a more quantitative measure of the proton dynamics that are within the energy resolution of the DCS (0.055 meV) through the width of the Lorentzian. Comparing this peak to the

Lorentzian for the DF22 hydrogel at 220K (Figure 5-13B) shows that the peak slightly narrows on cooling to 220K. This result demonstrates that motions of some protons in the

DF22 hydrogel remained within the energy window for the DCS QENS measurements at

220K. These relatively fast motions at 220K within this hydrogel are noteworthy as protons motions in supercooled water confined in mesoporous silica are not resolvable at temperatures below 250K.112 This difference suggests that the amorphous water within the DF22 hydrogel is significantly more mobile than supercooled water confined within nanopores of inorganic materials.

145

To enable comparison of the water dynamics in these hydrogels to other systems, the apparent self-diffusion constant of water was calculated from the Q dependent

FWHM (2Γ) obtained from the Lorentzian fits of the QENS data. At 270K, 2Γ increased linearly with Q2 at high Q with a plateau at low Q as shown in Figure 5-14B. The plateau is indicative of water confinement and modeled as a caged Fickian process,218

(5-3)

where 2Γ is FWHM, Dw is Fickian diffusion constant, and C is a constant. A cage radius218 can then be calculated,

(5-4) where 2Γ0 is the plateau FWHM and A is the cage radius. The prefactor constant results from the intermediate scattering function Fourier integral approximation of the elastic

146 line width in the low Q regime (associated with the plateau in DQ2).103 Experimentally, this linear relationship between 2Γ and Q2 was found for all temperatures examined for the DF22 hydrogel as shown in Figure 5-14. There is a reversal in the gel with the fastest motion as seen in Figure 5-14 from 295 to 260K. Where initially the DF5 has the fastest motions, largest FWHM, at 295K (Figure 5-14A), but on cooling to 260K (Figure 5-14C) this is reversed with DF22 having the fastest motions. Comparing the Q-dependent

FWHM trends in just the DF22 sample in Figure 5-14D, we see the motions were weakly dependent on temperature below 260K.

These motions can be quantified in terms of an effective diffusivity (based on Eqn

5-3) for the three hydrogel systems examined. Figure 5-15 illustrates that the water diffusivity within DF5 and DF15 decreased rapidly on cooling below 270K. For these two hydrogels, the motions became too slow to resolve at temperatures below 260K. We attribute this behavior to water crystallization. For DF15 and DF5 hydrogels, a small compositional dependence can be resolved with the diffusivity slightly greater in DF15 than DF5 at 260K (see Figure 5-15).

147

Figure 5-14. Analysis of DCS data provides insight into the water dynamics. FWHM at

295K (A), 270K (B), and 260K (C) for H2O swollen hydrogels of () DF5, () DF15 and () DF22 illustrates water motion is locally caged Fickian diffusion due to invariant

H2O is calculated from the linear fit of the FWHM (solid lines). The difference in the diffusivity of H2O within the hydrogels attributed to the variation in local hydration.

Error bars represent one standard deviation.

In contrast to DF5 and DF15 hydrogels, the self-diffusion coefficient for the mobile water within the DF22 hydrogel was only modestly impacted on cooling from

295K to 220K. Over this entire temperature range, the effective diffusivity decreases by less than a factor of 2. Even at 270K, the self-diffusion coefficient is greater for the DF22

148 hydrogel than the other hydrogels examined. When comparing to neat supercooled water

(in 300 m capillary tubes) in Figure 5-15, the self-diffusion coefficient for the water in the DF22 hydrogel near the normal melting point the diffusivity of water within this hydrogel is nearly an order of magnitude less than found for supercooled water.217, 219 At approximately 240K, the This behavior suggests that either the DMA segments, despite the low intrinsic mobility of the polymer (Figure 5-15), or confinement between glassy

FOSA domains leads to enhanced dynamics in H2O at these highly supercooled temperatures. This is believe to be the result of the water remaining in an amorphous state and not being locked into a crystal or tightly bound to the polymer chain.

220K in the DF22 hydrogel. Error bars in all figures represent one standard deviation.

Additionally, the compositional dependence of the freezing behavior of water within these DFx hydrogels suggests that the FOSA nanodomains act to confine water.

Estimates of the effective confinement length scale of the water were made from the

149 elastic incoherent structure factor (EISF) obtained from the QENS data. The water in the

DFx hydrogels was assumed to be confined between neighboring FOSA nanodomains, and the EISF was fit by a simple diffusion equation for water confined between two parallel surfaces220 (see Figure 5-16),

(5-5)

(5-6)

where n(T) is the fraction of protons participating in the measureable diffusive

221 motions, A0 describes the cage size relative to the EISF ratio, L is the length of confinement and j0 is a Bessel function of the first kind and zero order. The confinement length scale, L, decreased on cooling and the changes in L were greatest for DF5 and least for DF22, which are consistent with the results of the SANS analyses for the changes in the nanostructure of these hydrogels on cooling.

Figure 5-16. Representative fits of the EISF data from DCS measurements for the DF22 hydrogel at 295K and 270K.

150

For DF22, the calculated cage size at 270K and 295K was 3.2 Å and 4.2 Å, respectively. However, these cage sizes are significantly smaller than the average hydrated interstitial space between the hydrophobic FOSA domains calculated from

SANS (~10-30 Å at 270K).118, 133 Cage sizes calculated from EISF and FWHM assuming

Fickian diffusion are shown in Figure 5-17.

In addition to understanding the caging effort of the FOSA nanodomains, the

EISF also provides quantification of the fraction of protons in the hydrogel with motions in the measureable dynamic window of the spectrometer (n(T)). Figure 5-18 shows how n(T) changed on cooling. At 295K (above the normal melting point of ice), most of the protons in the system were measureable by the QENS experiments, but n(T) decreased with increasing FOSA content as the H2O decreased and the protons in the copolymer exhibit slow motions, even in the hydrated state. Protons with mobility slower than the resolution energy scatter elastically and act to decrease n(T). As the temperature decreased, n(T) decreased sharply at 260K for the DF5 and DF15 hydrogels due to

151 freezing of a significant fraction of the water, which suppressed the water motions to below the resolution of DCS. At 240K, only about 5% of the protons in DF5 were still mobile in DF5 and DF15, which is similar to the fraction of unfrozen water (5-6%) estimated from the enthalpy of fusion data in Figure 5-1B.

The behavior of water within DF22 was markedly different in that the decrease in n(T) with decreasing temperature is less pronounced (Figure 5-18). At 220K, n(T) was

~15%, compared with ~55% at 295K. Note that the dynamic window examined here was significantly faster than typically used to examine the dynamics of water in confinement

112, 222 at low temperatures. For slower motions, the HFBS measurements using H2O and

D2O swollen hydrogels (Figure 5-19) indicated coupling of the dynamics of the copolymer and water at low temperature. In this case, 15% of the protons in H2O were active in the energy window associated with HFBS for the DF22 hydrogel (Figure 5-20); these motions are in addition to the fast motions probed by DCS (Figure 6D). The combination of the QENS and HFBS data indicated that a significant fraction of the

152 protons within DF22 remains mobile at 220K, which is consistent with the high fraction of unfrozen water elucidated from DSC measurements.

Figure 5-20. The (A) cage radius and (B) fraction protons with resolvable motions by

HFBS are obtained from EISF fits of the HFBS data of DF22 swollen in H2O (blue squares) and D2O (grey circles). Error bars represent one standard deviation.

In order to The QENS measurements with DCS provided information about only relatively fast motions (< 75 ps) in the hydrogels. A lower energy instrument, the High

Flux Backscattering Spectrometer (HFBS) at the NIST Center for Neutron Research

(Gaithersburg, MD, USA)223, with an energy window of ± 0.00085 meV (HWHM) at Q

153

=1.01Å-1 corresponding to 0.14 ns <  < 3.8 ns, was used to probe the slower dynamics of the copolymer and the water in the hydrogels at low temperature. Figure 5-21 shows the mean square displacement (MSD) calculated from measuring the elastic intensity change as a function of temperature within the fixed energy window (± 0.00085 meV) for DF5 and DF22 hydrogels. In this case, the larger the MSD, the faster are the dynamics. For the non-hydrated DF5 copolymer, the MSD is less than 1 Å2 over the entire temperature range examined, which is consistent with motions in a glassy polymer (Figure 5-21).224

DFx hydrogels swollen with D2O enable the copolymer dynamics to be probed as protons on the copolymer dominate the QENS signal. The polymer MSD in Figure 5-21 at T >

266K for the D2O swollen DF5 was greater than that for D2O swollen DF22. This behavior is expected as the higher water content allows more bulk water to exist, thereby increasing the mobility of the solvated polymer chain segments. D2O plasticizes the system and the dynamics of the copolymers are intimately tied to their hydration and local environment. However upon cooling the DF5 hydrogel to 260K, the MSD decreased to nearly the same value as that of the dry glassy copolymer. This infers that the D2O is no longer plasticizing the copolymer and is consistent with ice formation that extracts the liquid D2O from the copolymer. The sharp decrease in MSD at ~265K for the

DF5 hydrogel in Figure 5-21 corresponds well with the freezing peak of water within the

DF5 hydrogel observed by DSC. Consistent with the inhibition of ice formation, the

MSD for the DF22 hydrogel did not exhibit the same abrupt decrease with the MSD for the copolymer in the DF22 hydrogel greater than that for the DF5 hydrogel. The MSD of the D2O swollen DF22 copolymer is continues to be greater than the dry copolymer for temperatures greater down to 220K.

154

Figure 5-21. Slower dynamics in the system were probed using HFBS with the Mean

Squared Displacement for DF22 and DF5 swollen in D2O, H2O, or dry (no water) during cooling at 2 K/min. Error bars represent one standard deviation.

To further understand the dynamics in the DF22 hydrogel, the H2O dynamics (in the same energy window) in the DF22 hydrogel were greater than that of the DMA segments of the copolymer for T > 235K. Upon further cooling to T < 230K, the water and the DMA dynamics at the timescales probed by HFBS were strongly coupled. These characteristics are consistent with the persistence of water capable of plasticizing the

DMA segments in the DF22 hydrogel down to very low temperatures.

These fast dynamics in the DF22 hydrogel are reversible in the supercooled state.

Figure 5-22A shows that the dynamics of the hydrogels were the same at 250K for DF22 when cooled from 295K to 220K and when heated from 220K to 250K. The reversibility of the water dynamics at 250K upon cooling and heating indicates the thermodynamic origin of the inhibition of ice formation described herein. The decrease in the water dynamics upon cooling DF22 from 250K to 220K (see the data in Figure 5-18) is not due to additional ice formation. Additionally, the EISF at 250K was indistinguishable between the cooling and heating experiments. Figure 5-22B demonstrates the

155 reversibility of the dynamics of liquid water upon reheating the DF22 hydrogel from the

“frozen state”. Different temperatures (270K for cooling and 275K for heating, Figure 5-

22B) were used for the comparison to avoid artifacts from the hysteresis on melting

(Figure 1C). The nanostructure appears to impact the H2O dynamics with higher self- diffusion coefficients for water in the lower FOSA content hydrogels (Figure 5-18), so this reversibility in dynamics on melting suggests that the water is in a similar environment.

Figure 5-22. The reversibility of both the dynamics and mechanical properties are shown

To further demonstrate the reversibility, the dynamic mechanical (DM) properties, using oscillatory shear, of DF22 hydrogels (Figure 5-23A) show that the impact of a freeze-thaw cycle on the structure of the viscoelastic behavior, and presumably the microstructure, of the hydrogel was small. The dynamic (G’) and loss

(G”) shear moduli of the pristine DF22 sample decrease with decreasing frequency due to the viscoelastic nature of the supramolecular hydrogel.212 The hydrogel was frozen and

156 annealed in liquid N2 for 60 min. The DM properties were then re-measured after heating to room temperature. The viscoelastic nature of the hydrogel was retained, though G’ and

G” decreased by < 30%. The catastrophic fracture behavior usually observed during freezing of a conventional, covalently crosslinked hydrogel was prevented by the ability of the DF22 network to rearrange and relax under the stress from ice formation and the low fraction of water that freezes in DF22. Visually the DF22 sample progressed from transparent to translucent and retained its shape with no crack formation after undergoing the freeze-thaw cycle (Figure 5-23B). In contrast, the DF5 hydrogel became opaque

(snow white) and brittle during a freeze-thaw experiment (Figure 5-23B). Both behaviors are consistent with the generation of large scale heterogeneities from ice crystallization indicated by the low Q scattering from SANS of the DF5 hydrogel at low temperature.

157

5.4 CONCLUSIONS

A supramolecular hydrophobic domain containing DMA-FOSA copolymer hydrogel with high FOSA content was shown to inhibit significant ice formation, through the confinement of water between hydrophobic nanodomains. At low FOSA content, the hydrogel microstructure was significantly altered as a consequence of the copolymer being excluded from the water crystals on freezing. However, at a sufficiently high

FOSA concentration (22 mol%), the nanodomain microstructure persisted with only 55% of the water freezing upon cooling the hydrogel to 205K. Dynamics representative of liquid-like water persisted in the DF22 hydrogel with less than a factor of two decrease in the effective self-diffusion coefficient of water on cooling from 295K to 220K.

Confinement of water between ~6 nm diameter hydrophobic supramolecular aggregates with <3 nm hydrated interstitial separation inhibits ice formation in this hydrogel. Ice suppression in a soft hydrogel is demonstrated using confinement by hydrophobic aggregates and may have biomedical applications, where freeze-thaw cycles may occur, as well as other applications where ice prevention is desirable, e.g., de-icing surfaces.

158

CHAPTER VI.

VISCOELASTIC CHARACTERIZATION OF SOFT HYDROGEL THIN FILMS

USING QUARTZ CRYSTAL MICROBALANCE (QCM-D): FROM THE

SAUERBREY LIMIT TO BEYOND FILM RESONANCE

6 .1 INTRODUCTION

For nearly 70 years, the piezoelectric oscillation of quartz has been used as an ultrasensitive mass sensor through the direct relationship between the frequency of the quartz oscillation and the mass of material adhered to the quartz in the elastic limit through the Sauerbrey expression.225 This relationship has enabled the quartz crystal microbalance (QCM) to be a mainstay of vacuum science.226, 227 The applicability of

QCM to many scientific questions was initially limited by the belief that the dissipative loading in a liquid environment would dampen the sensor oscillation to prevent accurate measurement, but in the 1980’s Kanazawa and coworkers demonstrated operation in liquids was possible157 and this opened opportunities for QCM to contribute to many electrochemical228 and biological229 investigations. Understanding the operation of the

QCM in liquid environments began with the development of simplified models for behavior in a viscous230 or viscoelastic mediums,231 but the oscillation frequency only provides a partial description of the impact of fluids on the QCM operation and understanding the quality (Q) factor for the quartz sensor is necessary for quantitative interpretation. Q is obtained from the full width at half maximum (Γ) of the resonance

159 peak,230 which is proportional to the energy transfer from the resonator to the system.

Operating the QCM in liquids232 leads to a decrease in the energy efficiency of the quartz resonator (Q factor) due to viscous energy dissipation. Thus, the Q factor allows rheological properties to be elucidated using QCM,161, 233 by measuring the width of the resonance oscillation frequency.234

The commercialization of a novel route developed by Kasemo and coworkers to obtain rheological properties from QCM measurements through intermittent driving potential and measuring the decay of the amplitude of the oscillation for multiple overtones (OT) or harmonics of the fundamental frequency, F,156 has lead to the proliferation of the rheological characterization of films. The QCM-D (Quartz Crystal

Microbalance with Dissipation operation mode) uses the dissipation factor (Ď), the time decay of the resonance amplitude, to assess the energy dissipation instead of Γ of the resonance peak typical of conductance measurements. The viscoelastic nature of many biological235, 236 or polymeric171, 231, 237 materials at surfaces has been determined with the

QCM-D, but these results are tied to the selection of the rheological model to describe the viscoelasticity.57, 169 The model is commonly a Voigt or Maxwell viscoelastic model, despite knowledge that these may not adequately describe the rheological behavior of complex fluids.238 This model dependence for the interpretation and analysis238 of the wealth of information obtained from QCM-D represents a serious challenge. For sufficiently rigid and/or thin adhered layers, the frequency will be directly related to mass through the Sauerbrey expression,12, 170, 239 such that details of the viscoelasticity of the material is not relevant to describe the QCM operation. The point for quantitative failure of the Sauerbrey expression can be described in terms of the ratio of the shear wavelength

160 in the medium (s) and the thickness (D) of the adhered viscoelastic layer and has been

60 predicted to occur at D/s = 0.041. This approach has been extended to examine highly lossy films near film resonance170 to predict an increase in frequency with increasing mass, counter to the even qualitative directionality of the Sauerbrey expression, but consistent with multiple models for QCM operation.57, 60, 240, 241 The film resonance behavior has been experimentally reported for thick films (> 0.5 μm) and thick highly swollen grafted polymer brushes.242-244 Film resonance generally is thought to require quite thick coatings; Shull and co-workers reported an increasing frequency reminiscent of film resonance for thin hydrogel films, but dismissed the result as anomalous behavior on account of limited thickness.245 With the combination of QCM-D and surface plasmon resonance, film resonance was reported for multiple layers of vesicles with a thickness of just 80 nm.56 As the vesicles did not fuse in this case, complexity of the layer response due to interstitial water during the buildup of vesicle layers may have contributed to the film resonance in this scenario.56 This later experiment demonstrates the power of using multiple measurements to assist in the interpretation of QCM data as interpretation of data near film resonance conditions can be challenging.

In this work, we illustrate the impact of the selection of the viscoelastic model on the interpretation of QCM-D data for a thermally responsive hydrogel. Simultaneous

QCM-D and spectroscopic ellipsometry (SE) measurements170 provides a robust route for independently determining D to assess the accuracy of the fits of the QCM-D data by the viscoelastic models: a simple Voigt model and a more rigorous rheological description developed by Shull and coworkers.58, 59 These results illustrate that film resonance can occur in thin coatings (< 200 nm), but both models examined can accurately describe the

161 thickness in this regime. Unexpectedly, the viscoelastic properties from the simple Voigt are in good agreement with the viscoelastic properties from the more rigorous rheological model, except when there is limited energy loss from the film (D/s <0.025). This behavior suggests this value of D/s <0.025 provides the lower limit to accurately extract viscoelastic data from QCM.

6.2 EXEPERIMENTAL

6.2.1 Materials and sample preparation

A random copolymer of N-isopropylacrylamide (NIPAAm) and 2-(N- ethylperfluorooctane sulfonamido)ethyl acrylate (FOSA) with 5 mol % FOSA as reported previously69, 70 was used as the polymer for the hydrogel as hydrophobic associations of

FOSA produce nanodomain FOSA aggregates that provide physical interchain crosslinks when hydrated.70 These hydrogels exhibit volume phase transition of an LCST-type with a midpoint of 17 °C in the bulk70 and 20 °C in thin films.246 Changing the temperature from 35 °C to 5 °C increased the swelling ratio (mass swollen polymer/mass dry polymer) of the films from 1.3 and 5. This large change in the water content of the hydrogel impacted its rheological properties.

The dried copolymer was dissolved in dioxane at 1.0 and1.75 wt% for spin coating on the silica-coated quartz sensors (QSX-335, Q-Sense). These sensors had a diameter of 14 mm, thickness of 0.37 mm, and nominal fundamental resonant F of 5

MHz. The sensors were cleaned by sonication in toluene, isopropanol, and DI water.

Immediately prior to film casting, the sensors were further cleaned by ultraviolet-ozone

(UVO CLEANER®, Model 42, Jelight Company Inc.) for 90 s. The copolymer-dioxane

162 solutions were spun coat at 2500 rpm for 30 s onto the quartz sensors. After 1 h at ambient conditions, the films on the sensors were annealed for 18 h at 150 °C and 27 torr.

6.2.2 Characterization:

The hydrogel films on the quartz sensor were characterized using QCM-D (Q-

Sense E1) and SE (J.A. Woollam, M-2000UI) at the same time using the ellipsometry module (Q-Sense, Biolin Scientific). As fitting of the QCM and SE data requires some knowledge about the sensor, the uncoated sensor characteristics in both air and water were determined with QCM and SE. The thickness of the dry copolymer was determined using F relative to the uncoated sensor in air with an assumed density of 1000 kg/m3.

Simultaneous measurement with SE provided an optical route to also determine the copolymer film thickness using an optical stack for the sensors consisting of gold

(optically opaque), titanium, TiO2, and silica from bottom to top. The series of measurements from the bare sensor to the swollen hydrogel film and models used to fit the SE data are shown in Tables 6-1 to 6-3 and Figures 6-1 to 6-3. The ellipsometric angles of the sensors were fit using the protocol described previously for these QCM-D sensors.154 These thicknesses were then fixed to fit for the thickness of the copolymer or hydrogel that was coated on the sensor with the optical properties of the copolymer and hydrogel both well described by the Cauchy model. The thickness determined from the

Sauerbrey expression based on the frequency shift from the blank crystal agreed within

30% of the thickness obtained from ellipsometry for the dry copolymer film.

163

Figure 6-1. The raw Psi and Delta for the bare sensor with fits from the model in Table 6-

1 shown as black dashed lines.

164

Table 6-2. Model used for fitting of coated QCM sensor. Layer 2, Layer 1, and

Substrate are defined as fit from the previous bare sensor fit for the identical sensor. The polymer layer is modeled as a Cauchy layer (Layer 3). The layer only requires thickness and A and B fit parameters to obtain a good fit of the raw psi and delta measured.

Figure 6-2. The raw Psi and Delta for the coated sensor with dry NF5 film with fits from the model in Table 6-2 shown as black dashed lines.

165

Table 6-3. Model used for fitting of coated QCM sensor in water in the liquid cell.

Layer 2, Layer 1, and Substrate are defined as fit from the previous bare sensor fit for the identical sensor. The polymer layer is modeled as a Cauchy layer (Layer 3) that is now swollen in water. The layer only requires thickness and A and B fit parameters to obtain a good fit of the raw psi and delta measured. Additional Ambient index now being water must be accounted for. Also, window corrections are provided by Delta offset, Window

#1. These are found by previous measurements of known substrates.

Figure 6-3. The raw Psi and Delta for the coated sensor with swollen NF5 film in liquid cell at 65˚ with fits from the model in Table 6-3 shown as black dashed lines.

166

For the swelling measurements, the coated sensor was first exposed to liquid water at 25 °C and allowed to equilibrate. The temperature was increased to 35 °C and allowed to equilibrate for 1 h. Then the temperature was decreased in 1-2 °C increments each hour until 5 °C was reached, with a total of 24 temperature steps. The frequency of the bare sensor in water was used as the zero reference for F. The effect of temperature on the resonator response in water was corrected to enable the direct calculation of the frequency and dissipation shifts associated with the hydration of the swollen hydrogel film as previously described,246 such that the changes in F and dissipation only contained the contribution from the adhered hydrogel film. A step by step example of the measurements taken with their respective changes to QCM resonance frequency and dissipation are shown in Chapter 3 along with the temperature corrections to illustrate the influence of correcting for temperature variation on the data.

The QCM-D data were fit using two different protocols that relate the viscoelastic properties of the film to the oscillator response. The Voigt Extended Viscoelastic model was used in conjunction with the standard Q-Tools software (Q-Sense), which is based on the model refined by Voinova and coworkers,57 which included frequency dependence to the viscoelastic properties.247 The frequency and dissipation changes were recursively fit to this model to minimize the mean square error between the prediction and measured values. Flow charts of these fit procedures are shown in Figures 6-4 and 6-5. The same input parameters/assumptions were used with a different viscoelastic model that is based on a self-consistent description of the complex modulus and the shear impedance as developed by Shull and coworkers.58, 59 The fits from both models can be used to

167 determine the shear modulus, viscosity and thickness of the hydrogel. Additional details about these two models are further explained in the subsequent Theory section.

Figure 6-4. Flow chart showing inputs and outputs for fitting of the QCM data with the viscoelastic (Voigt) model.

Figure 6-5. Flow chart illustrating the inputs and outputs associated with fitting of the

QCM data when applying the rheological model from Shull and coworkers.

168

6.3 THEORY AND MODELING OF QCM OPERATION

In order to understand the challenges with analyzing the behavior of a viscoelastically-loaded quartz crystal, the operation of the quartz crystal in an applied oscillating potential can be examined in terms of the shear wavelength, λs, which describes the propagation of the wave generated by the resonant frequency of the sensor through the medium near the sensor. As the shear wave propagates through the system, its wavelength is dependent upon the viscoelasticity and density of the local media. Here, the hydrogel is attached to the quartz crystal in water to ensure equilibrium hydration.

Both the hydrogel and the liquid environment typically impact λs. The description of the wave propagation can be expressed in 3 limiting cases as shown in Figure 6-6. If the hydrogel is sufficiently thick, the wave decays completely within the adhered layer of the hydrogel (Figure 6-6A). This case can be considered equivalent to bulk measurement248 of the hydrogel where the response does not depend upon the fluid in contact with the hydrogel. For thinner hydrogel films, the shear wave propagates into the bulk aqueous phase and thus both the hydrogel and the aqueous phase contribute to the response of the crystal. Two scenarios can arise under these conditions. In most cases at the hydrogel- water interface, a shear wave encountering this interface results in the wave energy being split between a transmitted wave and a reflected wave where the reflected wave impacts on the outbound shear wave (Figure 6-6B). The reflected wave generally decreases the oscillation frequency at the sensor.60 However, it is possible for the reflected shear wave to couple with the outbound shear wave to effectively increase the oscillation frequency at the sensor surface from the addition of these two transverse waves (Figure 6-6C). The

169 condition under which the shear wave is coupled to the reflected wave is known as film resonance and is a result of approaching the quarter wave condition.

Figure 6-6. Illustration of the propagation of the shear wave generated from the AC driven quartz-sensor surface into a sample consisting of a hydrogel film immersed in water. For sufficiently thick films, (A) the wave decays through the hydrogel without encountering an interface for reflection, which probes the “bulk” properties. For thinner films, the hydrogel-water interface is encountered by the shear wave, which can result in

The full decay of the shear wave within a single component yields information about the viscoelastic properties,249 but not about the film thickness. The bulk analysis of a viscoelastic material (thick film) has been examined previously157 and is not the focus of this work. A more common scenario is when the shear wave propagates through the adhered layer and then fully decays in the bulk solvent,250, 251 which will be the focus of this work. To quantify the oscillatory shear wavelength in the medium and the attenuation of the shear wave, we will follow the nomenclature from the analysis of White and

170

Schrag,25 where β and α, respectively, are used to relate the medium properties and resonant frequency of the sensor to the shear wave through equations (6-1) and (6-2):

, (6-1)

, (6-2) where ω is the sensor resonant frequency in radians, ρ is the density of the medium, ηm is the medium viscosity, and ϕ is the phase angle of the medium (0 = purely viscous, /2 = purely elastic). For this paper, we will describe the dimensionless quantity β in the simpler form as 1/λs:

(6-3) and also convert the liquids phase angle used previously to the more standard solids has angle ( related by ϕsolids = /2- ϕliquids). Where the phase angle ϕ is now described as ( /2

= purely viscous, 0 = purely elastic). From these equations, the behavior of a shear wave in an arbitrary medium can be described in terms of a dimensionless quantity, D/λs, where D is the thickness of the media normal to the wave. For the hydrogels, the bulk fluid is always water and thus the value for the shear wave λs, is that of the attached layer

(closest to the crystal).

60 Figure 6-7 illustrates how the real and imaginary parts of the impedance as D/λs increases. The real part of the impedance is directly related to the energy dissipation in the measured film, while the imaginary part of the impedance is inversely related to the frequency response of the film. At low D/λs, the energy losses induced by the attached film are sufficiently small that the system is essentially elastic and can be described by the Sauerbrey expression. The energy damping becomes sufficiently large at D/λs = 0.041

171 that the error in thickness associated with the Sauerbrey expression is observable

60 (>5%), which has been confirmed experimentally to occur between 0.025 < D/λs <

170 0.057. For larger D/λs, the dissipation or  (or multiple overtone frequencies) is necessary to accurately determine the layer thickness from QCM, but these calculations include the viscoelastic properties of the film, typically using a constitutive equation. In this viscoelastic regime, the dissipation increases and the frequency decreases as the film becomes thicker or more lossy. This viscoelastic regime is commonly experimentally reported using QCM-D.12, 97, 106, 154, 161, 169, 226, 233, 239, 252 However, the sensitivity of the

QCM response to the quantitative values for the viscosity or phase angle is not commonly discussed, but as the Sauerbrey limit is approached the sensitivity decreases and can lead to large errors in extracting viscoelastic properties as will be discussed later.

S S Figure 6-7. Impact of D/λs on the real, RM , (green lines) and imaginary, XM , (blue lines) parts of impedance for the fundamental frequency (5 MHz). Four regimes of operation:

172

At much larger D/λs as shown in Figure 6-7, the behavior of the QCM depends on the phase angle of the film. For a nearly perfect elastic film, the real part of the

S S impedance (RM ) goes through a maximum at approximately D/λs = 0.25 and RM decrease as the thickness increases or viscosity decreases from this point. This maximum is the result of the efficient coupling of the reflected wave with the incoming wave from the driving quartz surface. This is the film resonance condition. Prior to the maximum in

S S RM , the imaginary component of impedance, XM , which is directly related to -frequency, decreases with increasing D/λs (thicker films lead higher frequency at fixed viscoelastic

S properties). This decrease in XM , which is counter to the Sauerbrey expression, occurs at

60 D/λs ~0.11 for a fully viscous film and at higher D/λs with some elastic component.

Thus at D/λs <0.11, the frequency must decrease as mass is added (for fixed viscoelastic properties), but at larger D/λs (whose exact value depends on the phase angle), the frequency can increase when mass increases. At D/λs greater than the film resonance condition, the complex impedance is dominated by the intrinsic properties of the film and bulk properties of the film can be extracted from the QCM data. This shear wave based approach directly provides rheological properties through the phase angle (δ), where tan(δ /2)= α/β.60 The calculation of the complex impedance was calculated using the published shear wave propagation model60 with some minor corrections as noted in the

Mathematica code (see Appendix 1).

Outside of the Sauerbrey limit, the complex impedance is dependent upon the viscoelastic properties of the film and thus an appropriate description of these properties is necessary. From a perspective of simplicity, the Voigt model provides an expression that can describe a viscoelastic medium. The Voigt model has been used for the

173 interpretation of QCM data since the initial work of Johannsmann and coworkers233 and has been popularized by its inclusion with the software from the Q-Sense through the descriptions reported by Voinova et al.57 Here, we denote the Voigt model as V and properties determined from the QCM data with this model are described by the model and the two overtones used in each fit with the notation, ##V. For example, fits using the

3rd and 5th overtone associated with 15 MHz and 25 MHz nominal frequencies would be listed as 35V. However, the Voigt model, similar to the Maxwell model, does not generally describe most complex fluids when using conventional rheometers. As such,

Shull and coworkers58, 59 developed a solution to the QCM wave propagation equation based on the Kramers-Kronig relation to describe the sensor response from a rheological perspective when the dampening from the film to the sensor response is quite large. This model is self consistent and thus avoids some of the issues associated with the Voigt model as long as the dampening is sufficiently large, which occurs in the thick film limit.

Similar to the Voigt model, we will use R to denote this model, which requires two overtones to calculate the solution. The frequency response of both overtones and the dissipation for one overtone are used to solve the harmonic ratio, rh, and dissipation ratio,

59 rd. To clarify the solution set used, the sequence of overtones used will be denoted as

###R, where the first two numbers are the overtone frequencies and the last number is the overtone associated with the dissipation. For example, 353R would use the frequency of the 3rd and 5th overtone and the dissipation of the 3rd overtone. One notable difference between the two models is their limiting behavior. For the V model, a phase angle associated with an elastic film (ϕ ≈ 0) is obtained in the low dissipation limit, which corresponds to Sauerbrey behavior. In the low dissipation regime, the R model solution

174 fails to the viscous limit (ϕ ≈ /2). This difference in the limiting behavior from the models significantly impacts the viscoelastic properties predicted in the low dissipation regime. Because of this difference in phase angle for insensitivity to dissipation, the region where effective viscoelastic modeling can be found and corresponds quite well with previous work. The implications of these differences will be further discussed later.

6.4. RESULTS AND DISCUSSION

One challenge with the experimentally exploring the predictions associated with the shear wavelength is the wide parameter space in D/λs that requires widely varying thickness and/or rheological properties, which is difficult to obtain without a multitude of samples with varying thickness. Here a thermoresponsive hydrogel based on a random amphiphilic copolymer enables large changes in the swelling with temperature. The swelling impacts both the viscosity and the thickness to increase D/λs as temperature decreases. The swelling ratio is varied from 1.15 at 35 °C to 4.2 at 5 °C. Figure 6-8 illustrates the QCM-D data for two films with initial (dry) thicknesses of 52 nm and 100 nm as the temperature is decreased from 35 °C to 5 °C. The frequency shifts reported are relative to the uncoated sensor in water at the temperature of interest. This corrects for the temperature dependent water density and viscosity. The lines in Figure 6-8 illustrate the fits of these data to the V and R models with different sets of harmonics.

As shown in Figure 6-8A, the fit of the data to the V model is good for all harmonics examined. It is noteworthy that the slight upturn in frequency for the 9th overtone on cooling below 15°C can be captured using this simple viscoelastic model. As this model has been extensively used for fitting QCM-D data from thin films,12, 53, 97, 169,

175

231, 248, 253 the overall goodness of the fit is not unsurprising. Conversely, the fit from R model describes the data at high temperatures, but there is significant deviation at temperatures less than about 20 °C for all overtones examined. At high temperatures, the hydrogel material is relatively rigid as evidenced by the low dissipation (Figure 6-8B) and thus the good fit is expected as the model reduces to the Sauerbrey expression at zero dissipation. The quantitative failure of the fit at low temperatures with the R model may be a result of the assumptions associated with this model, which was developed for thick films with significant energy dissipation. However, the model qualitatively describes the frequency dependences of the overtones. For the 3rd overtone, the frequency is predicted to be monotonically decreasing as the temperature decreases, while the frequency is predicted to increase at low temperatures for the 9th overtone; these are qualitatively consistent with the measured data. The overtone dependent directional change in the frequency as the film swells is consistent with approaching the resonance conditions as

1/λs ~√, where  =2F. This infers that D/λs is approximately 1.7 times greater for the

9th than the 3rd overtone (neglecting any frequency dependence of the rheological properties), so the increase in frequency on swelling associated with the near resonance regime (Figure 6-7) should be observed first for the highest overtone. Figure 6-8B illustrates that the dissipation monotonically increases with decreasing temperature as would be expected as the hydrogel films swell. The fit of the dissipation using the V and

R model are nearly quantitative with more disagreement between prediction and measured dissipation with R model.

176

Figure 6-8. Temperature corrected QCM-D frequency and dissipation behavior for 52 nm

(A and B) and 100 nm (C and D) thick dry films for the hydrogel immersed in water

Increasing the initial thickness of the copolymer film to 100 nm accentuates the increasing frequency on decreasing temperature (film swelling) for the higher overtones

(Figure 6-8C). As expected for near film resonance conditions, the temperature where the frequency begins to increase on cooling occurs at progressively higher temperatures for higher overtones. This corresponds to a less swollen film at the higher overtones (less lossy and lower thickness. The V model describes the frequency changes well for the 3, 5, and 7 overtones. For the 9th overtone, the high energy dissipation decreased the intensity

177 of the overtone such that the uncertainty in the frequency was large. An example illustrating the change in the resonance peak at different overtones is shown in Figure S6 for hydrogel films at 25 °C (low dissipation) and 5 °C (high dissipation). Therefore, we have not included the 9th overtone in these fits. The quality of the fits with the R model are still inferior to that with the V model, but the agreement with the data is improved in comparison to the thinner film for the R model. The R model in general predicts the upturn in the frequency to occur at a slightly higher temperature (thinner film) than is experimentally observed. The improvement in the goodness of the fit with the R model as the thickness increases is consistent with the assumptions associated with this model drawn from the thick film limit. As the film thickness increases, the appropriateness of this R model to describe the viscoelastic nature of films from QCM improves. As shown in Figure 6-8D, both models can quantitatively describe the dissipation. However, the fit to the data only shows the models ability to find a solution to describe the frequency and dissipation, but do not ensure that the parameters of interest used in these models are correct.

To assess the results obtained from these two models, the film thickness from fits of the QCM-D data is directly compared with the thickness from SE obtained simultaneously during the temperature controlled swelling of the films. Figure 6-9 illustrates the film thicknesses obtained from QCM-D and SE as a function of temperature for these hydrogel films. Despite the differences in the fit quality of the

QCM-D data for the V and R models, as shown previously in Figure 6-8 the fits mostly agree and there is general agreement with the thickness obtained from SE. The residual is nearly model invariant for T > 20 °C for the thinner film (Figure 6-9A). The thickness

178 from fits of the QCM-D data is greater than the thickness from SE as is expected due to surface coupling of water.254 Additionally, the film density was assumed to be constant

(1000 kg/m3) for ease of the fits. This assumption will lead to an over prediction of the thickness as the density of the dry copolymer is approximately 1200 kg/m3, but the error introduced by this assumption decreases as the film swells and more water is absorbed into the film. The difference in the thickness between the QCM-D and SE measurements is slightly less dependent on the permutation of overtones used for the R model than the V model. The V model under predicts the film thickness from SE at the lowest temperatures for 57V and 79V. The thicknesses obtained from these 2 fits of the QCM-D data are most similar with the predicted film thickness from the Sauerbrey expression with the 5th overtone. This under prediction in thickness from QCM-D is inconsistent with the ideas of coupled water for acoustic measurements, which is generally used to rationalize the differences between acoustic and optical measurements.254

The thicknesses obtained from the thicker film (Figure 6-9B) are nearly model independent, similar to the thinner film, for T > 18 °C with the offset between QCM-D and SE being nearly constant. The total difference in thickness between techniques at higher temperatures is similar between the two film except when using the 5th and 7th overtone with the V model (57V) where the thickness is approximately 10 nm less than the other thicknesses from the QCM-D data. Nonetheless, the similarity in the offset for two different film thicknesses is consistent with the differences being driven primarily by coupled water that is included in the acoustic measurement, but not the optical measurement.

179

Figure 6-9. Comparisons in the temperature dependent film thickness of the hydrogels determined from SE and QCM-D using the Voigt (V) and rheological Kramers-Kronig

(R) model for films initially (A) 52 nm and (B) 100 nm thick. The legends provide the different overtone combinations used to calculate the thickness with these two models and the application of the Sauerbrey expression for the overtone combinations listed. To better illustrate the differences, the top panel shows a residual plot for the difference in the thickness (QCM-SE).

At lower temperatures, the film thickness difference decreases for the V model, consistent with the thinner film, with the thickness from SE being greater than that determined from QCM-D at the lowest temperature examined. This change at the low temperatures may be a result of the shear wave attenuation in the film that decreases the impact of the water on the sensor frequency relative to an uncoated sensor. As the frequency data were corrected for the operation in water at each temperature, this subtraction may be overcompensating the frequency to produce to a thinner film prediction. However, the difference between the fit thickness from the R model and the

180

SE thickness remains on average consistent between the different overtone combinations, but there is some scatter at low temperature. This scatter may be attributed to discontinuities in the fit of the data (Figure 6-9B). It is interesting to note that the thicknesses obtained from fits of the QCM-D data with the two models (Figure 6-9) are very similar despite the differences in the goodness of the fits (Figure 6-8).

In addition to the thickness, the viscoelastic properties were extracted from the fits of the QCM data. As both the V and R models yield similar thicknesses, it will be instructive to examine the viscoelastic properties determined from these fits to see if there is reasonable agreement in these properties as well. Figure 6-10 illustrates the temperature dependence of the shear moduli for the two hydrogel films examined as a function of the viscoelastic model applied to extract these data from the QCM data. For both films, there appears to be a divergence in the predicted shear modulus between the two models at high temperatures. As the dissipation at high temperature is low, this difference may be related to the sensitivity of the QCM, where the signal is dominated by the elastic mass of the sample with the limit for this rigid behavior being the Sauerbrey expression.159 For the thinner film (Figure 6-10A), this divergence in the behavior of the fits between the two models continues to lower temperatures than for the thicker film

(Figure 6-10B). This is consistent with arguments that there must be sufficient energy dissipation in the shear wave (D/λs) in order to extract meaningful viscoelastic properties from the QCM. We will return to this point later.

At lower temperatures, the dissipation is high and thus one would expect that physically meaningful viscoelastic properties can be obtained. As shown in Figure 6-

10A, there is good agreement in the shear modulus obtained from the R and V models

181 when T < 20 °C. The R model tends to yield a slightly larger shear modulus then the V model on average, but these moduli are within a factor of 2 for a given temperature. This good agreement is somewhat surprising given the differences in the assumptions in the development of the models. For the thicker film, the agreement between the models is not as good (Figure 6-10B). Counter to the film thickness, there is scatter from the V model in the predicted shear modulus, while the R model provides a smooth function for the shear modulus. This difference in the scatter is a direct consequence of the self- consistency of the R model. The qualitative trends for the shear modulus are similar between the two models, but the shear modulus is dependent on the selection of the overtones used for the V model, while there is no effect of the overtone selection with the

R model. As the films become thicker, one would expect the R model to yield more accurate viscoelastic properties, but these two films are the same material. As the fractional swelling is similar at the same temperature for both films, one would expect similar shear moduli for a given temperature. The shear moduli obtained from the V model is similar between the two film thicknesses (save for the scatter in the data), while there is an increase in shear modulus for the thicker hydrogel film when using the R model. For thin polymer films, a mobile surface layer255 has been attributed to decreasing moduli of glassy polymer films;256 a similar softer surface in these hydrogels could lead to a lower shear modulus for the thin films, consistent with the predictions from the R model.

182

Figure 6-11 illustrates how the viscosity of the hydrogels changes with temperature for both films. Similar to the shear modulus, the difference in the predictions between the two models examined is within a factor of 2, when there is sufficient dissipation (low temperatures). Similar to the shear modulus, the viscosity from the R model at high temperatures, i.e. low dissipation, is low, despite the limited hydration, whereas the viscosity from the V model increases to a plateau. We attribute the divergence in the viscosity predicted from the R and V model at T > 20°C for the thin film and T > 24 °C for the thick film to the limited dissipation of the film that challenges an accurate determination of the rheological properties of the film. As expected from the shear wave analysis, this transition shifts to higher temperatures (less lossy, lower 1/λs) as the film thickness (D) increases. In comparing the viscosity of the thin and thick film from the R model, the viscosity is increased in the thicker film, consistent with the differences between thick and thin films for the shear moduli. For the V model, the viscosity exhibits similar trends in the low temperature regimes. Interestingly, the

183 viscosity for the thicker film is in good agreement with the viscosity from the R model, while the predicted viscosity is greater for the V model than the R model for the thinner film. This is counter to the shear modulus where there was a greater difference between models for the thicker film. For the thick film, there is significant scatter in the temperature dependence of the viscosity using the V model with the higher overtones, 57.

These same overtones also lead to some data scatter with the R model; the likely cause is the sensor being highly dampened to increase the uncertainty in the frequency.

Figure 6-11. Shear viscosity of the hydrogel films predicted from the QCM-D data using the V (dashed lines) and R (open symbols) models for initial copolymer thickness of (A)

52 nm and (B) 100 nm.

The divergence in the viscosity and shear modulus between the two models appears to be related to the shear wavelength. Prior work has focused on the shear wavelength required for a measureable deviation from the Sauerbrey expression.60, 170

Experimentally, this has been found to be D/λs =0.041 ± 0.016 as the critical value to have observable (5%) differences from the Sauerbrey expression.170 With the use of the two different rheological models to fit the QCM-D data, it is possible to determine if the critical value for the failure of the Sauerbrey expression also corresponds to the location

184 where the rheological properties of the films cannot be accurately determined. The QCM-

D data presented can be reduced in terms of D/λs. The calculation of 1/λs requires some knowledge of the rheological properties:

(6-3) where ρ is the medium density, ηm is medium viscosity, ω is the angular frequency (ω =

2π*F resonance), and ϕ is the phase angle calculated from the viscoelastic ratio as:

(6-4) where μ is the shear modulus, η is the shear viscosity, and ω is the angular frequency (ω

= 2π*F resonance). Figure 6-12 illustrates the temperature dependence of D/λs for the two film thicknesses examined for the 3rd/5th overtone fit. The shape of the curves mimics the temperature dependence of the film thickness. In this case, we have used the phase angle obtained from the V model for the 3rd/5th overtone viscoelastic results. The gold circle corresponds to the values of D/λs where the viscosity and shear modulus begin to agree for the two viscoelastic models (V and R) examined.

185

From prior experimental work, deviations from the Sauerbrey model begin

170 between a D/λs value of 0.015 to 0.057. This divide between the Sauerbrey to viscoelastic regime (D/λs = 0.015-0.057) is shown by the rightwards ascending lines in

Figure 6. This corresponds to a window of 5°C and 4°C for the 52 nm and 100 nm film, respectively, as shown by the leftwards ascending lines. As shown in Figure 6-12, the gold circle associated with the agreement in viscoelastic properties with the V and R models falls near or within the cross hatched region defined by the D/λs values where deviations from Sauerbrey are first noticeable. This gold circle region where the viscoelastic properties become independent of the model occurs at D/λs = 0.15 ± 0.005 when considering both films examined. This value is slightly less than the expected D/λs

186 for noticeable deviations from the Sauerbrey expression, but is reasonable that the lower limit for deviations from Sauerbrey is where the viscoelastic properties of the adhered films can be extracted accurately from the QCM-D data. The phase angles calculated from the V model are shown in Figures 6-13 and 6-14.

Figure 6-13. Calculated phase angle from the fit results of the 52nm (Dry) NF5 film using the viscoelastic model.

187

In addition to providing a generalized guideline for where viscoelastic properties are likely questionable due to the limited energy dissipation, the D/λs representation also provides an opportunity to examine the film resonance condition. As the 100 nm film is nearly double that of the 52 nm film, there is an offset by a factor of approximately 2 in

D/λs. From the data shown in Figure 6-8C, the frequency reaches a minimum at approximately 14 °C for the 5th overtone and < 5 °C for the 3rd overtone for the thicker film. From Figure 6-12B, this corresponds to D/λs ~0.11 and ~0.14 . The frequency is approaching a minimum for the 52 nm film for the 3rd/5th overtone, but is not experimentally observed as D/λs < 0.088 at the lowest temperature examined (Figure

6-12A). However, examining the 5th/7th harmonic fit for the 52 nm film where a minimum is observed leads to a value of D/λs ~0.11 (Figure 6-15). This provides the

55, 56, 253, 257 requirement of D/λs > 0.11 for film resonance to become prominent. This value is consistent with expectations for these viscoelastic films as  at low temperatures is small (1.16), so these hydrogel films should be approaching the fully viscous limit

(D/λs = 0.11) for film resonance.

188

Figure 6-15. Temperature dependence of D/λs from the QCM-D measurements of the hydrogel film for 52 nm thick dry copolymer using film thickness from (●) SE or (■)

In order to use these predictions, some knowledge of the viscoelastic properties of the films is necessary in order to calculate D/λs . However, the QCM operates in the MHz frequency range, so the high frequency rheological behavior is important. As shown in

Figure 6-16, the storage modulus as determined these QCM measurements are nearly an order of magnitude greater than the storage modulus determined at frequencies assessable to a conventional rheometer for the hydrogels at 5°C. These results provide caution to using bulk rheological properties at much lower frequency to predict D/λs.

189

From the data in Figure 6-16, the approximate phase angle associated with the high frequency rheological cannot be inferred from standard rheological measurements for non-Newtonian fluids. This provides a challenge to determine the minimum film thickness necessary for sufficient divergence from the ideal mass thin layer as noted by

57 Voinova et al. However, the limiting D/λs = 0.024 ± 0.005 for the cut-off as shown in

Figure 6-12 provides a route to estimate the requisite minimum thickness for obtaining trustworthy viscoelastic properties from QCM. For example, a purely viscous film layer

(ϕ= 0) with a viscosity equal to that of water (η = 0.001 Pa*s) must be greater than 19 nm for D/λs > 0.024. If the film viscosity is increased by an order of magnitude to 0.01 Pa*s, the film thickness must be approximately 60 nm for rheological characterization with

QCM for a 3rd/5th overtone (15/25 MHz) fit. These rough calculations can provide some guidance for the rheological characterization of thin films using QCM.

190

In a similar manner, the use of the limiting case of D/λs = 0.11 for film resonance provides a route to understand if an unexpected increase in the resonator frequency during an experiment might be expected due to film resonance. Using the same limiting cases, a purely viscous film layer (ϕ= 0) with a viscosity equal to that of water (η = 0.001

Pa*s) must be greater than 86 nm for D/λs > 0.11, while increasing the viscosity by an order of magnitude (η = 0.01 Pa*s) necessitates a film thicker than 270 nm for film resonance.

6.5. CONCLUSIONS

A thermoresponsive supramolecular hydrogel provides a route to significantly alter the viscoelastic properties of thin hydrogel films as determined by QCM-D. The properties extracted with two distinct viscoelastic models, Voigt (V) and a Kramers-

Kronig consistent rheological (R) model, are in good agreement with each other at low temperatures where the dissipation is high. Fits of the QCM-D data with both models yields thicknesses that are consistent with that obtained from spectroscopic ellipsometry over the entire temperature range examined, even for the thicker film where the frequency of the QCM increased as the film swelled (film resonance). However at high temperature, the shear modulus and viscosity of the hydrogel films were found to be dependent on the model choice, which was attributed to the limiting behavior of these models at low energy dissipation. To understand this behavior, the propagation of the shear wave through the hydrogel was considered through the use of the D/λs calculation.

This allows the data for the hydrogel films of different thickness to be collapsed to a single curve – at sufficiently low D/λs (~0.032), the Sauerbrey model will provide an accurate prediction for the film thickness as noted in prior work.170 Interestingly by using

191 the two viscoelastic models and assuming where these agree is associated with sufficient energy loss in the crystal to assess the viscoelastic properties of the film, a slightly lower limit than that for the Sauerbrey failure is found experimentally (D/λs = 0.024 ± 0.005).

This value should provide some guidance for thin film studies to determine when the viscoelastic properties extracted from the QCM-D measurements are physically meaningful. Additionally, we find good agreement with the prior calculations for film resonance to occur only when D/λs > 0.11, which should assist in identification of the potential for film resonance even for thin films. The agreement in the fit viscoelastic properties of the films between the simple Voigt model and a more robust rheological model was somewhat surprising, but we attribute this agreement to the high frequency range of the QCM sensor. At high frequency, the relaxations of the components are limited to much shorter motions and length scales than typically encountered in standard rheological measurements.

192

CHAPTER VII.

OVERALL SUMMARY AND FUTURE STUDIES

7.1 CONCLUSIONS

The overall objective of this research was to study the nature of physically crosslinked hydrogels, specifically crosslinks formed through hydrophobic aggregation.

Combining the mechanisms employed in thermoplastic elastomers, physically crosslinked hydrogels allow ease of fabrication and resistance to damage. They do suffer from the crosslink being transient. As a result they can flow over time and change shape.

The goal of this work was to study the transient behavior of the physical crosslink in thin film lateral confinement. As the crosslinks are transient, they will obtain an equilibrium swelling ratio, overcoming the lateral confinement restrains found in covalently crosslinked thin films.

In the first part of this dissertation thin film lateral confinement of physically crosslinked thin film hydrogels was investigated. It is known that swollen covalently crosslinked thin films demonstrate a reduced swelling ratio due to lateral constraints.

Physical crosslinks being transient in nature are able to allow a confined thin film to overcome these lateral restraints and obtain a swelling ratio nearly identical to the bulk gel. This result suggests the osmotic pressure of the unsaturated hydrophilic chains is sufficient to overcome the physical bond strength and induce rearrangement to a new

193 equilibrium state of the physical network. Under the osmotic swelling stress the films showed an increased swelling ratio, surpassing the 1D Flory swelling theory. The thinnest films obtaining a swelling ratio surpassing that measured for bulk. The thickest films showed a slightly reduced swelling ratio, but this is ascribed to be due in part to the incomplete swelling of the films after the first cycle. Using a medium thickness film, the second swelling cycle on cooling of the film obtained a slightly increased swelling ratio that also overcame the bulk swelling ratio. This work demonstrates the promise for physical hydrogel networks in confined swelling spaces, as they are able to rearrange to obtain their equilibrium swelling ratio uniformly across the sample. In this work it was also found that the thin films showed indication of an increase in their LCST of volume phase transition. Previous work had shown this can be ascribed to orientation of the chains of the network. This prompted our second study and area of interest: the relaxation mechanism of nanodomains on rearrangement.

In the second part of this dissertation small angle neutron scattering with contrast matching was used to probe the nanostructure of the physical hydrogel on stress relaxation. The ability of physical hydrogels to dissipate energy by their reversible bond breakage gives them toughness, but also means the materials can be deformed permanently with sufficient stress and time. The nature of the physical crosslink rearrangements promoted by stress is important. This is especially important to be understood in depth before they can be used in application of tissue replacement and repair in the body. In this questions of how the physical crosslinks breakup and reform and how the microscopic processes can be related to the macroscopic as seen stress relaxation were answered. Stress relaxation experiments revealed that relaxation

194 processes measured in stress relaxation are directly related to the nanostructure rearrangements as measured by SANS. Interestingly, the physical crosslink domains have an almost step like sequence of events on stressing and stress relaxation. Where the interconnecting chains are first strained to the point of matching the physical crosslink energy, upon which physical crosslink segment pullout occurs and stress is relieved from the network. Upon relaxation of the interconnecting chain stress the domains show a rebound like effect as they spread in the transverse direction to the strain direction. It was also shown that the relaxation of physical crosslinks is dominated by segment pullout and not domain scission.

In the third part of this dissertation the ice inhibition effects in hydrophobically modified physical gels was studied. The typical means to study the supercooling of water has involved either hydrogen bonding effects, forming tightly bound hydration layers of water, or through means of physical confinement, where pore dimensions of less than 2 nanometers inhibits the formation of the hydrogen bond network on crystallization.

Physical crosslinks junctions of hydrophobic aggregates act as hard-wall like surfaces in the hydrogel. The physical domains are impermeable to water and confine the water to dimensions less than a few nanometers. It was shown that as the hydrophilic content decreased in the physical gels, the fraction of supercooled water increased. This is contrary to that expected for hydrogels and clearly demonstrates the ability of physical hydrogels to combine the effects of strongly bound water and confined water to prevent crystallization and allow supercooling of the water to <230K. From an analysis of the water dynamics using neutron scattering, it was also shown that the water in the DF22 hydrogel exhibits diffusion processes faster than that for bulk water at temperatures

195 below 240K. All three gels examines show diffusion characteristics at 295K that are an order of magnitude lower than for bulk water. Finally, it was also found that by increasing the physical crosslink content, the damage resistance to partial freezing of the water could be increased. On measuring the mechanical properties, a decreased of less than 10% in modulus was found upon immediate testing. This could only be expected to further improve as time or heat is applied to promote the reaggregation of physical crosslink moieties.

In the fourth part of this dissertation the operational regimes of the quartz crystal microbalance (QCM) were probed used the precise swelling of the physically crosslinked thermoresponsive hydrogel. Using the thermoresponsive NF5 copolymer, the swelling of two thin films were used to probe the operation of the QCM from a highly elastic

Sauerbrey regime to the non-linear film resonance regime. Using two different models for analysis it was found that at about the region of Sauerbrey model deviation of 5% error, the onset of sensitive viscoelastic modeling occurs. This study also revealed that both models can fit data through the film resonance regime where the relationship between mass and frequency change are reversed.

The study of the physical networks has many interesting projects yet to be fully understood. As the network formed by these physical bonds is quite uniform and offers a non-interacting nature (non-ionic, and glassy) it is a good model system for future study into physically crosslinks systems.

196

7.2 FUTURE WORK

7.2.1 In situ straining of physical gel

From the SANS we were able to obtain the stress relaxation information of the nanostructure. At very early times there appears to be a reversal of the chain orientation at short times that was not capture by the limited time resolution of SANS. As such from our gaining understanding of the change in components, we can now decipher the changes that would occur in a single scattering plot as obtained by x-ray. The proposed work would be study the straining process and initial relaxation. Also, with the short time resolution, multiple strain and strain rates can be investigated. As the physical crosslink should have a characteristic relaxation time at which the gel behaves in amore elastic manner versus viscous flow. From the SANS we believe this to be around 0.1 1/s strain rate.

7.2.2 Study of physical crosslink bond energy

As physical crosslinks are very useful, their complete understanding can give insight into future system design. One area that has always been of trouble to fully understand has been the strength of a physical crosslink and segment pullout. Using the neutron scattering results and recent small angle x-ray scattering results, it may be possible to calculate the energy to induce segment pullout of the hydrophobic crosslink.

Using the energy to deform a polymer chain in solution, one can then relate the domain spacing changes to that of the chain extension. Thereby a spring force constant of the chain can be found. From this it can be equated to the strain at which significant rearrangement begins to occur. With these two terms a relation of the strength of the physical bond may be able to obtained.

197

7.2.3 Ice inhibition versus hydration content

Work by Ito et al.133, 134 showed that the UFW remains the same as water content is decreased. Will this be the case for the physical gels as the nanodomains that induce confinement supercooling are brought in closer proximity? In typical hydrogels, the supercooled water fraction is intimately tied to the primary hydration layer. This layer is also the last bound layer to be lost on dehydrating the sample. In typical hydrogel systems, this primary hydration is the total amount of water which can be supercooled based on the content of polymer. In the physical system there is a second contributing component to the supercooled water as a result of the confinement effects from nanodomains. The proposed work would be to study the ratio of supercooled water to dry polymer mass as it scales with total hydration. This would be of special interest for the low physical crosslinking content systems, to understand at which domain to domain proximity the fraction of supercooled water increases, indicating the confinement effect induced supercooling of the water. It may be that there is no increase, as the confinement must also have some strength. As in the high FOSA content gels, the distance is an equilibrium balance of domain strength and swelling. While in a partially hydrated low

FOSA content system, the spacing could expand in locals and collapse in other much like it does for the fully hydrated state.

198

IIIX. REFERENCES

1. Alemán, J.; Chadwick, A. V.; He, J.; Hess, M.; Horie, K.; Jones, R. G.; Kratochvil, P.; Meisel, I.; Mita, I.; Moad, G. Definitions of terms relating to the structure and processing of sols, gels, networks, and inorganic-organic hybrid materials (IUPAC Recommendations 2007). Pure and Applied Chemistry 2007, 79 (10), 1801-1829.

2. Cartmell, J. V.; Sturtevant, W. R., Transparent hydrogel wound dressing. Google Patents: 1992; Vol. US5106629 A.

3. Kokabi, M.; Sirousazar, M.; Hassan, Z. M. PVA-clay nanocomposite hydrogels for wound dressing. European Polymer Journal 2007, 43 (3), 773-781

4. Quarfoot, A. J.; Hyla, P. H.; Patience, D., Hydrogel wound dressing. Google Patents: 1990.

5. Chaterji, S.; Kwon, I. K.; Park, K. Smart polymeric gels: Redefining the limits of biomedical devices. Progress in Polymer Science 2007, 32 (8-9), 1083-1122

6. Hoffman, A. S. Hydrogels for biomedical applications. Advanced drug delivery reviews 2012, 64, 18-23.

7. Bhattarai, N.; Gunn, J.; Zhang, M. Chitosan-based hydrogels for controlled, localized drug delivery. Advanced Drug Delivery Reviews 2010, 62 (1), 83-99

8. Deligkaris, K.; Tadele, T. S.; Olthuis, W.; van den Berg, A. Hydrogel-based devices for biomedical applications. Sensors and Actuators B-Chemical 2010, 147 (2), 765-774

9. Tokarev, I.; Minko, S. Stimuli-Responsive Porous Hydrogels at Interfaces for Molecular Filtration, Separation, Controlled Release, and Gating in Capsules and Membranes. Advanced Materials 2010, 22 (31), 3446-3462

10. Harmon, M. E.; Kuckling, D.; Pareek, P.; Frank, C. W. Photo-cross-linkable PNIPAAm copolymers. 4. Effects of copolymerization and cross-linking on the volume- phase transition in constrained hydrogel layers. Langmuir 2003, 19 (26), 10947-10956

199

11. Heller, A. Potentially implantable miniature batteries. Analytical and Bioanalytical Chemistry 2006, 385 (3), 469-473

12. Hook, F.; Kasemo, B.; Nylander, T.; Fant, C.; Sott, K.; Elwing, H. Variations in coupled water, viscoelastic properties, and film thickness of a Mefp-1 protein film during adsorption and cross-linking: A quartz crystal microbalance with dissipation monitoring, ellipsometry, and surface plasmon resonance study. Analytical Chemistry 2001, 73 (24), 5796-5804

13. Ajji, Z.; Othman, I.; Rosiak, J. M. Production of hydrogel wound dressings using gamma radiation. Nuclear Instruments & Methods in Physics Research Section B-Beam Interactions with Materials and Atoms 2005, 229 (3-4), 375-380

14. Eddington, D. T.; Beebe, D. J. Flow control with hydrogels. Advanced Drug Delivery Reviews 2004, 56 (2), 199-210

15. Zhang, X. B.; Pint, C. L.; Lee, M. H.; Schubert, B. E.; Jamshidi, A.; Takei, K.; Ko, H.; Gillies, A.; Bardhan, R.; Urban, J. J.; Wu, M.; Fearing, R.; Javey, A. Optically- and Thermally-Responsive Programmable Materials Based on Carbon Nanotube- Hydrogel Polymer Composites. Nano Letters 2011, 11 (8), 3239-3244

16. Sun, J. Y.; Zhao, X. H.; Illeperuma, W. R. K.; Chaudhuri, O.; Oh, K. H.; Mooney, D. J.; Vlassak, J. J.; Suo, Z. G. Highly stretchable and tough hydrogels. Nature 2012, 489 (7414), 133-136

17. Zheng, W. J.; An, N.; Yang, J. H.; Zhou, J. X.; Chen, Y. M. Tough Al- alginate/Poly(N-isopropylacrylamide) Hydrogel with Tunable LCST for Soft Robotics. Acs Applied Materials & Interfaces 2015, 7 (3), 1758-1764

18. Satarkar, N. S.; Biswal, D.; Hilt, J. Z. Hydrogel nanocomposites: a review of applications as remote controlled biomaterials. Soft Matter 2010, 6 (11), 2364-2371

19. Yuk, H.; Lin, S. T.; Ma, C.; Takaffoli, M.; Fang, N. X.; Zhao, X. H. Hydraulic hydrogel actuators and robots optically and sonically camouflaged in water. Nature Communications 2017, 8, 1-12

20. Costa, A. M. S.; Mano, J. F. Extremely strong and tough hydrogels as prospective candidates for tissue repair - A review. European Polymer Journal 2015, 72, 344-364

21. Nguyen, M. K.; Lee, D. S. Injectable biodegradable hydrogels. Macromolecular bioscience 2010, 10 (6), 563-579

200

22. Suh, J. K. F.; Matthew, H. W. T. Application of chitosan-based polysaccharide biomaterials in cartilage tissue engineering: a review. Biomaterials 2000, 21 (24), 2589- 2598.

23. Wojtowicz, A. M.; Templeman, K. L.; Hutmacher, D. W.; Guldberg, R. E.; Garcia, A. J. Runx2 Overexpression in Bone Marrow Stromal Cells Accelerates Bone Formation in Critical-Sized Femoral Defects. Tissue Engineering Part A 2010, 16 (9), 2795-2808

24. Zander, Z. K.; Hua, G.; Wiener, C. G.; Vogt, B. D.; Becker, M. L. Control of Mesh Size and Modulus by Kinetically Dependent Cross-Linking in Hydrogels. Advanced Materials 2015, 27 (40), 6283-6288

25. Jen, A. C.; Wake, M. C.; Mikos, A. G. Review: Hydrogels for cell immobilization. Biotechnology and Bioengineering 1996, 50 (4), 357-364

26. Zha, L.; Banik, B.; Alexis, F. Stimulus responsive nanogels for drug delivery. Soft Matter 2011, 7 (13), 5908-5916

27. Yu, C. J.; Yuan, P. X.; Erickson, E. M.; Daly, C. M.; Rogers, J. A.; Nuzzo, R. G. Oxygen reduction reaction induced pH-responsive chemo-mechanical hydrogel actuators. Soft Matter 2015, 11 (40), 7953-7959

28. Gong, J. P.; Katsuyama, Y.; Kurokawa, T.; Osada, Y. Double-network hydrogels with extremely high mechanical strength. Advanced Materials 2003, 15 (14), 1155-1158

29. Haque, M. A.; Kurokawa, T.; Kamita, G.; Gong, J. P. Lamellar Bilayers as Reversible Sacrificial Bonds To Toughen Hydrogel: Hysteresis, Self-Recovery, Fatigue Resistance, and Crack Blunting. Macromolecules 2011, 44 (22), 8916-8924

30. Bin Ihsan, A.; Sun, T. L.; Kuroda, S.; Haque, M. A.; Kurokawa, T.; Nakajima, T.; Gong, J. P. A phase diagram of neutral polyampholyte - from solution to tough hydrogel. Journal of Materials Chemistry B 2013, 1 (36), 4555-4562

31. Es-haghi, S. S.; Weiss, R. A. Fabrication of Tough Hydrogels from Chemically Cross-Linked Multiple Neutral Networks. Macromolecules 2016, 49 (23), 8980-8987

32. Hao, J. K.; Weiss, R. A. Mechanically Tough, Thermally Activated Shape Memory Hydrogels. Acs Macro Letters 2013, 2 (1), 86-89

33. Zhai, Y.; Meng, X.; Duan, H.; Ding, Z.; Liu, Y.; Lucia, L. Super Stable and Tough Hydrogel Containing Covalent, Crystalline, and Ionic Cross-Links. Macromolecular Chemistry and Physics 2016, 217 (1), 32-38

201

34. Agrawal, A.; Rahbar, N.; Calvert, P. D. Strong fiber-reinforced hydrogel. Acta Biomaterialia 2013, 9 (2), 5313-5318

35. Gong, J. P.; Katsuyama, Y.; Kurokawa, T.; Osada, Y. Double-network hydrogels with extremely high mechanical strength. Advanced Materials 2003, 15 (14), 1155-1158

36. Guan, Y.; Bian, J.; Peng, F.; Zhang, X. M.; Sun, R. C. High strength of hemicelluloses based hydrogels by freeze/thaw technique. Carbohydrate Polymers 2014, 101, 272-280

37. Hao, J. K.; Weiss, R. A. Viscoelastic and Mechanical Behavior of Hydrophobically Modified Hydrogels. Macromolecules 2011, 44 (23), 9390-9398

38. Naficy, S.; Brown, H. R.; Razal, J. M.; Spinks, G. M.; Whitten, P. G. Progress Toward Robust Polymer Hydrogels. Australian Journal of Chemistry 2011, 64 (8), 1007- 1025

39. Tanaka, Y.; Gong, J. P.; Osada, Y. Novel hydrogels with excellent mechanical performance. Progress in Polymer Science 2005, 30 (1), 1-9

40. Gong, J. P. Why are double network hydrogels so tough? Soft Matter 2010, 6 (12), 2583-2590

41. Es-haghi, S. S.; Leonov, A. I.; Weiss, R. A. On the Necking Phenomenon in Pseudo-Semi-Interpenetrating Double-Network Hydrogels. Macromolecules 2013, 46 (15), 6203-6208

42. Nakajima, T.; Furukawa, H.; Tanaka, Y.; Kurokawa, T.; Osada, Y.; Gong, J. P. True Chemical Structure of Double Network Hydrogels. Macromolecules 2009, 42 (6), 2184-2189

43. Sun, J.-Y.; Zhao, X.; Illeperuma, W. R. K.; Chaudhuri, O.; Oh, K. H.; Mooney, D. J.; Vlassak, J. J.; Suo, Z. Highly stretchable and tough hydrogels. Nature 2012, 489 (7414), 133-136

44. Cui, J. X.; del Campo, A. Multivalent H-bonds for self-healing hydrogels. Chemical Communications 2012, 48 (74), 9302-9304

45. Calvert, P. Hydrogels for Soft Machines. Advanced Materials 2009, 21 (7), 743- 756

46. Sun, T. L.; Kurokawa, T.; Kuroda, S.; Bin Ihsan, A.; Akasaki, T.; Sato, K.; Haque, M. A.; Nakajima, T.; Gong, J. P. Physical hydrogels composed of

202 polyampholytes demonstrate high toughness and viscoelasticity. Nature Materials 2013, 12 (10), 932-937

47. Li, J. Y.; Illeperuma, W. B. K.; Suo, Z. G.; Vlassak, J. J. Hybrid Hydrogels with Extremely High Stiffness and Toughness. Acs Macro Letters 2014, 3 (6), 520-523

48. Haraguchi, K.; Farnworth, R.; Ohbayashi, A.; Takehisa, T. Compositional effects on mechanical properties of nanocomposite hydrogels composed of poly(N,N- dimethylacrylamide) and clay. Macromolecules 2003, 36 (15), 5732-5741

49. Wang, W.; Troll, K.; Kaune, G.; Metwalli, E.; Ruderer, M.; Skrabania, K.; Laschewsky, A.; Roth, S. V.; Papadakis, C. M.; Muller-Buschbaum, P. Thin films of poly(N-isopropylacrylamide) end-capped with n-butyltrithiocarbonate. Macromolecules 2008, 41 (9), 3209-3218

50. Ito, K. Novel cross-linking concept of polymer network: Synthesis, structure, and properties of slide-ring gels with freely movable junctions. Polymer Journal 2007, 39 (6), 489-499

51. Ito, K. Slide-ring materials using topological supramolecular architecture. Current Opinion in Solid State & Materials Science 2010, 14 (2), 28-34

52. Flory, P. J.; Rehner, J. Statistical Mechanics of Cross‐Linked Polymer Networks II. Swelling. The Journal of Chemical Physics 1943, 11 (11), 521-526

53. Toomey, R.; Freidank, D.; Ruhe, J. Swelling behavior of thin, surface-attached polymer networks. Macromolecules 2004, 37 (3), 882-887

54. Tian, J.; Seery, T. A. P.; Ho, D. L.; Weiss, R. A. Physically cross-linked alkylacrylamide hydrogels: A SANS analysis of the microstructure. Macromolecules 2004, 37 (26), 10001-10008

55. Domack, A.; Prucker, O.; Ruhe, J.; Johannsmann, D. Swelling of a polymer brush probed with a quartz crystal resonator. Physical Review E 1997, 56 (1), 680-689

56. Graneli, A.; Edvardsson, M.; Hook, F. DNA-based formation of a supported, three-dimensional lipid vesicle matrix probed by QCM-D and SPR. Chemphyschem 2004, 5 (5), 729-733

57. Voinova, M. V.; Rodahl, M.; Jonson, M.; Kasemo, B. Viscoelastic acoustic response of layered polymer films at fluid-solid interfaces: Continuum mechanics approach. Physica Scripta 1999, 59 (5), 391-396

203

58. DeNolf, G. C.; Haack, L.; Holubka, J.; Straccia, A.; Blohowiak, K.; Broadbent, C.; Shull, K. R. High Frequency Rheometry of Viscoelastic Coatings with the Quartz Crystal Microbalance. Langmuir 2011, 27 (16), 9873-9879

59. Martin, E. J.; Mathew, M. T.; Shull, K. R. Viscoelastic Properties of Electrochemically Deposited Protein/Metal Complexes. Langmuir 2015, 31 (13), 4008- 4017

60. White, C. C.; Schrag, J. L. Theoretical predictions for the mechanical response of a model quartz crystal microbalance to two viscoelastic media: A thin sample layer and surrounding bath medium. Journal of Chemical Physics 1999, 111 (24), 11192-11206

61. Richter, A.; Paschew, G.; Klatt, S.; Lienig, J.; Arndt, K.-F.; Adler, H.-J. P. Review on hydrogel-based pH sensors and microsensors. Sensors 2008, 8 (1), 561-581

62. Luo, Y.; Shoichet, M. S. A photolabile hydrogel for guided three-dimensional cell growth and migration. Nature Materials 2004, 3 (4), 249-253

63. Zhao, X. H. Multi-scale multi-mechanism design of tough hydrogels: building dissipation into stretchy networks. Soft Matter 2014, 10 (5), 672-687

64. Lake, G.; Thomas, A. In The strength of highly elastic materials, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, 1967; The Royal Society: 1967; 108-119.

65. Simha, N. K.; Carlson, C. S.; Lewis, J. L. Evaluation of fracture toughness of cartilage by micropenetration. Journal of Materials Science-Materials in Medicine 2004, 15 (5), 631-639

66. Taylor, D.; O'Mara, N.; Ryan, E.; Takaza, M.; Simms, C. The fracture toughness of soft tissues. Journal of the Mechanical Behavior of Biomedical Materials 2012, 6, 139-147

67. Shibayama, M.; Karino, T.; Miyazaki, S.; Okabe, S.; Takehisa, T.; Haraguchi, K. Small-angle neutron scattering study on uniaxially stretched poly(N- isopropylacrylamide)-clay nanocomposite gels. Macromolecules 2005, 38 (26), 10772- 10781

68. Okumura, Y.; Ito, K. The polyrotaxane gel: A topological gel by figure-of-eight cross-links. Advanced Materials 2001, 13 (7), 485-487

69. Bae, S. S.; Chakrabarty, K.; Seery, T. A. P.; Weiss, R. A. Thermoprocessible hydrogels. I. Synthesis and properties of polyacrylamides with perfluoroalkyl side chains.

204

Journal of Macromolecular Science-Pure and Applied Chemistry 1999, A36 (7-8), 931- 948

70. Tian, J.; Seery, T. A. P.; Weiss, R. A. Physically cross-linked alkylacrylamide hydrogels: Phase behavior and microstructure. Macromolecules 2004, 37 (26), 9994- 10000

71. Sanabria-DeLong, N.; Agrawal, S. K.; Bhatia, S. R.; Tew, G. N. Impact of synthetic technique on PLA-PEO-PLA physical hydrogel properties. Macromolecules 2007, 40 (22), 7864-7873

72. Richards, R. Polyethylene‐structure, crystallinity and properties. Journal of Chemical Technology and Biotechnology 1951, 1 (8), 370-376.

73. Fawcett, E. W.; Gibson, R. O. 90. The influence of pressure on a number of organic reactions in the liquid phase. Journal of the Chemical Society (Resumed) 1934, (0), 386-395

74. Newman, S.; Cox, W. The glass temperature of semicrystalline polymers. Journal of Polymer Science 1960, 46 (147), 29-49.

75. Bailey, J. T.; Bishop, E. T.; Hendrick.Wr; Holden, G.; Legge, N. R. Thermoplastic elastomers - physical properties and applications. Rubber Age 1966, 98 (10), 69-77.

76. Deanin, R. D. Butadiene/styrene - thermoplastic elastomers. Spe Journal 1967, 23 (1), 45-&.

77. Teitelbaum, B. Y.; Yagfarov, M. S.; Anoshina, N. P.; Palikhov, N. A.; Menshova, R. K.; Apukhtina, N. P.; Myuller, B. E.; Panova, N. V. Crystallization of polyurethane thermoplastic elastomers based on polyethers and polyesters. Vysokomolekulyarnye Soedineniya Section A 1971, 13 (10), 2291-&.

78. Carvey, R. M.; Stetz, T. T. Thermoplastic polyurethane materials for adhesive systems. Adhesives Age 1968, 11 (9), 35-&.

79. Holden, G.; Bishop, E. T.; Legge, N. R. Thermoplastic elastomers. Journal of Polymer Science Part C-Polymer Symposium 1969, (26PC), 37-&.

80. Morton, M.; McGrath, J. E.; Juliano, P. C. Structure-property relationships for styrene-diene thermoplastic elastomers. Journal of Polymer Science Part C-Polymer Symposium 1969, (26PC), 99-&.

205

81. Harms, S.; Ratzke, K.; Faupel, F.; Egger, W.; Ravello, L.; Laschewsky, A.; Wang, W.; Muller-Buschbaum, P. Free Volume and Swelling in Thin Films of Poly(N- isopropylacrylamide) End-Capped with n-Butyltrithiocarbonate. Macromolecular Rapid Communications 2010, 31 (15), 1364-1367

82. Lee, J. A.; Kontopoulou, M.; Parent, J. S. Rheology and Physical Properties of Polyethylene/Polyethylene‐Ionomer Blends and their Clay Nanocomposites. Macromolecular rapid communications 2007, 28 (2), 210-214.

83. Hirasawa, E.; Yamamoto, Y.; Tadano, K.; Yano, S. Formation of ionic crystallites and its effect on the modulus of ethylene ionomers. Macromolecules 1989, 22 (6), 2776- 2780.

84. Makowski, H.; Lundberg, R.; Westerman, L.; Bock, J., Synthesis and properties of sulfonated EPDM. ACS Publications: 1980. 3-19

85. Urushizaki, F.; Yamaguchi, H.; Nakamura, K.; Numajiri, S.; Sugibayashi, K.; Morimoto, Y. Swelling and mechanical-properties of poly(vinyl alcohol) hydrogels. International Journal of Pharmaceutics 1990, 58 (2), 135-142.

86. Hernandez, R.; Sarafian, A.; Lopez, D.; Mijangos, C. Viscoelastic properties of poly(vinyl alcohol) hydrogels and ferrogels obtained through freezing-thawing cycles. Polymer 2004, 45 (16), 5543-5549

87. Fu, S.; Thacker, A.; Sperger, D. M.; Boni, R. L.; Buckner, I. S.; Velankar, S.; Munson, E. J.; Block, L. H. Relevance of Rheological Properties of Sodium Alginate in Solution to Calcium Alginate Gel Properties. Aaps Pharmscitech 2011, 12 (2), 453-460

88. Zhong, Q.; Metwalli, E.; Rawolle, M.; Kaune, G.; Bivigou-Koumba, A. M.; Laschewsky, A.; Papadakis, C. M.; Cubitt, R.; Muller-Buschbaum, P. Structure and Thermal Response of Thin Thermoresponsive Polystyrene-block- poly(methoxydiethylene glycol acrylate)-block-polystyrene Films. Macromolecules 2013, 46 (10), 4069-4080

89. Nykanen, A.; Nuopponen, M.; Laukkanen, A.; Hirvonen, S. P.; Rytela, M.; Turunen, O.; Tenhu, H.; Mezzenga, R.; Ikkala, O.; Ruokolainen, J. Phase behavior and temperature-responsive molecular filters based on self-assembly of Polystyrene-block- poly(N-isopropylacrylamide)-block-polystyrene. Macromolecules 2007, 40 (16), 5827- 5834

90. Bailey, J. T.; Hendrick.Wr; Holden, G.; Bishop, E. T.; Legge, N. R. Thermoplastic elastomers - physical properties and applications. Rubber Age 1965, 97 (5), 77-&.

206

91. Lu, Y.; Wang, D.; Li, T.; Zhao, X.; Cao, Y.; Yang, H.; Duan, Y. Y. Poly (vinyl alcohol)/poly (acrylic acid) hydrogel coatings for improving electrode–neural tissue interface. Biomaterials 2009, 30 (25), 4143-4151.

92. Brambilla, D.; Luciani, P.; Leroux, J. C. Breakthrough discoveries in drug delivery technologies: The next 30 years. Journal of Controlled Release 2014, 190, 9-14

93. Dong, L. C.; Hoffman, A. S. Synthesis and application of thermally reversible heterogels for drug delivery. Journal of Controlled Release 1990, 13 (1), 21-31

94. Oh, J. K.; Drumright, R.; Siegwart, D. J.; Matyjaszewski, K. The development of microgels/nanogels for drug delivery applications. Progress in Polymer Science 2008, 33 (4), 448-477

95. Ho, D. L.; Briber, R. M.; Jones, R. L.; Kumar, S. K.; Russell, T. P. Small-angle neutron scattering studies on thin films of isotopic polystyrene blends. Macromolecules 1998, 31 (26), 9247-9252

96. Jones, R. L.; Kumar, S. K.; Ho, D. L.; Briber, R. M.; Russell, T. P. Chain conformation in ultrathin polymer films. Nature 1999, 400 (6740), 146-149.

97. Patra, L.; Toomey, R. Viscoelastic Response of Photo-Cross-Linked Poly(N- isopropylacrylamide) Coatings by QCM-D. Langmuir 2010, 26 (7), 5202-5207

98. Suzuki, A.; Kojima, S. Phase-transition in constrained polymer gels. Journal of Chemical Physics 1994, 101 (11), 10003-10007

99. Hui, C. Y.; Jagota, A.; Bennison, S. J.; Londono, J. D. Crack blunting and the strength of soft elastic solids. Proceedings of the Royal Society a-Mathematical Physical and Engineering Sciences 2003, 459 (2034), 1489-1516

100. Lu, J.; Bates, F. S.; Lodge, T. P. Remarkable Effect of Molecular Architecture on Chain Exchange in Triblock Copolymer Micelles. Macromolecules 2015, 48 (8), 2667- 2676

101. Peters, A. J.; Lodge, T. P. Comparison of Gel Relaxation Times and End-Block Pullout Times in ABA Triblock Copolymer Networks. Macromolecules 2016. (49),7340- 7349

102. Angell, C. A. Insights into phases of liquid water from study of its unusual glass- forming properties. Science 2008, 319 (5863), 582-587

207

103. Bellissentfunel, M. C.; Chen, S. H.; Zanotti, J. M. Single-particle dynamics of water-molecules in confined space. Physical Review E 1995, 51 (5), 4558-4569

104. Bruni, F.; Ricci, M. A.; Soper, A. K. Water confined in Vycor glass. I. A neutron diffraction study. Journal of Chemical Physics 1998, 109 (4), 1478-1485

105. Chu, X. Q.; Liu, K. H.; Tyagi, M. S.; Mou, C. Y.; Chen, S. H. Low-temperature dynamics of water confined in a hydrophobic mesoporous material. Physical Review E 2010, 82 (2), 1-4

106. Eastman, S. A.; Kim, S.; Page, K. A.; Rowe, B. W.; Kang, S. H.; DeCaluwe, S. C.; Dura, J. A.; Soles, C. L.; Yager, K. G. Effect of Confinement on Structure, Water Solubility, and Water Transport in Nafion Thin Films. Macromolecules 2012, 45 (19), 7920-7930

107. Chu, X. Q.; Kolesnikov, A. I.; Moravsky, A. P.; Garcia-Sakai, V.; Chen, S. H. Observation of a dynamic crossover in water confined in double-wall carbon nanotubes. Physical Review E 2007, 76 (2), 1-6

108. Diallo, S. O.; Vlcek, L.; Mamontov, E.; Keum, J. K.; Chen, J. H.; Hayes, J. S.; Chialvo, A. A. Translational diffusion of water inside hydrophobic carbon micropores studied by neutron spectroscopy and molecular dynamics simulation. Physical Review E 2015, 91 (2), 1-9

109. Diallo, S. O.; Jazdzewska, M.; Palmer, J. C.; Mamontov, E.; Gubbins, K. E.; Sliwinska-Bartkowiak, M. Dynamics of nanoconfined water under pressure. Physical Review E 2013, 88 (2), 1-6

110. Gallo, P.; Ricci, M. A.; Rovere, M. Layer analysis of the structure of water confined in vycor glass. Journal of Chemical Physics 2002, 116 (1), 342-346

111. Gordillo, M. C.; Marti, J. Hydrogen bond structure of liquid water confined in nanotubes. Chemical Physics Letters 2000, 329 (5-6), 341-345

112. Liu, L.; Chen, S. H.; Faraone, A.; Yen, C. W.; Mou, C. Y.; Kolesnikov, A. I.; Mamontov, E.; Leao, J. Quasielastic and inelastic neutron scattering investigation of fragile-to-strong crossover in deeply supercooled water confined in nanoporous silica matrices. Journal of Physics-Condensed Matter 2006, 18 (36), S2261-S2284

113. Zanotti, J. M.; Bellissent-Funel, M. C.; Chen, S. H. Relaxational dynamics of supercooled water in porous glass. Physical Review E 1999, 59 (3), 3084-3093

208

114. Speedy, R. J. Stability-limit conjecture - an interpretation of the properties of water. Journal of Physical Chemistry 1982, 86 (6), 982-991

115. Zaworski, R. J.; Keenan, J. H. Thermodynamic properties of water in region of maximum density. Journal of Applied Mechanics 1967, 34 (2), 478-483

116. Sciortino, F.; Gallo, P.; Tartaglia, P.; Chen, S. H. Supercooled water and the kinetic glass transition. Physical Review E 1996, 54 (6), 6331-6343

117. Bigg, E. K. The supercooling of water. Proceedings of the Physical Society. Section B 1953, 66 (8), 688.

118. Paradossi, G.; Cavalieri, F.; Chiessi, E.; Telling, M. T. F. Supercooled water in PVA matrixes: I. An incoherent quasi-elastic neutron scattering (QENS) study. Journal of Physical Chemistry B 2003, 107 (33), 8363-8371

119. Frank, F. Supercooling of liquids. Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences 1952, 43-46.

120. Mossop, S. The freezing of supercooled water. Proceedings of the Physical Society. Section B 1955, 68 (4), 193.

121. Gibbard, H. F.; Gossmann, A. F. Freezing Points of Electrolyte Mixtures. I. Mixtures of Sodium Chloride and Magnesium Chloride in Water. Journal of Solution Chemistry 1974, 3 (5), 385-393

122. Liu, D. Z.; Zhang, Y.; Chen, C. C.; Mou, C. Y.; Poole, P. H.; Chen, S. H. Observation of the density minimum in deeply supercooled confined water. Proceedings of the National Academy of Sciences of the United States of America 2007, 104 (23), 9570-9574

123. Faraone, A.; Liu, L.; Mou, C. Y.; Shih, P. C.; Brown, C.; Copley, J. R. D.; Dimeo, R. M.; Chen, S. H. Dynamics of supercooled water in mesoporous silica matrix MCM- 48-S. European Physical Journal E 2003, 12, S59-S62

124. Faraone, A.; Liu, K. H.; Mou, C. Y.; Zhang, Y.; Chen, S. H. Single particle dynamics of water confined in a hydrophobically modified MCM-41-S nanoporous matrix. Journal of Chemical Physics 2009, 130 (13), 1-10

125. Sun, T. J.; Lin, F. H.; Campbell, R. L.; Allingham, J. S.; Davies, P. L. An Antifreeze Protein Folds with an Interior Network of More Than 400 Semi-Clathrate Waters. Science 2014, 343 (6172), 795-798

209

126. Deller, R. C.; Vatish, M.; Mitchell, D. A.; Gibson, M. I. Synthetic polymers enable non-vitreous cellular cryopreservation by reducing ice crystal growth during thawing. Nature Communications 2014, 5, 1-7

127. Hagiwara, Y.; Aomatsu, H. Supercooling enhancement by adding antifreeze protein and ions to water in a narrow space. International Journal of Heat and Mass Transfer 2015, 86, 55-64

128. Rasaiah, J. C.; Garde, S.; Hummer, G. Water in nonpolar confinement: From nanotubes to proteins and beyond. Annual Review of Physical Chemistry 2008, 59, 713- 740

129. Tsai, A. M.; Neumann, D. A.; Bell, L. N. Molecular dynamics of solid-state lysozyme as affected by glycerol and water: A neutron scattering study. Biophysical Journal 2000, 79 (5), 2728-2732.

130. Moore, E. B.; Allen, J. T.; Molinero, V. Liquid-Ice Coexistence below the Melting Temperature for Water Confined in Hydrophilic and Hydrophobic Nanopores. Journal of Physical Chemistry C 2012, 116 (13), 7507-7514

131. Takaiwa, D.; Koga, K.; Tanaka, H. Structures of filled ice nanotubes inside carbon nanotubes. Molecular Simulation 2007, 33 (1-2), 127-132

132. Hatakeyema, T.; Yamauchi, A.; Hatakeyema, H. Studies on bound water in polyvinyl-alcohol) - hydrogel by dsc and ft nmr. European Polymer Journal 1984, 20 (1), 61-64

133. Ito, K.; Yoshida, K.; Bellissent-Funel, M. C.; Yamaguchi, T. Dynamic Properties of Water Confined in Sephadex G15 Gel by Quasi-Elastic Neutron Scattering and Neutron Spin Echo Measurements. Bulletin of the Chemical Society of Japan 2014, 87 (5), 603-608

134. Ito, K.; Yoshida, K.; Ujimoto, K.; Yamaguchi, T. Thermal Behavior and Structure of Low-temperature Water Confined in Sephadex G15 Gel by Differential Scanning Calorimetry and X-ray Diffraction Method. Analytical Sciences 2013, 29 (3), 353-359.

135. Nakano, T.; Nakaoki, T. Coagulation size of freezable water in poly(vinyl alcohol) hydrogels formed by different freeze/thaw cycle periods. Polymer Journal 2011, 43 (11), 875-880

136. Tokarev, I.; Minko, S. Stimuli-responsive hydrogel thin films. Soft Matter 2009, 5 (3), 511-524

210

137. Puppi, D.; Chiellini, F.; Piras, A. M.; Chiellini, E. Polymeric materials for bone and cartilage repair. Progress in Polymer Science 2010, 35 (4), 403-440

138. Drury, J. L.; Mooney, D. J. Hydrogels for tissue engineering: scaffold design variables and applications. Biomaterials 2003, 24 (24), 4337-4351

139. Chu, L. Y.; Li, Y.; Zhu, J. H.; Chen, W. M. Negatively thermoresponsive membranes with functional gates driven by zipper-type hydrogen-bonding interactions. Angewandte Chemie-International Edition 2005, 44 (14), 2124-2127

140. Tanaka, T. Collapse of gels and critical endpoint. Physical Review Letters 1978, 40 (12), 820-823

141. Dusek, K.; Patterso.D. Transition in swollen polymer networks induced by intramolecular condensation. Journal of Polymer Science Part a-2-Polymer Physics 1968, 6 (7PA2), 1209-1216

142. Salehi, P.; Makhoul, G.; Roy, R.; Malhotra, M.; Mood, Z. A.; Daniel, S. J. Curcumin loaded NIPAAM/VP/PEG-A nanoparticles: physicochemical and chemopreventive properties. Journal of Biomaterials Science-Polymer Edition 2013, 24 (5), 574-588

143. Lowe, T. L.; Benhaddou, M.; Tenhu, H. Partially fluorinated thermally responsive latices of linear and crosslinked copolymers. Journal of Polymer Science Part B-Polymer Physics 1998, 36 (12), 2141-2152

144. Chen, C. W.; Akashi, M. Synthesis, characterization, and catalytic properties of colloidal platinum nanoparticles protected by poly(N-isopropylacrylamide). Langmuir 1997, 13 (24), 6465-6472

145. Plunkett, K. N.; Zhu, X.; Moore, J. S.; Leckband, D. E. PNIPAM chain collapse depends on the molecular weight and grafting density. Langmuir 2006, 22 (9), 4259-4266

146. Lee, S. G.; Pascal, T. A.; Koh, W.; Brunello, G. F.; Goddard, W. A.; Jang, S. S. Deswelling Mechanisms of Surface-Grafted Poly(NIPAAm) Brush: Molecular Dynamics Simulation Approach. Journal of Physical Chemistry C 2012, 116 (30), 15974-15985

147. Choi, B. C.; Choi, S.; Leckband, D. E. Poly(N-isopropyl acrylamide) Brush Topography: Dependence on Grafting Conditions and Temperature. Langmuir 2013, 29 (19), 5841-5850

211

148. Harmon, M. E.; Kuckling, D.; Frank, C. W. Photo-cross-linkable PNIPAAm copolymers. 2. Effects of constraint on temperature and pH-responsive hydrogel layers. Macromolecules 2003, 36 (1), 162-172

149. Harmon, M. E.; Jakob, T. A. M.; Knoll, W.; Frank, C. W. A surface plasmon resonance study of volume phase transitions in N-isopropylacrylamide gel films. Macromolecules 2002, 35 (15), 5999-6004

150. Jablonski, E. L.; Prabhu, V. M.; Sambasivan, S.; Fischer, D. A.; Lin, E. K.; Goldfarb, D. L.; Angelopoulos, M.; Ito, H. In Surface and bulk chemistry of chemically amplified photoresists: segregation in thin films and environmental stability issues, Conference on Advances in Resist Technology and Processing XXI, Santa Clara, CA, Feb 23-24, 2004; Spie-Int Soc Optical Engineering: Santa Clara, CA, 2004; pp 302-311.

151. Napolitano, S.; Wubbenhorst, M. The lifetime of the deviations from bulk behaviour in polymers confined at the nanoscale. Nature Communications 2011, 2, 260, 1-7

152. Seery, T. A. P.; Yassini, M.; Hogenesch, T. E.; Amis, E. J. Static and dynamic light-scattering characterization of solutions of hydrophobically associating fluorocarbon- containing polymers. Macromolecules 1992, 25 (18), 4784-4791

153. McCrackin, F. L.; Passaglia, E.; Stromberg, R. R.; Steinberg, H. L. Measurement of the thickness and refractive index of very thin films and the optical properties of surfaces by ellipsometry. Journal of Research of the National Institute of Standards and Technology 2001, 106 (3), 589-603.

154. Ramos, J. J. I.; Stahl, S.; Richter, R. P.; Moya, S. E. Water Content and Buildup of Poly(diallyldimethylammonium chloride)/Poly(sodium 4-styrenesulfonate) and Poly(allylamine hydrochloride)/Poly(sodium 4-styrenesulfonate) Polyelectrolyte Multilayers Studied by an in Situ Combination of a Quartz Crystal Microbalance with Dissipation Monitoring and Spectroscopic Ellipsometry. Macromolecules 2010, 43 (21), 9063-9070

155. Reimhult, E.; Hook, F.; Kasemo, B. Vesicle adsorption on SiO2 and TiO2: Dependence on vesicle size. Journal of Chemical Physics 2002, 117 (16), 7401-7404

156. Rodahl, M.; Hook, F.; Krozer, A.; Brzezinski, P.; Kasemo, B. Quartz-crystal microbalance setup for frequency and q-factor measurements in gaseous and liquid environments. Review of Scientific Instruments 1995, 66 (7), 3924-3930

157. Kanazawa, K. K.; Gordon, J. G. The oscillation frequency of a quartz resonator in contact with a liquid. Analytica Chimica Acta 1985, 175 (SEP), 99-105

212

158. Linstrom, P. J.; Mallard, W. G.; Eds, NIST Chemistry WebBook, NIST Standard Reference Database. http://webbook.nist.gov ed.; National Institute of Standards and Technology: Gaithersburg MD, 20899, 2013.

159. Sauerbrey, G. Verwendung von Schwingquarzen zur Wägung dünner Schichten und zur Mikrowägung. Zeitschrift für Physik 1959, 155 (2), 206-222

160. Kanazawa, K. K.; Gordon, J. G. Frequency of a quartz microbalance in contact with liquid. Analytical Chemistry 1985, 57 (8), 1770-1771

161. Kanazawa, K. K. Mechanical behaviour of films on the quartz microbalance. Faraday Discussions 1997, 107 (77-90), 77.

162. Cauchy, A.-L., Polytechnique, 1. 1821.

163. Shibayama, M.; Tanaka, T. Volume phase-transition and related phenomena of polymer gels. Advances in Polymer Science 1993, 109, 1-62.

164. Zhong, Q.; Metwalli, E.; Kaune, G.; Rawolle, M.; Bivigou-Koumba, A. M.; Laschewsky, A.; Papadakis, C. M.; Cubitt, R.; Mueller-Buschbaum, P. Switching kinetics of thin thermo-responsive hydrogel films of poly(monomethoxy-diethyleneglycol- acrylate) probed with in situ neutron reflectivity. Soft Matter 2012, 8 (19),

165. Wang, C. H.; Mukherjee, S.; Islam, A. K. M. M.; Yang, Y. W.; Mukherjee, M. Role of Interfacial Interaction in Orientation of Poly(N-isopropylacrylamide) Chains on Silicon Substrate. Macromolecules 2011, 44 (14), 5750-5757

166. Lutz, J.-F.; Akdemir, O.; Hoth, A. Point by point comparison of two thermosensitive polymers exhibiting a similar LCST: Is the age of poly(NIPAM) over? Journal of the American Chemical Society 2006, 128 (40), 13046-13047

167. Vogt, B. D.; Soles, C. L.; Jones, R. L.; Wang, C. Y.; Lin, E. K.; Wu, W. L.; Satija, S. K.; Goldfarb, D. L.; Angelopoulos, M. Interfacial effects on moisture absorption in thin polymer films. Langmuir 2004, 20 (13), 5285-5290

168. Voinova, M. V.; Jonson, M.; Kasemo, B. 'Missing mass' effect in biosensor's QCM applications. Biosensors & Bioelectronics 2002, 17 (10), 835-841

169. Eisele, N. B.; Andersson, F. I.; Frey, S.; Richter, R. P. Viscoelasticity of Thin Biomolecular Films: A Case Study on Nucleoporin Phenylalanine-Glycine Repeats Grafted to a Histidine-Tag Capturing QCM-D Sensor. Biomacromolecules 2012, 13 (8), 2322-2332

213

170. Vogt, B. D.; Lin, E. K.; Wu, W. L.; White, C. C. Effect of film thickness on the validity of the Sauerbrey equation for hydrated polyelectrolyte films. Journal of Physical Chemistry B 2004, 108 (34), 12685-12690

171. Lee, S. W.; Lee, D. Integrated Study of Water Sorption/Desorption Behavior of Weak Polyelectrolyte Layer-by-Layer Films. Macromolecules 2013, 46 (7), 2793-2799

172. Webber, R. E.; Creton, C.; Brown, H. R.; Gong, J. P. Large strain hysteresis and mullins effect of tough double-network hydrogels. Macromolecules 2007, 40 (8), 2919- 2927

173. Nakajima, T.; Kurokawa, T.; Ahmed, S.; Wu, W. L.; Gong, J. P. Characterization of internal fracture process of double network hydrogels under uniaxial elongation. Soft Matter 2013, 9 (6), 1955-1966

174. Tominaga, T.; Tirumala, V. R.; Lin, E. K.; Gong, J. P.; Furukawa, H.; Osada, Y.; Wu, W.-l. The molecular origin of enhanced toughness in double-network hydrogels: A neutron scattering study. Polymer 2007, 48 (26), 7449-7454

175. Hao, J.; Weiss, R. A. Mechanically Tough, Thermally Activated Shape Memory Hydrogels. ACS Macro Letters 2013, 2 (1), 86-89

176. Myung, D.; Waters, D.; Wiseman, M.; Duhamel, P. E.; Noolandi, J.; Ta, C. N.; Frank, C. W. Progress in the development of interpenetrating polymer network hydrogels. Polymers for Advanced Technologies 2008, 19 (6), 647-657

177. Nishi, K.; Shibayama, M. Probe-SAXS on hydrogels under elongation. Soft Matter 2016, 12 (24), 5334-5339

178. Wiener, C. G.; Tyagi, M.; Liu, Y.; Weiss, R. A.; Vogt, B. D. Supramolecular Hydrophobic Aggregates in Hydrogels Partially Inhibit Ice Formation. Journal of Physical Chemistry B 2016, 120 (24), 5543-5552

179. Hao, J. K.; Weiss, R. A. Tuning the Viscoelastic Behavior of Hybrid Hydrogels Composed of a Physical and a Chemical Network by the Addition of an Organic Solvent. Macromolecules 2016, 49 (17), 6687-6693

180. Xie, X. Y.; HogenEsch, T. E. Copolymers of N,N-dimethylacrylamide and 2-(N- ethylperfluorooctanesulfonamido)ethyl acrylate in aqueous media and in bulk. Synthesis and properties. Macromolecules 1996, 29 (5), 1734-1745

214

181. Stafford, C. M.; Guo, S.; Harrison, C.; Chiang, M. Y. M. Combinatorial and high- throughput measurements of the modulus of thin polymer films. Review of Scientific Instruments 2005, 76 (6), 062207, 1-5 182. Glinka, C. J.; Barker, J. G.; Hammouda, B.; Krueger, S.; Moyer, J. J.; Orts, W. J. The 30 m small-angle neutron scattering instruments at the National Institute of Standards and Technology. Journal of Applied Crystallography 1998, 31, 430-445

183. Kline, S. R. Reduction and analysis of SANS and USANS data using IGOR Pro. Journal of Applied Crystallography 2006, 39, 895-900

184. Hammouda, B., The SANS Toolbox. 2010; p 692.

185. Hirai, Y.; Terashima, T.; Takenaka, M.; Sawamoto, M. Precision Self-Assembly of Amphiphilic Random Copolymers into Uniform and Self-Sorting Nanocompartments in Water. Macromolecules 2016, 49 (14), 5084-5091

186. Wignall, G. D.; Melnichenko, Y. B. Recent applications of small-angle neutron scattering in strongly interacting soft condensed matter. Reports on Progress in Physics 2005, 68 (8), 1761-1810

187. Wu, B.; Li, X.; Do, C.; Kim, T. H.; Shew, C. Y.; Liu, Y.; Yang, J.; Hong, K. L.; Porcar, L.; Chen, C. Y.; Liu, E. L.; Smith, G. S.; Herwig, K. W.; Chen, W. R. Spatial distribution of intra-molecular water and polymeric components in polyelectrolyte dendrimers revealed by small angle scattering investigations. Journal of Chemical Physics 2011, 135 (14), 1-9

188. Nystrom, B.; Walderhaug, H. Dynamic viscoelasticity of an aqueous system of a poly(ethylene oxide)-poly(propylene oxide)-poly(ethylene oxide) triblock copolymer during gelation. Journal of Physical Chemistry 1996, 100 (13), 5433-5439

189. Schiessel, H.; Metzler, R.; Blumen, A.; Nonnenmacher, T. F. Generalized viscoelastic models: Their fractional equations with solutions. J. Phys. A-Math. Gen. 1995, 28 (23), 6567-6584

190. Shibayama, M.; Kawakubo, K.; Ikkai, F.; Imai, M. Small-angle neutron scattering study on charged gels in deformed state. Macromolecules 1998, 31 (8), 2586-2592

191. Mendes, E.; Schosseler, F.; Isel, F.; Boue, F.; Bastide, J.; Candau, S. J. A SANS study of uniaxially elongated polyelectrolyte gels. Europhysics Letters 1995, 32 (3), 273- 278

215

192. Mendes, E.; Oeser, R.; Hayes, C.; Boue, F.; Bastide, J. Small-angle neutron scattering study of swollen elongated gels: Butterfly patterns. Macromolecules 1996, 29 (17), 5574-5584

193. Oka, Y. I.; Sakohara, S.; Gotoh, T.; Iizawa, T.; Okamoto, K.; Doi, H. Measurements of mechanical properties on a swollen hydrogel by a tension test method. Polymer Journal 2004, 36 (1), 59-63

194. Kim, P.; Wong, T. S.; Alvarenga, J.; Kreder, M. J.; Adorno-Martinez, W. E.; Aizenberg, J. Liquid-Infused Nanostructured Surfaces with Extreme Anti-Ice and Anti- Frost Performance. Acs Nano 2012, 6 (8), 6569-6577

195. James, C.; Purnell, G.; James, S. J. A Review of Novel and Innovative Food Freezing Technologies. Food and Bioprocess Technology 2015, 8 (8), 1616-1634

196. Koza, M. M.; Geil, B.; Winkel, K.; Kohler, C.; Czeschka, F.; Scheuermann, M.; Schober, H.; Hansen, T. Nature of amorphous polymorphism of water. Physical Review Letters 2005, 94 (12), 125506, 1-4

197. Lobban, C.; Finney, J. L.; Kuhs, W. F. The structure of a new phase of ice. Nature 1998, 391 (6664), 268-270.

198. Loubeyre, P.; LeToullec, R.; Wolanin, E.; Hanfland, M.; Husermann, D. Modulated phases and proton centring in ice observed by X-ray diffraction up to 170 GPa. Nature 1999, 397 (6719), 503-506

199. Mochizuki, K.; Koga, K. Solid-liquid critical behavior of water in nanopores. Proceedings of the National Academy of Sciences of the United States of America 2015, 112 (27), 8221-8226

200. Bai, J. E.; Wang, J.; Zeng, X. C. Multiwalled ice helixes and ice nanotubes. Proceedings of the National Academy of Sciences of the United States of America 2006, 103 (52), 19664-19667

201. Koga, K.; Gao, G. T.; Tanaka, H.; Zeng, X. C. Formation of ordered ice nanotubes inside carbon nanotubes. Nature 2001, 412 (6849), 802-805

202. Koga, K.; Tanaka, H.; Zeng, X. C. First-order transition in confined water between high-density liquid and low-density amorphous phases. Nature 2000, 408 (6812), 564-567

216

203. Chen, L. B.; DeVries, A. L.; Cheng, C. H. C. Convergent evolution of antifreeze glycoproteins in Antarctic notothenioid fish and Arctic cod. Proceedings of the National Academy of Sciences of the United States of America 1997, 94 (8), 3817-3822

204. Peppas, N. A.; Hilt, J. Z.; Khademhosseini, A.; Langer, R. Hydrogels in biology and medicine: From molecular principles to bionanotechnology. Advanced Materials 2006, 18 (11), 1345-1360

205. Pathmanathan, K.; Johari, G. P. Relaxation and crystallization of water in a hydrogel. Journal of the Chemical Society-Faraday Transactions 1994, 90 (8), 1143- 1148

206. Azuah, R. T.; Kneller, L. R.; Qiu, Y. M.; Tregenna-Piggott, P. L. W.; Brown, C. M.; Copley, J. R. D.; Dimeo, R. M. DAVE: A Comprehensive Software Suite for the Reduction, Visualization, and Analysis of Low Energy Neutron Spectroscopic Data. Journal of Research of the National Institute of Standards and Technology 2009, 114 (6), 341-358

207. Takaiwa, D.; Hatano, I.; Koga, K.; Tanaka, H. Phase diagram of water in carbon nanotubes. Proceedings of the National Academy of Sciences of the United States of America 2008, 105 (1), 39-43

208. Beddoe, R. E.; Setzer, M. J. A low-temperature dsc investigation of hardened cement paste subjected to chloride action. Cement and Concrete Research 1988, 18 (2), 249-256

209. Suzuki, Y.; Steinhart, M.; Graf, R.; Butt, H. J.; Floudas, G. Dynamics of Ice/Water Confined in Nanoporous Alumina. Journal of Physical Chemistry B 2015, 119 (46), 14814-14820

210. Suzuki, Y.; Duran, H.; Steinhart, M.; Kappl, M.; Butt, H. J.; Floudas, G. Homogeneous Nucleation of Predominantly Cubic Ice Confined in Nanoporous Alumina. Nano Letters 2015, 15 (3), 1987-1992

211. Suzuki, Y.; Steinhart, M.; Butt, H. J.; Floudas, G. Kinetics of Ice Nucleation Confined in Nanoporous Alumina. Journal of Physical Chemistry B 2015, 119 (35), 11960-11966

212. Hao, J.; Weiss, R. A. Viscoelastic and Mechanical Behavior of Hydrophobically Modified Hydrogels. Macromolecules 2011, 44 (23), 9390-9398

213. Bartlett, P.; Ottewill, R. H. A neutron-scattering study of the structure of a bimodal colloidal crystal. Journal of Chemical Physics 1992, 96 (4), 3306-3318

217

214. Griffith, W. L.; Triolo, R.; Compere, A. L. Analytical scattering function of a polydisperse percus-yevick fluid with schulz-(gamma-) distributed diameters. Physical Review A 1987, 35 (5), 2200-2206

215. Percus, J. K.; Yevick, G. J. Analysis of classical statistical mechanics by means of collective coordinates. Physical Review 1958, 110 (1), 1-13.

216. Copley, J. R. D.; Cook, J. C. The Disk Chopper Spectrometer at NIST: a new instrument for quasielastic neutron scattering studies. Chemical Physics 2003, 292 (2-3), 477-485

217. Teixeira, J.; Bellissentfunel, M. C.; Chen, S. H.; Dianoux, A. J. Experimental- determination of the nature of diffusive motions of water-molecules at low-temperatures. Physical Review A 1985, 31 (3), 1913-1917

218. Volino, F.; Dianoux, A. J. Neutron incoherent-scattering law for diffusion in a potential of spherical-symmetry - general formalism and application to diffusion inside a sphere. Molecular Physics 1980, 41 (2), 271-279

219. Gillen, K. T.; Douglass, D. C.; Hoch, J. R. Self-diffusion in liquid water to -31 degrees C. Journal of Chemical Physics 1972, 57 (12), 5117-5119

220. Nogales, A.; Ezquerra, T. A.; Batallan, F.; Frick, B.; Lopez-Cabarcos, E.; Balta- Calleja, F. J. Restricted dynamics in poly(ether ether ketone) as revealed by incoherent quasielastic neutron scattering and broad-band dielectric spectroscopy. Macromolecules 1999, 32 (7), 2301-2308

221. Hall, P. L.; Ross, D. K. Incoherent neutron-scattering functions for random jump diffusion in bounded and infinite media. Molecular Physics 1981, 42 (3), 673-682

222. Faraone, A.; Fratini, E.; Garai, S.; Muller, A.; Tyagi, M.; Jenkins, T.; Mamontov, E.; Paul, R. L.; Copley, J. R. D.; Baglioni, P. Incoherent Quasielastic Neutron Scattering Study of the Relaxation Dynamics in Molybdenum-Oxide Keplerate-Type Nanocages. Journal of Physical Chemistry C 2014, 118 (24), 13300-13312

223. Meyer, A.; Dimeo, R. M.; Gehring, P. M.; Neumann, D. A. The high-flux backscattering spectrometer at the NIST Center for Neutron Research. Review of Scientific Instruments 2003, 74 (5), 2759-2777

224. Soles, C. L.; Douglas, J. F.; Wu, W. L.; Dimeo, R. M. Incoherent neutron scattering and the dynamics of confined polycarbonate films. Physical Review Letters 2002, 88 (3), 037401, 1-4

218

225. Sauerbrey, G. Verwendung von Schwingquarzen zur Wägung dünner Schichten und zur Mikrowägung. Z. Phys. 1959, 155 (2), 206-222.

226. Leskela, M.; Ritala, M. Atomic layer deposition (ALD): from precursors to thin film structures. Thin Solid Films 2002, 409 (1), 138-146

227. George, S. M. Atomic Layer Deposition: An Overview. Chemical Reviews 2010, 110 (1), 111-131

228. Kanazawa, K. K.; Melroy, O. R. The quartz resonator - electrochemical applications. Ibm Journal of Research and Development 1993, 37 (2), 157-171.

229. Marx, K. A. Quartz crystal microbalance: A useful tool for studying thin polymer films and complex biomolecular systems at the solution-surface interface. Biomacromolecules 2003, 4 (5), 1099-1120

230. Reed, C. E.; Kanazawa, K. K.; Kaufman, J. H. Physical description of a viscoelastically loaded at-cut quartz resonator. Journal of Applied Physics 1990, 68 (5), 1993-2001

231. Nunalee, F. N.; Shull, K. R.; Lee, B. P.; Messersmith, P. B. Quartz crystal microbalance studies of polymer gels and solutions in liquid environments. Analytical Chemistry 2006, 78 (4), 1158-1166

232. Kanazawa, K. K. QCM physical theory extended to include liquid immersion. Chemical Sensors Iv, Proceedings of the Symposium 1999, 99 (23), 306-313.

233. Johannsmann, D.; Mathauer, K.; Wegner, G.; Knoll, W. Viscoelastic properties of thin-films probed with a quartz-crystal resonator. Physical Review B 1992, 46 (12), 7808- 7815

234. Wolff, O.; Seydel, E.; Johannsmann, D. Viscoelastic properties of thin films studied with quartz crystal resonators. Faraday Discussions 1997, 107, 91-104

235. Claesson, M.; Cho, N.-J.; Frank, C. W.; Andersson, M. Vesicle Adsorption on Mesoporous Silica and Titania. Langmuir 2010, 26 (22), 16630-16633

236. Claesson, M.; Ahmadi, A.; Fathali, H. M.; Andersson, M. Improved QCM-D signal-to-noise ratio using mesoporous silica and titania. Sensors and Actuators B- Chemical 2012, 166, 526-534

219

237. Forrest, J. A.; Svanberg, C.; Revesz, K.; Rodahl, M.; Torell, L. M.; Kasemo, B. Relaxation dynamics in ultrathin polymer films. Physical Review E 1998, 58 (2), R1226- R1229

238. Reviakine, I.; Johannsmann, D.; Richter, R. P. Hearing What You Cannot See and Visualizing What You Hear: Interpreting Quartz Crystal Microbalance Data from Solvated Interfaces. Analytical Chemistry 2011, 83 (23), 8838-8848

239. Hovgaard, M. B.; Dong, M. D.; Otzen, D. E.; Besenbacher, F. Quartz crystal microbalance studies of multilayer glucagon fibrillation at the solid-liquid interface. Biophysical Journal 2007, 93 (6), 2162-2169

240. Kanazawa, K. K.; Melroy, O. R. The quartz resonator: Electrochemical applications. IBM Journal of Research and Development 1993, 37 (2), 157-171

241. Lucklum, R.; Hauptmann, P.; Ieee. Thin film shear modulus determination with quartz crystal resonators: A review. Proceedings of the 2001 Ieee International Frequency Control Symposium & Pda Exhibition 2001, 408-418

242. Granstaff, V. E.; Martin, S. J. Characterization of a thickness-shear mode quartz resonator with multiple nonpiezoelectric layers. Journal of Applied Physics 1994, 75 (3), 1319-1329

243. Martin, S. J.; Granstaff, V. E.; Frye, G. C. Characterization of a quartz crystal microbalance with simultaneous mass and liquid loading. Analytical Chemistry 1991, 63 (20), 2272-2281

244. Martin, S. J.; Frye, G. C.; Senturia, S. D. Dynamics and response of polymer- coated surface-acoustic-wave devices - effect of viscoelastic properties and film resonance. Analytical Chemistry 1994, 66 (14), 2201-2219.

245. Chen, W. L.; Shull, K. R.; Papatheodorou, T.; Styrkas, D. A.; Keddie, J. L. Equilibrium swelling of hydrophilic polyacrylates in humid environments. Macromolecules 1999, 32 (1), 136-144

246. Wiener, C. G.; Weiss, R. A.; Vogt, B. D. Overcoming confinement limited swelling in hydrogel thin films using supramolecular interactions. Soft matter 2014, 10 (35), 6705, 1-12

247. Wiener, C. G.; Weiss, R. A.; Vogt, B. D. Overcoming confinement limited swelling in hydrogel thin films using supramolecular interactions. Soft Matter 2014, 10 (35), 6705-6712

220

248. Johannsmann, D. Viscoelastic, mechanical, and dielectric measurements on complex samples with the quartz crystal microbalance. Physical Chemistry Chemical Physics 2008, 10 (31), 4516-4534

249. Cho, N.-J.; D'Amour, J. N.; Stalgren, J.; Knoll, W.; Kanazawa, K.; Frank, C. W. Quartz resonator signatures under Newtonian liquid loading for initial instrument check. Journal of Colloid and Interface Science 2007, 315 (1), 248-254

250. Dahlin, A. B.; Jonsson, P.; Jonsson, M. P.; Schmid, E.; Zhou, Y.; Hook, F. Synchronized Quartz Crystal Microbalance and Nanoplasmonic Sensing of Biomolecular Recognition Reactions. ACS Nano 2008, 2 (10), 2174-2182

251. Sundh, M.; Svedhem, S.; Sutherland, D. S. Formation of Supported Lipid Bilayers at Surfaces with Controlled Curvatures: Influence of Lipid Charge. Journal of Physical Chemistry B 2011, 115 (24), 7838-7848

252. Keller, C. A.; Kasemo, B. Surface specific kinetics of lipid vesicle adsorption measured with a quartz crystal microbalance. Biophysical Journal 1998, 75 (3), 1397- 1402.

253. Zhu, D. M.; Fang, J. J.; Wu, B.; Du, X. B. Viscoelastic response and profile of adsorbed molecules probed by quartz crystal microbalance. Physical Review E 2008, 77 (3), 1-7

254. Voinova, M. V.; Jonson, M.; Kasemo, B. 'Missing mass' effect in biosensor's QCM applications. Biosensors & Bioelectronics 2002, 17 (10), 835-841

255. Fakhraai, Z.; Forrest, J. A. Measuring the surface dynamics of glassy polymers. Science 2008, 319 (5863), 600-604 DOI: 10.1126/science.1151205.

256. Torres, J. M.; Stafford, C. M.; Vogt, B. D. Elastic Modulus of Amorphous Polymer Thin Films: Relationship to the Glass Transition Temperature. ACS Nano 2009, 3 (9), 2677-2685

257. Wingqvist, G.; Anderson, H.; Lennartsson, C.; Weissbach, T.; Yantchev, V.; Spetz, A. L. On the applicability of high frequency acoustic shear mode biosensing in view of thickness limitations set by the film resonance. Biosensors & Bioelectronics 2009, 24 (11), 3387-3390

221

APPENDIX 1. MATHEMATICA CODE FOR QCM MODEL

The model equations derived by White and Schrag (J.Chem. Phys., 1999, 111, 11192-

11206)60 had 3 sub-equations missing in the paper. These missing equations defined the variables: c, d, and A. These equations are included in the Mathematica code and also highlighted in gray below with all of the equations utilized to calculate the impedance.

One additional issue with the prior publication60 is that the trigonometric functions for the imaginary and real parts of impedance are reversed (these are highlighted yellow in the

equations below). The imaginary part of impedance ( ) (Xreduced, and X) is a cosine

function which relates to frequency and ( ) (Rreduced) is a sine function relates to dissipation.

Additionally to facilitate the incorporation of these equations into Mathematica, some changes in notation were made, as listed below:

Reduced Real part of impedance= Rreduced

Imaginary part of impedance = X

Reduced Imaginary part of impedance = Xreduced

Full impedance = Z

f denotes frequency

Δf is change in frequency measured by the sensor.

The units for the input parameters are as follows:

Density: g/cm3

Thickness: cm

222

Viscosity: centipoises

Frequency: Hz

The phase angle ϕ is bounded by the pure elastic (π/2) and pure viscous (0) cases.

223

224

APPENDIX 2. LIST OF PUBLICATIONS

“Viscoelastic characterization of thin thermoresponsive hydrogel coatings using quartz crystal microbalance: From the Sauerbrey limit through film resonance condition” Wiener, C.G.; Weiss, R.A.; White, C.C.; Vogt, B.D.; Preparing for submission to Langmuir, 2017

“Nitrogen Doping to alter pore hydrophilicity in mesoporous carbons and the induced water supercooling effects” Wiener, C.G.; Qiang, Z.; Xia, Y.; Tyagi, M.; Weiss, R.A.; Vogt, B.D.; Preparing for submission to Journal of Physical Chemistry Part B, 2017

“Nanostructure Evolution during Relaxation from a Large Step Strain in a Supramolecular Copolymer-Based Hydrogel: A SANS Investigation,” Wiener, C.G.; Wang, C.; Liu, Y.; Weiss, R.A.; Vogt, B.D.; Macromolecules 2017, 50 (4), 1672-1680

“Supramolecular hydrophobic aggregates in hydrogels partially inhibit ice formation,” Wiener, C.G.; Tyagi, M.; Liu, Y.; Weiss, R.A.; Vogt, B.D.; Journal of Physical Chemistry Part B. 2016, 120, 5543

“Overcoming confinement limited swelling in hydrogel thin films using supramolecular interactions,” Wiener, C.G.; Vogt, B.D.; Weiss, R.A.; Soft Matter, 2014, 10, 6705

“Modulation of the Mechanical Properties of Hydrophobically Modified Supramolecular Hydrogels by Surfactant-Driven Structural Rearrangement,” Wang, C.; Wiener, C.G.; Cheng, Z.; Vogt, B.D.; Weiss, R.A.; Macromolecules, 2016, 49 (23), 9228-9238

“Tough stretchable physically-crosslinked electrospun hydrogel fiber mats,” Yang, Y.; Wang, C.; Wiener, C.G.; Hao, J.; Shatas, S.; Weiss, R.A.; Vogt, B.D.; ACS Applied Materials and Interfaces, 2016, 8, 22774−22779

“Control of mesh size and modulus by kinetically dependent cross-linking in hydrogels,” Zander, Z.K.; Hua, G.; Wiener, C.G.; Vogt, B.D.; Becker, M.L.; Advanced Materials, 2015, 27 40, 6283

“Response of swelling behavior of weak branched poly(ethylene imine)/poly(acrylic acid) polyelectrolyte multilayers to thermal treatment,” Gu,Y.; Weinheimer, E. K.; Ji, X.; Wiener, C.G.; Zacharia, N.S.; Langmuir, 2016, 32 (24), 6020

225

“Understanding the decreased segmental dynamics of supported thin polymer films reported by incoherent neutron scattering,” Ye,C.; Wiener, C.G.; Tyagi, M.; Uhrig, D.; Orski, S.V.; Soles, C.L.; Vogt, B.D.; Simmons, D.S.; Macromolecules, 2015, 48, 801

“Large-scale solvent driven actuation of polyelectrolyte multilayers based on modulation of dynamic secondary interactions,” Gu, Y.; Huang, X.; Wiener, C.G.; Vogt, B.D.; Zacharia, N.S.; ACS Applied Material Interfaces, 2015, 7, 1848

226

APPENDIX 3. REPRINT PERMISSIONS

Reprinted (adapted) with permission from (Macromolecules, Vol. 36, No. 15, 2003). Copyright (2003) American Chemical Society.

Reproduced (“Adapted” or “in part”) from (Soft Matter, 2010, 6, 2583–2590) with permission of The Royal Society of Chemistry.

Reprinted (adapted) with permission from (Macromolecules 2016, 49, 8980-8987). Copyright (2016) American Chemical Society.

Reprinted (adapted) with permission from (Macromolecules 2013, 46, 6203-6208). Copyright (2013) American Chemical Society.

Reprinted by permission from Macmillan Publishers Ltd: [Nature Materials] (Nature Materials, Vol 12, October 2013), copyright (2013).

Reprinted (adapted) with permission from (Macromolecules, Vol. 42, No. 6, 2009). Copyright (2009) American Chemical Society.

Reprinted from Carbohydrate Polymers, 101, Guan Y., et al., High strength of hemicelluloses based hydrogels by freeze/thaw technique, 272-280, Copyright (2014), with permission from Elsevier.

Reproduced (“Adapted” or “in part”) from (Chem. Commun., 2012, 48, 9302–9304) with permission of The Royal Society of Chemistry.

Reprinted (adapted) with permission from (ACS Macro Lett. 2014, 3, 520-523). Copyright (2014) American Chemical Society.

Reprinted (adapted) with permission from (Macromolecules, Vol. 40, No. 22, 2007). Copyright (2007) American Chemical Society

Reprinted (adapted) with permission from (Macromolecules, Vol. 37, No. 3, 2004). Copyright (2004) American Chemical Society.

Reprinted from [The Journal of Chemical Physics 101, 10003 (1994)], with the permission of AIP Publishing.

227

Reprinted from Agrawal Acta Biomaterialia, 9, Agrawal A., Rahbar N., Clavert PD., Strong fiber-reinforced hydrogel, 5313-5318, Copyright (2013), with permission from Elsevier.

Reprinted by permission from Macmillan Publishers Ltd: [Nature Materials] (Nature, Vol 489, September (2012) 133-136), Copyright (2012).

Reprinted (adapted) with permission from (Macromolecules 49.19 (2016): 7340-7349). Copyright (2016) American Chemical Society.

228

APPENDIX 4. FINANCIAL AND GRANT SUPPORT

This work was partially financially supported by the Civil, Mechanical and

Manufacturing Innovation (CMMI) Division in the Directorate for Engineering of the National

Science Foundation, grant. CMMI-1300212 and the Chemical, Bioengineering, Environmental and Transport Systems (CBET) Division in the Directorate for Engineering of the National

Science Foundation, grant CBET-1606685. This work utilized facilities supported in part by the

National Science Foundation under Agreement No. DMR-1508249. We acknowledge the support of the National Institute of Standards and Technology, U.S. Department of Commerce, in providing the neutron research facilities used in this work.

229